Previous Page

RELIABILITY CONCEPTS IN RESERVOIR DESIGN 259

21. Laurenson, E.M. (1962). Hydrograph Synthesis by Runoff St ranges between zero and a maximum value c imposed
Routing. Report 66, Water research Laboratory, The Uni- by the reservoir size, which corresponds to the level of
versity of New South Wales, Manly Vale, Kennsington, New the spillway crest (or some specified level above it when
South Wales, Australia. the sluice gates are constructed over the spillway). During
22. Laurenson, E.M. (1964). A catchment storage model for runoff floods, excess water is routed through the spillway, which
routing. J. Hydrology 2: 141–163. causes temporary storage above the normal limit c. This
23. Nash, J.E. (1959). A note on the Muskingum flood routing is known as flood control storage. Water storage below
method. J. Geophys Res. 64: 1053–1056. the minimum level, known as dead or inactive storage,
24. Koussis, A.D. (1978). Theoretical estimation of flood routing serves two main purposes: It provides volume for sediment
parameters. J. Hydraul. Div., ASCE 104(HY1): 109–115. accumulation and environmental protection, as it protects
25. Koussis, A.D. (1980). Comparison of Muskingum difference the habitat of the reservoir during dry periods by hindering
schemes. J. Hydraul. Div., ASCE 106(HY6): 925–929. complete emptying. Associated with the last function is
also conservation of the quality of the landscape. This
article is focused on the design of the active storage of
RELIABILITY CONCEPTS IN RESERVOIR DESIGN a reservoir; some notes on additional storage zones are
contained in the last section of the article.
DEMETRIS KOUTSOYIANNIS
National Technical University Net Inflow Xt . It is the algebraic sum of cumulative
Athens, Greece
inflows to the reservoir from time t − 1 to time t, minus the
losses during the same time period. Inflows include runoff
from the catchment upstream of the reservoir (typically,
INTRODUCTION the main component of inflows), rainfall on the surface
area of the reservoir, and, possibly, water artificially
The reliability of a system is defined as the probability
conveyed from other sources (e.g., interbasin transfers
that the system will perform the required function for a
through tunnels or pipelines). Losses include evaporation
specified period of time under stated conditions. Reliability
from the surface area of the reservoir, possibly seepage to
is the complement of the probability of failure (or risk), the
groundwater, and leakage under or through the dam.
probability that the ‘‘loading’’ will exceed the ‘‘capacity.’’
Denoting α as the reliability, β the probability of failure, Water Demand, δt . It is the sum of all water
and P[ω] the probability of an event ω, the mathematical requirements for the different water uses served by the
expression of this definition is reservoir for the time period (t − 1, t). The demand may
vary with time (e.g., due to seasonal agricultural demand
α := P[L(t) < C(t); t ∈ ] =: 1 − β (1) or due to some rule, usually based on the quantity of water
in the reservoir).
where L(t) and C(t) represent the loading and capacity,
respectively, at time t, within a certain time period  Release, Rt . Also known as draft, withdrawal, or
(e.g., a year). The failure of a system can be classified as outflow, it is the actual amount of water taken from the
structural failure or performance failure. Structural failure reservoir to satisfy water demand during the time period
involves damage of the structure or facility, hindering (t − 1, t). When there is a sufficient amount of water in the
its ability to function as desired in the future, whereas reservoir, Rt equals demand δt ; otherwise Rt < δt .
performance failure does not necessarily involve structural
damage but rather inability of the system to perform as Spill, Wt . It is the excess water that, during times of
desired at some time within the period of interest, which floods and simultaneously high reservoir storage, cannot
results in temporary unfavorable consequences. be stored in the reservoir due to the upper reservoir storage
limit c.
RESERVOIR DYNAMICS Reservoir dynamics are easily expressed by the law
of mass conservation, or equivalently, the water balance
A reservoir’s function is to regulate natural inflows, equation. Considering that St is limited between 0 and c,
which vary irregularly, to provide outflows at a more the water balance equation is easily formulated as
regular rate that is determined by water demand for
one or more uses (water supply, irrigation, hydropower), St = max(0, min(St−1 + Xt − δt , c)) (2)
temporarily storing the surplus, when inflows exceed
In addition, the release is determined as
outflows. Reservoir dynamics are more conveniently
expressed in discrete rather than continuous time. Rt = min(St−1 + Xt , δt ) (3)
The quantities that are necessary to describe dynamics
are the following: and the spill as

Storage St . More precisely known as active storage, it is Wt = St−1 − St + Xt − Rt = max(0, St−1 + Xt − δt − c) (4)
the volume of water stored, at time t, above the minimum
level, which is determined either technically (i.e., as the Equations 2–4 apply when the inflow and withdrawal
level of the lowest valve of off-take) or legally by a decree occur at constant rates throughout the period (t, t −
imposing rules for a reservoir’s operation. Active storage 1)—this could be called the ‘‘steady’’ model. A simple
260 RELIABILITY CONCEPTS IN RESERVOIR DESIGN

modification of the equations allows for the case where The purpose of the sum in Equation 8 (or 9) is to count the
the inflow (or withdrawal) is highly seasonal and can (in periods when storage is not zero (or release rt equals
the limit) be modeled as a sudden occurrence; this could the demand δt ). Thus, reliability is expressed as the
be called the ‘‘sudden’’ model. These two models, called proportion of time steps when the system performs as
‘‘simultaneous’’ and ‘‘staggered’’ by Pegram (1), bound all desired. For this reason, α has also been termed time-
the behavior observed in real reservoirs. based reliability. Here, it must be observed that although
Equations 5–7 are mathematically equivalent to each
DEFINITION OF RELIABILITY APPLIED TO A RESERVOIR other, as are Equations 8 and 9, when applied to historical
time series, they may result in different estimates. For
Now, the previously stated general definition of reliability example, a city’s water supply may not be allowed to
[also known as dependability; e.g., (2) p. 312] can be empty, as restrictions on releases are applied before this
applied to a reservoir. It is observed that the failure of situation is reached. In such a case, Equation 8 may result
a reservoir’s function is a performance failure, a failure to in the erroneous estimate α = 1, whereas Equation 9 will
meet the water demand. At time t, the loading is the water estimate the reliability correctly if the desired demand,
demand δt , and the capacity is the sum of St−1 (storage before restrictions, is entered into the calculations. Thus,
at time t − 1) and Xt (inflow from time t − 1 to t). Thus, Equation 9 is preferable when dealing with historical time
application of Equation 1 yields series, but both are equivalent in synthetic simulations;
moreover, application of Equation 8 is faster as it does not
α = 1 − β = P[δt < St−1 + Xt ] (5) require simulating releases at all (only Equation 2 needs
to be applied).
Considering Equations 2 and 3, the following equivalent The stationarity assumption that was inherent in the
and more convenient expressions are found: previous analysis is satisfactory when the time step is a
year (either calendar or hydrologic). However, the annual
α = P[St > 0], β = P[St = 0] (6) time step is usually too large and hides the variation of
both inflows and demand within a year, which may result
and in a failure some time within the year that is recovered at
α = P[Rt = δt ], β = P[Rt < δt ] (7) the end of the year. Therefore, a smaller time step (e.g.,
monthly) is usually chosen, so that one year corresponds
In this context, the demand δt is regarded as to k > 1 (e.g., 12) time steps. On this finer timescale, all
a known quantity at any time instant t. All other processes depend on the time step in a periodic manner,
quantities involved, namely, St , Xt , Rt , and Wt , are that is, they are cyclostationary. Yet, the reliability and
regarded as random variables. Given the storage capacity failure probability are usually expressed on the annual
c, the demand δt , and the probability distribution and scale, in which stationarity is redeemed. To shift from
autocorrelation functions of the input Xt , in theory, the the finer scale to the annual scale, the rule adopted is
probability distributions of output variables St , Rt , and that a failure occurring in one or more finer scale time
Wt can be determined in terms of that of Xt ; this, steps is regarded as a failure for the year. Using this rule,
however, is not an easy task due to the nonlinearity of the Equation 6 becomes
dynamics expressed in Equations 2–4. Theoretically, once
! # ! k #
the distribution function of St or Rt has been determined, "
k $
the reliability α is obtained directly from Equation 6 or 7, α =P (St−i > 0) , β = P St−i = 0 (10)
respectively. However, although the theoretically based i=1 i=1
calculations unavoidably involve complete knowledge of
where the symbols α and β were used instead of α
the distribution function of St and Rt , the reliability α can
and β to distinguish
% from time-based reliability, whereas
be determined in an alternative, much simpler, manner.
the symbol ‘ ’ indicates that all & of the following events
Under the assumptions of stationarity and ergodicity,
should occur simultaneously and ‘ ’ indicates that any of
α can be estimated from a historical (for an existing
the following events should occur. Similarly, Equation 7
reservoir) or synthesized (via simulation) time series
becomes
of storage st or release rt of adequate length n. (Here
lower case symbols were used for values of the random ! k # ! k #
" $
variables St and Rt .) Specifically, the estimate of α based α =P (Rt−i = δt−i ) , β = P Rt−i < δt−i (11)
on Equation 6 is i=1 i=1

1 or alternatively,
n
α= [1 − U(−st )] (8)
n ! # ! #
t=1 
k 
k 
k 
k
α =P Rt−i = δt−i , β = P Rt−i < δt−i
where U(x) is Heaviside’s unit step function, where i=1 i=1 i=1 i=1
U(x) = 1 for x ≥ 0 and U(x) = 0 for x < 0. Correspondingly, (12)
the estimate of α based on Equation 7 is Likewise, Equations 8 and 9 become

k
n/k
1
n
α= U(rt − δt ) (9) α = min{[1 − U(−st )]; t = k(p − 1) + 1, . . . , kp} (13)
n n
t=1 p=1
 RELIABILITY CONCEPTS IN RESERVOIR DESIGN 261

k
n/k reliability level or, equivalently, a certain return period
α = min{U(rt − δt ); t = k(p − 1) + 1, . . . , kp} (14) for the design flood. Traditionally, however, this has not
n
p=1 been the case in reservoir design, which has rarely been
based on sound probability. This is obvious even from the
respectively. The sums in Equations 13 and 14 count terminology traditionally used. For example, the use of the
the number of years in which no failure has occurred. term firm yield implies a nonprobabilistic, or failure-free
Apparently, α and β provide information on the concept. Specifically, the firm yield of a reservoir has been
occurrence of a failure within a year and not in the defined as the draft or withdrawal that lowers the water
time period during which the failure lasted. Therefore, content in a reservoir from a full condition to its mini-
they have been known as occurrence-based reliability and mum allowable level just once during a critical historical
failure probability, respectively. An overall indication of drought. (3, p. 27.8; 4, p. 534). It has been characterized as
the duration of failures within an average year can be essentially the no-failure yield (3, p. 27.8). In a probabilis-
obtained by applying Equation 8 or 9 and estimating β, tic context, however, any draft has a nonzero probability
the time-based probability of failure. of failure (unless the demand is less than the hypothet-
Apart from occurrence-based and time-based reliability, ical lower bound of the inflow distribution, which can be
an additional reliability measure has been often used, plausible only for perennial streams; this is unusual).
which is not expressed in terms of probability (and thus, Several procedures have been widely used in reservoir
literally does not comply with the general definition of design, which are rather deterministic and not consistent
reliability). This is the so-called volumetric or quantity- with the reliability concept. The most common has
based reliability, expressed as the ratio of the average been mass curve analysis and its variations. A mass
release to demand: curve is a plot of cumulative inflow volumes (typically
! k #' k based on historical discharge records) as a function of
 
αV = 1 − βV = E Rt δt (15) time. Using this plot, the firm yield, as well as the
i=1 i=1 required reservoir storage to attain this firm yield, can
be determined graphically. In addition, the method can
Given that a failure that occurs in a year does not determine the required storage for a smaller target release.
extend over the whole year, and, in addition, the release This graphical method was developed 120 years ago (5)
during the failure is not necessarily zero but some positive and has been widely used until now, although criticized (6)
quantity smaller than demand, it is easily concluded that for not providing information on the probability of failure
and for the fact that the reservoir capacity determined by
α ≤ α ≤ αV (16) this method increases as the arbitrary length of available
observed inflow data increases. As shown by Feller (7),
Among the three measures of reliability, the most this increase is asymptotically proportional to the square
important and most frequently used is the severest, root of the length of the record.
the occurrence-based reliability α . Another means for A first variation of the method is its application using
expressing virtually the same concept is the return synthetic, rather than observed, data (6). This eliminates
period or recurrence interval of emptiness, T. This is the the drawback of the arbitrary length of the record and
mean time between two consecutive empty states of the also provides some measure of uncertainty by applying
reservoir, and it is none other than the reciprocal of the the same procedure using different generated synthetic
probability of failure β (1): series. However, this kind of description of uncertainty is
not consistent with a rational definition of reliability (e.g.,
T := 1/β = 1 (1 − a ) (17) that of the previous section).
A more theoretical flavor for the method has been
which is expressed in years (given that α and β are given by the so-called range analysis, commenced by
expressed on an annual timescale). The concept of the the work of Hurst (8; see also 9, p. 184). Range is
return period of emptiness of a reservoir is similar to that essentially the algebraic difference of the maximum and
typically used for design floods. The difference is that in minimum departures of the mass curve from the straight
design floods, failure is the exceedance of the magnitude of line that joins its starting and ending points. The range
the design flood, whereas failure in a reservoir is emptying concept has greatly contributed to the understanding
of the reservoir. Typical design values of reliability and and description of the so-called Hurst phenomenon in
return period for reservoir design are α = 99% (T = hydrology, climatology, and other geophysical sciences.
100 years) for municipal water supply reservoirs, α = Applied to a reservoir, the range represents the required
70 − 85% (T = 3.3 − 6.7 years) for irrigation reservoirs in storage of a reservoir operating without any spill or other
subhumid climates, and α = 80 − 95% (T = 5 − 20 years) loss and providing a constant outflow equal to the mean
for irrigation reservoirs in arid climates (2, p. 313). flow. Obviously, this is an oversimplification of a real
reservoir. On the other hand, the range concept involves
TRADITIONAL RESERVOIR DESIGN PROCEDURES complexity in estimation, and simpler and more efficient
methods have been proposed that can be used instead of
Most hydraulic structures, such as flood protection works range analysis (10).
and drainage networks, whose load varies randomly, have An additional design method is the so-called sequent-
been designed on a probabilistic basis, adopting a certain peak analysis (11, p. 274; 12, p. 400). Essentially, it is
262 RELIABILITY CONCEPTS IN RESERVOIR DESIGN

a tabulated version of mass-curve analysis and can also for various combinations of demand, reservoir size, and
incorporate in the calculations, apart from runoff, the skewness of gamma distributed inflows:
effects of precipitation, evaporation, and leakage. The
method does not involve the reliability concept, nor does it (ε + 0.15)[κ + d1 (α, γ )] = d2 (α, γ ) (18)
consider spills from the reservoir.
where d1 and d2 are coefficients depending on reliability α
and skewness γ and are given by nomographs (see also 2,
SIMPLIFIED RELIABILITY-BASED PROCEDURES FOR
p. 323). McMahon and Mein (16) adapted this formula to
RESERVOIR DESIGN
indicate reliability more explicitly; this can be estimated
from the standardized normal variate zα corresponding to
As already mentioned, the analytical determination of α, using the equation
reliability in the general case of a reservoir fed by
inflows with seasonality, that have arbitrary probability 
zα = 2 ε(κ + d(α)σ/µ) (19)
distribution, and autocorrelation functions is a very
difficult, if not impossible, task (1). Therefore, existing where d is a coefficient depending on reliability α and is
analyses have been based on several simplifications. given by a table (see also 3, p. 27.14).
However, the results of such analyses are very useful, A more accurate and rigorous theoretical methodology
at least for the initial stage of reservoir design. The typical to estimate the reservoir size–yield–reliability relation-
simplifying assumptions are to ship was developed by Pegram (1). This was based on
finite-difference and integral equations, which employ
• neglect secondary inflows (precipitation) and losses reservoir dynamics (Equation 2) in a probabilistic con-
(evaporation, leakage); text to determine the return period of emptiness. Pegram
• neglect seasonality by the adoption of an annual applied his methodology for normal, lognormal, and dis-
time step; crete inputs both independent and serially correlated. His
• neglect autocorrelation and assume that inflows are results, when compared to those of the Gould method
independent in time; (Equations 18–19) indicate that the latter underestimates
• use a specific distribution function for inflows, the reservoir size required to attain a certain reliability
typically two-parameter such as normal, lognormal, level. Using Pegram’s results for normally distributed
or gamma. inflows, which were verified and expanded here with
extended simulations, the following approximate relation-
ship has been established:
The objective of such probability-based theoretical
analyses is to determine the relation of the following
ln(T − 1) = 2(ε + 0.25)(κ + 0.5)0.8 (20)
three quantities:
This is valid for T > 2(α > 0.5) and can be alternatively
• reservoir size c; it is usually standardized as κ := c/σ ,
written as
where σ is the standard deviation of annual net inflow
Xt ;
ln(T − 1) = − ln(1/α − 1)
• demand δ, which is assumed constant for all years; it
is usually standardized as ε := (µ − δ)/σ , where µ is = (2/σ 1.8 )(µ + 0.25σ − δ)(c + 0.5σ )0.8 (21)
the mean of annual net inflow Xt ; ε has been termed
the standardized inflow (8; 3, p. 27.7) or the drift (1); For known mean µ and standard deviation σ of inflows,
• probability, expressed either as reliability α, prob- Equation 21 can directly yield either the reliability α for a
known reservoir size and demand, the reservoir size c for
ability of failure β, or return period T; because
a given demand and reliability, or the demand δ that can
of the annual time step used, time-based reliabil-
be met at a given reliability for a known reservoir size.
ity (Equations 6, 7) is identical to occurrence-based
Equation 20 is graphically depicted in Fig. 1 in comparison
reliability (Equations 10–12).
with Pegram’s exact results. This equation is suggested
for preliminary estimates, but it should be applied with
The first among the probabilistic approaches used in
caution for the reasons explained in the next section.
such analyses is discretization of reservoir storage into
several zones, each representing a certain state, and the
Effects of Inflow Characteristics on Reservoir Size
use of a Markov chain model to represent transitions from
state to state (13,14; see also 9, p. 264). As explained earlier, simplified design procedures such
A second method is stochastic (Monte Carlo) simulation, as that using Equation 20 are based on a number
in which a long synthetic series of inflows is generated of abridging assumptions about inflows. Significant
from the appropriate distribution function and then differences may appear if these assumptions are not valid.
transformed into a series of storage values using More specifically, what may cause significant departures
Equation 2; reliability is then easily determined from from Equation 20 are hydrologic persistence, especially
the storage time series using Equation 8. Gould (15) used long-term, and the seasonal distribution of inflow and
this method to propose a reservoir size–yield–reliability demand. Less significant differences are caused by the
formula, fitted to 240 sets of Monte Carlo simulations skewness of inflows and secondary inflows and losses.
 RELIABILITY CONCEPTS IN RESERVOIR DESIGN 263

Design reliability level, a, % 6

80 90 95 98 99
8
1
05 0. 5

Standardized reservoir size, k

0. 0.2
7 =
Standardized reservoir size, k

e
6 4

5
0.4 3
4

3 0.6 2

2 0.8
1
1 1.0

0 0
5 10 20 50 100
0 0.2 0.4 0.6 0.8 1
Design return period, T, years
Coefficient of skewness of inflows, Cs
Figure 1. A simple representation of reservoir size–yield–reli-
ability: Standardized reservoir size (κ = K/σ ) required to achieve Figure 3. Effect of the skewness of inflows on the standardized
a certain drift (ε = (µ − δ)/σ ) at a certain reliability level for reservoir size κ required to achieve a drift ε = 0.8 at a reliability
independent inputs normally distributed. Lines are constructed level α = 98% (continuous line), a drift ε = 0.2 at a reliability
from Equation 20, whereas plotted points are theoretical results level α = 90% (dashed line), and a drift ε = 0.2 at a reliability
by Pegram (1) for ε = 0.2 (triangles), 0.4 (diamonds), 0.6 (squares), level α = 98% (dotted line). Results are obtained by simulation
0.8 (circles), and 1.0 (stars). using independent two-parameter gamma distributed inflows.

demand (low drift ε). Thus, the required reservoir size for
10 ε = 0.2 and α = 90% is c = 2.4σ when ρ = 0 (H = 0.5) and
9 Short-term persistence becomes 4 times larger (c = 9.6σ ) when ρ = 0.4 (H = 0.74).
To quantify the effect of seasonal variation of inflow
Standardized reservoir size, k

(AR(1))
8 Long-term persistence and demand, it is observed that in the worst case,
7 (FGN) the total annual inflow comes before the beginning of
6 withdrawal (the ‘‘sudden’’ model). The reservoir dynamics
in Equation 2 assumes that inflow and withdrawal are
5
distributed evenly during a year (the ‘‘steady’’ model). If
4 inflow precedes withdrawal, Equation 2 should be modified
3 to read St = max[0, min(St−1 + Xt , c) − δt ]. It is then easily
determined that extra storage equal to δ is required in
2
addition to that estimated from Equation 21. This worst
1 case, however, is not very realistic, except in arid regions.
0 In the real world, the extra storage capacity required
0 0.1 0.2 0.3 0.4 0.5 (in addition to that estimated from Equation 21) is a
Lag-one autocorrelation, r percentage of δ. This can be as high as 50% for water
supply reservoirs and 80% for irrigation reservoirs in
Figure 2. Effects of short-term and long-term hydrologic persis-
tence of inflows on the standardized reservoir size κ required to semiarid regions.
achieve a drift ε = 0.8 at a reliability level α = 98% (continuous The effect of the skewness of inflows is demonstrated
lines) and a drift ε = 0.2 at a reliability level α = 90% (dashed in Fig. 3, which depicts the standardized reservoir
lines). Results are obtained by simulation. size κ required to achieve three combinations of drift
and reliability (ε = 0.8, α = 98%; ε = 0.2, α = 90%; and ε =
0.2, α = 98%) versus the coefficient of skewness, Cs . It can
The effect of hydrologic persistence is demonstrated be observed that the effect of skewness is not significant;
in Fig. 2, which depicts the standardized reservoir size for low draft (high drift ε), this effect can be beneficial
κ required to achieve two combinations of drift and (lowering of required storage), but for high draft, the
reliability (ε = 0.8, α = 98% and ε = 0.2, α = 90%) versus required storage is practically insensitive to skewness.
the lag-one autocorrelation coefficient, ρ. Two cases
of hydrologic persistence have been examined, short- GENERALIZED SIMULATION PROCEDURE FOR
term and long-term. In short-term persistence, it was RELIABILITY-BASED RESERVOIR DESIGN
assumed that the inflows follow the autoregressive process
of order 1 [AR(1) or Markov], whereas in long-term The simplest general procedure for estimating the reser-
persistence, it was assumed that they follow the fractional voir size–yield–reliability relationship in an accurate and
Gaussian noise (FGN) process with Hurst exponent H = detailed manner and for any arbitrary inflow character-
ln(2 + 2ρ)/ ln 4. Obviously, the effect of persistence is very istics is stochastic (Monte Carlo) simulation. A simplified
significant, especially for long-term persistence and high simulation procedure is outlined as follows:
264 RELIABILITY CONCEPTS IN RESERVOIR DESIGN

1. Generate a series of inflows Xt at an appropriate DESIGN OF ADDITIONAL STORAGE ZONES OF A

timescale (e.g., monthly) using an appropriate RESERVOIR
stochastic model (e.g., 17, 18; see also the entry
STOCHASTIC SIMULATION OF HYDROSYSTEMS). The design of flood control storage is typically reliability-
based but in a very different context from that described
2. Assume a reservoir size c.
for active storage. A failure of the flood routing function
3. Calculate a series of reservoir storages using of a reservoir is not a performance failure but a structural
Equation 2. one: an overtopping of the dam due to a severe flood
4. Estimate the reliability using Equation 13. can result in collapse of the dam. Therefore much lower
5. Repeat steps 3–4 for different reservoir sizes. levels of probability of failure are adopted, of the order of
10−3 − 10−6 . The typical steps here are (1) estimation of a
design storm, based on statistical analysis of rainfall,
This procedure is very simple to execute even in a
for an appropriate return period (103 − 106 years); (2)
tabulated form on a spreadsheet.
estimation of the inflow hydrograph using an appropriate
The drawback of stochastic simulation is that it
rainfall–runoff model; and (3) routing this hydrograph
requires a vast simulation length to find accurate results.
through the spillway and estimation of the outflow
It can be shown that the required number of simulated
hydrograph and the maximum water level. Implicit
time steps (e.g., months) to estimate the occurrence-based
assumptions in the entire procedure, like the assumption
failure probability β at an acceptable error ±εβ and that the reservoir is full at the beginning of the flood,
confidence γ is decrease the risk further. It must be noted that several
procedures have been proposed that are supposedly risk-
n = k(z(1+γ )/2 /ε)2 (1/β − 1) (22) free. These are based on the so-called probable maximum
precipitation concept. However, it has been argued that
where k is the number of time steps per year and zp a risk-free procedure is an illusion and the value of
is the p-quantile of the standard normal distribution. probable maximum precipitation can be exceeded by a
For instance, for k = 12, γ = 95% (z(1+γ )/2 = 1.96), ε = certain probability (e.g., of the order of 10−5 ; 21).
10%, and β = 0.01, this yields n = 456, 000 months The sizing of dead storage has been typically based on
(38,000 years). Today, this is not a major problem as the the expected sediment accumulation in the reservoir for
required computer time for such a simulation length can a certain design period (e.g., of the order of 102 years).
be less than one second in a common PC. Because of the large design period and the accumulative
The simplified procedure described can be extended character of this process, the approach to this problem is
to a detailed simulation procedure, which includes, in very different. Expected values, rather than probabilities,
addition to runoff, the precipitation, evaporation, and are involved in the calculations. The additional objectives
leakage of the reservoir. In this case, level–area–volume that the dead storage serves, environmental protection
and level–leakage relationships are required to establish (protection of the habitat of the reservoir during dry
the functions a(S) and l(S) which yield reservoir area a periods) and conservation of the quality of the landscape,
and leakage l for any storage S (usually using interpolation have not been given special attention, until now, and have
from arrays of tabulated values). In this case, the runoff not been considered in the design procedure. One would
Q, has to be expressed in equivalent depth units, as are expect that, some years after construction, these additional
precipitation P and evaporation E. If f is the catchment objectives would not be served adequately because the
area, the net inflow becomes water in the dead volume would be reduced due to
sediment accumulation. Fortunately, however, this has
Xt = Qt [f − a(St−1 )] + (Pt − Et )a(St−1 ) − l(St−1 ) (23) not been the case: the implicit assumption that sediments
will reach the bottom of reservoir near the dam is not
verified. Thus, in large reservoirs, sediment accumulation
where it was implicitly assumed that variations in the
occurs mainly in the active zone of reservoir (near the
reservoir area and leakage within a time step are not large,
entrance of the river to the reservoir) and much less
so that a(St−1 ) and l(St−1 ) can be assumed representative
in dead storage. The unfavorable consequence is the
of the entire period within a time step. When the detailed
reduction of active storage. These problems need to be
simulation procedure is employed, three time series,
further investigated in future reservoir designs.
instead of one, have to be synthesized. Obviously, Qt
and Pt are cross-correlated, and therefore they cannot Acknowledgment
be generated independently of each other; a bivariate The review by the editor and the comments by G.G.S. Pegram are
stochastic model is needed in this case. On the other hand, gratefully appreciated.
Et can be generated independently of the other two series.
Obviously, the simulation procedure, either simplified BIBLIOGRAPHY
or detailed, can be applied directly using historical, rather
than synthesized inputs. However, due to short record 1. Pegram, G.G.S. (1980). On reservoir reliability. J. Hydrol. 47:
length, the accuracy of results will not be satisfactory. 269–296.
Extension of the simulation method, combined with 2. Raudkivi, A.J. (1979). Hydrology. An Advanced Introduction
optimization, for a multiple reservoir system can be found to Hydrological Processes and Modelling. Pergamon Press,
in References 19 and 20. Oxford.
 LAKES 265

3. McMahon, T.A. (1993). Hydrologic design for water use, Ch. the bottom. Stratified lakes are deep enough to be
27. In: Handbook of Hydrology. D. Maidment (Ed.). McGraw- vertically divided into a lighted surface-near euphotic
Hill, New York. zone, where photosynthesis of green plants is possible,
4. Chow, V.T., Maidement, D.R., and Mays, L.W. (1988). Applied and a deep-water aphotic zone, which is too dark for
Hydrology. McGraw-Hill, New York. photosynthetic activities. In many lakes, thermal vertical
5. Rippl, W. (1883). The capacity of storage reservoirs for water stratification is observed as a warm layer near the
supply. Proc. Inst. Civ. Eng. 71: 270–278. surface, the epilimnion, and a colder layer below, the
6. Schultz, G.A. (1976). Determination of deficiencies of the hypolimnion. The epilimnion is wind-exposed and well
Rippl-diagram method for reservoir sizing by use of mixed, whereas the deep water layer of the hypolimnion
synthetically generated runoff data, Proceedings XIIth
is not included in the mixis process. Introductions to the
Congress of ICOLD (International Commission on Large
ecology, hydrology, and limnology of lakes are available in
Dams). March/April 1976, Mexico City.
textbooks, handbooks (1,2), and reviews (3–9).
7. Feller, W. (1951). The asymptotic distribution of the range of
sums of independent random variables. Ann. Math. Stat. 22:
427.
ORIGIN OF LAKE BASINS
8. Hurst, H.E. (1951). Long term storage capacities of reservoirs.
Trans. ASCE 116: 776–808.
9. Kottegoda, N.T. (1980). Stochastic Water Resources Technol- Lake basins containing standing waterbodies were formed
ogy. Macmillan Press, London. by different processes; tectonic forces; volcanic, glacial,
10. Koutsoyiannis, D. (2002). The Hurst phenomenon and fluviatile activities; or by solution of easily dissolvable
fractional Gaussian noise made easy. Hydrol. Sci. J. 47(4): rock such as limestone. Hutchinson (1) ordered the lakes
573–596. by their mode of origin into about 80 categories. Typical
11. Mays, L.W. and Tung, Y.-K. (1992). Hydrosystems Engineer- examples are listed by Meybeck (8,10), who also gives
ing and Management. McGraw-Hill, New York. numbers of the types of origin, of size, and of age.
12. Mays, L.W. (2001). Water Resources Engineering. John Wiley Tectonic lakes are found in geological rift valleys; the
& Sons, New York, NY. deepest are Lake Baikal (11) and Tanganyika (12). The
13. Moran, P.A.P. (1959). The Theory of Storage. Methuen, Caspian Sea and the Aral Sea (13) are rest areas of
London. the ancient Thetis Sea which was closed by tectonic
14. Zsuffa, L. and Gálai, A. (1987). Reservoir Sizing by Transi- movements, and they are the largest inland waters in area
tion Probabilities: Theory, Methodology, Application. Water (374,000 km2 , 64,500 km2 ). The oldest inland waters are
Resources Publications, Littleton, CO, p. 186. found among tectonic lakes, some of these, it is estimated,
15. Gould, B.W. (1960). Water supply headworks storage estima- are older than 20 to 30 million years.
tion. Aust. Civ. Eng. Constr. 1(9): 46–52. Volcanic processes led to the origin of different
16. McMahon, T.A. and Mein, R.G. (1986). River and Reservoir lake types. Maar lakes emerged after singular volcanic
Yield. Water Resources Publications, Littleton, CO. events (steam explosions in contact zones of magma and
17. Koutsoyiannis, D. (2000). A generalized mathematical frame- groundwater). Lava dams or emerging volcanoes resulted
work for stochastic simulation and forecast of hydrologic time in reservoir-like lakes, such as L. Kivu and Laguna del
series. Water Resour. Res. 36(6): 1519–1533. Laja (Chile). Large lakes were formed on lava plateaus
18. Koutsoyiannis, D. (2001). Coupling stochastic models of (L. Myvatn, Iceland) and in caldera basins, whereas crater
different time scales. Water Resour. Res. 37(2): 379–392. lakes are usually small. Lakes at active volcanoes are
19. Nalbantis, I. and Koutsoyiannis, D. (1997). A parametric rule often heavily acidified by acidic brines rinsing into the
for planning and management of multiple reservoir systems. lake water (14).
Water Resour. Res. 33(9): 2165–2177. The main types of glacially formed lakes are (1) small
21. Koutsoyiannis, D. and Economou, A. (2003). Evaluation of kar lakes at the slopes of mountains at different altitudes,
the parameterization-simulation-optimization approach for (2) piedmont lakes where glaciers from valleys reached the
the control of reservoir systems. Water Resour. Res. 39(6):
foreland at the foot of the mountains, and (3) extended
1170, 1–17.
lake districts from low altitude glaciations as on the
21. Koutsoyiannis, D. (1999). A probabilistic view of Hershfield’s
Canadian Shield and the circum-Baltic area. Postglacial
method for estimating probable maximum precipitation.
lakes of temperate climatic zones are only 8 to 12 thousand
Water Resour. Res. 35(4): 1313–1322.
years old.
In the lowland area of large rivers, the flow of the
LAKES running rivers often changes, leaving standing waters,
oxbows or side waters, as fluvial lakes, which were parts
WALTER GELLER of the rivers in the past or are permanently in open
UFZ-Centre for Environmental connection with the river system.
Research Lakes are more numerous as size decreases. Mey-
Magdeburg, Germany beck (8,10) gives the estimated worldwide number of lakes,
ordered by size classes and by the origin of lake basins.
A list of the deepest and largest lakes, by area and by
Lakes are waterbodies of considerable size that are larger water volume, is given by Herdendorf (15,16), including
than ponds. Shallow lakes are mixed by the wind from many uncertainties on the basic data of even the 50 most
the surface to the ground, and the daylight reaches important inland waters of the world.
266 LAKES

WATER BUDGET AND SALINITY OF LAKES system of neutral freshwaters. This is caused by rainwater
that is weakly acidic from CO2 and, thereby, dissolves the
The oceans contain 83.5% of the world’s water, the carbonatic minerals of soils and rock (23,24). In areas poor
inland waters only 0.015%. Lakes contain a total water in carbonates, the water contains low concentrations of
volume of 205,000 km3 ; about 50% of this total volume dissolved salts, and in other areas of dominating limestone,
is freshwater, on the one side, and saline water on the the water is rich in dissolved carbonates. The qualitatively
other side. The deepest and oldest lake, Baikal, has a similar composition of freshwaters is described as a
volume of about 20% of the total freshwater volume of worldwide ‘‘standard water composition’’ (25,26).
lakes and is the worldwide largest singular freshwater The gases in air, 78.09% N2 , 20.95% O2 , and 0.03%
resource. The annual fluxes through the pools of the CO2 , are in physical equilibrium between air and water. At
atmosphere, rivers, and groundwater are 496, 38, and low temperatures, higher concentrations of gases dissolve
12 thousand km3 ; exchange times are 10 days, 16 days, in the water than at high temperature. The maximum
and 700 years, respectively. The volume of freshwater content of dissolved oxygen and CO2 in water is 14.5
lakes is 100 thousand km3 ; exchange times are 1 month to (8.9) mg O2 /L and 1.005 (0.51) mg CO2 /L at 0 ◦ C (20 ◦ C).
500 years [data after (17,18)].
Lakes are standing waters that are usually connected
with and included in the flow system of river catchments. STRATIFICATION AND MIXING
The lakes within the catchment systems can be ordered by
their degree of connection to the overall water flows (19). Lakes are usually stratified into an upper layer, the
Lakes with intensive through-flow have high rates of epilimnion, which is well mixed by the wind, and a
flushing and water exchange and short retention times. deep waterbody, the hypolimnion, which is not included
The lake volume compared with the volume of all inflows in the mixing process. Stratification occurs seasonally in
per year gives the virtual filling time. The relation of the temperate climatic zones, or in a diurnal–nocturnal
the outflow and the lake volume gives the net exchange rhythm in the tropical zones (Fig. 1). Ice covers block
rate of the lake water. Including groundwater exchange, energy inputs by wind into the water column during the
precipitation on the lake surface, and evaporation, the winter. The regime of stratification and mixis of the water
relation of all inputs to all losses of water gives the gross column, as given by the climate, defines the types of lakes
exchange rate; its reciprocal value is the retention time or as (1) monomictic, (2) dimictic, or (3) polymictic, according
average age of the water. to the number of periods of total mixing (holomixis) per
Depending on the given regional water balance, lake year. Tropical lakes at low altitude are polymictic; they
basins are permanently flushed, have an outflow, are mix during the night and stratify during the day. Dimictic
in a steady state of the flow of water and of dissolved lakes have a stable stratification with a warm epilimnion
salts, and, thereby, contain freshwater. In areas with a during the summer and an ice cover during the winter;
negative water balance, the river catchments are endorheic they show holomixis during the two seasonal periods of
systems, and terminal lake basins occur without outflows; homothermy. One holomixis occurs in spring after ice
here, the water finally evaporates, and the inflowing melting when the surface water has the same temperature
salts accumulate, leaving saline lakes. Saline lakes are as the hypolimnion, usually 4 ◦ C, and a second during
typical of (semi)arid areas and contain nearly half of late autumn before freezing. Monomictic lakes mix only
the world’s lake water volume (20). Saline lakes (with once per year. The warm-monomictic lakes of oceanic-
salinities >3%) (21,22) are found in all climatic zones, temperate regions never freeze and show holomixis during
warm, temperate, and cold regions, where negative water the winter. Cold-monomictic lakes are found in polar
balances are given by large-scale climatic zones or by climatic regions; they are covered by ice during large
regional orographic conditions (rain-lee east of the Rocky parts of the year and melt only in summer when the
Mountains in North America, east of the Andes in South water column reaches 4 ◦ C and mixes from the surface to
America, or in the Iranian basin). Such endorheic river the ground.
catchments, in which the water flow ends in terminal The stability of stratification of a warm epilimnion and
lakes, are the Jordan River system with the Dead Sea, the a cold hypolimnion determines whether wind forces mix
system of Lake Titicaca with Lake Poopó as the terminal only the uppermost layers, or the entire water column from
saline lake, and the system of Lake Tahoe and Pyramid the surface to the bottom. Stratification stability depends
Lake in North America. In Central Australia, Lake Eyre on the density gradient between the epilimnetic and the
is a large, endorheic saline basin. The highest possible salt hypolimnetic partial waterbodies. Indicative numbers are
concentrations are found in the Dead Sea (∼300 g/L) and the Wedderburn Number or the Lake Number, which, at
in Deep Lake (Antarctica) which never freezes; the water given threshold values, predict optional holomixis with the
reaches −14.5 ◦ C during the winter (22). next wind-forcing (27).

Climatic Zones and Lake Temperatures

STANDARD COMPOSITION OF FRESHWATERS
Lakes are integrators of the regional climate and weather
Freshwaters, soft waters as well as hard waters, are development (28). The surface temperatures follow the
chemically dominated by the carbonate system, CO3 2− actual air temperature with a time delay, and lakes of
and HCO3 − as anions, and Ca2+ and Mg2+ as cations; the same region show time courses of similar surface
together they are the main constituents and the buffering temperatures. The deep water of lakes in polar and
 LAKES 267

Wind-induced partial mixis of

Epi- the warmer epilimnion in
stratified lakes, as occurring in
temperate zones during summer,
Hypolimnion and during the day in tropical
(cold water) climatic zones. The colder water
of the hypolimnion is not
included in the mixing process.

Wind-induced holomixis of the

entire water column from
Holomixis surface to bottom in
unstratified, homothermic lakes,
occurs in spring, fall, or winter Figure 1. Seasonal or diurnal succession of
in temperate zones and stratification into warm epilimnion and cold
during the night in tropical hypolimnion and unstratified water column when
climatic zones. the water is homothermic with complete overturn
of the water from surface to ground (holomixis).

temperate continental climatic zones is at 4 ◦ C, the larger zooplankton, and fish. The phytoplankton is
temperature of water at its maximum density. The deep autotrophic and limited to the euphotic layers near
water of warm-monomictic lakes of temperate oceanic the surface where light intensities are sufficient for
climatic zones in the Southern Hemisphere, in New photosynthesis and oxygen production. The limit between
Zealand and South America and also in western and the euphotic and the aphotic zone is usually at depths
southern Europe, has a temperature above 4 ◦ C. The where the light intensity decreases to <1% of that at the
hypolimnetic temperatures remain from the last holomixis lake surface.
at the end of winter, and changes reflect long-term climatic The pelagic zone is the body of free water of the lake,
conditions (29). the littoral zone is the euphotic part at the lakeshore,
and the profundal is the aphotic dark layer of the
Waves and Oscillations benthic zone below the littoral. The benthic community
outside the littoral zone, the profundal, consists of the
Driven by wind, waves emerge on the surfaces of lakes;
organisms living in and on the sediments, microorganisms,
their heights can be approximated by h = 0.105 (fetch)0.5
macrozoobenthos, and benthic fish species. Below the
(wave height and fetch given in cm; the fetch is the
light limit for photosynthesis, the species of the benthic
maximum length of the wind blowing across an open
community, macro- and microorganisms, are heterotrophic
lake surface) (9). Less peculiar are oscillations along the
and consumers of oxygen. All particles in the water
lake axis, which have amplitudes of few centimeters at
eventually sink to the lake bottom, form annual layers
the lake surface (seiches). Their frequencies depend on the
of growing sediments (varves), and leave an archive of
length of the lake axis, between minutes in small lakes
residuals. Therefore, the history and the development of
and one day in large lakes. At the boundary layer between the lake can be reconstructed from the microlayers of a
the epi- and hypolimnion, however, internal oscillations sediment core, with respect to eutrophication and nutrient
can reach wave amplitudes of 10 m or more. The periods of levels and impacts of climatic changes.
these internal oscillations are slower and reach 1 month The littoral zone is the euphotic part of the benthic
in the largest lakes, such as Baikal or Tanganyika (1,27). zone. The living community consists of autotrophic
macrophytes and algae and of heterotrophic animals and
ECOSYSTEMS OF LAKES microorganisms. The belt of macrophytes appears as an
interface between land and water; wetland plants are on
The ecosystems of lakes belong to the best described the land side of the belt, followed by plants rooting in
and understood ecosystems because the composition of the water and emerging in the air, swimming plants, and
the community is limited in number of species and lastly submerged plants that cover the light bottom of the
the populations are limited by the lake size. The deeper littoral zone. The partial ecosystem of the littoral
seasonal development of the populations, their change zone connects the land and water, and simultaneously, as
in species composition, and the growth and decrease of the a biologically reactive zone, retains and changes the flow
populations are well investigated and analyzed in terms of nutrients and other allochthonous inputs of matter into
of numbers of individuals as well as in terms of energetics the lake’s pelagic zone.
and species composition (30).
Trophic State and Productivity
Lakes have characteristic compartments as partial
habitats of their ecosystems. The ecosystem of the pelagic Trophy is a measure of the level of nutrients and of the
zone of lakes is dominated by planktonic microorganisms, biological productivity resulting from nutrients supplies.
268 LAKES

The organismic productivity of aquatic ecosystems, herbivores, zooplankton species that feed on these algae,
measurable as the formation of biomass, is limited and carnivores that feed on herbivorous zooplankton. The
by the availability of the rarest nutrient element efficiency of energy transfer from the lower to the higher
(C,N,P,S,Fe, . . .). In temperate zones, phosphorus is trophic levels is about 10% from plants to animals and
usually the least available element because there is up to 30% from animal prey to their predators. Because
no gaseous transport form through the air and P is there are losses of unused particles and dissolved organic
tightly sorbed on soil particles and retained on the land. matter (DOM) on all levels, recycling by bacteria as
The production of organic substance is limited by the decomposers of DOM is important. Based on the bacteria
availability of phosphorus, so the trophy (productivity) of as food, leading via bacteria-feeding protozoans to larger
lakes can be related to the P-content of the water at the predators, the ‘‘lost’’ DOM is recycled again into the
start of the growing season (31,32). The process of growth ‘‘normal’’ food chain by the ‘‘microbial loop.’’ This secondary
needs a proportion of 106C:16N:1P as the atomic ratio of food chain is important for the overall efficiency of the
biomass (33,34). The primary production of phytoplankton ecosystem as a processor of energy and its ability to recycle
and macrophytes—depending on P—reaches annual rates nutrient elements.
of 10 to 50 g C/m2 of organic carbon in nutrient-poor lakes In oligotrophic systems, the organisms, prey and
and more than 500 g C/m2 in very nutrient-rich systems. predators, are present in low concentrations, and a
The biomass produced sediments to the lake bottom, predator has to find and to grasp each single prey
where this dead organic substance is decayed by oxygen- individually. The prey in such food chains is usually
consuming bacteria. Depending on the trophic state, the smaller than its predator by a factor of 10 to 20 by
oxygen conditions in the hypolimnic water body reach body length. Therefore, in oligotrophic planktonic systems,
different end points in late summer and fall. In oligotrophic the food chain from green algae, via zooplankton to fish
lakes, the nutrients and productivity are low, as well as the needs four to five subsequent trophic levels. The overall
following decomposition and oxygen losses in deep water. transfer efficiency with its losses at any transfer is small.
In eutrophic lakes, however, the nutrient supply and the In the more productive systems of meso- to eutrophic
resulting productivity are high. After decomposition in lakes, however, the concentrations of organisms from
deep water layers, the oxygen content in the hypolimnion algae and bacteria to zooplankton and fish are high,
may reach critically low limits in late summer. The
and additional, new types of filtering predators appear.
consequences of high nutrient supplies are mitigated in
Filtrators use a filtering system, comb-like sieves in
deep lakes with high hypolimnic volumes but are more
cladocerans (Daphnia), fine gill structures in filter-feeding
severe in shallower lakes with small hypolimnic volumes
fish (Coregonids), to collect large numbers of very small
and small oxygen reserves therein (Fig. 2).
prey particles. The linear size relation between prey and
filter-feeding predators is 100 × to 1000×, and, thereby,
Food Chains, Filtrators, and Size Spectra in plankton
the length of the food chain becomes short, three levels
Within the pelagic zone of lakes, the species of the from algae to fish. In temperate lakes, we find Daphnia as
plankton community can be structured by the food the keystone filtrator on the level of herbivores all over the
chain from the primary producers, the green algae, to world. The filtering, zooplankton-feeding fish of the genus

Algal production in Algal decomposition

epilimnion in hypolimnion

Lake Esrom:
eutrophic
PPR: 240 g C/m2/year
Figure 2. The relationship between the sup-
ply and demand of oxygen in deep water is
determined by the intensity of the respec-
tive decomposition rate and the volume of
the hypolimnetic waterbody of a given lake.
L. Maggiore:
The oxygen demand with algal decomposition
oligotrophic
consumes a large percentage of the supply in
PPR: 250 g C/m2/year
the small hypolimnic waterbody of L. Esrom,
it is medium in L. Maggiore with its big
hypolimnion, and it is unimportant in Thing-
vallavatn because the mass of algae is small
and the big volume of the hypolimnion has a big
oxygen reserve. L. Esrom is eutrophic, L. Mag-
giore has the same annual primary produc-
Thingvallavatn:
tion (PPR) of algae, but is oligo-/mesotrophic oligotrophic
because of its morphology. Thingvallavatn is PPR: 95 g C/m2/year
a typical oligotrophic lake because of low pri-
mary production and its depth (redrawn from
Reference 35).
 LAKES 269

Coregonus are found only in the Northern Hemisphere. In of the circum-Baltic countries have been eroded by glacia-
the ancient Lake Tanganyika, an evolution of a ‘‘new’’ tion that left ten thousands of large and small lake basins.
filtrator occurred, Limnotrissa, where a zooplankton- The large basins of the Hudson Bay and the Baltic Sea are
feeding sardine, led to very high fish production. In corresponding central basins. The Baltic Sea was inter-
other ancient tropical lakes, herbivorous Tilapia species mittently an inland lake and then became a brackish shelf
appeared that filter-feed on algae. sea. The surrounding areas from Finland to northern Ger-
The food web of any ecosystem might be described on many have extended lake districts that cover large parts of
the level of species composition and populations, according the land. The ‘‘limnic ratio’’ is a measure of the percentage
to its taxonomic structure. The system, however, might be of the recent lake area of the deglaciated land area (8):
alternatively viewed with an ataxonomic approach and world total deglaciated area: 6.8%; Scandinavian Shield:
explained functionally as a system of trophic levels linked 12.2%; Finland: 9.4%; Norway: 13.9%; Sweden: 8.55%;
by energy transfers: the ‘‘trophic-dynamic system’’ (34,36). South Baltic Platform: 2.2%; Canadian Shield:10.3%; and
Another way to describe the planktonic food web and to Patagonia: 3.0%. The lakes of this origin are the largest in
explain some of its functions and properties is to view its Europe, Lakes Ladoga and Onega in western Russia and
size structure and the continuity of sizes as a spectrum. the large southern Swedish lakes, and in North America,
Investigations of the entire planktonic system revealed the Laurentian Great Lakes and the large lakes of Canada
the plankton as a spectrum of size classes, in which the
(Table 1).
biomass of all size classes is equal. Many small organisms
Formerly glaciated mountain districts are found on
are present with the same sum of biomass, as are the fewer
all continents (Table 1). The fjord lakes and foothill
organisms of larger sizes. The regular pattern of body
lakes in the deglaciated mountain areas are found in
sizes over all organisms of plankton (37,38) corresponds
the Northern and the Southern Hemispheres of the
to growth and physiological rates in a reciprocal way, fast
temperate climatic regions in Scandinavia, the loghs/lochs
overturn in the pools of small organisms and slow overturn
in the big ones. of Ireland and Scotland, the English Lake District, the
Alps, New Zealand, the Patagonian mountain areas of
REGIONAL LAKE LIMNOLOGY Chile and Argentina, and the Central and Northern Asian
mountains. Fjord lakes are the deepest glacial lakes.
Glacial Lakes Their mountain valley erosion basins have peculiarly steep
The glaciated areas of the Canadian Shield, the Scandina- borders and flat bottoms with depths of 500 m or more.
vian mountains, and the adjacent sedimentary platforms Where the glacial erosions from valleys extend into the

Table 1. Large Glacial Lakes on Shields, Sedimentary Platforms, and in Mountain Areas
Surface Max. Water
Area, Depth, Volume, Retention Age Data
Glacial Lakes km2 m km3 Time Years Origin, 103 y Source

Canadian Shield lakes (8,10,49)

Superior 82,100 405 12,200 180 8, glacial
Huron 59,500 223 3,500 48 8, glacial
Michigan 57,750 281 4,800 31 8, glacial
Great Bear 31,326 446 2,200 glacial
Great Slave 28,568 614 2,088 tectonic, glacial
Andean lakes (39,40)
Argentino 1,410 500
Gral. Carrera, Chile/Buenos Aires, Arg. 1,892 586 >500 tectonic, glacial
Nahuel Huapi, Arg. 464
Llanquihue, Chile 871 317 159 74
New Zealand (74)
Taupo 616 165 59 10.6 volcanic, glacial
Waikitipu 289 380
Wanaka 180 311
Baltic lakes (8,10,72)
Ladoga, Russia 18,130 908 >8, baltic
Onega, Russia 9,700 292 >8, baltic
Vänern, Sweden 5,648 152 8.8 6.5, baltic
Vättern, Sweden 1,912 128 74 58 6.5, baltic
Mälaren, Sweden 1,096 14 2.3 6.5, baltic
Alpine piedmont lakes (28,54)
Geneva, W.-Alps 580 310 89.9 piedmont
Constance, N.-Alps 472 253 47.7 4.15 piedmont
Garda, S.-Alps 370 346 49 26.6 piedmont
Maggiore, S.-Alps 213 370 37.5 4.1 piedmont
Como, S.-Alps 146 410 22.5 4.5 piedmont
270 LAKES

foothill areas of the foreland, the characteristic piedmont long, covering 0.5 to 1 million years (44). The oldest and
lakes cover large areas, but they are shallow compared deepest of the ancient lakes gave time to their biota for
with fjord lakes or fjord-like parts of the same lake (e.g., evolutionary processes which led to adaptive radiations
L. Garda, L. Argentino, L. Buenos Aires, show both, deep of—mostly endemic—species, found 10 to 100× more
fjord-like parts and large, shallow foothill areas). numerously than in young lakes. There are >400 species
Lakes in the (sub-)polar regions are covered by ice of Cichlid fish species in Tanganyika and Malawi and
permanently or during the largest part of the year. Lake >200 species of Gammarid crustacean species in Baikal.
Vanda (Antarctica) has a permanent, clear, ice cover. The ecosystem structure is severely changed by new
The deepest water layers are stabilized by saline inputs keystone species in the food web and new top carnivores:
and trap the down-welling light energy, leading to a in Lake Baikal the pelagic Gammarid Macrohectopus and
temperature of 25 ◦ C above the bottom (63 m) (22). Some Cottoidei fish and the freshwater seal Phoca sibirica. In
large lakes in the Antarctic were found by satellite radar Lake Tanganyika, the species Limnotrissa directs the food
imaging under several thousand meters of glacial ice cover. chain to high production of fish, whereas in the other
The largest of these lakes, L. Vostok, is the size of Lake large African lakes the insect larva Chaoborus plays
Ontario (39). the role of the main zooplankton feeder. This species
leaves the lake when it becomes an adult insect, so
TROPICAL AND ANCIENT LAKES the branch of the food chain is then lost for the lake
ecosystem.
The lakes of the (sub-)tropical climatic zone are described
in textbooks by Beadle (42) and Serruya (43). The MAN-MADE LAKES: RESERVOIRS AND PIT LAKES
most important difference from temperate lakes is the
diurnal variability of temperatures that are as large as Man-made lakes are built in great numbers, ten thousands
the seasonal variability between summer and winter of reservoirs and pit lakes in the voids of surface mining.
in temperate lakes. The restricted time of nocturnal They show limnological peculiarities that come primarily
holomixis presumably limits the maximum depth of from the management regime for reservoirs and from
mixis in deep tropical lakes. Therefore, the very deep geogenic impacts on water chemistry for pit lakes in
lakes Tanganyika and Malawi are meromictic, and, coal and ore mining districts. The state of reservoir
consequently, have anoxic, deep waterbodies. The deep building and the problems are presented by the World
African Rift Valley Lakes are ancient lakes and, therefore, Commission on Dams (http://www.dams.org). Reviews and
have peculiarities beyond their tropical sites (12). other comprehensive scientific publications on mining
The largest and deepest lakes of the world are also the lakes are in an initial phase (50).
oldest lakes: Baikal, Tanganyika, Malawi, and Issik-Kul.
The most important ancient lakes are listed in Table 2. REGIONAL PROGRAMS FOR PROTECTING LAKES AND
The old lakes originate from the Tertiary and have a COLLECTING DATA
long development with changing water levels that could
be reconstructed by paleolimnological investigations. The The protection of lakes and other surface waters
best investigated ancient lake is L. Biwa, whose history is different regionally. The Environmental Protection
is based on sediment cores that are 200 and 1000 m Agency (EPA) has set standards in North America, and the

Table 2. Basic Data on Ancient Lakesa

Surface Area, Maximum Water Volume, Age & Origin,
km2 Depth m km3 Years

African Rift Valley

Tanganyika 32,900 1471 18,900 20 mio, tectonic
Malawi 22,490 709 6,140 2 mio, tectonic
Victoria 68,460 92 2,700 20,000, tectonic
Asia
Dead Sea 1,020 433 188 tectonic
Caspian 374,000 1025 78,200 >5 mio, tectonic
Baikal 31,500 1637 23,000 20 mio, tectonic
Aral 64,100 1,020 >5 mio, tectonic
Issyk-Kul 6,240 702 1,730 25 mio, tectonic
Biwa 681 104 27.6 2 mio, tectonic
N. America
Tahoe 500 501 156 2 mio, tectonic

S. America
Maracaibo 13,010 280 36 mio, tectonic
Titicaca 8,562 284 903 2.8 mio, tectonic
a
References (8,44–49).
 LAKES 271

European Union in the EU countries. The regulations of 22. Hobbie, J.E. (1984). Lakes and Reservoirs. F.B. Taub (Ed.).
the EU-Water Framework Directive have initiated many Ecosystems of the World. Elsevier, Amsterdam, pp. 63–105.
activities to register, describe, and improve surface waters 23. Stumm, W. and Schnoor, J.L. (1995). Physics and Chemistry
and their state in Europe (see documents of the European of Lakes. A. Lerman, D.M. Imboden, and J.R. Gat (Eds.).
Environmental Agency (EEA) and EU-Commission). Springer, Berlin, New York. pp. 185–215.
There are data collections on world lakes for some 24. Stumm, W. and Morgan, J.J. (1995). Aquatic Chemistry 3.
regions and for selected countries: ILEC, EEA and EU- John Wiley & Sons, New York.
documents, EPA-documents (8,12,13,15,16,21,42,51–71) 25. Rodhe, W. (1949). Verh. Int. Verein. Limnol. 10: 377–386.
(31,32,72–75). For a series of important single lakes, there 26. Livingstone, D.A. (1973). Data of Geochemistry 6. Prof. Pap.
are monographic descriptions (11,76–80). U.S. Geological Survey, Denver, CO.
27. Imboden, D.M. and Wüest, A. (1995). Physics and chemistry
BIBLIOGRAPHY of lakes. A. Lerman, D.M. Imboden, and J.R. Gat (Eds.).
Springer, Berlin, New York, pp. 83–138.
1. Hutchinson, G.E. (1957). A Treatise on Limnology. 1: Geog-
raphy, Physics, and Chemistry. John Wiley & Sons, New 28. Livingstone, D.M. (1993). Verh. Int. Verein. Limnol. 25:
York. 75–81.
2. Hutchinson, G.E. (1967). A Treatise on Limnology. 2: Intro- 29. Ambrosetti, W. and Barbanti, L. (1999). J. Limnol. 58: 1–9.
duction to Lake Biology and the Limnoplankton. John Wiley 30. Reynolds, C.S. (1997). Vegetation Processes in the Pelagic: A
& Sons, New York. Model for Ecosystem Theory. Excellence in Ecology, No. 9.
3. Hutchinson, G.E. (1975). A Treatise on Limnology. 3: Limno- Ecology Institute, Oldendorf, Germany.
logical Botany. John Wiley & Sons, New York. 31. Vollenweider, R.A. (1980). OECD Eutrophication Programme.
4. Hutchinson, G.E. (1993). A Treatise on Limnology. 4: The The Canadian Project. OECD Eutrophication Programme,
Zoobenthos. John Wiley & Sons, New York. Burlington, VT.
5. Le Cren, E.D. and Lowe-McConnell, R.H. (1980). The Func- 32. Vollenweider, R.A. and Kerekes, J. (1980). OECD Eutrophi-
tioning of Freshwater Ecosystems. International Biological cation Programme. Synthesis Report. OECD Eutrophication
Programme; 22. Cambridge University Press, Cambridge, Programme, Burlington, VT.
UK. 33. Parsons, T.R., Takahasi, M., and Hargrave, B. (1977). Bio-
6. Margalef, R. (1983). Limnologı́a. Omega (Ed.). Barcelona, logical Oceanographic Processes 2. Pergamon Press, New
Spain. York.
7. Taub, F.B. (1984). Lakes and Reservoirs. Ecosystems of the 34. Odum, E.P. (1971). Fundamentals of Ecology 3. W.B.
World. Elsevier Science, Amsterdam. Saunders, Philadephia.
8. Meybeck, M. (1995). Les lacs et leur bassins. In: Limnologie 35. Jónasson, P. (1996). Verh. Internat. Verein. Limnol. 26: 1–33.
générale. R. Pourriot and M. Meybeck (Eds.). Masson, Paris, 36. Lindeman, R.L. (1942). Ecology 23: 339–418.
pp. 6–59.
37. Sheldon, R.W., Prakash, A., and Sutcliffe, W.H. (1972). Lim-
9. Wetzel, R.G. (2001). Limnology—Lake and River Ecosystems, nol. Oceanogr. 17: 327–340.
3rd Edn. Academic Press, San Diego.
38. Gaedke, U. (1992). Limnol. Oceanogr. 37: 1202–1220.
10. Meybeck, M. (1995). Physics and Chemistry of Lakes.
39. Bell, R.E. and Karl, D.M. (1998). Lake Vostok Work-
A. Lerman, D.M. Imboden, and J.R. Gat (Eds.). Springer,
shop—Final Report. NSF, Washington, DC.
Berlin, New York, pp. 1–36.
40. Geller, W. (1992). Aquatic Sci. 54: 37–57.
11. Kozhova, O.M. and Izmest’eva, L.R. (1998). Lake Baikal, 2nd
Edn. Biology of Inland Waters, Backhuys, Leiden, Germany. 41. Murdie, R.E., Pugh, D.T., and Styles, P. (1999). Geo-Marine
Lett. 18: 315–320.
12. Johnson, T.C. and Odada, E.O. (1996). The Limnology,
Climatology and Paleoclimatology of the East African Lakes. 42. Beadle, L.C. (1981) The Inland Waters of Tropical Africa.
OPA, Amsterdam. Longman, London.
13. Mandych, A.F. (1995). Enclosed Seas and Large Lakes of 43. Serruya, C. and Pollingher, U. (1983). Lakes of the Warm Belt.
Eastern Europe and Middle Asia. SPB, Amsterdam. Cambridge University Press, New York.
14. Varekamp, J.C., Pasternack, G.B., and Rowe, G.L., Jr. (2000). 44. Horie, S. (1984). Lake Biwa. Monographiae Biologicae. W.
J. Volcanology Geothermal Res. 97: 161–179. Junk, Dordrecht, The Netherlands.
15. Herdendorf, C.E. (1982). J. Great Lakes Res. 8: 379–412. 45. Kawanabe, H., Coulter, G.W., and Roosevelt, A.C. (1999).
16. Herdendorf, C.E. (1990). Large Lakes, Ecological Structure Ancient Lakes. Kenobi Productions, Ghent, UK.
and Function. M.M. Tilzer, and C. Serruya (Eds.). Springer, 46. Kira, T. (1993). Data Book of World Lake Environments. 1:
Berlin, New York, pp. 3–38. Asia and Oceania. International Lake Environment Commit-
17. Schwarz, H.E., Emel, J., Dickens, W.J., Rogers, P., and tee; UNEP, Kusatsu.
Thompson, J. (1990). The Earth as Transformed by Human 47. Kira, T. (1994). Data Book of World Lake Environments. 2:
Action. B.L. Turner, W.C. Clark, R.W. Kates, J.F. Richards, Africa and Europe. International Lake Environment Com-
J.T. Mathews, and W.B. Meyer (Eds.). Cambridge University mittee; UNEP, Kusatsu.
Press, Cambridge, UK, pp. 253–270. 48. Kira, T. (1995). Data Book of World Lake Environments. 3:
18. Lvovitch, M.I. (1973). Trans. Am. Geophys. Union 54: 28–42. The Americas. International Lake Environment Committee;
19. Magnuson, J.J. and Kratz, T.K. (2000). Verh. Int. Verein. UNEP, Kusatsu.
Limnol. 27: 74–87. 49. Kozhova, O.M. and Izmest’eva, L.R. (1998). Lake Baikal, 2nd
20. Williams, W.D. (1996). Verh. Int. Verein. Limnol. 26: 61–79. Edn. Biology of Inland Waters. Backhuys, Leiden.
21. Hammer, U.T. (1986). Saline Lake Ecosystems of the World. 50. Castro, J.M. and Moore, J.N. (2000). Environ. Geol. 39:
Junk, Dordrecht, The Netherlands. 1254–1260.
272 THE THEORY OF ALTERNATIVE STABLE STATES IN SHALLOW LAKE ECOSYSTEMS

51. Allan, R.J., Dickman, M., Gray, C.B., and Cromie, V. (1994). 76. Dejoux, C. and Iltis, A. (1992). Lake Titicaca. Monographiae
The Book of Canadian Lakes. Monograph Series No. 3. The Biologicae. Kluwer Academic, Dordrecht, The Netherlands.
Canadian Association on Water Quality, Gloucester, Ontario. 77. Jónasson, P. (1992). Thingvallavatn. OIKOS, Odense.
52. Boulton, A.J. and Brock, M.A. (1999). Australian Freshwater 78. Létolle, R. and Mainguet, M. (1996). Der Aralsee. Springer,
Ecology. Gleneagles, Glen Osmond, Australia. Berlin, New York.
53. Evans, M.S. (2000). Aquatic Ecosystem Health Manage. 3: 79. Saijo, Y. and Hayashi, H. (2001). Lake Kizaki. Backhuys,
65–79. Leiden.
54. Fricker, Hj. (1980). OECD Eutrophication Programme. 80. Wood, R.B. and Smith, R.V. (2002). Lough Neagh. Kluwer
Regional Project Alpine Lakes. Swiss Federal Board for Envi- Academic, Dordrecht, The Netherlands.
ronmental Protection, Bern.
55. IARH (2003). National data book of Argentine lakes and
lagoons and their environment. (preliminary report). Istituto THE THEORY OF ALTERNATIVE STABLE STATES
Argentino de Recursos Hidrocos, Buenos Aires.
56. ILEC (1988). Data Book of World Lake Environments. A
IN SHALLOW LAKE ECOSYSTEMS
Survey of the State of World Lakes. Otsu, Japan.
ADRIAN E. WILLIAMS
57. ILEC (1994). The Pollution of Lakes and Reservoirs. UNEP
APEM Ltd.
Environment Library 12. UNEP, Nairobi.
Manchester, United Kingdom
58. Jin, X. (1995). Lakes in China. China Open Press, Beijing.
59. Jolly, V.H. and Brown, J.M. (1975). New Zealand Lakes.
Auckland University Press, Oxford University Press, Auck- OVERVIEW
land, Oxford.
60. Kawanabe, H., Coulter, G.W., and Roosevelt, A.C. (1999).
The idea of alternative stable states was first proposed
Ancient Lakes. Kenobi Productions, Ghent, UK.
in the late 1960s (1), and within ecological communities,
61. Kira, T. (1993). Data Book of World Lake Environments. 1:
it was described mathematically in the 1970s (2,3).
Asia and Oceania. International Lake Environment Commit-
tee; UNEP, Kusatsu. The contemporary view of shallow lakes is that two
alternative stable states can exist over a range of
62. Kira, T. (1994). Data Book of World Lake Environments. 2:
Africa and Europe. International Lake Environment Com- nutrient concentrations: a clear water state dominated
mittee; UNEP, Kusatsu. by submerged macrophytes and a turbid phytoplankton-
63. Kira, T. (1995). Data Book of World Lake Environments. 3: dominated state (4–8). Unique states may be found at
The Americas. International Lake Environment Committee; either end of the nutrient continuum, but over a wide
UNEP, Kusatsu. range of nutrient concentrations (<50 to several thousand
64. Kristensen, P. and Hansen, H.O. (1994). European Rivers µg l−1 TP), either of these two states can exist, stabilized
and Lakes. EEA Environmental Monographs; 1. European by a number of buffer mechanisms (5,6,9–11).
Environmental Agency, Copenhagen. A ‘‘ball and cup’’ diagram is a simple way to visualize
65. Maitland, P.S., Boon, P.J., and McLusky, D.S. (2002). The this idea (Fig. 1). The position of the ecological system
Freshwaters of Scotland. John Wiley, Chichester, UK. (the ball) can change between the two states and settles
66. Moriarty, Ch. (2002). Studies of Irish Rivers and Lakes. at the bottom of the cup (stable equilibrium). The slope
Marine Institute, Dublin. of the cup determines both the speed and the direction of
67. Munawar, M. and Hecky, R.E. (2001). The Great Lakes of the the system change to a stable state. At either low or high
World (GLOW) Food-Web, Health and Integrity. Backhuys,
Leiden.
68. Rast, W. and Lee, G.F. (1978). OECD Eutrophication Pro-
gramme. The North American Project. OECD Eutrophication
Programme. EPA, Corvallis, OR.
69. Robertson, A. and Scavia, D. (1984). Lakes and Reservoirs.
Ecosystems of the World. F.B. Taub (Ed.). Elsevier, Amster-
dam, pp. 135–176.
70. Tilzer, M.M. and Serruya, C. (1990). Large Lakes—Ecological
Structure and Function. Springer, Berlin, New York.
71. Taub, F.B. (1984). Lakes and Reservoirs. Ecosystems of the Low
World. Elsevier Science, Amsterdam.
72. Willén, E. (1984). Lakes and Reservoirs. Ecosystems of the
World. F.B. Taub (Ed.). Elsevier, Amsterdam, Nutrients
pp. 107–134.
73. Williams, W.D. (1984). Lakes and Reservoirs Ecosystems
of the World. F.B. Taub (Ed.). Elsevier, Amsterdam, High
pp. 499–520. Macrophyte Algal
74. Viner, A.B. (1987). Inland Waters of New Zealand. DSIR, clear state turbid state
Wellington. Figure 1. ‘‘Ball and cup’’ schematic of the potential for two
75. Zhadin, V.I. and Gerd, S.V. (1963). Fauna and Flora of the stable states (clear water macrophyte-dominated versus turbid
Rivers, Lakes and Reservoirs of the U.S.S.R. Translation 1963. algal-dominated) to exist within a shallow lake under a wide
Israel Progr. Sci. Transl. Ltd., Jerusalem. range of nutrient conditions.
 THE THEORY OF ALTERNATIVE STABLE STATES IN SHALLOW LAKE ECOSYSTEMS 273

nutrient levels, one of the two stable states exists—the may even release chemicals toxic to macrophytes (28).
ball can only be in one location because of the slope of In addition, phytoplankton-dominated open water has
the horizon. However, between these points the system few refugia for grazing Cladocera; thus, any Cladocera
can rest in either cup (state). The ease with which one venturing forth are typically removed by zooplanktivorous
stable state can be switched to the other is indicated by fish, which may be compounded by the fact that, with
the position of the two balls with respect to one another reduced macrophytes, few large macroinvertebrates would
and the slopes that separate them. Therefore, if the be available to large fish, favoring smaller sized fish
starting position is a pristine, low-nutrient (oligotrophic), feeding largely on zooplankton (6). Finally, in poorly
macrophyte-dominated, clear water lake, the nutrient flushed systems, phytoplankton species can consist of
levels can increase substantially before the only possible large filamentous or colonial blue-green algal species
scenario is a turbid algal-dominated lake. However, to that are inedible to zooplankton, which along with the
reverse the situation takes more than just reducing the above mechanisms, all help to maintain phytoplankton-
nutrients as the ball can remain in the turbid state dominated water states over a wide range of nutrient
even at relatively low nutrient levels. To switch the state concentrations.
prematurely requires that a force be applied to the ball By understanding the mechanisms that buffer and
to push it over the edge of the algal cup and into the thus maintain the alternate stable states, it is possible
macrophyte state. to attempt to force a shift or switch from one state
Therefore, each stable state has a number of buffer sys- to another by manipulating top-down and bottom-up
tems that help maintain that state, and through top-down processes. Such attempts are often made during lake
and bottom-up processes, fish, plankton, invertebrates, restoration projects when a desire exists to shift an algal-
macrophytes, and nutrients are all entwined in reciprocal dominated turbid lake into a clear water macrophyte-
feedback mechanisms that not only determine the state dominated state. However, the long-term success of lake
that is present (12,13) but can also be manipulated to force restoration projects, although they use knowledge of top-
a shift. down and bottom-up processes, is not guaranteed and
Macrophyte buffer mechanisms that help maintain unexpected results are common (29–31). The uncertainty
and stabilize a clear water state in shallow lake sys- that surrounds these outcomes often stems from a lack of
tems are numerous (4,14–16). Macrophytes act as habi- understanding about the buffer systems that stabilize the
tats and refugia for macroinvertebrates and cladocerans alternate states within the specific lake under restoration
that reduce the epiphyton and phytoplankton commu- and an oversimplification of the top-down and bottom-up
nities by grazing (6,17). Davis (18) suggested that the processes involved (13).
plant bed environment, often deoxygenated, favors graz-
ers by discouraging their predators. Macrophytes may
also release allelopathic chemicals (19); partake in the BIBLIOGRAPHY
‘‘luxury uptake’’ of nutrients, removing available nutri-
ents for phytoplankton (20–22); and reduce available 1. Lewontin, R.C. (1969). The meaning of stability. Brookhaven
nitrate by anaerobic decomposition processes such as de- Symp. Biol. 22: 13–23.
nitrification (23). Moreover, certain species of macrophytes 2. Noy-Meir, I. (1975). Stability of grazing systems: An applica-
have been shown to oxidize the sediment and reduce tion of predator-prey graphs. J. Ecol. 63: 459–481.
the release of phosphorus. With less available phospho- 3. May, R.M. (1977). Thresholds and breakpoints in ecosystems
rus in the water column, phytoplankton populations are with a multiplicity of states. Nature 269: 471–477.
reduced (24,25). In addition, stands of macrophytes reduce 4. Blindow, I., Andersson, G., Hargeby, A., and Johansson, S.
water movement within them. In consequence, suspended (1993). Long term pattern of alternative stable states in two
sediment resettles and turbidity falls, aiding the growth shallow eutrophic lakes. Freshwater Biol. 30: 159–167.
of macrophytes (26). Moss (10) suggests that even if light 5. Scheffer, M. et al. (1993). Alternative equilibria in shallow
penetration is falling, submerged macrophytes may switch lakes. T.R.E.E. 8(8): 275–279.
from low growing forms to taller species and that plants 6. Timms, R.M. and Moss, B. (1984). Prevention of growth of
heavily encrusted with epiphyton may be able to shed potentially dense phytoplankton populations by zooplankton
grazing, in the presence of zooplanktivorous fish, in a shallow
their leaves and produce new growth. These processes
wetland ecosystem. Limnol. Oceanogr. 29(3): 472–486.
competitively disadvantage phytoplankton and can buffer
7. Hosper, S.H. (1989). Biomanipulation, new perspective for
the macrophyte-dominated state.
restoring shallow, eutrophic lakes in The Netherlands.
Once established, on the other hand, phytoplankton can Hydrolog. Bull. 23: 5–11.
buffer the switch back to a macrophyte state by growing
8. Irvine, K., Moss, B., and Balls, H. (1989). The loss of
much earlier in the season than macrophytes in temperate submerged plants with eutrophication II. Relationships
regions. In doing so, they can curtail the development of between fish and zooplankton in a set of experimental ponds,
turions, rhizomes, or seeds by shading in spring or reduce and conclusions. Freshwater Biol. 22: 89–107.
the formation of propagules by shading and competition for 9. Moss, B. and Leah, R.T. (1982). Changes in the ecosystem of a
CO2 in late summer (10). Submerged plants require carbon guanotrophic and brackish shallow lake in eastern England;
for growth, but algae have shorter CO2 and HCO3 diffusion potential problems in its restoration. Internationale Revue
pathways owing to their small size, thus removing the der Gesamten Hydrobiologie 67: 625–639.
carbon available to the bulkier macrophytes (27). Some 10. Moss, B. (1990). Engineering and biological approaches to
researchers have suggested that certain blue-green algae the restoration from eutrophication of shallow lakes in
274 NOAA LAKE LEVEL FORECAST FOR LAKE MICHIGAN RIGHT ON TARGET

which aquatic plant communities are important components. 29. Marsden, M.W. (1989). Lake restoration by reducing external
Hydrobiologia 200/201: 367–377. phosphorus loading: The influence of sediment phosphorus
11. Scheffer, M. (1990). Multiplicity of stable states in freshwater release. Freshwater Biol. 21: 139–162.
systems. Hydrobiologia 200/201: 475–486. 30. De Melo, R., France, R., and McQueen, D.J. (1992). Bioma-
12. Williams, A.E., Moss, B., and Eaton, J. (2002). Fish induced nipulation: Hit or myth? Limnol. Oceanogr. 37(1): 192–207.
macrophyte loss in shallow lakes: Top-down and bottom-up 31. Moss, B. et al. (1997). Vertically challenged limnology;
processes in mesocosm experiments. Freshwater Biol. 47: contrasts between deep and shallow lakes. Hydrobiologia
2216–2232. 342/343: 257–267.
13. Williams, A.E. and Moss, B. (2003) Effects of different fish
species and biomass on plankton interactions in a shallow
lake. Hydrobiologia 491: 331–346. NOAA LAKE LEVEL FORECAST FOR LAKE
14. Canfield, Jr., D.E. et al. (1984). Prediction of chlorophyll MICHIGAN RIGHT ON TARGET
a concentrations in Florida lakes: Importance of aquatic
macrophytes. Canad. J. Fisheries Aquat. Sci. 41: 497–501.
15. Coops, H. and Doef, R.W. (1996). Submerged vegetation NOAA Great Lakes
development in two shallow, eutrophic lakes. Hydrobiologia. Environmental Research Lab
340: 115–120.
16. Romo, S., Van Donk, E., Gylstra, R., and Gulatis, R. (1996).
A multivariate analysis of phytoplankton and food web
changes in a shallow biomanipulated lake. Freshwater Biol.
36: 683–696.
17. Carpenter, S.R. and Lodge, D.M. (1986). Effects of submerged
macrophytes on ecosystem processes. Aquatic Botany 26:
341–370.
18. Davis, J.C. (1975). Minimal dissolved oxygen requirements of
aquatic life with emphasis on Canadian species: a review. J.
Fisheries Res. Board Canada 32: 2295–2332.
19. Wium-Andersen, S., Cristophersen, A.C., and Houen, G.
(1982). Allelopathic effects on phytoplankton by substances
isolated from aquatic macrophytes (Charales). Oikos 39:
187–190.
20. Phillips, G.L., Eminson, D., and Moss, B. (1978). A mech-
anism to account for macrophyte decline in progressively
eutrophicated freshwaters. Aquatic Botany 4: 103–126.
21. Van Donk, E., Gulati, R.D., and Grimm, M.P. (1989). Food
web manipulation in Lake Zwemlust: positive and negative
effects during the first two years. Hydrolog. Bull. 23: 19–34.
22. Ozimek, T., Van Donk, E., and Gulati, R.D. (1993). Growth
and nutrient uptake by two species of Elodea in experimental
conditions and their role in nutrient accumulation in a
macrophyte dominated lake. Hydrobiologia. 251: 13–18.
July 17, 2003—How high or low the water level is in the
23. Takii, S. and Fukui, M. (1996). Comparison of anaerobic
five Great Lakes has long been of interest to those who
mineralization processes in sediments between littoral
use the lakes for commercial and recreational boating and
reed and offshore sites in a shallow hypertrophic lake.
Hydrobiologia. 319: 37–45.
fishing, energy, and as a source of drinking water. The
levels have been measured and forecast monthly for many
24. Carpenter, S.R., Elser, J.J., and Olson, K.M. (1983). Effects
of roots of Myriophyllum verticillatum L. on sediment redox
years, and NOAA scientists hit the bull’s-eye with their
conditions. Aquatic Botany 17: 243–250. July forecast of the water level of Lake Michigan.
‘‘We were within inches of the actual measurement,’’
25. Van Donk, E., Grimm, M.P., Gulati, R.D., and Klien
Breteler, J.P.G. (1990). Whole lake food web manipulation said Cynthia Sellinger, deputy director of the NOAA
as a means to study community interactions in a small Great Lakes Environmental Research Laboratory in Ann
ecosystem. Hydrobiologia 200/201: 275–289. Arbor, Mich.
26. Schiemer, F. and Prosser, M. (1976). Distribution and GLERL uses meteorological forecasts from the NOAA
biomass of submerged macrophytes in Neusiedlersee. Aquatic National Weather Service to forecast the lake levels six
Botany 2: 289–307. months out. Back in January, GLERL’s Lake Michigan
27. Simpson, P.S. and Eaton, J.W. (1986). Comparative studies forecast was 176.61 meters—about 579 feet; in July the
of the photosynthesis of the submerged macrophyte Elodea actual level was 176.65 meters—approximately 2 inches
canadensis and the filamentous algae Cladophora glomerata higher than the forecast. Sellinger meets monthly with her
and Spirogyra sp. Aquatic Botany 24: 51–12.
28. Van Vierssen, J. and Prins, Th.C. (1985). On the relationship
between the growth of algae and aquatic macrophytes in This article is a US Government work and, as such, is in the
brackish water. Aquatic Botany 21: 165–179. public domain in the United States of America.
 SUBMERGED AQUATIC PLANTS AFFECT WATER QUALITY IN LAKES 275

NOAA Great Lakes Water Levels

NOAA Great Lakes Satellite Information

MEDIA CONTACT
Jana Goldman, NOAA Research, (301) 713–2483.
(Photo courtesy Greg Lang of the NOAA Great Lakes
Environmental Research Laboratory.)

SUBMERGED AQUATIC PLANTS AFFECT WATER

QUALITY IN LAKES
KARL E. HAVENS
South Florida Water Mgmt.
colleagues from Environment Canada and the U.S. Army District
Corps of Engineers to forecast the levels. West Palm Beach, Florida
Since 1997, Lake Michigan has dropped 4.1 feet, the
largest drop in that short of time span since records have
been kept. Lakes Michigan and Huron are often considered BACKGROUND INFORMATION
one body of water as they are connected by the Straits of
Mackinac, located between the upper and lower peninsulas A principal cause of water quality impairment in lakes and
of the state of Michigan. reservoirs is enrichment by excessive levels of nutrients,
The other lakes are more difficult to forecast, because resulting in undesirable conditions that include noxious
the water flow of Superior and Ontario are regulated by and/or toxic algae blooms (1), and taste and odor problems
power companies and Erie gets 80 percent of its water (2). The nutrient content of lake water is determined
from Michigan and Huron, Sellinger explained. by a myriad of factors, including external loading rate,
The Great Lakes—Superior, Michigan, Huron, Ontario, water residence time and depth, geochemical interactions
and Erie—are the major bodies of water within the Great in the water body, and biological processes. Biological
Lakes Basin. The basin is located along the international processes are particularly important in shallow lakes.
boundary between Canada and includes portions of eight Fish and benthic animals recycle nutrients from bottom
states and the provinces of Ontario and Quebec. Com- sediments, buoyant algae carry nutrients upward in the
bined, these lakes make up the largest surface fresh water water column, settling rates of algae determine loss rates
body in the world—they contain 90 percent of the U.S.’s for water column nutrients, and plants facilitate various
fresh water and 20 percent of the world’s fresh water. nutrient-scavenging processes (3,4). Submerged plants, in
The Great Lakes carry thousands of recreational and particular, can have major influences on water quality
commercial ships annually. Lake levels are important to in terms of clarity, biomass of noxious algae, taste and
shippers, who must load the lake carriers with fewer goods odor, organics, and suspended solids—factors of major
when the levels are low and to recreational boaters, who relevance to society’s use of the water.
often have to find dock space in deeper water. The Lakes This chapter addresses four questions:
provide drinking water to the 40 million residents as well
as water to power and cool electric generating plants. 1. What types of submerged plants occur in lakes,
Lake levels have been dropping for the past six years and how does type of plant relate to water
and lake evaporation has increased significantly. quality benefits?
‘‘The only good thing that comes out of lower lake levels 2. What factors control the biomass and spatial extent
is more beach for beach-goers,’’ according to Sellinger. of submerged plants?
However, that also exposes and destroys some vital 3. What are the processes whereby plants affect lake
breeding habitat for lake wildlife. water quality, in particular, the concentrations
Lake levels have been forecast before GLERL was of nutrients, turbidity, and the occurrence of
formed in 1974. Part of the laboratory’s mandate is to noxious algae?
provide the lake level forecasts, a task that was given to 4. How can lakes be managed to promote the
GLERL when the research arm of the U.S. Army Corps of development of submerged plants that will help to
Engineers was transferred to the laboratory. improve water quality?
NOAA is dedicated to enhancing economic security
and national safety through the prediction and research For a detailed overview of the ecological role of sub-
of weather and climate-related events and providing merged plants in shallow lakes, the reader may consult
environmental stewardship of the nation’s coastal and Jeppesen et al. (5) or Scheffer (3). For identification of dif-
marine resources. NOAA is part of the U.S. Department ferent types of plants in lakes and for general information
of Commerce. about their biology and distribution, there are a number
of web-based resources provided by major universities; for
RELEVANT WEB SITES
example: http://aquat1.ifas.ufl.edu (University of Florida,
NOAA Great Lakes Environmental Research Lab Center for Aquatic and Invasive Plants).
276 SUBMERGED AQUATIC PLANTS AFFECT WATER QUALITY IN LAKES

TYPES OF SUBMERGED AQUATIC VEGETATION RELATIVE can be difficult to tell whether plants are responding to
TO WATER QUALITY BENEFITS variations in underwater irradiance or whether they are
the cause of high irradiance in certain areas of a lake
Lakes can support rooted vascular plants (true plants because they have various functions that reduce turbidity.
that have tissues for gas and materials transport); plant- It is a classic ‘‘chicken versus egg’’ paradox. Some shallow
like macroalgae, such as Chara (shrimp grass) that do lakes display zones of clear water immediately adjacent
not have vascular tissues or roots; and filamentous algae to turbid water of the same depth (8), the only difference
that can form dense mats on the lake bottom, sometimes is that the clear areas have dense beds of submerged
covering rocks and even other plants. Filamentous algae aquatic vegetation (SAV). In such instances, one might
are least desirable from a water quality standpoint, conclude that the plants have caused the clear water. In
because at high biomass, they can kill other plants by Lake Okeechobee, Florida, clear water expanded over an
shading and they form floating noxious mats at the ever-increasing area during summer 2000, coincident with
lakeshore when they decompose (3). Chara generally is the expansion of large beds of Chara. Areas adjacent to
a pioneer species that rapidly invades lake habitats after the plant beds had more turbid water, but they also had
a disturbance (6,7), and because it does not have roots, slightly greater depths (6). It was not possible to establish
all of its accumulated nutrient content may be recycled a cause–effect relationship. Scheffer (3) provides a good
back into the water column if unfavorable conditions (e.g., overview of the issue of correlation versus causation as it
reduced light, high wind, or cold water) cause the plants pertains to SAV and turbidity in lakes, for those interested
to be lost. Certain vascular plants, such as Ceratophyllum in more information on this subject.
(coontail), also do not have roots and would have this same In addition to depth and turbidity, sediment type
problem. In regard to Chara, note that clear water and may influence the distribution and biomass of submerged
low nutrient levels are most often associated with dense plants. Soft sediments that are readily suspended by wind,
beds of this plant in temperate (8,9) and subtropical (7) waves, and fish may not support plants, especially types
lakes and that processes other than direct incorporation like Chara that do not have roots. In Lake Okeechobee,
of nutrients into biomass may be most important for which has areas of sand, peat, and mud sediments, Chara
water quality improvement by Chara beds. Rooted does not occur at locations that have soft organic mud, even
vascular plants—common examples in North America are when depth and light conditions appear to be favorable
Vallisneria (eelgrass) and Potamogeton (pondweed)—are (7). A simple empirical model, based on water depth and
more permanent than Chara, and because they can sediment type, was able to predict at 75% accuracy the
transfer nutrients to their underground roots prior to spatial distribution of Chara over a 170 km2 landscape
seasonal ‘‘die-off,’’ they have potential to provide a better in that lake. The sediment type found also was the
long-term nutrient sink. major factor controlling where plants could occur in Lake
Kinneret, Israel (17), whereas water depth was the factor
FACTORS CONTROLLING THE BIOMASS AND EXTENT OF controlling the timing of plant occurrence.
SUBMERGED VEGETATION Other factors that can influence the distribution and
biomass of submerged plants include the degree of
Scientists and lake managers have given considerable exposure to wind and waves (13,18), the slope of the
attention to the factors that control the biomass and lake bottom (19), and as mentioned before, overgrowth by
spatial extent of submerged plants in lakes. Obviously, filamentous algae (20).
water depth plays a key role because ultimately it
determines the amount of light that reaches the lake PROCESSES WHEREBY SUBMERGED VEGETATION
bottom where plants germinate from seeds (or oospores, AFFECTS WATER QUALITY
in the case of algae) and begin to grow. However, light
availability at any given depth is heavily influenced by When one compares lakes with and without dense
the turbidity of the water, which is determined by the submerged plants or follows the conditions in a lake over
amount of suspended solids, algal cells, and dissolved time as plants increase, striking changes in water quality
organic matter. Hence, a number of authors have identified are apparent. Water can change from murky green (algae
significant relationships between the maximal depth for blooms) to crystal clear (Fig. 1), and at the same time,
submerged plant occurrence in lakes and simple indicators nutrient levels in the water can dramatically decline. Even
of turbidity, such as Secchi disk transparency (10–13). The a relatively low biomass of submerged plants can attenuate
relationships vary from lake to lake. For example, in Lake peak concentrations of suspended algae (phytoplankton)
Okeechobee, Florida, the published models predict that in a lake’s water column (21; Fig. 2). The benefits of clear
submerged plants can occur at depths between 0.7 and water and plants for the fish, wildlife, and humans who
1.8 m, whereas plants actually occur at depths as great use the resource are obvious.
as 2.0 m. This may reflect the occurrence of plants that Submerged plants mediate changes in water quality by
are adapted to low light levels in this turbid shallow lake a variety of mechanisms. The focus here is on processes
(14). In general, Chara can occur at greater depths than that control nutrient and light availability, rather than
vascular plants, for any given level of turbidity (15). food web interactions. However, it should be recognized
Experimental studies have confirmed the causal effect that a potentially important function of submerged plants
of underwater light availability on submerged plant is to provide refuge for large zooplankton (22). During the
growth (14,16). However, when evaluating field data, it daytime, large zooplankton can escape fish predation by
 SUBMERGED AQUATIC PLANTS AFFECT WATER QUALITY IN LAKES 277

(a) and reduced risk of nuisance algal blooms, although

uncertainty remains about the magnitude of this effect.
One of the complicating issues is that plants may also
provide refuge for the small fish that eat zooplankton (22).
In a conceptual framework (Fig. 3), one mechanism
whereby submerged plants can benefit water quality is
the uptake (U) of dissolved nutrients that otherwise could
fuel algal blooms and the array of adverse impacts shown
in the diagram. Benthic algal mats can also take up
nutrients from the water, further reducing availability for
the bloom-forming phytoplankton. Much of the nutrient
uptake by plants actually is associated with attached
algae (periphyton) that grow on plant surfaces (23,24).
Vascular plants and benthic algae also take up nutrients
from lake sediments; it has been documented that benthic
algal mats can substantially reduce the diffusive flux (D)
(b)
of nutrients from sediments to the overlying water in both
shallow freshwater and estuarine systems (25–27). All of
the plants and algae mentioned here can at times excrete
(E) dissolved nutrients back into the sediment or water
column. Conditions that favor net growth of plants and/or
benthic algal mats favor a net uptake of nutrients, which
translates into better water quality.
Nutrients present in the water column or sediments
in a dissolved form can also be transformed (T) into
particulates that ultimately settle (S) to the lake bottom
at a rate that depends on water column mixing (or lack
thereof), the density of the water, and the characteristics of
the particles. Plants can stimulate both the transformation
of dissolved nutrients to particulates in the water and the
rate of settling. In regard to transformation (T), when
Figure 1. Photographs of typical conditions in turbid, phyto-
plants carry out intense photosynthesis, the pH of the
plankton-dominated lakes that have high levels of nutrients
in the water column and blooms of algae (a), versus clear,
surrounding water can be substantially elevated. At high
plant-dominated lakes that have low nutrient levels in the water pH (e.g., >10), much of the dissolved inorganic carbon
and no algal blooms. These photographs were taken by the author (C) occurs as carbonate ions (CO3 )−2 (28), which can
at Lake Okeechobee, Florida. be precipitated to the sediments, or onto the surface of
plants, as calcium carbonate (CaCO3 ). More important,
from a water quality standpoint, is that this process can
60 be accompanied by coprecipitation of phosphorus (29), as
No plants
insoluble hydroxyapatite [Ca10 (PO4 )6 (OH)2 ]. Phosphorus
50 Plants
Phytoplankton, ug L−1

is a nutrient that often is limiting to phytoplankton in

40 freshwater lakes (30). Therefore, coprecipitation could help
reduce phytoplankton growth and suppress formation of
30 blooms. However, it must be noted that at high pH, the
release of dissolved phosphorus from lake sediments (D)
20
is sometimes enhanced, even under oxidized conditions
10 (31). Depending on circumstances, this could counteract
the effects on particle transformation.
0 Plants play an important role in settling of particles (S)
1999 2000 2001 2002 by slowing the rate of water movement (32), and acting
Figure 2. Data from the shoreline region of Lake Okeechobee, as sediment ‘‘traps’’ by intercepting particles using their
Florida, showing the concentrations of phytoplankton chlorophyll leaves and stems; the particles subsequently sink to the
(µgL−1 ) at sampling stations with plants versus stations lake bottom (33). The uptake of nutrients by periphyton,
without plants. The presence of plants dampens the peaks in
which ultimately provide particulate material (dead cells)
phytoplankton, as a result of various processes described in the
text. Modified from a figure in Havens (20).
to the sediments is another loss process mediated by
plants. Once the particles settle within a plant bed, they
are less likely to be re-suspended (R) because dense beds of
migrating into the plant zone, and at night, when they submerged plants significantly reduce the amount of wind
are not susceptible to those visual predators, they can and wave energy that reaches the sediment surface (34).
migrate out into the open water to graze on phytoplankton. Finally, it has been suggested that submerged plants
The result is theoretically a reduction in algal biomass might inhibit the growth of suspended algae by releasing
278 SUBMERGED AQUATIC PLANTS AFFECT WATER QUALITY IN LAKES

A
SAV Blooms
B.O.D.
toxins
Planktonic solids
algae organics

U,E M
Water

U,E Dissolved T Particulate

nutrients nutrients
& solids

U,E S
D R

Algae

U,E U,E M
T
Dissolved Particulate
nutrients nutrients Sediment
M

Figure 3. A simple conceptual model, illustrating how submerged aquatic plants and benthic
algal mats influence the dynamics of phosphorus in a typical eutrophic lake. The arrows
correspond to various processes: A = release of allelopathic chemicals; U = uptake of dissolved
nutrients; E = excretion of dissolved nutrients; T = transformation of nutrients from dissolved
to particulate form, or vice versa; D = diffusion of dissolved nutrients from sediments to water,
or vice versa; M = mortality; S = settling of particles; and R = resuspension of particles by
wind, waves, or bioturbation. B.O.D. = Biological Oxygen Demand. The two large arrows point
to impacts of algal blooms and solids on the water resource. This figure is modified from a similar
model presented in Havens et al. (23).

allelopathic chemicals (A) into the water column (35). This likely to be light availability, which as noted before, varies
effect has been documented in controlled tank studies with with the depth and turbidity of the water. One of the
a high biomass of plants and restricted water volume, factors controlling turbidity is the nutrient content of the
but uncertainty remains regarding its importance in water, which determines the abundance of phytoplankton
lake ecosystems. cells that absorb light. Scheffer (38) noted that there is
a complex relationship between nutrients and turbidity
MANAGING LAKES TO PROMOTE SUBMERGED in lakes that varies with the presence or absence of
VEGETATION plants. This relationship (Fig. 4) is highly relevant when
considering how to establish plants in a nutrient-rich,
The total biomass of primary producers in a lake is turbid lake.
controlled by nutrient input to the ecosystem (36), As a lake becomes enriched with nutrients, the turbidity
whereas partitioning of that biomass into phytoplankton of the water increases. In lakes with plants, the turbidity
and plants is determined by a complex set of factors. at any given nutrient level is lower than that in lakes
Understanding how particular environmental conditions without plants due to the mechanisms described before
influence submerged plants is critical to facilitating plant (Fig.3). Once a ‘‘critical turbidity’’ is reached, where light
survival and growth, and in turn, maintaining good water does not allow net growth to occur, plants are lost,
quality (see below). Lakes dominated by soft organic and turbidity quickly increases (upward arrow), as the
sediments or heavily influenced by wind-driven waves are nutrients formerly sequestered by plants go into the
not well suited for submerged plants. However, even under biomass of phytoplankton in the water. To restore plants
these conditions, options may exist, such as dredging to to a lake that has reached this state, nutrients must be
remove soft mud or construction of wave barriers to provide reduced to a lower level than where the switch previously
protected areas for plants to grow (37). In lakes that have occurred (downward arrow). Both the clear-water, plant-
favorable sediments and low to moderate wave energy dominated state and the turbid-water, phytoplankton-
along the shoreline, the key factor controlling plants is dominated state are highly stable—it takes a large
 SUBMERGED AQUATIC PLANTS AFFECT WATER QUALITY IN LAKES 279

nts
o pla Shallow
N

Turbidity
Turbidity

nts
o pla
Critical N Deep
turbidity
nts nts
h pla pla
Wit Wit
h

Nutrient level Nutrient level

Figure 4. Relationship between turbidity and nutrient level in Figure 5. Effects of water depth on the occurrence of plants
lake water, in lakes with versus without plants. Lakes with in turbid lakes. By substantially lowering the water level of a
plants have a lower turbidity for any given nutrient level than lake, the critical turbidity is increased (upper dashed line), and
lakes without plants. When nutrient levels increase in lakes with this can allow plants to recover (downward arrow), even though
plants, the plants can be lost once a critical turbidity is reached, the nutrient level remains high. If the water level increases
at which point the net growth of plants is zero. When this occurs, substantially (lower dashed line), the plants may be lost, even at
turbidity can rapidly increase (upward arrow). To restore the a lower level of nutrients than that where they were recovered
plants in such a lake (downward arrow), the nutrient level must (upward arrow). Unless there is capability for strong control of
be substantially reduced. This figure is modified from a diagram water levels and no other adverse impacts of keeping the lake low
in Scheffer (3). (e.g., impacts on water supply), the ultimate solution to keeping
plants and clear water in a lake is to reduce external nutrient
inputs. This figure is based on information in Scheffer (3).
input of nutrients to switch a plant-dominated lake to
a phytoplankton lake and an even larger reduction of
nutrients to switch the lake back (3,38). This stability spatial extent of submerged plants increased from less
of states reflects the biological feedback loops that than 10 to over 170 km2 (16). Water in shoreline areas,
occur in the system. Plants cause clear water, which which was highly turbid from algae blooms in the late
favors more plants, which cause even clearer water. . .etc. 1990s, became crystal clear in 2000–2002. Recently, it
Phytoplankton cause turbid water, which reduces plants, has somewhat deteriorated, as higher water levels led to
which favors more phytoplankton, which further increases reduced biomass of plants. Ultimately, reduced nutrient
turbidity and so on. input is the solution to the water quality problem in this
Note that at extremes along the nutrient axis and other nutrient-enriched lakes (39), because as long
(extremely low or high nutrient levels), plants can exist as alternative states exist, uncontrollable events, such as
in only one state—clear or turbid, respectively. However, marked increases in water level or wind storms, could
for lakes that have moderate to high levels of nutrients potentially shift a plant-dominated lake back to a turbid
and the possibility of alternative stable states (38), one lake with phytoplankton blooms.
method for switching a lake from phytoplankton to plant
dominance is to reduce inputs of limiting nutrients, such
as phosphorus and nitrogen (3,39). Ultimately, this is the SUMMARY
path to lake restoration. However, an interim management
strategy that can sometimes be used in tandem with Water quality in shallow lakes and reservoirs can be
nutrient reduction is to lower the lake’s water level. strongly influenced by biological processes, including those
Critical turbidity is a function of depth, so that in a deep associated with fish, invertebrates, algae, and plants. The
lake, relatively low turbidity might be sufficient to prevent most striking influences often are those associated with
plant growth. However, in shallow water, enough light submerged aquatic plants, which include rooted vascular
may reach the lake bottom to allow plants to germinate plants, loosely attached macroalgae, such as Chara, and
and grow. By lowering the water level, it is possible to filamentous benthic algal mats. Where conditions favor
raise the critical turbidity (Fig. 5), so that plants can high biomass and spatial extent of these plants, water
become reestablished at a higher level of nutrients than turbidity generally is low, biomass of phytoplankton is
that when the lake had deeper water. At this point, the low, and quality of water for drinking and other uses is
lake has essentially been reset onto the lower curve. good. The plants maintain these conditions by removing
A switch from turbid to clear water happened in nutrients from the water; supporting attached algae that
the shoreline region of Lake Okeechobee, Florida, in also sequester nutrients; stabilizing bottom sediments;
summer 2000, after an intentional lowering of the water and a variety of other physical, chemical, and biological
level was followed by a major drought (40). The water processes. Actions to favor development of submerged
level in the lake dropped by more than 2 m, and the plants in shallow, nutrient-enriched lakes can help to
280 SUBMERGED AQUATIC PLANTS AFFECT WATER QUALITY IN LAKES

maintain good water quality, although the long-term Lake Kinneret, Israel: Taking advantage of a window of
solution still is to reduce external nutrient inputs. opportunity. Aquatic Bot. 62: 249–267.
18. Chambers, P.A. (1987). Nearshore occurrence of submersed
aquatic macrophytes in relation to wave action. Can. J. Fish
BIBLIOGRAPHY Aquatic Sci. 44: 1666–1669.
19. Duarte, C.M. and Kalff, J. (1986). Littoral slope as a predictor
1. Paerl, H.W. (1988). Nuisance phytoplankton blooms in of maximum biomass of submerged macrophyte communities.
coastal, estuarine, and inland waters. Limnol. Oceanogr. 33: Limnol Oceanogr. 31: 1072–1080.
823–847. 20. Havens, K.E. (2003). Submerged aquatic vegetation correla-
2. Smith, V.H. et al. (2002). Managing taste and odor problems tions with depth and light attenuating materials in a shallow
in a eutrophic drinking water reservoir. Lake Reservoir subtropical lake. Hydrobiologia 493: 173–186.
Manage 18: 319–323. 21. Weisner, S.E.B., Strand, J.A., and Sandsten, H. (1997).
3. Scheffer, M. (1998). The Ecology of Shallow Lakes. Chapman Mechanisms regulating abundance of submerged vegetation
and Hall, London, UK, p. 357. in shallow eutrophic lakes. Oecologia 109: 592–599.
4. Havens, K.E. and Schelske, C.L. (2001). The importance of 22. Jeppesen, E., Lauridsen, T.L. Kairesalo, T., and Perrow, M.R.
considering biological processes when setting total maximum (1998). Impact of submerged macrophytes on fish-zooplankton
daily loads (TMDL) for phosphorus in shallow lakes and interactions in lakes. In: The Structuring Role of Sub-
reservoirs. Environ. Pollut. 113: 1–9. merged Macrophytes in Lakes. E. Jeppesen, M. Sondergaard,
5. Jeppesen, E., Sondergaard, M., Sondergaard, M., and M. Sondergaard, and K. Christoffersen (Eds.). Springer,
Christoffersen, K. (Eds.). (1998). The Structuring Role of Berlin, pp. 353–368.
Submerged Macrophytes in Lakes. Springer, Berlin, p. 423. 23. Havens, K.E. et al. (2001). Complex interactions between
6. Havens, K.E. et al. (2001). Hurricane effects on a shallow lake autotrophs in shallow marine and freshwater ecosystems:
ecosystem and its response to a controlled manipulation of Implications for community responses to nutrient stress.
water level. Sci. World J. 1: 44–70. Environ. Pollut. 113: 95–107.
7. Havens, K.E. et al. (2002). Large-scale mapping and predic- 24. Vadeboncoeur, Y. and Steinman, A.D. (2002). Periphyton
tive modeling of submerged aquatic vegetation in a shallow function in lake ecosystems. Sci. World J. 2: 1449–1468.
eutrophic lake. Sci. World J. 2: 949–965. 25. Hansson, L.A. (1988). Effects of competitive interactions on
8. Scheffer, M. et al. (1994). Vegetated areas with clear water in the biomass development of planktonic and periphytic algae
turbid shallow lakes. Aquatic Bot. 49: 193–196. in lakes. Limnol Oceanogr. 33: 121–128.
9. Van den Berg, M.S., Coops, H., Meijer, M.L., Scheffer, M., 26. Hansson, L.A. (1989). The influence of a periphytic biolayer
and Simons, J. (1998). Clear water associated with a dense on phosphorus exchange between substrate and water. Archiv
Chara vegetation in the shallow and turbid Lake Veluwe- fur Hydrobiologie 115: 21–26.
meer, The Netherlands. In: The Structuring Role of Sub- 27. McGlathery, K.J., Krause-Jensen, D., Rysgaard, S., and
merged Macrophytes in Lakes. E. Jeppesen, M. Sondergaard, Christensen, P.B. (1997). Patterns of ammonium uptake
M. Sondergaard, and K. Christoffersen (Eds.). Springer, within dense mats of the filamentous macro-alga Chaeto-
Berlin, pp. 339–352. morphalinum. Aquatic Bot. 59: 99–115.
10. Canfield, D.E., Langeland, K.A., Linda, S.B., and Haller, T.T. 28. Golterman, H.L. (1969). Methods for Physical and Chemical
(1985). Relations between water transparency and maximum Analysis of Fresh Waters. IBP Handbook No. 8. Blackwell
depth of macrophyte colonization in lakes. J. Aquatic Plant Scientific, Oxford, UK, p. 213.
Manage. 23: 25–28. 29. Murphy, T., Hall, K., and Yesaki, I. (1983). Co-precipitation
11. Chambers, P.A. and Kalff, J. (1985). Depth distribution and of phosphate and calcite in a naturally eutrophic lake.
biomass of submerged aquatic macrophyte communities in Limnology Oceanogr. 28: 58–67.
relation to Secchi depth. Can. J. Fish Aquatic Sci. 42: 30. Schindler, D.W. (1977). Evolution of phosphorus limitation in
701–709. lakes. Science 195: 260–262.
12. Middleboe, A.L. and Markager, S. (1997). Depth limits and 31. James, W.F., Barko, J.W., and Field, S.J. (1996). Phosphorus
minimum light requirements of freshwater macrophytes. mobilization from littoral sediments of an inlet region in Lake
Freshwater Biol. 37: 553–568. Delavan, Wisconsin. Archiv fur Hydrobiologie 128: 245–257.
13. Hudon, C., Lalonde, S., and Gagnon, P. (2000). Ranking the 32. Fonseca, M.S., Fisher, J.S., Zieman, J.C., and Thayer, G.W.
effects of site exposure, plant growth form, water depth, and (1982). Influence of the sea grass, Zostera marina L., on
transparency on aquatic plant biomass. Can. J. Fish Aquatic current flow. Estuarine Coastal Shelf Sci. 15: 351–364.
Sci. 57(Suppl. 1): 31–42. 33. Carpenter, S.R. (1981). Submersed vegetation: an internal
14. Grimshaw, H.J. et al. (2002). The effects of shading on factor in lake ecosystem succession. Am. Naturalist 118:
morphometric and meristic characteristics of wild celery, 372–383.
Vallisneria americana, transplants from Lake Okeechobee, 34. Vermaat, J.E., Santamaria, L., and Roos, P.J. (2000). Water
Florida. Archiv fur Hydrobiologia 155: 65–81. flow across and sediment trapping in submerged macrophyte
15. Blindow, I. (1992). Long and short term dynamics of beds of contrasting growth form. Archiv fur Hydrobiologia
submerged macrophytes in two shallow lakes. Freshwater 148: 549–562.
Biol. 28: 15–27. 35. van Donk, E. and van de Bund, W.J. (1998). Impact of
16. Van Dijk, G.M. and Van Vierssen, W. (1991). Survival of submerged macrophytes including charophytes on phyto-
a Potamogeton pectinatus L. population under various light and zooplankton communities: Allelopathic versus other
conditions in a Shallow Eutrophic Lake Lake Veluwe in The mechanisms. Aquatic Bot. 72: 261–274.
Netherlands. Aquatic Bot. 39: 121–130. 36. Liboriussen, L. and Jeppesen, E. (2003). Temporal dynamics
17. Gafny, S. and Gasith, A. (1999). Spatially and temporally in epipelic, pelagic, and epiphytic algal production in a clear
sporadic appearance of macrophytes in the littoral zone of and a turbid shallow lake. Freshwater Biol. 48: 418–431.
 LAKES—DISCHARGES TO 281

37. Canfield, D.E. Bachmann, R.W., and Hoyer, M.V. (2000). A oils and invasive species. The entry of the zebra mussel
management alternative for Lake Apopka. Lake Reservoir into the Great Lakes is a good example of how an aquatic
Manage 16: 205–221. ecosystem can be significantly changed by the introduction
38. Scheffer, M. (1989). Alternative stable states in eutrophic of just one new species. Now, regulations require ships to
shallow freshwater systems: A minimal model. Hydrobiologi- dump their ballast water in the open ocean before entering
cal Bull. 23: 73–85. the Great Lakes. Used oil from ship engines has also been
39. Moss, B., Madgwick, J., and Phillips, G. (1997). A Guide to dumped into lakes for years. Sewage from ships has also
the Restoration of Nutrient-Enriched Shallow Lakes. W.W. been dumped untreated into lakes, although cruise ships
Hawes, UK, p. 180.
have recently been pressured to install onboard sewage
40. Steinman, A.D., Havens, K.E., Carrick, H.J., and Van Zee, treatment systems so that only treated effluent is dumped.
R. (2002). The past, present, and future hydrology and Atmospheric deposition from smokestacks is becoming
ecology of Lake Okeechobee and its watersheds. In: The
a more serious problem with increased global shipping.
Everglades, Florida Bay, and Coral Reefs of the Florida Keys:
An Ecosystem Sourcebook. K.G. Porter and J. Porter (Eds.).
Ships burn low-grade diesel fuel and these carcinogenic
CRC Press, FL, pp. 19–37. compounds then enter lakes.
Boats used for recreation also pollute lakes. Boat
owners may dump their sewage straight into the lake
LAKES—DISCHARGES TO rather than pumping it out dockside at sewage receptacle
facilities. Gas spills often happen at docks when gas
LAUREL PHOENIX pumping is not done with care, and some gas and oil
Green Bay, Wisconsin are emitted into the water from outboard motors.
Smaller recreational equipment like jet skis pollute
water with oil and gasoline at a far higher rate than
Lakes have suffered from a long history of receiving outboard motors. Jet skis also damage littoral habitat, kill
pollutants from various sources. These pollutants degrade aquatic animals, or frighten aquatic species and disturb
water quality, with implications for aquatic and terrestrial mating or nesting behaviors. Chromosomal damage in fish
ecosystems as well as for a variety of issues for has been linked to the fossil fuel discharges from jet skis.
human populations dependent on that water and those
ecosystems. The degree to which these pollutants degrade Terrestrial Point Sources
the lake depends on the amount and types of pollutants The most common point sources polluting lakes are
as well as the physical characteristics of the lake. For sewage systems and industrial effluent pipes. Sewage
an introductory overview of how discharges to lakes are can enter lakes either through intentional discharge of
harmful, the pollutant sources and how pollutants are raw or treated sewage, or through occasional discharges
transported into the lake are described. The physical from combined sewer overflows (CSOs). Depending on the
characteristics that determine the assimilative capacity of city and on the country, one can find cities discharging
a lake are then covered. Finally, the effects that different raw sewage into lakes rather than spending the money
pollutants have on lakes and their dependent ecosystems to treat it properly. Sewage treatment plants in other
are explained as well as the ramifications for aquatic cities may have primary, secondary, or tertiary sewage
species, birds, wildlife, and humans. treatment, resulting in various types and concentrations
of pollutant discharge into lakes or their tributaries.
POLLUTANT SOURCES AND TRANSPORT Combined sewer overflows are found in cities where
stormwater and sewage flow down the same pipes, rather
Polluted discharges into lakes can come from watercraft, than flowing through separate stormwater and sanitary
terrestrial sources, and atmospheric deposition. Water- sewers. In rain events, the additional stormwater entering
craft are categorized into ships, boats, and recreational sewage pipes may be more than a sewage treatment plant’s
equipment. Terrestrial pollutant sources can be catego- designed capacity. When this happens, sewer pipe outlets
rized into point and nonpoint sources. Point sources of are designed to open and discharge raw sewage into a
local stream or lake to prevent too much sewage from
pollution emanate from a single location, like a factory’s
reaching the treatment plant; passing through the plant
pipe dumping effluent into a stream or lake. Nonpoint
without sufficient treatment; and/or backing up into the
sources of pollution, often called ‘‘polluted runoff,’’ can
city. Consequently, the raw or partially treated sewage
come from all over a landscape and from many locations at
from some cities contributes a daily material loading
once, such as fertilizer runoff from suburban lawns or oil
to a lake, whereas combined sewer overflows discharge
washing off city streets. Atmospheric deposition has char-
sporadically as a result of heavy rain or rapid snowmelt.
acteristics of both point and nonpoint source pollution and
Some of the constituents of sewage are human waste,
so is treated separately here. Atmospheric deposition of
bacteria, viruses, chemicals, detergents, heavy metals, and
pollutants originating from smokestacks travels through
pharmaceutical and personal care products (PPCPs) (1)
the air and then falls into the water.
such as antibiotics, steroids, birth control pills, shampoo,
hair dye, and perfume.
Watercraft
Industrial pollutants enter a tributary or lake by a
Ships discharge a variety of pollutants into lakes. The pipe, either intentionally through a permitting program
ballast water that ships discharge into the water contains or by accidental release. In the United States, any
282 LAKES—DISCHARGES TO

industrial facility, including sewage treatment plants, backfired; new unexpected problems arise from the exotic
must have a National Pollutant Discharge Elimination introduction.
System (NPDES) permit to discharge anything into a
water body. The type, quantity, and concentration of Atmospheric Deposition
pollutants from various industrial processes are written
into the permit, thus limiting the effluent discharge to the Although some scientists may regard atmospheric depo-
permit limits. Pollutants may be chemical, biological (e.g., sition as a type of nonpoint source pollution, it is catego-
bacteria, viruses, or organic matter that deoxygenates the rized here as distinct from terrestrial point or nonpoint
water), thermal, radioactive, metals, and so on. Because source pollution because it comes from discrete sources
there is a permitting system, and since the name implies (smokestacks), but from numerous locations. The type of
a reduction or elimination of pollutant discharges, it is airborne pollutant discharged from smokestacks depends
widely assumed that permits issued are still within the on the type of industrial facility, the type of energy they
assimilative capacity of the water body, and that the use, and their chemical processes. For example, electri-
appropriate government agencies monitor and enforce cal generation plants burning coal, or other industries
that use coal for their energy source, can emit lead, mer-
permits to ensure that the water is not becoming degraded.
cury, zinc, arsenic, uranium, sulfur dioxide, and nitrogen
Unfortunately, this is not the case, and rivers and lakes
oxides. All of these can enter the water surface either
are receiving rising material loads.
through wet (rain or snow) or dry deposition. The heavy
metals, radioactive uranium, and acid rain constituents
Terrestrial Nonpoint Sources have different effects on lakes. Downwind lakes close to
these smokestacks will receive a heavier loading of heavy
Nonpoint source pollutants move into a lake from its entire
metals, whereas the SO2 and NOx that contribute to acid
drainage area. Nonpoint sources are grouped into urban
rain can travel much farther and degrade more distant
runoff, mining, agriculture, silviculture, groundwater,
lakes. Other industries emitting other kinds of chemicals
and human transport. Urban runoff includes fertilizers,
or metals into the air will affect lakes according to how
pesticides, and pet waste from lawns; oil, rubber, deicing
those metals or chemicals interact with lake chemistry or
chemicals, and metals from roads; and sediment from
lake biota.
construction.
Atmospheric deposition of mercury is bioconverted by
Mining contributes acid mine drainage to tributaries or
bacteria into methylmercury, a fat-soluble neurotoxin.
lakes or can put cyanide into the water when it is used to
It then moves up through and beyond the aquatic food
leach gold out of gold ore. Strip mining transports sediment
chain. Regions receiving deposition of both mercury vapor
into streams. Agriculture contributes a significant portion
and acid rain will result in an even higher conversion of
of the overall nonpoint source pollutant load to lakes.
elemental mercury in methylmercury, moving first into
These pollutants include sediment from erosion, pesticides
the water and then into the fish, wiping out entire fish
mobilized from the soil by rain, and overapplication of
species. Acid rain also mobilizes aluminum, manganese,
commercial fertilizer or manure.
and zinc from soils into the lake.
Silviculture primarily contributes pesticides and sed-
iment to lakes. Pesticides used on tree farms or areas
sprayed for infestations (e.g., gypsy moth) are transported PHYSICAL CHARACTERISTICS OF LAKES
through rain and snowmelt into water bodies. Sediment
from road building for timber trucks or from clear-cutting What is the assimilative capacity for a lake for certain
or harvesting on steep slopes is also discharged into types or amounts of pollutants? A lake’s ability to absorb
streams and lakes. pollutants without degrading its biological and chemical
There are two primary types of pollutants moving processes is a function of its physical characteristics. Some
through groundwater. Leaking septics contribute organic limnological characteristics are described here. (For more
sewage material, Escherichia coli, viruses, and toxic information into lake processes and response to pollutants,
household cleaners. Acid rain filters into soils, and see Reference 2.) What is the lake’s volume and depth? The
the resulting chemical reaction then mobilizes mercury, greater the lake volume, the more it can dilute pollutants
manganese, aluminum, and zinc into the groundwater down to less harmful concentrations. Vertical temperature
and then into lakes through subsurface flow. Humans differences found in deep lakes can stratify water into
can transport invasive species unintentionally or with layers. This stratification, and whether or not the lake
purpose. For example, zebra mussels can ‘‘hitchhike’’ on ‘‘turns over’’ (a mixing of the layers by turbulence in the
a pleasure boat being hauled from one water body to water) in spring and fall, controls the amount of oxygen,
another, or canals have been built that connect previously nutrients, and other chemical cycling occurring in the
distinct ecosystems and allow migration of species into a upper or lower waters of the lake.
new lake. Intentional transfer of invasive species happens Therefore, stratification influences which types of
when people get rid of pet fish by flushing them down pollutants may stay or have an effect only in the upper
the toilet or dumping them directly into streams or lakes, strata, only in the lower strata, or whether pollutants have
and this fish will now be an exotic species for that lake. access and impact throughout the entire water column
Other exotic species can be placed in a lake with the and the sediment below. Because stratification controls
intent of using them to control other aquatic animals or temperature, it controls the rates of biochemical reactions,
plants, but several of these experiments (e.g., carp) have and these rates may also provide a buffering effect against
 LAKES—DISCHARGES TO 283

some pollutants. Similarly, because stratification controls Sediments

oxygen levels, pollutant loadings of organic materials (e.g.,
Pollutant deposition into lake sediments can seriously
paper pulp, sewage effluent, manure) will have more
alter existing water chemistry. Chemical pollutants
negative impact during the times of year that oxygen
deposited into the lake sediment alter the chemical balance
levels are lowest. Consequently, the same daily amounts and exchange of chemicals between the sediment and
of material loadings to a lake can have different effects on water. Sediment becomes a long-term storage site for
the lake depending on the season. pollutants. If sediment gets stirred up later on because
The flushing rate of a lake describes how long it takes of turnover, floodflows from tributaries, or dredging, even
for new water inputs to the lake to entirely replace more pollutants are released into the water and the food
existing lake water with new water. How often the lake chain and once again chemical exchanges between the
theoretically refills with new water affects how fast some lake bottom and the water are altered. For example, lead
of the pollutants entering a lake may be flushed out before in sediments deposited up through the 1970s from vehicle
they can harm lake processes very much. The higher the exhaust is still found in lake sediments today. Altering
flushing rate, the more diluted some pollutants can be water chemistry of a lake affects all biota in the lake.
because some pollutants will not be able to stay in the lake
for long. Plants
The typology of a lake, whether it is naturally more
oligotrophic or eutrophic, affects its ability to absorb Plants are vulnerable in several ways to pollutant
some pollutants without changing lake characteristics discharge. If the discharge into the lake is sediment from
significantly. Over hundreds or thousands of years, a lake agricultural, silvicultural, or construction activities, this
naturally receives more nutrient inputs and naturally new sediment layer can smother existing plants. Chemical
undergoes eutrophication, or nutrient enrichment, and pollutants can kill off some aquatic plants. Thermal
slowly changes from oligotrophic to eutrophic. When pollution from some industrial plants or their cooling
rapid nutrient inputs are due to human action, this towers can kill plants by altering water temperature.
is called cultural eutrophication. A more oligotrophic Invasive species introduced by ballast water or by
lake is typically deep and has very clear green or blue ‘‘hitchhiking’’ on pleasure boats hauled from one lake to
water, high oxygen, few nutrients or calcium, and limited another can wipe out plants by increased depredation.
phytoplankton or other plant species. At the other end of Acid rain reduces phytoplankton as well as zooplankton
the spectrum is a eutrophic lake, which is shallow and has populations, undermining the aquatic food chain.
opaque green or brownish green water, very low oxygen in
summer, high nutrients and calcium, and frequent algal Fish
blooms in reaction to the high nutrient concentration. So, Fish are also very vulnerable to discharges into lakes.
for example, a naturally oligotrophic lake would become Some fish can lose their food source if aquatic plants are
more eutrophic in response to high sewage effluent or farm harmed or eliminated by pollutants. Thermal pollution
manure inputs, and a more eutrophic lake would have less from industrial cooling water affects fish metabolism and
assimilative capacity for high loadings of sewage effluent, reproduction. It also stresses fish by increasing oxygen
since the lake is already heavily loaded with nutrients. demand in the water and makes fish more vulnerable to
disease. Reproduction can also be harmed by contact or
ingestion of some chemicals. There might be fewer eggs,
POLLUTANT FATE AND CONSEQUENCES damaged eggs, or no eggs. Recently, estrogen in animal
manure, human hormones, caffeine, and wood-pulp plant
The various sources and transportation mechanisms of effluents are among some of the suspected causes of male
pollutants entering lakes of varying characteristics affect fish producing eggs in their sex organs (4). Chemicals like
lakes in different ways and to different degrees. This chlorine are added in sewage treatment or flushed through
section gives a brief sampling of some of the negative cooling tower pipes to kill bacterial slimes and then are
effects discharges to lakes can have on the water, sediment, discharged into the lake and kill fish. Biomagnification in
and biota. (For a more comprehensive explanation of the the food chain of substances like mercury, polychlorinated
various effects of different pollutant discharges into lakes, biphenyls (PCBs), or dioxins can pass from the sediments
see Reference 3.) through phytoplankton and zooplankton into smaller and
then bigger fish. High acid rain amounts not only wipe
Water out entire fish species, but also cause lakes and streams to
become fishless.
Lake water can be degraded in numerous ways. For Fish can even serve as a transport mechanism for
example, heavy nutrient loading from sewage, manure, these chemicals from one region to another. For example,
and runoff from agricultural soils high in phosphorus salmon have been found to contaminate lakes when they
reduce the amount of oxygen in the water. Moreover, go upstream for spawning and then die in streams and
various chemical compounds and metals entering the lake lakes and release the PCBs or dioxins into the lake water
change the natural rates of chemical cycling in the lake and sediments. Beluga whales in the St. Lawrence River
by changing the ion concentrations in the lake. Acid rain downstream of the Great Lakes have suffered impaired
forces calcium out of solution as manganese, zinc, and fertility and death from the DDT, PCBs, lead, mercury,
aluminum are brought in. and insecticides they have been exposed to from discharges
284 LASERS SCAN LEVEES FROM THE AIR

into the river as well as the Great Lakes. Once they have only affect food chains within the lake but can also harm
been dissected, their bodies must be disposed of according birds, wildlife, and humans that depend on the aquatic
to hazardous waste regulations. Invasive species can harm species for food. The externalities of industrial production,
fish. A northern snakehead, perhaps a former aquarium sewage or stormwater treatment, and other land uses are
pet dumped by its owners, was recently caught in Chicago’s then spread to new populations as well as rebounding
Burnham Harbor on Lake Michigan (5). Snakeheads are on polluter populations themselves. Their quality of life
considered ‘‘voracious eaters’’ and can consume many of degrades along with their drinking water, recreational
the smaller fish in a lake. Invasive species may also start options, and fish and wildlife populations.
competing with existing fish by preying on the same food
sources. Another exotic, the zebra mussel, affects fish by
BIBLIOGRAPHY
clearing the water and thus making fish more visible
to prey.
1. For more information, see: http://www.epa.gov/esd/chemistry/
pharma/index.htm.
Shellfish 2. Wetzel, R. (2001). Limnology: Lake and River Ecosystems,
Like fish, shellfish can also experience reduced reproduc- 3rd Edn. Academic Publishers, Burlington, MA, 2001;
tion from chemicals and hormones. For example, spawning or S.W. Effler (Ed.) (1996). Limnological and Engineering
Analysis of a Polluted Urban Lake. Springer, New York.
behavior is altered by exposure to some antidepressants.
Shellfish may face increased competition for food from 3. Chiras, X., Reganold, X., and Owen, X. (2002). Natural
Resource Conservation, 8th Edn. Prentice-Hall, Upper Saddle
invasive species. Shellfish also absorb and concentrate
River, NJ, Chap. 11.
pollutants in their flesh, making them dangerous to eat.
4. Fahrenthold, D.A. (2004). ‘‘Male bass in Potomac producing
eggs,’’ Washington Post. October 15, p. A01.
Wildlife
5. ‘‘Snakehead found in Lake Michigan,’’ Chicago Tribune.
Wildlife such as bear or cormorants can be harmed by October 14, 2004.
biomagnification of PCBs, mercury, and so on when they
eat the polluted fish.
LASERS SCAN LEVEES FROM THE AIR
Humans
JOHN HALL
Human populations are also harmed in numerous ways. New Orleans District
Drinking water is degraded when a lake has industrial U.S. Army Corps of Engineers
and agricultural chemicals, PPCPs, heavy metals, and so
on in its source water. We know from painful experience
how mercury causes Minimata disease or how cadmium, a New Orleans District is taking to the air to learn more
byproduct of metal refinery, causes a disease that softens about the ground.
the bones. Algal blooms from cultural eutrophication give For good reason—for large jobs flying surveys are faster
drinking water a bad taste and smell. and cheaper than ground surveys. And the eye-in-the-sky
Shoreline use and value can also be diminished. is particularly handy in New Orleans District’s 30,000
Some lake pollution will force beaches to close, reducing square miles of coastal Louisiana, where ‘‘ground’’ is a
recreational resources for locals and possibly keeping relative term.
tourists away and out of the water because of E. coli, the The district is using laser technology called lidar, for
smell and slime of algal blooms, and the toxins some blue- LIght Detection And Ranging. It is analogous to radar,
green algae produce. Heavy nutrient inputs into a lake can which stand for RAdio Detection And Ranging, except that
cause eutrophication that results in water that looks like lidar uses light beams instead of radio waves. A scanning
pea soup and has a horrific smell. These can also cause a laser rangefinder emits a beam of coherent light from a
reduction in recreational use, when boats get caught up helicopter flying 45 miles per hour at an altitude of 220
in algal mats or weedy macrophytes and tourists avoid an feet. (The laser beam is invisible to the naked eye, and not
area, and lower the value of lakeshore properties. If fish intense enough to cause harm.) The beam bounces back
or bear are unsafe to eat, then recreational opportunities and the lidar system’s instruments gather vast amounts
are diminished. Tourists wanting to hunt and fish may go of data which can be sliced, diced, and served up in many
elsewhere, and those that continue to eat unsafe fish can forms. The process is almost effortless, compared to the
harm their own health or have babies with birth defects. labors of ground surveys in wetlands and bayous.
New Orleans District’s first use of lidar was on the Lake
Pontchartrain and Vicinity Hurricane Protection Project.
PRESENT AND FUTURE CHOICES
Total cost was $124,000, with most of the dollars going to
the contractor, John E. Chance and Associated, Inc., which
Pollutant sources, transport, and fate have serious conse-
specializes in corridor mapping (electric lines, highways,
quences for lakes. Cities will either have more degraded
and levees).
drinking water, with potential health consequences, or
they will have to spend enormous amounts of money to
purify the water enough to make it safe for human con- This article is a US Government work and, as such, is in the
sumption. The effects of pollutants on aquatic life can not public domain in the United States of America.
 LASERS SCAN LEVEES FROM THE AIR 285

Mike Brennan, civil engineer, projects lidar information on his

computer. (Photo courtesy of New Orleans District)

A laser scanning rangefinder make a hefty instrument package On the computer screen, these pictures may be viewed as
for a helicopter (Photo courtesy of New Orleans District)
an overlay precisely placed amid the lidar data. Or, you
can just turn the video loose and watch the levee roll by
as if you were on the helicopter.
‘‘We can rapidly determine levee heights without To illustrate, Mark Huber of Survey Section pulled
doing a ground survey,’’ said Al Naomi, senior project up on his computer a lidar aerial view of the Cause-
manager of the hurricane project. ‘‘You can get a lot of way toll plaza in Metairie, La. Bright colors marked
data rapidly.’’ the elevations. In mid-screen, he overlaid a video
Indeed. The system produces 8,000 to 20,000 data picture with the tollbooths, parked cars, roadways,
points per second. But what does all this data really mean? and even the bike path that circles underneath, all
‘‘It provides all three dimensions, the X, Y, and Z clearly visible.
coordinates,’’ said Terral Broussard, a district retiree ‘‘Lidar may not be quite as accurate as ground surveys,’’
who works for the contractor. This is trigonometry in Naomi said. ‘‘But it’s faster, cheaper and requires no
action—three coordinates used to locate a point in space. landowner interface.’’
The X and Y axis are length and width; the Z axis is ‘‘Other (Corps) districts have used lidar for land
the elevation. mapping,’’ Huber said. ‘‘We’re the first to pull the final
Lidar surveys have been completed for all 125 product from the raw data, creating levee cross sections
miles of the Lake Pontchartrain and Vicinity project and centerline profiles.’’
(www.mvn.usace.army.mil, click on hurricane box). These How extensively New Orleans District will use lidar
levees and floodwalls are in St. Bernard, Orleans, is unclear. While lidar cannot retrieve data from water
Jefferson, and St. Charles parishes. But lidar isn’t limited surfaces, vegetation does reflect its signals. That makes it
to looking at levees. In New Orleans District, it is being a candidate for wetland surveys where ground surveying
considered for the Comite River Diversion Project, a flood is inherently difficult.
control project to be built near Baton Rouge. Lidar apparently has a lot of use as a quick look,
For land surveys, the laser is not the only sensor. and for studies. The question is whether it will prove
Chance’s helicopter also carries a video camera, and loca- accurate enough for project design. But, according to
tion is provided by the satellite-based Global Positioning Huber, no errors exceed six inches compared with ground
System. The video provides high-resolution color images. surveys.
286 LEVEES FOR FLOOD PROTECTION

LEVEES FOR FLOOD PROTECTION — additional information on soil profiles

— undisturbed strengths of foundation materials
STEFANO PAGLIARA
— more detailed information on borrow areas and
SIMONE POZZOLINI other required excavations.
Università di Pisa
Pisa, Italy
4. Using the information obtained in step 3,
— determine both embankment and foundation
soil parameters, and refine preliminary sections
Levees were among the first river training structures built where needed.
to protect human settlements and activities from floods. — compute rough quantities of suitable material,
A levee is an embankment whose primary purpose is to and refine borrow area locations.
furnish protection from high water due to river floods. 5. Divide the entire levee into reaches of similar
Levees are similar to small earth dams, but there are foundation conditions, embankment height, and
important differences: fill material, and assign a typical trial section to
each reach.
— Levees are subject to high water levels only for a 6. Analyze each trial section for
limited period of time (of the order of hours or rarely — foundation and embankment seepage;
days), so that a levee embankment is generally not — slope stability;
subject to saturation conditions.
— settlement;
— Levee alignment is dictated primarily by flood
— trafficability of the levee surface;
protection requirements; this often results in
construction on poor foundations. — maintenance requirements.
— Borrow is generally obtained from areas adjacent 7. Design special treatment to preclude any problems
to the levee site, which produce fill material that is determined from step 6.
often far from ideal. 8. Based on the results of step 7, establish final sections
for each reach, and define construction methods.
Embankments that are subject to water loading for 9. Design embankment slope protection.
prolonged time periods or permanently should be designed
using earth dam criteria. Field Investigation and Laboratory Testing
Levee construction requires both field and laboratory
DESIGN OF LEVEES investigation.

Levee design involves numerous factors, dealing with Field Investigation. The extent of field investigation
both hydraulic and geotechnical considerations (Fig. 1). depends on several factors (1). Conditions requiring
Design factors may vary, and no general procedure can extensive field investigation include poor experience in
be established. However, it is possible to present a the area with respect to levee performance, significant
general stepwise design procedure (1) based on successful levee height, bad foundation conditions, long high water
experience that can be used as a starting point in levee duration, low quality of levee materials, high risk
design. Such a procedure may be summarized by the conditions in case of levee failure, and the presence of
following steps: a concrete structure in the levee alignment.
Field investigation and testing are generally performed
1. Conduct a geologic study based on a thorough review in three successive steps:
of available data; initiate preliminary subsurface
explorations. 1. preliminary office geologic study;
2. Analyze preliminary exploration data and establish 2. field geologic survey;
preliminary soil profiles, borrow locations, and 3. field testing (preliminary phase and final phase).
embankment sections.
3. Initiate final exploration to provide Each step is described in Table 1.

Freeboard Crown Landside

Riverside Left levee

Levee embankment Levee toe Channel Floodplain

Foundation soils
Right levee

Figure 1. Sketch of a levee (cross section and plan view).

 LEVEES FOR FLOOD PROTECTION 287

Table 1. Office Study and Field Investigation

Office study Collection and study of

— topographic, soil, and geologic maps

— aerial photographs
— existing information on soil and ground characteristics

Field survey Observations and geologic analysis of the area, including such features as

— riverbank slopes, rock outcrops, earth and rock cuts or fills

— surface materials
— poorly drained areas
— evidence of instability of foundations and slopes
— emerging seepage.
— Physiographic features

Field testing Preliminary phase

— Widely spaced disturbed sample borings and test pits

— Geophysical surveys or cone penetrometer test, used to
interpolate between widely spaced borings
— Borehole geophysical tests
— Preliminary estimates of foundation strength

Final phase

— Additional disturbed sample borings

— Undisturbed sample borings in critical reaches
— Field vane shear tests, useful if undisturbed samples are
unavailable.
— Field pumping tests, providing information on large scale
permeability of in situ deposits
— Water table observations by piezometers in foundations and
borrow areas

Laboratory Testing. Laboratory testing programs for riverside slopes are about 1V:3H÷1V:3.5H; landside slopes
levees vary from minimal to extensive, depending on are about 1V:2H÷1V:2.5H (V = Vertical; H = Horizontal).
the importance of the project and on the site conditions.
Soil tests to be performed will vary according to the soil Freeboard. Freeboard (the difference between the levee
classification (Table 2). crown elevation and the high water stage) is used to
account for uncertainties in levee performance. Many
Embankment Design levees have been designed that have a 1-meter freeboard.
A levee cross section is generally trapezoidal berms may Nowadays, the freeboard concept is less used in levee
be added to ensure stability or seepage control. Berms are design. Risk-based analysis directly accounts for hydraulic
always used for high levee design. uncertainties, and the geotechnical uncertainties are
Embankment design requires detailed analysis for considered in the deterministic analysis of foundations
levees of significant height, bad foundation conditions, and embankment behavior.
or unsuitable embankment materials.
Crown Width. The width of the levee crown depends
Embankment Geometry primarily on roadway requirement; it is in general in
Slope. Mechanical stability is largely dependent on the range of 2.5–5 meters. To provide access for normal
the side slopes of the levee. The slope steepness is maintenance operations and flood fighting operations,
generally selected by considering the types of compaction, minimum widths of 3.0–3.5 m are commonly used.
fill material characteristics, and foundation strength. For
high levees, a soil mechanics stability analysis is required. Stability Analysis. The principal mechanisms of levee
For lower levees laid on proven foundations, recommended failure are
288 LEVEES FOR FLOOD PROTECTION

Table 2. Laboratory Testing — Earthquake. Earthquake loadings are not normally

Soil tests of fine-grained cohesive materials considered in analyzing the stability of levees
because of the low probability of coincidence with
— Visual classification and water content floods. Depending on the seismic characteristics
determination of the study area, determination of liquefaction
susceptibility may instead be required.
— Atterberg limits
— Permeability Shear strength selection under various loading conditions
— Consolidation must consider soil type (free draining or low permeability)
— Compaction and pore water pressure conditions. Minimum factors of
— Shear strength safety for the various stability analyses should be selected
according to local regulations and to sound engineering
Soil tests of pervious materials judgement, considering failure consequences.

— Visual classification Measures to Increase Stability

— In situ density and relative density determination Methods of Improving Embankment Stability. Levee
— Granulometry stability can be enhanced by changes in embankment
— Permeability section, such as
— Consolidation
Flattening Embankment Slopes. Flattening embankment
— Shear strength
slopes will usually increase the stability of an embankment
against a shallow foundation type failure.

Stability Berms. Berms concentrate additional weight

— overtopping where it is most needed and force a substantial
— piping increase in the failure path. Berms can be an effective
— embankment or foundation slides means of stabilization for shallow foundation and
embankment type failures and also for more deeply seated
— surface erosion.
foundation failures.
Common methods used to analyze levee embankment Methods of Improving Foundation Stability. Levees
stability are (2) located on foundation soils of inadequate shear strength
require some type of foundation treatment or the use of
— methods assuming definite boundary stress values; alternative design features in the levee section, such as a
— methods based on the various soil characteristics wider levee crest and flatter slopes. Foundation deposits
that determine the failure state; that are prone to cause problems include very soft clays,
— methods considering the degree of safety for force sensitive clays, loose sands, natural organic deposits, and
and momentum equilibrium; debris deposited by humans.
— methods comparing deformation work with the Means of dealing with inadequate foundation soils
potential energy of the structure. include

Excavation and Replacement. This procedure, though

Primary attention should be devoted to the problems of
extremely effective, is economically feasible only where
defining the shear strengths, unit weights, geometry, and
deposits of unsuitable material are not excessively deep.
limits of possible sliding surfaces.
The more severe loading conditions for a levee are (1)
Stage Construction. A levee is built in successive
intervals of time where the strength of the foundation
— End of construction. Levee embankment and
material is inadequate to support embankment weight, if
foundation stability must be analyzed using drained
built continuously at a pace faster than the foundation
conditions for free draining soils and undrained
material can drain. Rest periods during levee construction
conditions for low permeability soils.
permit dissipation of pore water pressures.
— Sudden drawdown. Excess pore water pressure can Stage construction is appropriate when pore water
develop if a prolonged flood stage saturates the pressure dissipation is reasonably rapid. The rate of
upstream embankment portion and then falls faster consolidation can be increased by using vertical drains.
than the soil can drain. This can negatively affect The disadvantages of this method are the delays
riverside slope stability. in construction operations and uncertainty as to its
— Steady seepage from full flood stage. Flood water scheduling and efficiency.
levels can sometimes fully saturate levee embank-
ments. This is an uncommon condition for ordi- Densification of Loose Sands. Sands is generally not den-
nary levees but may be critical for landside sified due to economic constrictions; possible exceptions
slope stability. are the presence of important structures in a levee system.
 LEVEES FOR FLOOD PROTECTION 289

Levee Settlement. Evaluation of postconstruction set- it is not economically feasible in the presence of thick
tlement is important when settlement can result in loss pervious strata.
of freeboard of the levee or damage to structures in the
embankment. Detailed settlement analyses by theoretical Riverside Blankets. Levees are frequently situated in
analysis should be made when significant consolidation alluvial areas on foundations that have natural surface
is expected, as under high embankment loads, embank- covers of fine-grained impervious to semipervious soils
ments of highly compressible soil or built on compressible overlying pervious sand and gravel. If these surface strata
foundations, and near structures in levee systems founded (or blankets) are continuous and extend riverward for a
on compressible soils. considerable distance, they can effectively reduce seepage
Where foundation and embankment soils are pervious flow and pore pressures landside of the levee.
or semipervious, most of the settlement will occur The effectiveness of the blanket depends on its
during construction. For impervious soils, it is usually thickness, length, distance to the levee riverside toe, and
conservatively assumed that all the calculated settlement permeability. Protection of the riverside blanket against
of a levee built in a normal sequence of construction erosion is an important feature.
operations will occur after construction.
Sometimes a levee is overbuilt by a given percentage Landside Seepage Berms. Landside berms provide addi-
of its height (varying from 0–10%) to take into account tional weight useful for counteracting uplift seepage forces.
anticipated settlement of the foundation and within the Berms also cause an increase in the hydraulic length of
levee fill itself, but overbuilding increases the severity seepage flows, providing a reduction in pressure forces at
of stability problems and may be impracticable for some the landside toe.
foundations. Seepage berms can improve the performance of existing
impervious or semipervious top stratam but may also be
Seepage Control placed directly on pervious deposits. Moreover, seepage
Seepage flows develop when river waters flow through the berms increase the levee’s landside slope stability.
embankment body or through foundation soils. Seepage is
a very slow process, and the establishment of a seepage line Pervious Toe Trench. Partially penetrating toe trenches
generally requires a long time (2). As levees are exposed to can improve seepage conditions close to the levee toe in a
high water levels for relatively short periods, the danger levee situated on deposits of pervious material overlain by
from seepage is more limited than that for earth dams. little or no impervious material.
Pervious toe trenches are useful for controlling shallow
Foundation Underseepage. Without control, underseep- underseepage. To collect deeper seepage flows, relief well
age flows in pervious foundations beneath levees may systems may be used in conjunction with pervious toe
result in excessive pore water pressure, rupturing phe- trenches. Trench dimensions will vary according to the
nomena, and piping beneath the levee embankment. The volume of expected underseepage, desired reduction in
principal seepage control measures are described in the fol- uplift pressure, and construction practicality. The trench
lowing paragraphs (1). The effectiveness of seepage control backfill must be designed as a filter material. To improve
measures can be evaluated by determination of flow-net, drainage efficiency, trenches are often provided with
empirical methods, or mathematical solutions of seepage perforated pipes to collect seepage.
flow equations (Fig. 2).
Pressure Relief Wells. Pressure relief wells are used to
Cutoffs. A cutoff beneath a levee stratum is one of the reduce uplift seepage forces by providing controlled outlets
most positive means for blocking seepage through pervious for seepage that would otherwise emerge uncontrolled
soils. Cutoffs may consist of excavated trenches backfilled landward of the levee. They are installed where pervious
with compacted earth or slurry trenches usually located strata underlying a levee are too deep to be penetrated by
near the riverside toe. Steel sheet piling is generally not cutoff or toe drains and when the construction of landside
entirely watertight due to interlock leakage, but its use can seepage berms is not feasible due to space limitations.
significantly reduce the possibility of foundation piping. A The design of pressure relief well systems involves the
cutoff must penetrate approximately 95% or more of the determination of well spacing, size, and penetration in
thickness of pervious strata to be effective, so constructing the underlying strata. Factors to be considered include

Internal impervious core

Pervious toe trench Seepage berm
Cutoff

Pervious toe Horizontal drainage layer Inclined drainage layer Riverside impervious layer

Figure 2. Seepage control devices.

290 LEVEES FOR FLOOD PROTECTION

depth, stratification and permeability of foundation soils, sections as zoning is usually used in earth dams or in
distance to the effective source of seepage, characteristics major levees.
of the landside top stratum, and the degree of pressure
relief desired.
PROTECTION OF RIVERSIDE SLOPES

Seepage Through Embankments. Embankment seepage The protection needed on a riverside slope to withstand the
emerging on the landside slope results in high seepage erosional forces of waves and stream currents depends on
forces that decrease slope stability. Moreover seepage can a number of factors such as floodwater stage duration,
soften fine-grained fill in the vicinity of the landside bank slope, erosion susceptibility of the embankment
toe or lead to fine materials piping. Generally flood materials, presence of abrupt transitions, short-radius
water stages do not act against a levee long enough to bends, or structures riverside of the levee that increase
make seepage flows exit on the landside slope, but the flow turbulence.
combination of the effects of flood stages with a period of Several types of slope protection have been used.
heavy precipitation may lead to consistent seepage flows. In general, high-class slope protection, such as riprap,
Various means to manage seepage trough embankments articulated mat, soil cement, or paving should be provided
are described hereafter. on riverside slopes beneath bridges and adjacent to
structures passing through levee embankments.
Pervious Toe Drain. Pervious toe drains provide a ready
exit for seepage through an embankment and can lower
the phreatic surface sufficiently so that no seepage will LEVEE CONSTRUCTION
emerge on the landside slope. They can also be combined
with partially penetrating toe trenches as a method for Almost any soil is suitable for constructing levees, except
controlling shallow underseepage. very wet, fine-grained soils or highly organic soils. Levees
of homogeneous cross section are generally made of
Horizontal Drainage Layers. Horizontal drainage layers silt, silty sand, or sandy clay. Fine-grained soils act as
essentially serve the same purpose as pervious toes but sealing materials due to their low permeability, but coarse
can extend further under the embankment requiring a components increase levee stability. The choice of material
relatively small amount of additional material. They can is often governed by the extent of local soils and by the
also protect the base of the embankment against high technological experience of the contractor.
uplift pressures where shallow foundation underseepage is Levee embankments can be built with various degree
occurring. Sometimes horizontal drainage layers also carry of compaction. The central portion of the embankment
away seepage from shallow foundation drainage trenches. may be compacted or semicompacted, but riverside and
landside berms (for seepage or stability purposes) may
Inclined Drainage Layers. An inclined drainage layer is be constructed of uncompacted or compacted fill. When
very effective for controlling internal seepage and is used foundations have adequate strength and where space
extensively in earth dams, but it is rarely used in levee is limited in urban areas both with respect to the
construction because of its cost. It might find use for short quantity of borrow and levee geometry, compacted levee
levee reaches in urban areas. An inclined drainage layer fill construction by earth dam procedures is frequently
completely intercepts embankment seepage regardless of selected. This involves using select material, water content
the degree of stratification in the embankment or the control, and compaction procedures. Where foundations
material type riverward or landward of the drain. Inclined are weak and compressible, high-quality fill construction
drains must be tied into horizontal drainage layers to is not justified because these foundations can support
provide an exit for the collected seepage. only levees with flat slopes. In such cases, uncompacted
The design of pervious toe drains and horizontal and or semicompacted fill is appropriate. Uncompacted fill is
inclined drainage layers must ensure that such drains generally used where the only available borrow is very
have adequate thickness and permeability to transmit wet and frequently has high organic content and where
seepage without appreciable head loss and, at the same rainfall is very high during the construction season.
time, prevent migration of finer soil particles. Moreover, Except in seismically active areas or other areas
the design of drainage layers must satisfy the criteria for requiring a high degree of compaction, compaction by
filter design. vibration other than that afforded by tracked bulldozers is
not generally necessary.
Landside Berms. Landside berms can also be very useful
in embankment seepage control. For a brief description of LEVEE REHABILITATION
their characteristics, refer to the previous paragraphs.
As many levee systems are very old, the rehabilitation
Embankment Zoning. To counteract seepage flows, a of existing levees is gaining greater prominence. Many
central impervious core or covering a riverside slope with techniques have been studied to rehabilitate existing
an impervious layer can be useful. The latter is generally levees. Recently, new working techniques that use geosyn-
more economical than a central impervious core and, in thetic materials extensively, have been introduced. Table 3
most cases, is entirely adequate. As a matter of fact, shows conventional and innovative levee rehabilitation
levee embankments are often constructed as homogeneous techniques (3).
 LIMNOLOGY 291

Table 3. Levee Rehabilitation Techniques

Conventional Innovative
Rehabilitative Rehabilitative
Problem Method Method

Overtopping protection Rebuild Cellular confinement system

Vegetation Reinforced grass
Concrete slabs Soil cement
Current and wave attack Vegetation Reinforced grass
Revetment Concrete block system
Gabions Soil cement
Rainfall surface erosion Vegetation Turf reinforcement mat
Chemical stabilization
Embankment seepage Toe drain Biopolymer chimney drain
Conventional chimney drain Biopolymer toe trench
Foundation seepage Conventional toe trench Jet grouted cutoff
Conventional cutoff
Riverside blanket
Landside seepage berm
Pressure relief well
Slope instability Drainage Reinforced soil slope
Slope flattening and benching Soil nailing
Conventional restraint Pin pile
structure
Chemical treatment Stone-filled trench
Geosynthetic drainage system
Anchored geosynthetic system

BIBLIOGRAPHY on limnology that present lake systems in detail,

running waters, however, are treated only in a few
1. U.S. Army Corps of Engineers. (2000). Design and Construction chapters (13–16). In 1922, the International Association
of Levees. Engineering Manual No. 1110-2-1913 30.
of Theoretical and Applied Limnology (SIL) was founded,
2. Peter, P. (1982). Canal and River Levees. Elsevier Scientific and limnology became the science of aquatic ecosystems.
Publishing Company, Amsterdam.
The volumes of the book series Die Binnengewässer
3. Perry, E.B. (1998). Innovative Methods for Levee Rehabilita- (edited by A. Thienemann since 1925) present the basics
tion. Prepared for U.S. Army Corps of Engineers, Technical
of limnological knowledge. The first concepts to compare
report REMR-GT-26.
lakes (17–20) were directed to lake types, referring to
size and morphology, climate, the behavior to stratify
and mix seasonally or diurnally (21), and different
LIMNOLOGY
nutrient contents resulting in different productivity and
WALTER GELLER oxygen regimes (temperate, tropical and polar lakes,
mono-, dimictic, and polymictic lakes, oligotrophic and
ANNETTE GELLER
eutrophic lakes, clear-water and dystrophic brown-water
UFZ Center for Environmental
Research lakes). Regional limnology was included by Hutchinson (7)
Magdeburg, Germany and (22) with comprehensive surveys over the lakes of the
world, their mode of origin, size, and number.
Studies of hydrobiologists first focused on an inventory
Limnology is the science of inland waters, also called epi- of plants and animals, and then on the microorganisms in
continental waters (1), including fresh and saline, stand- standing and running waters. A more general concept
ing (lenitic) and running (lotic) surface waters, estuaries, of productivity and energy flow in food chains was
wetlands (2,3), and hyporheic (subterranean) waters (4) as developed to describe the input and fate of matter and
ecosystems. Limnology integrates hydrological, physical, energy in ecosystems (23). Lakes all over the world were
and chemical factors as framework constraints for eco- investigated according to Lindeman’s ‘‘Trophic-Dynamic
logical structures and processes. The first comprehensive Concept’’ on a large-scale comparison of the functioning of
publication coining the term ‘‘Limnology’’ was written by lake ecosystems (24), being a part of the International
F. A. Forel (5) on Lake Geneva (Lac Léman) with chapters Biological Programme (IBP). The nonbiological fields
on geography, hydrography, geology, climatology, hydrol- of limnology were treated by physicists (25–27) and
ogy, hydraulics, thermics, optics, acoustics, chemistry, chemists (28,29) in basic reviews and the textbooks
biology, history, navigation, and fisheries (see Ref. 6). Aquatic Chemistry and Physics and Chemistry of Lakes.
Limnology was traditionally focused on lake systems, Institutions for limnological research, all focusing on
as seen by the classic textbook of Hutchinson, A lake limnology, were founded during the early decades
Treatise on Limnology, in Refs. 7–12. Other textbooks of the twentieth century in Austria, England, France,
292 LIMNOLOGY

Germany (including a first station on tropical limnology Human use of land and waters changed the quality of
in Manaos, Brazil), Italy, the Netherlands, Sweden, inland waters with increasing population densities and
Switzerland, Russia at Lake Baikal, the United States, by increasing loads of waste and nutrients impacting
and Canada. The only early stations for river ecology were lakes and rivers. The limnology of creeks and rivers
at the Amazon and Volga rivers. was predominantly developed within applied sciences,
The limnology of creeks and rivers as part of aquatic aiming at fisheries, water pollution by wastewater, the
ecology developed later than lake limnology (30,31). As limits of self-purification of running waters, and solution
running waters are more difficult to characterize, the of problems caused by pollution. Rivers have received
development of concepts considered to suitably cover urban wastewater for a long time. The pollution effects
the relevant processes still proceeds. Typologies of were first scientifically described at the beginning of
rivers can be based on abiotic features as hydrological the twentieth century. Kolkwitz and Marsson (42,43)
characteristics, water flow and its variability, the amount characterized rivers from the source of pollution along
of dissolved and suspended matter, salt content and stretches of self-purification by the benthic community’s
turbidity, and the course of temperature along the changes of species composition, being the base of the
river and through time (32). The Rivers Handbook ‘‘Saprobic System.’’ The saprobic system is still used as a
of Calow and Petts (33) has chapters on hydrology, biological tool to discriminate polysaprobic, mesosaprobic,
physical and chemical characteristics, the biota, the and oligosaprobic conditions (44).
ecosystem, and perturbations and biological impacts, Wastewater pollution is a more severe threat for
and it addresses monitoring programs; modeling of lakes than for rivers, because the organic loads, being
hydrological development, of water quality, of sediment nutrients for the aquatic organisms, remain in the
transport, and of biological responses; and the evolution lakes for years according to their water renewal time.
of habitats. The cumulative enrichment of nutrients, especially of
phosphorus and nitrogen, with its biological consequences
First, in the early twentieth century fisheries, biologists
in lake waters (e.g., algal blooms) and the increasing
divided European rivers along their stretch into distinct
of oxygen deficiencies in deep water layers, called
zones according to slopes, flow velocities, and the
‘‘eutrophication,’’ was observed in most lakes in densely
communities of fish (zones of salmonids, of cyprinids) (34).
settled areas of Europe and in North America. The increase
Benthos biologists amended this approach by including
of total phosphorus in lakes depends on the population
benthic invertebrates (rhitron, potamon) (35). In 1980,
density in the lake’s catchment area. The limits for
Vannote et al. (36) established the ‘‘River Continuum
acceptable P-loadings of morphologically different lakes
Concept’’ using the continuous shift along river stretches
have been treated in comparative studies of the OECD in
(1) of the relation between autotrophy and heterotrophy,
Europe and in Canada (45–48).
and (2) of the relation between autochthonous nutrients,
Systematic measures to reduce the inputs of P and N
that are produced in the river, and allochthonous,
into surface waters have been taken for several decades
terrestrial-borne nutrients. Rare or seasonal events as (e.g., purification of wastewater in treatment plants, ban-
floods temporarily include the floodplains into the riverine ning of detergents containing phosphate, abstraction of
system. The pulsed changes of dry and high-water wastewater from lakes by ring-pipes). The results in many
conditions are essential for these systems adapted to lakes clearly show that these efforts have been success-
regular floods, which affect the floodplains and offer new ful. The present state of nutrient loads of rivers and
ecological niches (e.g., fruit-eating fish species in flooded lakes in Europe and the development during the past is
Amazon forests) (37). The hereby derived ‘‘Flood Pulse documented in reports, which can be downloaded from the
Concept’’ sees the river as part of a larger ecosystem European Environmental Agency (http://www.eea.eu.int/).
including the floodplains as system components being Eutrophication of European lakes, in terms of phospho-
temporarily dry land or shallow flooded side-parts of rus concentrations, generally is decreasing. The actual
the stream. The ecological continuity from water to land phosphorus concentrations are highest in the Eastern
is approached by the ‘‘Concept of Ecotones’’ (38). The European countries and lowest in the Nordic countries.
scale of integrated characteristics and management of The state of surface waters in North America is docu-
the landscape results from enclosing the whole catchment mented in reports accessible from the U.S. Environmental
area of the river system, based on the aquatic ecosystems, Protection Agency (http://www.epa.gov).
both rivers and lakes as partial systems. To characterize Geological regions poor in carbonates are sensitive to
the structure and the hydrological processes within acid rain as it changes the chemical composition of seeping
landscapes with respect to water flows, the integrated runoff and ground- and surface-waters (49). Rain-acidified
approach of the concept of ‘‘Ecohydrology’’ (39) was inland waters are carbonate-free, enriched in sulfate and
developed. The scale of catchment areas is the basis for aluminum, and acidic within a range between 4.5 and 5.5
the European Union (EU) ‘‘Water-Framework-Directive’’ pH units. Reports on long-term development and large-
that was established to keep or to reach a good/maximum scale geographic distribution in EU-countries (50) showed
chemical state and a good/maximum ecological state that in Norway 6000 lakes (51) and in Sweden 17,000
in all running waters and lakes of the EU-member lakes (52) were acidified. In southern Norway, 9630 fish
states (40,41). Furthermore, special commissions have populations were lost (53). Measures against acid rain and
been established to deal with trans-boundary problems acidification of soils, lakes, and rivers started by reducing
of water protection of rivers and lakes (Table 1). the emissions of acidic smoke (54) and direct treatment
 LIMNOLOGY 293

Table 1. Commissions Established to Manage Inland Waters, River Basins, and Lake Catchments Including
Trans-boundary Problems Within Their Catchment Areas. The Table Exemplarily Shows National and International
Commissions and the Respective Lakes and Rivers
River, Lake,
Basin Commission Web Page

Danube International Commission for the Protection of the http://www.icpdr.org/pls/danubis/

Danube River (ICPDR) DANUBIS.navigator
Rhine International Commission for the Protection of the Rhine http://www.iksr.org/
River (IKSR)
Elbe International Commission for the Protection of the River http://www.ikse-
Elbe (IKSE, SMOL) mkol.de/html/ikse/ikse/deutsch/
indexd.htm
Oder International Commission for the Protection of the River
Oder (signed by Germany, Czech Republic and Poland
on 11 April 1996)
Dnieper International DNIPRO Fund (IDF)—National Program of http://greenfield.fortunecity.com/
Environmental Sanitation of River Dnipro Basin and hunters/228/toppage1.htm
Drinking Water Quality Improvement
Bodensee, Lake Constance International Commission for the Protection of the Lake http://www.igkb.de
Constance
Lake Geneva Commission Internationale pour la Protection des Eaux http://www.cipel.org
du Lac Léman contre la pollution (CIPEL)
Great Lakes Basin The International Joint Commission prevents and http://www.ijc.org
resolves disputes between the United States of America http://www.epa.gov/glnpo
and Canada under the 1909 Boundary Waters Treaty.
It assists the two countries in the protection of the
trans-boundary environment, including the
implementation of the Great Lakes Water Quality
Agreement.
Lake Tahoe Lahontan Regional Water Quality Control Board http://www.epa.gov
Lake Washington The Cedar River Basin Plan combines a traditional King http://dnr.metrokc.gov/wlr/watersheds/
County Basin Plan with a Non-point Source Pollution cedar-lkwa.htm/
Action Plan. The Plan was prepared under the policy
direction of the Watershed Management Committee
composed of key tribal, state and local government
agencies and non-govern-mental organizations, and a
Citizens Advisory Committee of area residents.
Mississippi River Basin The Mississippi River Initiative began 1997 and has http://www.epa.gov
developed into a coordinated federal effort to keep illegal
pollution, ranging from raw sewage to industrial waste,
out of the river and to restore the river and surrounding
communities. The Initiative employs cooperative efforts
of the Dept. of Justice, EPA, Customs Service, Coast
Guard, Fish and Wildlife Service, FBI, and institutions
of 26 U.S. states of the Mississippi River Basin.
Aral Sea basin Inter-State Commission for Water Coordination (ICWC):
Water ministers of the five states in the basin,
Kazakhstan, Kyrgyzstan, Turkmenistan, Tajikistan and
Uzbekistan; the ICWC was established 1992 with joint
responsibility for water management with the two river
basin agencies (Amu-Darya and Syr-Darya).
Note: See IWAC (www.iwac-riza.org) for complete list of trans-boundary co-operations in Europe.

of acidic waters. In Sweden, 200,000 tons of lime were from coal or ore mining is determined by geogenic impacts,
spread every year over 20 years, and 6000 lakes were mostly acidification and contamination by heavy metals.
treated. About half of the acidified area was restored until Reviews and other comprehensive scientific publications
the early 1990s (55–57). on mining lakes are in an initial state (60).
Manmade lakes worldwide amount to tens of thou- Some traditional institutions for limnology investigated
sands of reservoirs and pit lakes from surface mining ‘‘their’’ lakes over a long period. Eutrophication and
on coal, gravel, and ores. Reservoirs show limnological acidification of surface waters by acid rain were recognized
peculiarities (58,59) that primarily result from the man- because of these long-term observations. The effects
agement regime. The state of reservoir building and its of countermeasures, stepwise reoligotrophication, and
problems are presented by the World Dam Commission neutralization were also monitored hereby. Only long
(http://www.wdc.org). The water chemistry of pit lakes datasets by traditional long-term investigations (e.g.,
294 LIMNOLOGY

‘‘Long Term Ecological Research’’) render possible the 24. Le Cren, E.D. and Lowe-McConnell, R.H. (1980). The Func-
analyses of the responses of lakes and rivers (61). tioning of Freshwater Ecosystems. International Biological
Presently, these datasets are reevaluated to track Programme, 22. Cambridge Univ. Press, Cambridge.
the effects of long-term changes of climate on the 25. Mortimer, C.H. (1974). Mitt. Internat. Verein. Limnol. 20:
physical, chemical, and ecological behavior of lake systems 124–197.
and rivers (62,63). In addition, paleolimnologists get 26. Imberger, J. and Patterson, J.C. (1990). Advances in Applied
information for previous changes of climate and of nutrient Mechanics 27. J.W. Hutchinson and T.Y. Wu (Eds.). Academic
Press, Cambridge.
inputs altering in the course of historical and geological
times, by investigating the fossil rests of organisms and 27. Imboden, D.M. and Wüest, A. (1995). Physics and Chemistry
of Lakes. A. Lerman, D.M. Imboden, and J.R. Gat (Eds.).
other suitable proximate-indicators in laminated lake
Springer, Berlin.
sediments (64–66). In most lakes of temperate regions,
28. Stumm, W. and Morgan, J.J. (1995). Aquatic Chemistry, 3rd
the sediments show data reaching back to the end of the
Edn. John Wiley & Sons, New York.
last glaciation, and to some million years in the oldest,
29. Lerman, A., Imboden, D., and Gat, J. (1995). Physics and
tertiary lakes like Lake Biwa, Japan (67).
Chemistry of Lakes. Springer-Verlag, Berlin.
30. Behning, A. (1928). Das Leben der Wolga. Schweizerbart’sche
BIBLIOGRAPHY Verlagsbuchhandlung, Stuttgart.
31. Hynes, H.B.N. (1970). The Ecology of Running Waters.
University of Toronto Press, Toronto.
1. Margalef, R. (1994). Limnology Now. Elsevier, Amsterdam.
32. Meybeck, M. (1996). Verh. Internat. Verein. Limnol. 26:
2. Mitsch, W.J. (1994). Global Wetlands—Old World and New.
81–96.
Elsevier, Amsterdam.
33. Calow, P. and Petts, G.E. (1992). The Rivers Handbook.
3. Roggeri, H. (1995). Tropical Freshwater Wetlands. Kluwer
Blackwell, Oxford, pp. 523, 526.
Academic Publishers, Dordrecht.
34. Huet, M. (1949). Schweiz. Z. Hydrol. 11: 332–351.
4. Schwoerbel, J. (1961). Arch. ‘‘Hydrobiol.’’ Suppl. 25: 182–214.
35. Illies, J. and Botosaneanu, L. (1963). Mitt. Internat. Verein.
5. Forel, F.A. (1892). Le Léman: Monographie Limnologique. F.
Limnol. 12: 1–57.
Rouge, Lausanne, pp. 543, 651, and 715.
36. Vannote, R.L. et al. (1980). Can. J. Fish. Aquat. Sci. 37:
6. Edmondson, W.T. (1994). Limnology Now: A Paradigm of
130–137.
Planetary Problems. R. Margalef (Ed.). Elsevier, Amsterdam.
37. Junk, W.J., Bayley, P.B., and Sparks, R.E. (1989). Canadian
7. Hutchinson, G.E. (1957). A Treatise on Limnology. 1:
Special Publication of Fisheries and Aquatic Sciences 106:
Geography, Physics, and Chemistry. John Wiley & Sons, New
110–127.
York.
38. Naiman, R.J., Decamps, H., and Fournier, F. (1989). The Role
8. Hutchinson, G.E. (1967). A Treatise on Limnology. 2:
of Land/Inland Water Ecotones in Landscape Management
Introduction to Lake Biology and the Limnoplankton. John and Restoration—a Proposal for Collaborative Research.
Wiley & Sons, New York. UNESCO, Paris.
9. Hutchinson, G.E. (1975). A Treatise on Limnology. 3: 39. Zalewski, M., Janauer, G.A., and Jolankaj, G. (1997). Eco-
Limnological Botany. John Wiley & Sons, New York. hydrology: A New Paradigm for the Sustainable Use of
10. Hutchinson, G.E. (1993). A Treatise on Limnology. 4: The Aquatic Resources. Technical Documents in Hydrology No. 7.
Zoobenthos. John Wiley & Sons, New York. UNESCO, Paris.
11. Ruttner, F. (1962). Grundriß der Limnologie. 3. Aufl. Walter 40. EU. (2000). EU-Water-Framework-Directive (WFD): Directive
de Gruyter & Co., Berlin. 2000/60/EC of the European Parliament and of the Council
12. Pourriot, R. and Meybeck, M. (1995). Limnologie générale. of 23 October 2000 establishing a framework for Community
Masson, Paris. action in the field of water policy. Official Journal L 327,
13. Margalef, R. (1983). Limnologı́a. Omega, Barcelona. 22/12/2000.
14. Horne, A.J. and Goldman, C.R. (1994). Limnology. McGraw- 41. WATECO. (2000). Economics and the Environment—The
Hill, New York. Implementation Challenge of the Water Framework Directive,
a Guidance Document. Brüssel. Available: http://www.forum.
15. Wetzel, R.G. (2001). Limnology—Lake and River Ecosystems,
europa.eu.int
3rd Edn. Academic Press, San Diego.
42. Kolkwitz, R. and Marsson, M. (1908). Ber. Bot Ges. 26:
16. Schönborn, W. (2003). Lehrbuch der Limnologie. E. Schweit-
505–519.
zerbart’sche Verlagsbuchhandlung, Stuttgart.
43. Kolkwitz, R. and Marsson, M. (1909). Int. Rev. Ges. Hydrobiol.
17. Birge, E.A. and Juday, C. (1911). Bull. Wisc. Geol. Nat. Hist.
2: 126–159.
Surv. 22: 259.
44. Liebmann, H. (1951). Handbuch der Frischwasser- und
18. Naumann, E. (1919). Lunds Univ. Arsskrift N.F.II 17: 1–22.
Abwasser-Biologie. Oldenbourg, München.
19. Thienemann, A. (1915). Verh. Naturh. Ver. Preuss. Rheinl.
45. Vollenweider, R.A. (1968). Die wissenschaftlichen Grundla-
71: 281–389.
gen der Seen- und Fließgewässereutrophierung unter beson-
20. Thienemann, A. (1918). Arch. Hydrobiol. 12: 1–65. derer Berücksichtigung des Phosphors und des Stickstoffs als
21. Hutchinson, G.E. and Löffler, H. (1956). Proc. Nat. Acad. Sci. Eutrophierungsfaktoren. 69. Report DAS/CSJ. OECD.
Wash. 42: 84–86. 46. Vollenweider, R.A. (1980). OECD Eutrophication Programme.
22. Meybeck, M. (1995). Physics and Chemistry of Lakes. The Canadian Project. OECD Eutrophication Programme.
A. Lerman, D.M. Imboden, and J.R. Gat (Eds.). Springer, 47. Vollenweider, R.A. and Kerekes, J. (1980). OECD Eutrophi-
Berlin. cation Programme. Synthesis Report. OECD Eutrophication
23. Lindeman, R.L. (1942). Ecology. 23: 339–418. Programme. OECD, Burlington.
 ADSORPTION OF METAL IONS ON BED SEDIMENTS 295

48. Fricker, H.J. (1980). OECD Eutrophication Programme. The optimum contact time needed to reach equilibrium
Regional Project Alpine Lakes. Swiss Federal Board for was approximately 60 minutes and was independent
Environmental Protection, Bern. of the initial concentration of cadmium ions. The
49. Stumm, W. and Schnoor, J.L. (1995). Physics and Chemistry adsorption curves are smooth and continuous leading
of Lakes, 2nd Edn. A. Lerman, D. Imboden, and J. Gat (Eds.). to saturation, suggesting a possible monolayer coverage
Springer, Berlin. of cadmium ions on the surface of the adsorbent. The
50. Luckewille, A., Jeffries, D.S., Johannessen, M., Raddum, G., extent of adsorption increases with pH. Furthermore,
Stoddard, J.L., and Traaen, T. (1997) The nine year report: the adsorption of cadmium increases as adsorbent doses
acidification of surface water in Europe and North Amer-
increase and as adsorbent particle size decreases. The
ica—long term developments (1980s and 1990s). NIVA Report
serial no. 3637-97.
two geochemical phases, iron and manganese oxides,
probably support the adsorption of cadmium ions. The
51. Henriksen, A., Skjelkvale, B.L., Mannio, J., Wilander, A.,
Jensen, J.P., Moiseenko, T., Harriman, R., Traaen, T., Fjeld,
adsorption data have been analyzed by using the
E., Vuarenmaa, J., Kortelainen, P., and Forsius, M. (1997) Langmuir and Freundlich adsorption models to determine
Regional Lake Surveys 1995 in Finland, Norway, Sweden, the mechanistic parameters of the adsorption process.
Denmark, Russian Kola, Russian Karelia, Scotland and Isotherms have been used to obtain thermodynamic
Wales—Results. NIVA Report serial no. 3645-97. parameters, free energy change (
Go ), enthalpy change
52. Leonard, J. and Crouzet, P. (1999) Acidification. Topic (
H o ) and entropy change (
So ). The negative values
Report. Copenhagen. Available: http://europa.eu.int. p. 85 of free energy change (
Go ) indicate the spontaneous
53. Hesthagen, T., Sevaldrud, I.H., and Berger, H.M. (1999). nature of the adsorption of cadmium on bed sediments, and
Ambio 28: 112–117. the positive values of enthalpy change (
H o ) suggest the
54. Likens, G.E., Driscoll, C.T., and Buso, D.C. (1996). Science endothermic nature of the adsorption process. The uptake
272: 244–246. of cadmium is controlled by bulk as well as intraparticle
55. Olem, H. (1991). Liming Acidic Surface Waters. Levis Publ., diffusion mechanisms.
Chelsea.
56. Henrikson, L. and Brodin, Y.-W. (1995). Liming of Acidified
Surface Waters. (Eds.). Springer, Berlin. INTRODUCTION
57. Erikson, F., Hörnström, E., Mossberg, P., and Nyberg, P.
(1983). Hydrobiologia 101: 145–164. Heavy metals added to a river system by natural and
58. Straskraba, M., Tundisi, J.G., and Duncan, A. (1993). Com- man-made sources during their transport are partitioned
parative Reservoir Limnology and Water Quality Manage- among different compartments of aquatic ecosystems,
ment. Developments in Hydrobiology. Kluwer Academic Pub- such as water, sediment, and biota (1). The added heavy
lishers, Dordrecht.
metals then undergo many different chemical changes,
59. Tundisi, J.G. and Straskraba, M. (1999). Theoretical Reser- whereby a high degree of variation in metal concentrations
voir Ecology and its Applications. Backhuys Publishers,
occurs (1,2). Metals that do not remain soluble in water
Leiden.
are adsorbed and accumulate on bottom sediments that
60. Castro, J.M. and Moore, J.N. (2000). Environmental Geology
act as a sink (3). The distribution of heavy metals in
39: 1254–1260.
the bottom sediments is affected by the mineralogical
61. Goldman, C.R. (1988). Limnol. Oceanogr. 33: 1321–1333.
and chemical composition of the suspended material,
62. George, D.G. and Harris, G.P. (1985). Nature 319: 536–539. anthropogenic influences, and in situ processes such as
63. Ambrosetti, W. and Barbanti, L. (1999). J. Limnol. 58: 1–9. deposition, sorption, and enrichment in organisms (4).
64. Anderson, N.J. (1993). Trends Ecol. Evol. 8: 356–361. Thus, in the natural conditions of the river water,
65. Frey, D.G. (1969). Mitt. Internat. Verein. Limnol. 17: 7–18. suspended and riverbed sediments play an important
66. Smol, J.P. (1992). J. Aquat. Ecosystem Health 1: 49–58. role in buffering higher metal concentrations in water,
67. Horie, S. (1984). Lake Biwa. Monographiae Biologicae. W. particularly by adsorption and/or precipitation. Therefore,
Junk, Dordrecht. the study of sediments and their sorptive properties can
provide valuable information relating to the tolerance
of the system to the added heavy metal load and may
ADSORPTION OF METAL IONS ON BED determine the fate and transport of pollutants in the
SEDIMENTS aquatic environment.
The tremendous increase in the use of heavy metals
CHAKRESH K. JAIN during the past few decades has resulted in an
National Institute of Hydrology increased concentration of metals in waterbodies and
Roorkee, India has great environmental significance due to their toxicity
and nonbiodegradable nature (5). Cadmium, one of
the most toxic metals, finds its way to watercourses
The adsorption characteristics of cadmium on bed through wastewater discharges from metal plating and
sediments of the River Ganga (India) have been studied cadmium–nickel battery industries, phosphate fertilizers,
to determine the tolerance of the river system for mining, pigments, stabilizers, and alloys (6). The toxic
the heavy metal load. The effect of various controlling effects of cadmium include high blood pressure, kidney
parameters, initial concentration, solution pH, sediment damage, and destruction of testicular tissue and red
dose, contact time and particle size, have been evaluated. blood cells (7). Even low concentrations of cadmium
296 ADSORPTION OF METAL IONS ON BED SEDIMENTS

cause hypertensive diseases. The International Agency Alaknanda that originates in Sapta Tal Glacier. After
for Research on Cancer classified cadmium as one covering a distance of about 220 km in the Himalayas,
of the chemicals that is carcinogenic to humans (8). it enters the plains at Hardwar, and after meandering
Therefore, understanding the environmental conditions across a distance of about 2290 km in the plains of Uttar
and mechanisms that regulate the mobilization and Pradesh, Bihar, and West Bengal, it joins the Bay of
distribution of cadmium in the environment is essential. Bengal through a large number of branches flowing in
Gardiner (9) studied the effect of various parameters on India and Bangladesh (Fig.1).
the adsorption of cadmium on river mud and other natural Physiographically, the area is generally flat, except
solids. Koelmans and Lijklema (10) studied the adsorption for the Siwalik Hills in the north and northeast of the
of cadmium onto sediment and suspended solids in catchment. The area is devoid of any relief features of
Lake Volkerak in The Netherlands and reported that provinces except from deep gorges cut by drains and
cadmium is bound almost completely to geochemical iron, rivers flowing through the area. The drainage pattern
manganese, and organic phases. Wiley and Nelson (11) is dominated by the River Ganga, which is the only major
examined the influence of various factors on the adsorption river flowing through the area. The meltwater of the
of cadmium onto the sediments of Sturgeon Lake in glaciers in the upper Himalayas maintains a perennial
Oregon and reported that pH is the most critical supply of water in the River Ganga.
parameter affecting cadmium adsorption. Christensen The region is characterized by a moderate type
(12) and Palheiros et al. (13) also reported almost similar of subtropical monsoonal climate. It has a cool, dry,
findings. Bajracharya et al. (14) studied the effect of zinc winter season from October to March, a hot, dry,
and ammonium ions on the adsorption of cadmium on sand summer season from April to June, and a warm rainy
and soil and reported that both ions suppress adsorption season from July to September. The average annual
capacity significantly. Fu and Allen (15) studied the rainfall across the Ganga Basin varies from 780 mm
adsorption of cadmium by oxic sediments using a multisite- in the upper part to 1040 mm in the middle course
binding model. The model has been used satisfactorily and 1820 mm in the lower delta of Bangladesh (21). It
to predict the extent of adsorption across the pH range has been observed that the rainfall is heaviest in the
4.5–7.0. northern region of Hardwar district, close to the foothills
Despite the apparent wealth of information on adsorp- of the Himalayas, and becomes less southward. The
tion processes, little is known about quantitatively describ- major land use is agriculture, and there is no effective
ing adsorption by coarser sediment. The important com- forest cover.
ponents of the suspended load for geochemical transport Geologically, the area is a part of the west Indogangetic
are silt, clay, hydrous iron and manganese oxides, and plain, which is composed mainly of Pleistocene and
organic matter. Generally, adsorption studies for sedi- subrecent alluvium brought down by the river action
ment less than 50 µm in size predominate (13,16); those from the Himalayan region. At Hardwar, the sequence
for sediment more than 75 µm in size are lacking because of sandstone and shales along with gravel beds and clays
sorption decreases as particle size increase. Although clay are grouped within the upper and middle Siwaliks. The
and silt adsorb metal ions much better than coarser frac- soils of this region do not form a compact block. They
tions of sediment, one should take into account that most differ from valley to valley and slope to slope according to
river sediments contain 90–95% sand and only 0–10% different ecological conditions. The bed of the river is rocky
clay and silt. Therefore, in river systems that have a high down to Hardwar. The soils of the area are predominantly
sand percentage and low clay and silt content, the overall loam to silty loam and are normally free from carbonates.
contribution of the sand to adsorption of metal ions could The pollution status of the River Ganga has been described
be comparable to or even higher than that of the clay and earlier (22,23).
silt fraction. Our earlier efforts to establish this hypothesis
have been quite successful (17–19).
EXPERIMENTAL METHODOLOGY
Recently, we studied the adsorption characteristics of
zinc on bed sediments of the River Ganga (20). In this
Freshly deposited sediments from shallow water near
article, we report the adsorption of cadmium on bed
the bank of the River Ganga at Hardwar in the state
sediments of the River Ganga to assess the tolerance
of Uttaranchal were collected in polyethylene bags and
of the river system to the added cadmium load. The
brought to the laboratory. Samples were taken from the
importance of geochemical phases has been investigated.
upper 5 cm of the sediments where flow rates were low
Adsorption data have been analyzed by using adsorption
and sedimentation was assumed (24,25).
models to determine the mechanistic parameters of the
The size distribution of the sediment samples was
adsorption process, and isotherms have been used to obtain
analyzed by using nylon sieves to obtain various fractions.
thermodynamic parameters.
The textural features of the sediments were observed,
and a preliminary classification made according to grain
THE RIVER SYSTEM size and distinctive geochemical features. The important
geochemical phases for the adsorption process are organic
The River Ganga is a perennial river formed by the matter, manganese oxides, iron oxides, and clays. The
confluence of two smaller rivers at Deoprayag. Bhagirathi contents of manganese oxide and iron oxide were measured
is one of them originating at Gaumukh in the Gangotri as total manganese and total iron, respectively, and
Glacier, 3129 m above mean sea level; the other is were extracted from the sediment samples using acid
 ADSORPTION OF METAL IONS ON BED SEDIMENTS 297

Afghanistan
China

Pakistan Bhutan
G Nepal
a nga
Bangladesh

Arabian Burma
Sea Bay of
India
Bengal

Sri Lanka

at h
i
o m da
ir

g
Bha

Deoprayag Ala k n
Rishikesh
Hardwar
N
E
P
A
Ya m u

L
Gh

na
ag

ha
ra
G

al
an

Ka i

mb
da

ha nd
s
k

C Si Ga
ng a
en

B ha

K n
So Aj
ar ay
Dam od
gira
a

al
tw

th
b
C ha m

Be

a s ai

l
Hu g
i

Bay of Bengal
500 km

Figure 1. The Ganga Basin showing the locations of the sediment sampling sites.

digestion (HF + HClO3 + HNO3 ) in an open system. cleaned by soaking in detergent followed by soaking in
Organic matter was determined by oxidation with 10% nitric acid for 48 h and finally rinsed with deionized
hydrogen peroxide. water several times prior to use.
All chemicals used in the study were obtained from Adsorption experiments were conducted in a series of
Merck, India, and were of analytical grade. Aqueous Erlenmeyer flasks of 100 mL capacity covered with Teflon
solutions of cadmium were prepared from cadmium sheet to prevent introduction of any foreign particles.
nitrate. Deionized water was used throughout the study. Fifty mL of cadmium ion solution (200–2000 µg/L) was
All glassware and other containers were thoroughly transferred in the flasks together with desired adsorbent
298 ADSORPTION OF METAL IONS ON BED SEDIMENTS

doses (Ws in g/L), and placed in a water bath shaker of cadmium ions on this fraction (150–210 µm) and to
maintained at 30 ◦ C. A pH of 6.5 ± 0.1 was maintained compare it to the clay and silt fraction (<75 µm particle
throughout the experiment by using dilute HNO3 and size) to demonstrate the importance of the coarser fraction
NaOH solutions. Aliquots were retrieved periodically in controlling metal pollution.
and filtered through 0.45 µm cellulose nitrate membrane
filters. The filters were soaked in 1% v/v HNO3 for 1 h and OPERATING VARIABLES
thoroughly rinsed with deionized water prior to use.
The concentration of cadmium ions was determined Equilibrium Time (t )
by flame atomic absorption spectrometry, using a Perkin-
Elmer Atomic Absorption Spectrometer (Model 3110) with To determine the equilibrium time for the adsorption
air–acetylene flame. The detection limit for the cad- process, adsorption experiments were performed for the
mium ion was 0.0005 mg/L. Operational conditions were uptake of cadmium ions for different contact times and
adjusted to yield optimal determinations. Quantification for a fixed adsorbent dose of 0.5 g/L and an initial
of metals was based on calibration curves of standard cadmium concentration of 1000 µg/L for the two particle
solutions of cadmium ion. These calibration curves were sizes of adsorbent (0–75 and 150–210 µm) at pH 6.5
determined several times during the period of analysis. (Fig. 2). The solution pH for the experiments was close
to that encountered in the river water. These plots
indicate that the concentration of cadmium ions in solution
RESULTS AND DISCUSSION becomes asymptotic to the time axis such that there is no
appreciable change in the remaining concentration after
The content of important geochemical phases (iron and 60 minutes in both fractions. This time is presumed to
manganese) in different fractions of the sediment along represent the time at which an equilibrium concentration
with weight percentages are given in Table 1. The is attained. All additional experiments were conducted for
sediment has a coarse texture and is composed of more 60 minutes.
than 90% sediment of size >75 µm and <10% silt and According to Weber and Morris (26), the uptake for
clay. The organic content of the sediment was of the order most adsorption processes varies almost proportionately
of 0–1%. The background cadmium level in the various as t1/2 rather than as contact time. Therefore, plots of
fractions of the sediments was negligible in the unpolluted
zone, compared to the amount of adsorbate added for
the adsorption tests. This confirms the absence of any 0–75 micron 150–210 micron
cadmium particulate attached to the sediment particles.
85
It is evident from the data that the amount of
manganese and iron in the various fractions of the
83
sediment decreases as particle size increases. This
Adsorption, %

indicates the possibility that the two geochemical phases

act as active support material for the adsorption of 81
cadmium ions. However, the relative contribution of
the individual components could not be obtained from 79
the present studies because the type and composition
of a sediment’s mineral and organic fractions vary 77
simultaneously in natural systems, and the effect of
individual constituents cannot be isolated. The content 75
of iron in the sediment fractions is relatively higher and 0 30 60 90 120 150 180
indicates the possible presence of iron minerals other than Contact time, min
hydroxides. However, this result should be confirmed by
0–75 micron 150–210 micron
further investigations. It is further evident from Table 1
that the sediment fraction of 150–210 µm particle size 85
constitutes 68.6% of the total sediment load. Therefore,
it was considered appropriate to study the adsorption 83
Adsorption, %

81
Table 1. Characteristics of Sediments
Sediment Fraction, Weight, Total Mn, Total Fe, 79
µm % mg/g mg/g

<75 9.8 1.570 68.2 77

75–150 13.5 1.420 62.0
150–210 68.6 1.199 58.0 75
210–250 6.6 0.844 55.1 0 5 10 15
250–300 0.7 0.830 52.1 t1/2, min1/2
300–425 0.6 0.823 52.0
425–600 0.2 0.824 52.0 Figure 2. Effect of contact time (t and t1/2 ) on percent adsorption
of cadmium ions.
 ADSORPTION OF METAL IONS ON BED SEDIMENTS 299

cadmium adsorbed, Ct versus t1/2 , are also presented Comparing the two plots, it is clearly evident that the
for the two particle sizes of adsorbent in Fig. 2. It is affinity of cadmium is greater for the <75-µm fraction,
clearly evident that adsorption of cadmium ions follow clay and silt, compared to the coarser fraction. It is
three phases: (1) an early extremely fast uptake (Phase I), also evident from Table 1 that the <75 µm fraction
(2) a transition phase (Phase II), and (3) an almost flat contains more iron and manganese than the 150–210 µm
plateau section (Phase III). Phase I is attributed to the fraction, indicating the possibility of an association of
instantaneous use of the most readily available adsorbing these substrate (iron and manganese) with clay and silt
sites on the adsorbent surface (bulk diffusion). Phase II, particles. These findings illustrate the possible role of clay
exhibiting additional removal, is attributed to the diffusion and silt as sites for cadmium adsorption. However, due
of the adsorbate from the surface film into the macropores to the paucity of data, correlation statistics could not be
of the adsorbent (pore diffusion or intraparticle diffusion), attempted to confirm this statement. The clay and silt
stimulating further migration of adsorbate from the liquid constitute <10% of the total sediment load; comparing the
phase onto the adsorbent surface. Phase III, a plateau weight percentages of the <75 µm to the 150–210 µm and
section, represents the equilibrium state. their corresponding adsorption capacities for cadmium
ions, it is clear that the coarser sediment contributes
Adsorption Isotherm more than the clay and silt fraction in controlling
cadmium pollution.
The adsorption isotherms for cadmium adsorption on the
bed sediments, shown in Fig. 3, have a fixed adsorbent
Effect of pH
dose of 0.5 g/L at a pH of 6.5 ± 0.1. The adsorption
data indicate linear distribution in the range 0 to The adsorption of cadmium on the riverbed sediments
1000 µg/L. It is evident that for the same equilibration was studied across the pH range 2–7 for a fixed
time, the amount of cadmium adsorbed is greater for initial concentration of cadmium (Ci = 1000 µg/L) and
greater initial concentrations of cadmium ions. In addition, an adsorbent dose of 0.5 g/L at particle sizes of 0–75
the percentage adsorbed is greater for lower initial and 150–210 µm (Fig. 4). The pH of the solution was
concentrations of cadmium and decreases as the initial adjusted by using dilute hydrochloric acid and sodium
concentration increases (Fig. 3). This is obvious because hydroxide solutions. The pH was measured before and
more efficient use of the adsorptive capacities of the after the solution had been in contact with the sediment;
adsorbent is expected from a greater driving force. the difference between the two values was generally less
than 0.1 pH unit. A general increase in adsorption as
the pH of solution increased was observed up to pH 6.0
0–75 micron 150–210 micron for both fractions of the sediment. From the results, it is
3500 evident that the pH for maximum uptake of cadmium ion is
6.0. Further, it is apparent that the adsorption of cadmium
3000 rises from 6.5% at pH 2.0 to 83.3% at pH 6.0 for the clay and
Adsorption, µg/g

2500 silt fraction (0–75 µm) and from 3.9% at pH 2.0 to 81.5%
at pH 6.0 for the coarser sediment fraction (150–210 µm).
2000
A similar trend for pH was reported by Palheiros et al.
1500 (13) for the adsorption of cadmium on riverbed sediment.
1000
Adsorbent Dose (Ws )
500
The effect of adsorbent dose on the adsorption properties
0 of the bed sediments of the River Ganga was studied at
0 200 400 600 800
Equil. concn, µg/L

0–75 micron 150–210 micron 0–75 micron 150–210 micron

100 100

90 80
Adsorption, %

Adsorption, %

60
80
40
70
20
60
0 200 400 600 800 0
0.0 2.0 4.0 6.0 8.0
Equil. concn, µg/L
pH
Figure 3. Adsorption of cadmium ions at different concentra-
tions. Figure 4. Effect of pH on adsorption of cadmium ions.
300 ADSORPTION OF METAL IONS ON BED SEDIMENTS

pH 6.5 using different adsorbent doses varying from 0.5 adsorption is a surface phenomenon; the smaller par-
to 2.5 g/L and at a fixed initial cadmium concentration ticle sizes offered a comparatively larger surface area
of 1000 µg/L (Fig. 5). The experiments were conducted in and hence higher adsorption occurred at equilibrium. The
a water bath shaker to disperse the sediment particles higher content of iron and manganese in the 0–75 µm
in the aqueous medium. It was observed that for a sediment fraction also accounts for greater adsorption of
fixed initial concentration of cadmium (Co = 1000 µg/L), cadmium in this fraction compared to larger size fractions
the adsorption of cadmium per unit weight of adsorbent of the sediment.
decreases as adsorbent load increases. On the other hand,
the percent adsorption increases from 83.4 to 93.7% for the
0–75 µm fraction as the adsorbent load increases from 0.5 ADSORPTION MODELS
to 2.5 g/L. The adsorption of cadmium was higher for the
0–75 µm fraction compared to the 150–210 µm fraction. Adsorption data for a wide range of adsorbate
Because of the higher surface area as well as higher concentrations are most conveniently described by
content of iron and manganese in the 0–75 µm fraction; adsorption models, such as the Langmuir or Fre-
these are the main driving forces for the adsorption of undlich isotherm, which relate adsorption density qe
cadmium ions. (metal uptake per unit weight of adsorbent) to equi-
librium adsorbate concentration in the bulk fluid
Particle Size (dp ) phase, Ce . The adsorption data for cadmium on the
bed sediments of the River Ganga was analyzed
The effect of adsorbent particle size on cadmium adsorp-
using Langmuir and Freundlich models to evaluate
tion is shown in Fig. 5 for a fixed initial concentration of
the mechanistic parameters of the adsorption pro-
cadmium (Ci = 1000 µg/L), an adsorbent dose of 0.5 g/L,
cess.
and at pH 6.5. These plots reveal that for a fixed adsorbent
dose, smaller particles adsorb more cadmium. Further, it
is observed that the percentage of cadmium adsorbed Langmuir Model
decreases from 83.4% on the 0–75 µm fraction to 77.4% Langmuir’s isotherm model is valid for monolayer
on the 425–600 µm fraction. This result occurs because adsorption onto a surface containing a finite number of
identical sites. The Langmuir treatment is based on the
assumption that maximum adsorption corresponds to a
0–75 micron 150–210 micron
saturated monolayer of solute molecules on the adsorbent
95 surface, that the energy of adsorption is constant, and
that there is no transmigration of adsorbate in the plane
of the surface.
The linear form of Langmuir isotherm equation is
Adsorption, %

90 represented by the following equation:

1 1 1
= o + (1)
qe Q bQo Ce
85
where qe is the amount adsorbed at equilibrium time
(µg/g), Ce is the equilibrium concentration of the adsorbate
ions (µg/L), and Qo and b are Langmuir constants related
80
0.0 0.5 1.0 1.5 2.0 2.5 3.0 to maximum adsorptive capacity (monolayer capacity) and
Adsorbent dose, g/L
energy of adsorption, respectively. When 1/qe is plotted
against 1/Ce , a straight line of slope 1/bQo and intercept
85 1/Qo is obtained (Fig. 6), which shows that the adsorption
of cadmium ions follows the Langmuir isotherm model.
83 The Langmuir parameters, Qo and b, are calculated from
the slope and intercept of the graphs and are given in
Adsorption, %

Table 2. These values may be used for comparison and

81 correlation of the sorptive properties of the sediments.

79 Freundlich Model
The Freundlich equation, widely used for many years, is
77 applicable to isothermal adsorption. This is a special case
of heterogeneous surface energies in which the energy
75 term, b, in the Langmuir equation varies as a function
0 100 200 300 400 500 600 of surface coverage, qe , strictly due to variations in the
Adsorbent size, µm heat of adsorption (27). The Freundlich equation has the
Figure 5. Effect of adsorbent dose and size on adsorption of general form,
cadmium ions. qe = KF C1/n (2)
 ADSORPTION OF METAL IONS ON BED SEDIMENTS 301

(a) 0–75 micron 150–210 micron Table 3. Freundlich Parameters for the Adsorption of
Cadmium
0.004
Sediment Fraction, Adsorptive Capacity, Adsorptive Intensity,
µm KF mg/g 1/n
0.003
0–75 0.0045 1.765
150–210 0.0043 1.726
1/qe

0.002

adsorptive intensity (28). The Freundlich parameters for

0.001
the adsorption of two metal ions are given in Table 3.
The Freundlich type adsorption isotherm is an
0 indication of the surface heterogeneity of the adsorbent,
0 0.01 0.02 0.03 0.04 0.05 and the Langmuir type isotherm corresponds to the surface
1/Ce homogeneity of the adsorbent. This leads to the conclusion
that the surface of the bed sediments of the River Ganga
(b) 0–75 micron 150–210 micron is made up of small heterogeneous adsorptive patches
that are similar to each other in respect to adsorptive
4.0
phenomena (29).

3.5
THERMODYNAMICS
log qe

3.0 The effect of temperature on the adsorption of cadmium

on bed sediments of the River Ganga is shown in Fig. 7 for
a fixed adsorbent dose of 0.5 g/L and an initial cadmium
2.5 concentration of 1000 µg/L at pH 6.5 for the two sediment
fractions (0–75 µm and 150–210 µm). The temperature
2.0 range used in the study was from 20 to 40 ◦ C. It is evident
1.0 1.5 2.0 2.5 3.0 from the plots that the adsorption of cadmium increases
log Ce as temperature increases for both fractions, which may be
mainly due to an increase in the number of active sites
Figure 6. Graphical representation of adsorption isotherms:
caused by the breaking of some bonds.
(a) Langmuir isotherm and (b) Freundlich isotherm.
Thermodynamic parameters such as free energy change
(
Go ), enthalpy change (
H o ), and entropy change (
So )
were determined by using the following equations (30–32).
The Freundlich equation is basically empirical but is
often useful to describe data. Data are usually fitted to the
Go = −RT ln Kc (4)
logarithmic form of the equation,

G =
H − T
S
o o o
(5)
1
log qe = log KF + log Ce (3) where
n

Go = change in free energy, kJ/mol;
where qe is the amount adsorbed (µg/g), Ce is the
H o = change in enthalpy, kJ/mol;
equilibrium concentration of the adsorbate ions (µg/L), and
So = change in entropy, J/mol/K;
KF and n are Freundlich constants related to adsorptive T = absolute temperature, K;
capacity and adsorptive intensity, respectively. When log R = gas constant = 8.314 × 10−3 ;
qe is plotted against log Ce , a straight line of slope 1/n and Kc = equilibrium constant, defined as
intercept log KF is obtained (Fig. 6). This plot reflects
satisfaction of the Freundlich isotherm model for the CAe
adsorption of cadmium ions on Ganga River sediment. Kc = (6)
Ce
The intercept of the line, log KF , roughly indicates
the adsorptive capacity, and the slope, 1/n, indicates where CAe and Ce are the equilibrium concentrations
(µg/L) of the metal ions on the adsorbent and in the
solution, respectively.
Table 2. Langmuir Parameters for the Adsorption of Equations 4 and 5 can be combined and rewritten as
Cadmium

So
H o
Sediment Fraction, Adsorption Maxima, Binding Energy log Kc = − (7)
µm Qo , mg/g Constant, b, mg/L 2.303R 2.303RT

0–75 5.0 3.306 When log Kc is plotted against 1/T, a straight line of
150–210 10.0 1.206
H o
So
slope and intercept is obtained (Fig. 8).
2.303R 2.303R
302 ADSORPTION OF METAL IONS ON BED SEDIMENTS

(a) 20 °C 30°C 40°C (a)

1.0
1900

1800 0.8
Adsorption, µg/g

1700 0.6

log Kc
1600 0.4

1500 0.2
y = −1.7423x + 6.4816
1400
0 30 60 90 120 150 180 0.0
3.1 3.2 3.3 3.4 3.5
Contact time, min
(1/T)×1000, 1/K
(b) 20 °C 30°C 40°C
(b)
1800 1.0

1700 0.8
Adsorption, µg/g

log Kc 0.6
1600
0.4
1500
0.2
y = −1.4263x + 5.3903
1400
0 30 60 90 120 150 180 0.0
3.1 3.2 3.3 3.4 3.5
Contact time, min
(1/T)×1000, 1/K
Figure 7. Effect of contact time and temperature on adsorption
of cadmium ions: (a) sediment fraction = 0–75 µm; (b) sediment Figure 8. Van’t Hoff plot for adsorption of cadmium ions:
fraction = 150–210 µm. (a) sediment fraction = 0–75 µm; (b) Sediment fraction = 150–
210 µm.

The values of
H o and
So were obtained from the slope

and intercept of the van’t Hoff plots of log Kc vs 1/T (Fig. 8). Table 4. Thermodynamic Parameters for the Adsorption
The thermodynamic parameters for the adsorption process of Cadmium
are given in Table 4. Sediment Temperature,
Go ,
H o ,
So ,
Positive values of
H o suggest the endothermic nature Fraction, µm ◦
C Kc kJ/mol kJ/mol J/mol/K
of the adsorption. Negative values of
Go indicate the
spontaneous nature of the adsorption process. However, 0–75 20 3.545 −3.083 33.36 124.1
the negative value of
Go decreased as temperature — 30 5.024 −4.066 — —
— 40 8.524 −5.576 — —
increased, indicating that the spontaneous nature of
150–210 20 3.444 −3.012 27.31 103.2
adsorption is inversely proportional to the temperature. — 30 4.494 −3.786 — —
The positive values of
So show increased randomness at — 40 7.064 −5.087 — —
the solid/solution interface during the adsorption process.
The adsorbed water molecules, which are displaced by the
adsorbate species, gain more translational energy than is
lost by the adsorbate ions, thus allowing the prevalence of where q and qe are the amounts of metal adsorbed
randomness in the system. The enhancement of adsorptive (µg/g) at time t(min) and at equilibrium, respectively,
capacity at higher temperatures may be attributed to the and kad is the Lagergren rate constant for adsorption
enlargement of pore size and/or activation of the adsorbent (1/ min). The straight line plots of log(qe − q) versus
surface (33). t for different concentrations (Fig. 9) and temperatures
(Fig. 10) indicate the applicability of this equation. Values
Adsorption Dynamics of kad were calculated from the slope of the linear
The rate constant of adsorption is determined by using the plots and are presented in Tables 5 and 6 for different
Lagergren first-order rate expression, (34–36): concentrations and temperatures, respectively. The values
of Lagergren rate constants (kad ) were almost the same
kad at both concentrations, whereas the value increases as
log(qe − q) = log qe − ×t (8) temperature increases.
2.303
 ADSORPTION OF METAL IONS ON BED SEDIMENTS 303

(a) 0.5 mg/L 1.0 mg/L (a) 20 °C 30°C 40°C

2.5 2.5
y1.0 = −0.0201x + 2.0148
2.0 2.0
log(qe−q)

1.5

log(qe−q)
1.5
1.0
1.0 y40 = −0.0209x + 2.1319
0.5 y30 = −0.0201x + 2.0148
y0.5 = −0.0202x + 1.7762 0.5
0.0 y20 = −0.0191x + 1.9011
0 10 20 30 40 50 0.0
Time, min 0 10 20 30 40 50
Time, min
(b) 0.5 mg/L 1.0 mg/L
(b) 20 °C 30 °C 40°C
2.5
y1.0 = −0.0179x + 2.0516 2.5
2.0
log (qe−q)

2.0
1.5

log (qe−q)
1.0 1.5

0.5 1.0 y40 = −0.0189x + 2.0976

y0.5 = −0.0182x + 1.7639
y30 = −0.0179x + 2.0516
0.0 0.5
0 10 20 30 40 50 y20 = −0.017x + 1.8998
Time, min 0.0
Figure 9. Lagergren plots at different initial concentra- 0 10 20 30 40 50
tions: (a) sediment fraction = 0–75 µm; (b) sediment fraction Time, min
= 150–210 µm.
Figure 10. Lagergren plots at different temperatures: (a) sedi-
ment fraction = 0–75 µm; (b) sediment fraction = 150–210 µm.
Intraparticle Diffusion
The rate constant for intraparticle diffusion (kid ) is given Table 5. Rate Constants at Different Concentrations
by Weber and Morris (37):
Lagergren Intraparticle
q = kid t1/2 (9) Sediment Concentration, Rate Constant Rate Constant
Fraction, µm µg/L kad (1/ min) kid , µg/g/min
where q is the amount adsorbed (µg/g) at time t (min). Plots 0–75 500 4.65×10−2 3.84
of q versus t1/2 are shown in Figs. 11 and 12 for different — 1000 4.63×10−2 9.88
initial concentrations and temperatures, respectively. All 150–210 500 4.19×10−2 5.85
plots have the same general features, an initial curved — 1000 4.12×10−2 8.86
portion followed by a linear portion and a plateau. The
initial curved portion is attributed to bulk diffusion,
the linear portion to intraparticle diffusion, and the Table 6. Rate Constants at Different Temperatures
plateau to the equilibrium. This indicates that transport
of cadmium ions from the solution through the particle Lagergren Intraparticle
solution interface, into the pores of the particle as well Sediment Temperature Rate Constant Rate Constant
◦
Fraction, µm C kad (1/ min) kid , µg/g/min
as the adsorption on the available surface of sediment,
are both responsible for the uptake of cadmium ions. The 0–75 20 4.40×10−2 8.11
deviation of the curves from the origin also indicates that – 30 4.63×10−2 11.80
intraparticle transport is not the only rate limiting step. – 40 4.81×10−2 14.29
The values of the rate constants (kid ) were obtained from 150–210 20 3.92×10−2 7.50
the slope of the linear portion of the curves for each – 30 4.12×10−2 10.00
concentration of metal ions (Table 5) and temperature – 40 4.35×10−2 12.35
(Table 6). The values of intraparticle rate constants (kid )
increase as concentration and temperature increase.
ions have more affinity for the clay and silt fraction
CONCLUSION of the sediment, the overall contribution of the coarser
fraction to adsorption is greater compared to the clay
The study has demonstrated that river sediments provide and silt fraction. The pH of the solution is the most
good buffering capacity for metal loads. Although cadmium important parameter in the control of cadmium pollution.
304 ADSORPTION OF METAL IONS ON BED SEDIMENTS

(a) 0.5 mg/L 1.0 mg/L (a) 20 °C 30 °C 40 °C

1800 1850

1600
1750
Adsorption, µg/g

Adsorption, µg/g
1400

1200 1650
1000

800 1550

600
0 5 10 15 1450
1/2, 0 5 10 15
t min1/2
t1/2, min1/2
(b) 0.5 mg/L 1.0 mg/L
(b) 20 °C 30°C 40°C
1700
1800
1500
Adsorption, µg/g

1300 1700
Adsorption, µg/g
1100
1600
900

700
1500
500
0 5 10 15
t1/2, min1/2 1400
0 5 10 15
Figure 11. Intraparticle diffusion plots at different initial
t1/2, min1/2
concentrations: (a) sediment fraction = 0–75 µm; (b) sediment
fraction = 150–210 µm. Figure 12. Intraparticle diffusion plots at different tempera-
tures: (a) sediment fraction = 0–75 µm; (b) sediment fraction
= 150–210 µm.
Iron and manganese oxides support the adsorption of
cadmium. The adsorption data follow both the Langmuir
and Freundlich isotherm models. The kinetic data suggest Sr) in recent lacustrine sediments, Resumes des Publications
that the adsorption of cadmium is an endothermic process, IX Congress Intl. De Sedimentologie, Nice, Theme II, p. 6.
which is spontaneous at low temperature. The uptake of 5. Zuane, J.D. (1990). Hand Book of Drinking Water Quality,
cadmium is controlled by both bulk as well as intraparticle Van Nostrand Reinhold, New York.
diffusion mechanisms. It is stated that in natural river 6. Low, K.S. and Lee, C.K. (1991). Cadmium uptake by the moss
systems, salinity and pH greatly affect the speciation calympers delesertii, besch. Bioresour. Technol. 38: 1–6.
of the metal ions and thus greatly interfere with their 7. (1980). Encyclopedia of Environmental Sciences, 2nd Edn.
fixation on clay minerals. Adsorption/desorption equilibria McGraw-Hill, New York, p. 354.
and complexation with fulvic and humic acids also play 8. IARC (International Agency for Research on Cancer). (1976).
an important role in speciation. Further studies are being Mongraphs on the evaluation of carcinogenic risk of chemicals
planned to understand better the processes that affect the to man, In: Cadmium and Cadmium Compounds, 11, p. 3974.
sorption of cadmium in natural systems. 9. Gardiner, J. (1974). The chemistry of cadmium in natural
water—II. Adsorption of cadmium on river muds and
BIBLIOGRAPHY naturally occurring solids. Water Res. 8: 157–164.
10. Koelmans, A.A. and Lijklema, L. (1992). Sorption of tetra-
1. Moore, J.W. and Ramamoorthy, S. (1984). Heavy Metals in chlorobenzene and cadmium to sediments and suspended
Natural Waters, Applied Monitoring and Impact Assessment. solids in lake Volkerak/Zoom. Water Res. 26: 327–337.
Springer-Verlag, New York. 11. Wiley, J.O. and Nelson, P.O. (1984). Cadmium adsorption
2. Forstner, W. and Wittman, G.T.W. (1979). Metal Pollution in by aerobic lake sediments. J. Environ. Eng., ASCE, 110(1):
the Aquatic Environment. Springer-Verlag, New York. 226–243.
3. Hakanson, L. (1980). An ecological risk index for aquatic 12. Christensen, T.H. (1984). Cadmium soil sorption at low
pollution control-A sedimentological approach. Water Res. 14: concentrations: I. Effect of time, cadmium load, pH, and
975–1001. calcium, Water, Air, and Soil Poll. 21: 105–114.
4. Forstner, W. and Muller, G. (1975). Factors controlling the 13. Palheiros, B., Duarte, A.C., Oliveira, J.P. and Hall, A. (1989).
distribution of minor and trace elements (heavy metals V, Li, The influence of pH, ionic strength and chloride concentration
 MICROBIOLOGY OF LOTIC AGGREGATES AND BIOFILMS 305

of the adsorption of cadmium by a sediment. Water Sci. Tech. 34. Singh, D.B., Prasad, G., Rupainwar, D.C., and Singh, V.N.
21: 1873–1876. (1989). As(III) removal from aqueous solution by adsorption.
14. Bajracharya, K., Barry, D.A., Vigneswaran, S., and Gupta, Water Air and Soil Pollution 42: 373–386.
A.D. (1996). Heavy metal adsorption in soil: Comparison 35. Namasivayam, C. and Ranganathan, K. (1995). Removal of
of bisolute adsorption models and laboratory experiments. Cd(II) from waste water by adsorption on waste Fe(III)/Cr(III)
In: Water Quality Hydrology. V.P. Singh and B. Kumar hydroxide. Water Res. 29: 1737–1744.
(Eds.). Vol. 3, Kluwer Academic Publishers, The Netherlands, 36. Zhang, L., Zhao, L., Yu, Y., and Chen, C. (1988). Removal of
pp. 19–26. lead from aqueous solution by non-living rhizopus nigricans.
15. Fu, G. and Allen, H.E. (1992). Cadmium adsorption by oxic Water Res. 32(5): 1437–1444.
sediments. Wat. Res. 26: 225–233. 37. Weber, W.J. Jr. and Moris, J.C. (1962). Advances in Water
16. Gagnon, C., Arnac, M., and Brindle, J.R. (1992). Sorption Pollution Research. Proc. 1st International Conference on
interactions between trace metals (Cd and Ni) and phenolic Water Pollution Research. Vol. 2, Pergamon Press, Oxford,
substances on suspended clay minerals. Water Res. 26: UK, p. 231.
1067–1072.
17. Jain, C.K. and Ram, D. (1997). Adsorption of lead and zinc on
bed sediments of the river Kali. Water Res. 31(1): 154–162. MICROBIOLOGY OF LOTIC AGGREGATES AND
18. Jain, C.K. and Ram, D. (1997). Adsorption of metal ions BIOFILMS
on bed sediments. Hydrological Sciences Journal 42(5):
713–723. ERIC A. STRAUSS
19. Jain, C.K. and Ali, I. (2000). Adsorption of cadmium on Upper Midwest Environmental
riverine sediments: quantitative treatment of the large Sciences Center
particles. Hydrological Processes 14: 261–270. La Crosse, Wisconsin
20. Jain, C.K. (2001). Adsorption of zinc onto bed sediments of the
river Ganga: Adsorption models and kinetics. Hydrological
Sciences Journal 46(3): 419–434.
INTRODUCTION
21. Krishna Murti, C.R. (1991). Some physical features of the
Ganga basin. In: The Ganga: A Scientific Study. Krishna The microbial community is a vital component of lotic
Murti, C.R., Bilgrami, K.S., Das, T.M., and Mathur, R.P. (flowing water) ecosystems and is important for nutri-
(Eds.). The Ganga Project Directorate, New Delhi, India.
ent cycling, organic matter decomposition, downstream
22. Jain, C.K. (1998). Adsorption of Zinc on Bed Sediments of transport of materials, and transfer of energy to higher
River Ganga, Technical Report, CS(AR)9/98-99, National
trophic levels. Some of this microbial community is trans-
Institute of Hydrology, Roorkee, India.
ported downstream as unattached cells; however, much
23. Jain, C.K. (2002). A hydro-chemical study of a mountainous
of it is associated with suspended organic particles (i.e.,
watershed: the Ganga, India. Water Res. 36(5), 1262–1274.
aggregates) or attached to available substrate in biofilms.
24. Sakai, H., Kojima, Y., and Saito, K. (1986). Distribution of
Research on microbial activity within organic aggregates
metals in water and sieved sediments in the Toyohira river.
Water Res. 20: 559–567.
has mostly been limited to marine and lake systems. As
a result, more is known about the microbial communities
25. Subramanian, V., Grieken, R.V., and Vant, D.L. (1987).
Heavy metal distribution in the sediments of Ganges and and microbial processes in marine and lake aggregates
Brahmaputra rivers. Environ. Geol. Water Sci. 9(2): 93–103. compared with aggregates in lotic ecosystems. Conversely,
26. Weber, W.J. Jr. and Moris, J.C. (1963). J. Sanit. Eng. biofilms have been studied in great detail in many systems,
Div., ASCE, SA2, 31, cited in Faust S.D. and Aly O.M., including lotic ecosystems. Many excellent and detailed
1987. Adsorption Processes for Water Treatment, 65–122, reviews are available on the topics of aggregates and
Butterworth, London, pp. 65–122. biofilms (1–11). Included here is a general overview of the
27. Adamson, A.W. (1967). Physical Chemistry of Surfaces, 2nd microbiology and ecology of these topics and how they are
Edn. Interscience Publishers, Inc., New York. relevant to lotic ecosystems.
28. Weber, W.J. Jr. (1972). Physico-chemical Processes for Water
Quality Control. John Wiley & Sons, New York, NY, LOTIC AGGREGATES
pp. 208–210.
29. Ajmal, M., Khan, A.H., Ahmad, S., and Ahmad, A. (1998). Origins and Terminology
Role of sawdust in the removal of copper (II) from industrial
Microorganisms readily colonize substrates in lotic
wastes. Water Res. 32(10): 3085–3091.
ecosystems to form diverse and active communities.
30. Singh, A.K., Singh, D.P., Pandey, K.K., and Singh, V.N.
Suspended organic matter is often rich in nutrients and
(1988). Wollastonote as adsorbent for removal of Fe(II) from
water. J. Chem. Technol. Biotechnol. 42: 39–49. easily metabolized carbon, providing an especially good
substrate for microbial colonization and growth and for the
31. Catena, G.C. and Bright, F.V. (1989). Thermodynamic study
on the effects of β-cyclodextrin inclusion with anilino- subsequent formation of lotic aggregates. Lotic aggregates
naphthalene sulfonates. Anal. Chem. 61: 905–909. can be defined as the interactive association between a
32. Frajii, L.K., Hayer, D.M., and Werner, T.C. (1992). Static and microbial community and suspended organic matter in
dynamic fluorescence quenching experiments for the physical a flowing water system. The organic matter that is a
chemistry laboratory. J. Chem. Edu. 69: 205–215. central component of lotic aggregates may have a variety
33. Vishwakarma, P.P., Yadava, K.P., and Singh, V.N. (1989). of origins including allochthonous and autochthonous
Nickel(II) removal from aqueous solutions by adsorption on particles and byproducts of animal and microbial activity.
fly ash. Pertanika 12: 357–366. Allochthonous organic matter (e.g., wood, leaves, and
306 MICROBIOLOGY OF LOTIC AGGREGATES AND BIOFILMS

terrestrial debris) readily enters lotic systems through Table 1. Common Terms, Abbreviations (in parentheses),
various pathways. Larger particles tend to quickly settle and Sizes Used to Classify Suspended Matter in Flowing
onto the channel bed or get trapped in channel structures, Water Systems
whereas the smallest particles often become entrained Term Size
in the flowing water. Over time, physical and biological
processes fragment the large particles into smaller detrital Dissolved organic matter (DOM) <0.45 µm
Particulate organic matter (POM) >0.45 µm
particles, which can then become resuspended into the
Course particulate organic matter (CPOM) >1 mm
water column. Organic matter particles of autochthonous Fine particulate organic matter (FPOM) 0.45 µm to 1 mm
origin (e.g., biofilm slough and aquatic vegetation) also Aggregate >0.45 µm
may become fragmented and entrained in the current Macroaggregate >500 µm
in the same manner as those particles of allochthonous Microaggregate <500 µm
origin. Organic byproducts of animal activity such as Colloid 1.5 nm to 0.45 µm
fecal pellets and discarded invertebrate exoskeletons are Floc >0.45 µm
common in highly productive systems and can be nutrient- Seston >0.45 µm
rich substrates for microbial colonization (12). Aggregates The size ranges listed here are generally accepted definitions, but they can
also may form in conjunction with dissolved organic vary somewhat depending on application.
matter (13) or colloidal organic polymers that are small
(1.5 nm to 0.45 µm), high-molecular-weight byproducts of
organic matter decomposition that remain suspended in Microbial Community
the water (14). Bacteria and other microorganisms secrete a viscous, slimy
Aggregates have the potential to enlarge as they mixture of exopolymers (i.e., chains of polysaccharides)
travel downstream by combining with other aggregates called glycocalyx. In general, the properties of exopolymers
(aggregation) and to gather additional materials primarily can contribute many beneficial functions to organisms
through impaction (Fig. 1) (15). However, the physical including assisting in locomotion and adhesion to surfaces,
forces exerted by flowing and turbulent water usually and protection from abrasion, dehydration, and nutrient
fragment larger aggregates (disaggregation), maintaining loss. More specifically, in lotic aggregates, the exopolymers
the aggregates at a smaller size compared with those found bind the contents of aggregates together and assist in the
in nonflowing systems (16). Although relatively small, the capture of additional nutrients, organic matter, and other
size of lotic aggregates are still highly variable and can microorganisms. Much of the exopolymers associated with
be further classified into microaggregates (<5 to 500 µm) lotic aggregates are likely produced by microorganisms
and macroaggregates (>500 µm) (7). Larger aggregates within the aggregate, but they also may accumulate via
are also sometimes called ‘‘river snow’’ in keeping impaction with suspended exopolymers.
with terminology used to describe similar aggregates Microbial communities thriving within the fragile
present in marine (marine snow) and lake (lake snow) matrix of exopolymers and particles consist of a diverse
ecosystems (17,18). In the field of stream ecology, lotic heterogeneous assemblage of bacteria, algae, fungi,
aggregates may also be called seston, which can be protozoa, and sometimes zooplankton. Compared with
defined as particles entrained in flowing water. A list marine and lake ecosystems, relatively few studies
of common terms and abbreviations used to describe have examined the microbial community structure of
suspended matter in lotic systems is provided in Table 1. aggregates in lotic systems. However, many studies
on marine aggregates have demonstrated that the
concentration of algae, zooplankton, protozoa, and bacteria
within aggregates is much greater, often 2–3 orders of
magnitude, than in the surrounding water (7). Larger
aggregates often support larger microbial communities,
but smaller particles may be relatively more densely
populated per unit surface area (19). Detailed microscopic
analyses have shown that bacteria are not uniformly
distributed within or on the surface of aggregates but
form microcolonies (7). These microbial communities have
temporally dynamic compositions that change quickly in
response to conditions both inside and outside of the
aggregate and can also exhibit seasonal patterns (17,20).
Inorganic (nitrogen, phosphorus, and silica) and organic
(carbohydrates, amino acids, and dissolved organic carbon)
nutrients are often 1–2 orders of magnitude more enriched
in aggregates compared with surrounding water (7). In
addition, many heavy metal and organic contaminants
also readily adsorb to the aggregate matrix. As a result
Figure 1. Conceptual diagram illustrating processes affect- of these enriched conditions, downstream movement of
ing lotic aggregate size and composition. DOM = dissolved aggregates is an important mechanism of transport for
organic matter. nutrients, organic materials, and contaminants. Moreover,
 MICROBIOLOGY OF LOTIC AGGREGATES AND BIOFILMS 307

these nutrient-rich substrates result in high growth Feedback

efficiencies of heterotrophic bacteria associated with the
lotic aggregates, which suggests that these aggregates

Nitrification
N fixation
O2

assimilation
are hot spots for microbial processes and the transfer

Depth into bioflim

Nitrate
of organic matter into bacterial biomass (20). Enzymes NO3−

Denitrification
Ammonium assimilation
produced by aggregate-bound bacteria solubilize organic

Ammonification
matter 1–2 orders of magnitude faster than the bacterial
cells can assimilate the dissolved matter (21). This NH4+
decomposition results in aggregate mass loss and leaching

DNRA
of dissolved organic matter back into the flowing water.
Finally, consumption of lotic aggregates by many species
of detritus- and filter-feeding organisms is a vital link
N 2O
of energy transfer between microbial and higher trophic
levels (Fig. 1).

Zone of high activity Relative concentration

BIOFILMS
Figure 2. General patterns of nitrogen cycling (left) and
Biofilm Structure relative concentrations of nitrogen constituents (right) in a lotic
biofilm [redrawn and modified from Paerl and Pinckney (3)].
Microbial biofilms are the thin slime layers that Feedback between the chemical and biological components
are observed on surfaces and substrates (e.g., rocks, facilitates the formation and equilibrium of the gradient
sediment, vegetation, leaves, wood) in aquatic systems. (i.e., the chemical environment within the biofilm affects
Biofilms (sometimes called microbial mats) are similar microbial processes and vice versa). DNRA = dissimilatory
to aggregates and possess many of the same general nitrate reduction to ammonium.
structural and functional attributes. For example, biofilms
(like aggregates) contain microbial communities that are
matrix and the matrix itself. These differences suggest
vital to lotic ecosystem function and are responsible
that such structures may facilitate water movement
for many key production, decomposition, and nutrient
and nutrient delivery within the biofilm. Microsensors
cycling processes (22). Highly developed biofilms are
have also been used to quantify vertical chemical
generally a few millimeters thick and can develop on
gradients in biofilms that have important implications
virtually any available surfaces in natural lotic systems,
for microbial community structure and biogeochemical
provided the system has sufficient nutrients. Biofilm
processing (Fig. 2). Chemical gradients will be discussed
development is especially rapid on surfaces that contain
further in the next section.
or leach nutrients (e.g., particulate organic matter,
The physical stress caused by the turbulent flow of
POM) (2). Available surfaces are still colonized in nutrient-
water in lotic systems can have significant effects on
poor or oligotrophic systems, but biofilms are generally
biofilm structure and organization. Many studies have
less developed. Under oligotrophic conditions, biofilm
observed that biofilms subjected to fast flowing water
bacteria are often able to obtain limited nutrients by
tend to be thinner, smoother, denser, more stable, and
concentrating organics on the exopolymer matrix, using
have lower surface sinuosity than biofilms in slower
waste products of nearby microorganisms, and by pooling
or nonflowing water (23,24). Cells occupying the surface
their biochemical resources (i.e., enzymes) to break down
of lotic biofilms have been observed to be arranged
organic particles (11).
in formations that can withstand the higher shear
Historically, biofilms were thought to be a relatively
stress; examples of these formations include hexagon-
homogeneous distribution of bacteria in an exopolymer
shaped structures and ripples positioned parallel to flow
matrix and nutrient and material delivery throughout the
direction (24). Freely oscillating streamer-like structures
biofilm would be dominated by diffusion (2). However,
are also sometimes observed adhered to the surface
recent advances in the analytical tools used to study
of lotic biofilms (8). Streamers are predominately single
biofilms have allowed scientists to more accurately
species colonies that elongate over time in the absence of
characterize the structure or architecture of biofilms.
disturbance.
For example, the application of confocal scanning laser
microscopy has allowed nondestructive inspection of
Microbial Processes
intact biofilm samples. This technique has revealed the
physical structure of established biofilms to consist of a The rates and types of microbial processes at work in any
diverse microbial assemblage arranged in microcolonies particular biofilm are largely regulated by the physical and
and interspersed within a heterogeneous exopolymer chemical characteristics of the lotic environment. Several
matrix containing protrusions and a network of pores types of algae and cyanobacteria flourish in biofilms if
and channels (2,4). Microsensors, another relatively recent sufficient nutrients are available and the overlying water
research tool, have been used to examine chemical is shallow and clear enough to allow adequate light
properties in spatially explicit areas within biofilms penetration to the substrate. Oxygen (O2 ) generated by
and have found distinct differences in dissolved oxygen these primary producers can reach very high levels and
concentrations between the pores and channels in the is a key factor supporting aerobic biofilm processes. In
308 MICROBIOLOGY OF LOTIC AGGREGATES AND BIOFILMS

fact, under high-light conditions, the surficial area of anaerobic process that occurs in the lowest areas of the
well-lit biofilms can produce O2 concentrations 2–3 times biofilm where NO3 − is still available.
greater than atmospheric concentrations (25). Much, if It is generally accepted that bacterial activity and pro-
not all, of the oxygen is respired as it is transported and duction of biofilm biomass is resource limited in lotic
diffused deeper into the biofilm creating a vertical gradient systems. Thus, nutrient availability and the quantity
of oxygen concentrations throughout the biofilm (Fig. 2). and quality of the dissolved organic carbon in the water
Chemical gradients are common in biofilms and control have strong positive effects on bacterial activity in the
which microbial processes are able to occur at certain biofilm (26,27). Total bacterial activity also often corre-
depths. The opposite is also true; microbial processes lates well with primary production in biofilms because
control the availability of certain ions. The feedback of the influx of oxygen and high-quality organic carbon
relationship between these two factors quickly forms a provided by algae (27,28). Finally, vertical hydrodynamics
relatively stable gradient equilibrium that is particularly of stream channels (i.e., upwelling and downwelling) can
important for the transformations and cycling of certain be an important means of nutrient and dissolved organic
nutrients, primarily nitrogen, carbon, and sulfur. matter (DOM) delivery through the biofilm, which can
The nitrogen cycle is a complex group of microbial subsequently increase microbial activity (29).
processes that transform, oxidize, or reduce nitrogen Under certain chemical or physical conditions, spe-
cialized biofilms that are dominated by unique groups
ions among the various oxidation states. These processes
of microorganisms may develop. For example, orange-
occur under different environmental conditions, all of
colored iron-oxidizing biofilms are prevalent in areas
which can be located in distinct locations along the
where groundwater with high concentrations of reduced
chemical gradients found in biofilms (Fig. 2). Nitrogen
iron (Fe2+ ) is discharged into oxygenated waters (30).
fixation, which is the conversion of atmospheric dinitrogen
Another example is the colorful biofilms of thermophilic
gas (N2 ) to biologically available ammonium (NH4 + ), is
bacteria and cyanobacteria that are commonly associated
predominately done by cyanobacteria; thus, it occurs in
with water flowing from hot springs (Fig. 3). The colors
the upper, high-light areas of the biofilm. Nitrification, of these biofilms depend primarily on the chlorophyll con-
which is the oxidation of NH4 + to nitrate (NO3 − ), also tent in the cells, but also on the ratio of chlorophyll to
occurs in the upper areas because it is an aerobic process. carotenoids (yellow to red pigments) (31).
Bacteria that have the enzymes to reduce NO3 − for
assimilation likely occur throughout the biofilm profile,
but they can only use NO3 − where it is present. Organic BIBLIOGRAPHY
matter is ubiquitous throughout the biofilm; therefore,
1. Johnson, B.D., Kranck, K., and Muschenheim, D.K. (1990).
ammonification (the release of NH4 + from organic matter)
Physico-chemical factors in particle aggregation. In: The
and the resulting NH4 + are also widespread. Ammonium Biology of Particles in Aquatic Systems. R.S. Wotton (Ed.).
assimilation into microbial biomass is not oxygen- CRC Press, Boca Raton, FL, pp. 57–81.
dependent and can take place wherever NH4 + is present. 2. Costerton, J.W. et al. (1995). Microbial biofilms. Ann. Rev.
Denitrification, which is the anaerobic reduction of NO3 − Microbiol. 49: 711–745.
to nitrite (NO2 − ), nitrous oxide (N2 O), and N2 , occurs 3. Paerl, H.W. and Pinckney, J.L. (1996). A mini-review of
deeper in the biofilm where oxygen is low and NO3 − is microbial consortia: their roles in aquatic production and
high. Dissimilatory NO3 − reduction to NH4 + is strictly an biogeochemical cycling. Microbial Ecol. 31: 225–247.

Figure 3. Orange-colored biofilm associated with

hot spring outflow draining into the Firehole River,
Yellowstone National Park, Wyoming. The bright
orange color is because of pigments in thermophilic
cyanobacteria and bacteria in the biofilm.
 MICROORGANISMS IN THEIR NATURAL ENVIRONMENT 309

4. Davey, M.E. and O’Toole, G.A. (2000). Microbial biofilms: 24. Battin, T.J. et al. (2003). Effects of current velocity on the
From ecology to molecular genetics. Microbiol. Molecular Biol. nascent architecture of stream microbial biofilms. Appl.
Rev. 64: 847–867. Environ. Microbiol. 69: 5443–5452.
5. Kiørboe, T. (2001). Formation and fate of marine snow: Small- 25. Fenchel, T., King, G.M., and Blackburn, T.H. (1998). Bacte-
scale processes with large-scale implications. Scientia Marina rial Biogeochemistry: The Ecophysiology of Mineral Cycling.
65: 57–71. Academic Press, San Diego, CA, p. 307.
6. Paerl, H.W. and Kuparinen, J. (2002). Microbial aggregates 26. Tank, J.L. and Webster, J.R. (1998). Interaction of substrate
and consortia. In: Encyclopedia of Environmental Micro- and nutrient availability on wood biofilm processes in
biology. G. Bitton (Ed.). John Wiley & Sons, New York, streams. Ecology 79: 2168–2179.
pp. 160–181. 27. Fischer, H. (2003). The role of biofilms in the uptake and
7. Simon, M., Grossart, H-P., Schweitzer, B., and Ploug, H. transformation of dissolved organic matter. In: Aquatic
(2002). Microbial ecology of organic aggregates in aquatic Ecosystems: Interactivity of Dissolved Organic Matter.
S.E.G. Findlay and R.L. Sinsabaugh (Eds.). Academic Press,
ecosystems. Aquat. Microbial Ecol. 28: 175–211.
San Diego, CA, pp. 285–313.
8. Stoodley, P., Sauer, K., Davies, D.G., and Costerton, J.W.
28. Espeland, E.M., Francoeur, S.N., and Wetzel, R.G. (2001).
(2002). Biofilms as complex differentiated communities. Ann.
Influence of algal photosynthesis on biofilm bacterial pro-
Rev. Microbiol. 56: 187–209.
duction and associated glucosidase and xylosidase activities.
9. Turley, C.M. (2002). The importance of ‘marine snow’. Microbial Ecol. 42: 524–530.
Microbiol. Today 29: 177–179.
29. Battin, T.J. (2000). Hydrodynamics is a major determinant of
10. Zimmermann-Timm, H. (2002). Characteristics, dynamics streambed biofilm activity: From the sediment to the reach
and importance of aggregates in rivers—an invited review. scale. Limnol. Oceanog. 45: 1308–1319.
Int. Rev. Hydrobiol. 87: 197–240. 30. Emerson, D. and Revsbech, N.P. (1994). Investigation of
11. Prakash, B., Veeregowda, B.M., and Krishnappa, G. (2003). an iron-oxidizing microbial mat community located near
Biofilms: A survival strategy of bacteria. Curr. Sci. 85: Aarhus, Denmark: Field studies. Appl. Environ. Microbiol.
1299–1307. 60: 4022–4031.
12. Wotton, R.S. and Malmqvist, B. (2001). Feces in aquatic 31. Brock, T.D. (1994). Life at High Temperatures. Yellowstone
ecosystems. BioScience 51: 537–544. Association for Natural Science, Yellowstone National Park,
13. Kerner, M. et al. (2003). Self-organization of dissolved organic Wyoming, p. 31.
matter to micelle-like microparticles in river water. Nature
422: 150–154.
14. Thurman, E.M. (1985). Organic Geochemistry of Natural MICROORGANISMS IN THEIR NATURAL
Waters. Martinus Nijhoff/Dr. W. Junk Publishers, Dordrecht, ENVIRONMENT
The Netherlands, p. 497.
15. Wotton, R.S. (2004). The ubiquity and many roles of MATTHIAS KOSCHORRECK
exopolymers (EPS) in aquatic systems. Scientia Marina 68: Center of Environmental
13–21. Research (UFZ)
16. Wotton, R.S. (1996). Colloids, bubbles, and aggregates—a Magdeburg, Germany
perspective on their role in suspension feeding. J. N. Amer.
Bentholog. Soc. 15: 127–135.
17. Böckelmann, U., Manz, W., Neu, T.R., and Szewzyk, U. GENERAL REMARKS
(2000). Characterization of the microbial community of lotic
organic aggregates (‘river snow’) in the Elbe River of Germany In their natural aquatic environment, microbes face condi-
by cultivation and molecular methods. FEMS Microbiol. Ecol. tions that principally differ from the picture of a mixture of
33: 157–170. different microbes swimming around in a kind of liquid cul-
18. Neu, T.R. (2000). In situ cell and glycoconjugate distribution ture. The study of microorganisms in laboratory cultures
in river snow studied by confocal laser scanning microscopy. often gives limited information about the role and behavior
Aquat. Microbial Ecol. 21: 85–95. of microorganisms in their natural habitat. Our under-
19. Berger, B., Hoch, B., Kavka, G., and Herndl, G.J. (1996). standing of microbes in their natural aquatic habitat was
Bacterial colonization of suspended solids in the River considerably increased by the development of in situ meth-
Danube. Aquat. Microbial Ecol. 10: 37–44. ods in the last decades. The introduction of microsensors
20. Grossart, H.P. and Ploug, H. (2000). Bacterial production allowed characterizing the microbial habitat on the size
and growth efficiencies: Direct measurements on riverine scale of microorganisms. The recent advances in micro-
aggregates. Limnol. Oceanog. 45: 436–445. scopic techniques (confocal laser scanning microscopy,
21. Smith, D.C., Simon, M., Alldredge, A.L., and Azam, F. (1992). atomic force microscopy) offer the possibility of real 3-D
Intense hydrolytic enzyme activity on marine aggregates examination of microbial aggregates or biofilms. Molecular
and implications for rapid particle dissolution. Nature 359: techniques such as in situ hybridization to study the phy-
139–142. logeny and physiology of microorganisms in their natural
22. Battin, T.J., Kaplan, L.A., Newbold, J.D., and Hansen, C.M. environment are rapidly, improving.
(2003). Contributions of microbial biofilms to ecosystem A characteristic of the natural habitat is its hetero-
processes in stream mesocosms. Nature 426: 439–442. geneity. Hot spots of high activity may be surrounded by
23. Liu, Y. and Tay, J.-H. (2001). Metabolic response of biofilm a ‘‘desert.’’ Natural habitats are characterized by gradi-
to shear stress in fixed-film culture. J. Appl. Microbiol. 90: ents. The co-occurrence of various, sometimes opposing,
337–342. gradients creates a high diversity of ecological niches.
310 MICROORGANISMS IN THEIR NATURAL ENVIRONMENT

Depending on transport velocities and reaction rates, these

gradients can occur on a meter scale (in open water)
down to a micrometer (in sediments) or even submicrom-
eter scale (in microbial aggregates) (Fig. 1). The natural
environment is not static but subject to fast temporal
fluctuations. Temporal fluctuations occur on different
timescales from seasonal effects down to very short-term
changes within seconds or minutes. A phototrophic micro-
bial mat, for example, can shift from oxygen oversaturation
to anoxia within minutes when it is darkened (1). Microor-
ganisms have developed different strategies to cope with
changing conditions. The largest part of the bacteria in
aquatic systems is in a metabolically inactive dormant
state, waiting for favorable conditions. Motility, often in
combination with chemotaxis, is an important strategy
for following changing gradients. Microorganisms can also
alter their phenotypic properties in response to chang-
ing conditions. On the scale of microorganisms, surface
effects are important and diffusion becomes the domi-
nant transport process. Solid surfaces are covered by a
diffusive boundary layer (DBL) (2) (Fig. 1).
The term microbes means bacteria, algae, fungi, and
protozoa. This article focuses on the ecology of bacteria Figure 2. Confocal laser scanning micrograph of a lotic aggregate
in aquatic environments. A general overview of aquatic from the River Saale, Germany. Maximum intensity projection of
microbiology can be found in Ford (3), Hurst et al. (4), or 31 sections at 1 µm distance from four channels. Color allocation:
reflection = white, autofluorescence of algae = blue, nucleic acid
Sorokin (5).
staining of bacteria = green, lectin stained glycoconjugates = red
(courtesy of Thomas R. Neu) This figure is available in full color
HABITATS online.
Aquatic systems offer different habitats for microbes to
live in: free water, sediment, and hyporheic interstitial and activities are usually higher in the sediment and at
(6). Of special importance are interfaces such as the surfaces than in free water. Sediments typically contain
surface of water, sediment, rocks (epilithon), and aquatic about 109 bacteria per mL. The fraction of metabolically
plants (epiphyton). Microbes that live in free water active cells is also higher in sediment compared to free
(pelagic organisms) can be dispersed as single cells or water. In a Swedish lake, only 4% of the pelagic bacteria,
concentrated in aggregates (Fig. 2). Organic aggregates, 37% of epiphytic bacteria, and 46% of sediment bacteria
which are larger than 500 µm, are called lake, river, or were metabolically active (8). This can have several
marine ‘‘snow’’ (7). They are hot spots of microbial life in causes such as higher nutrient availability, higher habitat
aquatic systems and are enriched in nutrients compared diversity, or protection against predation. Microbial
to the surrounding water. The formation of aggregates numbers and activities usually decrease exponentially
allows direct syntrophic interactions such as the exchange as sediment depth increases. The absence of turbulent
of metabolic intermediates. mixing allows the patchiness and spatial (mostly vertical)
Natural waters typically contain between 0.5 and gradients to form in the sediment. The situation is
5 × 106 bacteria per mL (5). Microbial numbers, diversity, especially complex in the rhizosphere of aquatic plants.

0 −2
Water
2
Depth in water column, m

Depth in sediment, mm

−1
4
DBL
6
0
8 Sediment
1
10

12
2
14
3
16
Figure 1. Vertical gradient of oxygen (a) in a lake with
0 20 40 60 80 100 0 20 40 60
anoxic hypolimnion and (b) at the sediment water interface.
DBL = diffusive boundary layer. Oxygen, % saturation Oxygen, % saturation
 MICROORGANISMS IN THEIR NATURAL ENVIRONMENT 311

Surfaces are often covered by biofilms (9,10) or on organic substrates. The availability of electron accep-
microbial mats (11–13). Living attached to a surface tors (O2 , NO3 2− , FeIII , SO4 2− , CO2 ) is a major determinant
allows microbes to stay in favorable places under of which physiological groups of microorganisms are pre-
conditions of moving water. A biofilm is a ‘‘surface dominant. Surface waters are mostly oxic, and aerobic
accumulation, which is not necessarily uniform in time or microbes dominate. In sediments and under certain cir-
space, that comprises cells immobilized at a substratum cumstances also in free water (groundwater, hypolimnion),
and frequently embedded in an organic polymer matrix of conditions are anoxic, and anaerobic organisms dominate
microbial origin’’ (10). The solid surface on which a biofilm microbial populations (Fig. 1). Anoxic bacteria are also
develops is called the ‘‘substratum.’’ Living conditions in frequently found in oxic environments (15). These can be
biofilms may differ substantially from conditions in free facultative anaerobes such as denitrifiers and also oxy-
water. Biofilms are enriched in nutrients, are a place gen resistant anaerobes or strict anaerobes that live in
of intensive biotic interactions, and may offer protection anoxic microsites (e.g., inside microbial aggregates). The
against unfavorable conditions. Biofilms are not stable oxic–anoxic interface is a place of high microbial activity
but undergo different successional stages. Biofilm growth and special adaptations (16). At the oxic–anoxic interface,
starts due to adsorption of cells to a substratum. As the redox cycling of mobile electron carriers such as FeII /FeIII
biofilm growth thickers, its composition changes, transport or H2 S/SO4 2− can occur on a small spatial scale.
processes influence metabolic activities in the biofilm, and Heterotrophic microbes depend on organic substrates
it becomes more and more affected by shear stress which as a carbon source; autotrophs such as photosynthetic
finally leads to detachment and washout of parts of the organisms or nitrifying bacteria can use CO2 that is
biofilm. available in most aquatic systems which are buffered by a
The liquid–gas interface is characterized by high carbon buffer system. Exceptions are acidic waters where
concentrations of oxygen and hydrophobic substances and inorganic carbon is available only as physically dissolved
high light intensities. It is colonized by a special microflora CO2 and potentially limits autotrophic growth.
called neuston (14). Concentrations of nutrients are usually low in natural
Microbes play an important role in artificial aquatic environments. Consequently, cells in natural water are
systems such as wastewater treatment plants, biotech- usually much smaller than cells grown in culture media,
nological reactors, and drinking water supplies. Living and generation times are longer. Microbes have developed
conditions in such systems can differ substantially from several strategies to survive starvation. Most obvious is
natural habitats and are often characterized by high the formation of metabolically inactive resting stages like
nutrient content and homogeneity (in reactors) or very spores or the increased affinity of nutrient uptake systems.
low nutrient content (in drinking water). The general All nutrients have to cross the diffusive boundary layer
habitat characteristics and the adaptations of microor- before they can be taken up by an organism. The diffusion
ganisms, however, are similar in natural and artificial flux to a cell depends on its surface to volume ratio, so
aquatic systems. smaller organisms are more efficient in diffusive substrate
uptake (17). Microbial growth in natural environments is
ABIOTIC FACTORS often very slow (18).

Water Temperature
Per definition, aquatic bacteria live in habitats where Temperature is the second most important parame-
water is available. Many natural waterbodies, however, ter besides nutrient availability that controls microbial
temporarily dry out due to water level fluctuations. growth in natural aquatic systems. Aquatic microbes can
Organisms in such habitats develop strategies to survive live in liquid water at all temperatures between freez-
periods of dessication. They may form spores or other ing and boiling. The world record for hyperthermophilic
resting stages. bacteria is currently 113 ◦ C, and many extreme ther-
Even if microbes are swimming in liquid water, it might mophiles cannot growth at ‘‘normal’’ temperatures (19).
not be available to them because of low water activity. Such extreme temperatures occur at submarine hydrother-
Water activity expresses the ratio of water in the vapor mal vents where water remains liquid at temperatures
phase to the amount in vapor saturated air (ranging from above 100 ◦ C due to the hydrostatic pressure. Most nat-
0 to 1.0). Low water activity can be caused by high salt ural aquatic habitats, however, are characterized by low
concentration. Most bacteria need a water activity higher temperatures. In deep lakes and in the oceans, great parts
than 0.98 which is the activity of sea water at 25 ◦ C are colder than 5 ◦ C, and microbial processes are slow.
(14). Especially resistant halobacteria can live at water The temperature in the natural environment is usually
activities as low as 0.6. They are found in salt lakes or salt below the optimum growth temperature of bacteria. Even
saturated evaporation ponds. polar or alpine ice is colonized by microorganisms, and
the boundary between surface ice and water is a place of
Substrates microbial life (20).
Microbes need different types of substrates. These include
Light
electron donors, electron acceptors, macronutrients (car-
bon, nitrogen, and sulfur), micronutrients (e.g., different Microbes are affected by light in different ways, both
metals and trace elements). Many organisms also depend directly or indirectly. Light is the energy source of
312 MICROORGANISMS IN THEIR NATURAL ENVIRONMENT

photosynthesis which can take place under aerobic and a complex network of different reactions carried out by
anaerobic conditions. Anoxic photosynthesis typically different organisms.
takes place at the chemocline where anoxic conditions
meet maximum light intensities. High abundances of Syntrophy
purple sulfur bacteria can be concentrated in only a
In anoxic environments, the degradation of organic mat-
few centimeter thick layer of water (21). It is especially
ter resulting in the accumulation of uncompletely oxidized
important for phototrophic organisms to maintain their
intermediates such as acetate or hydrogen gas is energeti-
vertical position in the water column. To counteract
cally unfavorable. Methanogens and sulfate reducers can,
sedimentation, microbes may use active movement or
for example, consume hydrogen gas and lower its partial
regulate buoyancy by intracellular gas vesicles (22).
pressure so that the hydrogen producing process becomes
In moving water, pelagic organisms are continuously
thermodynamically favorable. This ‘‘interspecies hydrogen
transported from regions of high light near the surface
transfer’’ allows the degradation of compounds which are
to darker conditions further down and back (23).
refractory to degradation by the pure culture of a single
At the water surface or on mud surfaces, solar radiation
species (33). A prerequisite of such syntrophic interaction
can be inhibiting for microbial life. One strategy to
is that the organisms are close spatially.
prevent photodamage is migration of motile organisms
into the sediment during periods of high illumination
(24). Microorganisms have pigments to protect them from Grazing
photodamage. UV radiation can alter nutrient conditions Bacterial numbers in pelagic systems and in surface
by mobilizing dissolved organic matter (25) or by reducing sediment are often controlled by grazing. Microorganisms
iron (26). have developed different strategies against grazing. They
can change cell size, change cell shape to longer forms, or
pH form multicellular aggregates (34). Some bacteria are not
digested by higher organisms but live as endosymbionts
Natural aquatic systems cover the whole pH range from
inside bigger aquatic organisms.
0 to 14, and all of them are inhabited by microbes (27).
Acidophilic bacteria (28) live in extremely acidic waters.
Basophiles and alkaliphiles live in hypersaline waters. Bioturbation
The activity of microorganisms can modify the pH of their The activity of macroorganisms can greatly modify the
habitat. The oxidation of iron or sulfur compounds can sediment environment (35). Organisms and particles can
lower the pH to extreme values below pH 2. Photosynthetic be moved by bioturbation, and porewater chemistry might
activity can create local pH values higher than pH 10 in be changed. Bioturbation can increase both homogeneity
phototrophic biofilms (29). or patchiness of the sediment and introduce short-term
changes of conditions.
Shear Stress
Water movement affects microbes mechanically in aquatic Bacteriophages
habitats and is an important regulator of biofilm growth. Viruses are the most abundant biological entities in
High current velocities cause detachment and washout of aquatic systems. Typical abundances of these 20 to 200 nm
attached microorganisms (9). On the other hand, drift is particles are 1010 per liter in surface water. They probably
an import mechanism for colonizing new habitats. infect all organisms and influence biogeochemical and
ecological processes as well as gene transfer among
microorganisms (36).
BIOTIC INTERACTIONS

Food Web CONCLUDING REMARK

Microbes play an important role in the food web in aquatic
environments. Bacteria, flagellates, and microzooplankton Due to the development of new methods for studying
consume dissolved organic matter (DOM) which is released microbes in their natural environment, we now have a
by phytoplankton. By this microbial loop (30), energy first impression of the complexity of the natural habitat.
and nutrients are returned to the main food chain. In Many issues such as biotic interactions in aggregates
unproductive aquatic systems, bacterial production can and biofilms, the reaction of microbes to quickly changing
exceed phytoplankton production and make these systems environmental conditions, and their distribution in very
net sources of CO2 (31). heterogeneous environments remain important topics for
Bacteria compete with eukaryotic algae and phago- further research. The study of the distribution, role, and
trophs for resources. Usually, bacteria play a larger role function of microbes in their natural aquatic environment
in low productivity (oligotrophic) ecosystems compared is one of the most exciting areas in aquatic ecology.
to highly productive (eutrophic) systems. Oligotrophic
systems that have low nutrient concentrations and higher BIBLIOGRAPHY
proportions of soluble organic C favor heterotrophic
bacteria to phytoplankton and phagotrophic heterotrophs 1. Revsbech, N.P. and Jørgensen, B.B. (1986). Microelectrodes:
(32). The microbial mineralization of organic matter is Their use in microbial ecology. Adv. Microb. Ecol. 9: 293–351.
 CALIBRATION OF HYDRAULIC NETWORK MODELS 313

2. Boudreau, B.P. and Jørgensen, B.B. (Eds.). (2001). The Ben- 25. Karentz, D. et al. (1994). Impact of UV-B radiation on pelagic
thic Boundary Layer—Transport Processes and Biogeochem- freshwater ecosystems, report of working group on bacteria
istry. Oxford University Press, Oxford. and phytoplankton. Arch. Hydrobiol. 43: 31–69.
3. Ford, T.E. (1993). Aquatic Microbiology—An Ecological 26. Herzsprung, P. et al. (1998). Vertical and annual distribution
Approach. Blackwell, Boston. of ferrric and ferrous iron in acidic mining lakes. Acta
4. Hurst, C.J., Knudsen, G.R., McInerney, M.J., Stetzen- Hydrochim. Hydrobiol. 26: 253–262.
bach, L.D., and Walter, M.V. (Eds.). (1997). Manual of Envi- 27. Seckbach, J. (Ed.). (1999). Enigmatic Microorganisms and
ronmental Microbiology. ASM Press, Washington, D.C. Life in Extreme Environments. Kluwer, Dordrecht.
5. Sorokin, Y.I. (1999). Aquatic Microbial Ecology. Backhuys 28. Johnson, D.B. (1998). Biodiversity and ecology of aciophilic
Publishers, Leiden. microorganisms. FEMS Microb. Ecol. 27: 307–317.
6. Pusch, M. et al. (1998). The role of micro-organisms in the 29. Revsbech, N.P., Nielsen, J., and Hansen, P.K. (1988). Benthic
ecological connectivity of running waters. Freshwater Biol. primary production and oxygen profiles. In: Nitrogen Cycling
40: 453–495. in Coastal Marine Environments. T.H. Blackburn and J.
Sörensen (Eds.). John Wiley & Sons, New York, NY,
7. Simon, M., Grossart, H.P., Schweitzer, B., and Ploug, H.
pp. 69–83.
(2002). Microbial ecology of organic aggregates in aquatic
ecosystems. Aquatic Microb. Ecol. 28: 175–211. 30. Azam, F. et al. (1983). The ecological role of water-column
microbes in the sea. Mar. Ecol. Prog. Ser. 10: 257–263.
8. Haglund, A.L., Törnblom, E., Boström, B., and Tranvik, L.
(2002). Large differences in the fraction of active bacteria in 31. del Giorgio, P.A., Cole, J.J., and Cimbleris, A. (1997). Respi-
plankton, sediments and biofilm. Microb. Ecol. 43: 232–241. ration rates in bacteria exceed phytoplankton production in
unproductive aquatic systems. Nature 385: 148–151.
9. Characklis, W.G. and Marshall, K.C. (Eds.). (1990). Biofilms,
Ecological and Applied Microbiology. John Wiley & Sons, 32. Cotner, J.B. and Biddanda, B.A. (2002). Small players, large
New York, NY. role, microbial influence on biogeochemical processes in
pelagic aquatic ecosystems. Ecosystems 5: 105–121.
10. Bryers, J.D. (Ed.). (2000). Biofilms II—Process Analysis and
Application. John Wiley & Sons, New York, NY. 33. Schink, B. (1991). Syntrophism among Prokaryotes. In: The
Procaryotes. A. Balows, H.G. Trüper, M. Dworkin, W. Harder,
11. Cohen, Y. and Rosenberg, E. (1989). Microbial Mats—Physio- and K.H. Schleifer (Eds.). Springer, Berlin, pp. 276–299.
logical Ecology of Benthic Microbial Communities. American
34. Jürgens, K. and Sala, M. (2000). Predation-mediated shifts
Society for Microbiology, Washington.
in size distribution of microbial biomass and activity during
12. Stal, L.J. and Caumette, P. (Eds.). (1994). Microbial Mats— detritus decomposition. Oikos 91: 29–40.
Structure, Development and Environmental Significance.
35. Aller, Y., Woodin, S.A., and Aller, R.C. (Eds.). (2001).
NATO ASI Series. Springer, Berlin.
Organism-Sediment Interactions. University of South Car-
13. Krumbein, W.E., Paterson, D., and Stal, L.J. (Eds.). (1994). olina, Columbia.
Biostabilization of Sediments. BIS Verlag, Oldenburg.
36. Fuhrman, J.A. (1999). Marine viruses and their biogeochem-
14. Schlegel, H.G. and Jannasch, H.W. (1991). Prokaryotes and ical and ecological effects. Nature 399: 541–548.
Their Habitats. In: A. Balows, H.G. Trüper, M. Dworkin, W.
Harder, and K.H. Schleifer (Eds.). The Prokaryotes. Springer,
Berlin, pp. 75–125.
CALIBRATION OF HYDRAULIC NETWORK
15. Peters, V. and Conrad, R. (1995). Methanogenic and other
strictly anaerobic bacteria in desert soil and other oxic soils. MODELS
Appl. Environ. Microb. 61: 1673–1676.
SRINIVASA LINGIREDDY
16. Brune, A., Frenzel, P., and Cypionka, H. (2000). Life at the
oxic-anoxic interface, microbial activities and adaptations. LINDELL E. ORMSBEE
FEMS Microb. Rev. 24: 699–710. DON J. WOOD
17. Koch, A. (1996). What size should a bacterium be? A question University of Kentucky
of scale. Ann. Rev. Microbiol. 50: 317–348. Lexington, Kentucky
18. Koch, A.L. (1997). Microbial physiology and ecology of slow
growth. Microbiol. Molec. Biol. Rev. 61: 305–318.
19. Rothschild, L.J. and Mancinelli, R.L. (2001). Life in extreme INTRODUCTION
environments. Nature 409: 1092–1101.
20. Thomas, D.N. and Diekmann, G.S. (2002). Antarctic sea Availability of reliable network modeling software coupled
ice—a habitat for extremophiles. Science 295: 641–644. with affordable computing hardware technology has led
21. Overmann, J., Beatty, J.T., Krouse, H.R., and Hall, K.J. to rapid growth in the use of hydraulic network mod-
(1996). The sulfur cycle in the chemocline of a meromictic els for design, planning, and operational studies of water
salt lake. Limnol. Oceanogr. 41: 147–156. distribution systems. Friendly graphical user interfaces
22. Konopka, A. (1982). Buoyancy regulation and vertical migra- (GUIs) and integrated computer-aided design (CAD) and
tion by Oscillatoria rubescens in Crooked Lake, Indiana. Br. geographic information systems (GIS) modeling environ-
Phycol. J. 17: 427–442. ments helped drastically reduce model development time
23. Reynolds, C.S. (1997). Vegetation processes in the pelagic, and the associated costs. The validity of these models, how-
a model for ecosystem theory. In: Excellence in Ecology. ever, depends largely on the accuracy of input data and
O. Kinne (Ed.). Ecology Institute, Oldendorf. the assumptions made in developing the model. Although
24. Kingston, M.B. (1999). Effect of light on vertical migration carefully developed models tend to have greater control on
and photosynthesis of Euglena proxima (Euglenophyta). J. much of the data associated with the model, certain model
Phycol. 35: 245–253. parameters exist that are either not readily available or
314 CALIBRATION OF HYDRAULIC NETWORK MODELS

difficult to obtain. Such parameters typically include pipe an actual field survey. Such considerations obviously influ-
roughness factors and nodal demands and their spatial and ence the methods used to collect the necessary model data
temporal distribution. As a result of the difficulty of obtain- and the subsequent calibration steps.
ing economic and reliable measurements of both of these
parameters, final model values are normally determined
IDENTIFICATION OF CALIBRATION MODEL PARAMETERS
through the process of model calibration. Model calibration
AND THEIR INITIAL ESTIMATES
involves adjustment of these and other uncertain network
model parameters until the model results closely approx- The second step in calibrating a hydraulic network model
imate actual observed conditions as measured from field is to determine initial estimates of the primary model
data. In general, a network model calibration effort should parameters. Although most models will have some degree
encompass seven basic steps: of uncertainty associated with several model parameters,
the two model parameters that normally have the greatest
1. Identification of intended use of the model degree of uncertainty are the pipe roughness coefficients
2. Identification of calibration model parameters and and the demands to be assigned to each junction node.
their initial estimates
3. Model studies to determine calibration data sources Pipe Roughness Values
4. Data collection Initial estimates of pipe roughness values may be obtained
5. Macrocalibration using average literature values or directly from field mea-
6. Sensitivity analysis surements. Various researchers and pipe manufacturers
have developed tables that provide estimates of pipe rough-
7. Microcalibration
ness as a function of various pipe characteristics, such as
pipe material, pipe diameter, and pipe age. Table 1 shows
These steps have been successfully tested by the typical Hazen–William roughness factors (2). Although
authors on several model calibration studies. Each of these such tables may be useful for new pipes, their specific
steps is discussed in the following sections. applicability to older pipes decreases significantly as the
pipes age. In addition, the variation in roughness with
IDENTIFICATION OF INTENDED USE OF THE MODEL pipe age depends on other factors, such as water chem-
istry, and will vary from location to location. As a result,
Before calibrating a hydraulic network model, it is initial estimates of pipe roughness for all pipes other than
important to first identify its intended use (e.g., pipe relatively new pipes should normally come directly from
sizing for master planning, operational studies, design field testing. Even when new pipes are being used, it is
projects, rehabilitation studies, water quality studies) and helpful to verify the roughness values in the field because
the associated type of hydraulic analysis (steady-state the roughness coefficient used in the model may actually
versus extended-period). Usually, the type of analysis is represent a composite of several secondary factors such as
directly related to the intended use. For example, water fitting losses and system skeletonization.
quality and operational studies require an extended-period
Adjusting Pipe Roughness vs. Pipe Diameter
analysis, whereas some planning or design studies may be
performed using a steady-state analysis. In the latter, The approach of adjusting pipe roughness values assumes
the model predicts system pressures and flows at an that the pressure drop is entirely because of frictional
instant in time under a specific set of operating conditions and other minor losses within the pipeline. If the
and demands (e.g., average or maximum daily demands). intended use for the calibrated model does not include
In extended-period analysis, the model predicts system water quality simulations, this approach of adjusting
pressures and flows over an extended period (typically 24 pipe roughness value will have little or no impact on
hours). model results. On the other hand, this approach might
Both the intended use of the model and the associated seriously affect results from water quality simulations,
type of analysis provide some guidance about the type which is particularly true when a substantial portion of
and quality of collected field data and the desired level the distribution system comprises fairly old pipelines. Loss
of agreement between observed and predicted flows and of carrying capacity in such pipes is not only because of
pressures (1). Models for steady-state applications can increased roughness but also because of the reduced flow
be calibrated using multiple static flow and pressure area resulting from tuberculation, scaling, sedimentation,
observations collected at different times of day under and biofilm growth. Flow area reductions as high as
varying operating conditions (e.g., fire hydrant flow test 50% are not uncommon in certain distribution systems.
data). On the other hand, models for extended-period Although adjusting pipe roughness to reflect the reduction
applications require field data collected over an extended in carrying capacity will not affect hydraulic gradeline and
period (e.g., one to seven days). pressure calculations, significant deviations in simulated
In general, a higher level of model calibration is water quality parameters are possible as the erroneous
required for water quality analysis or an operational study velocities are used in water quality calculations. Therefore,
than for a general planning study. For example, determin- for more accurate water quality predictions, it might be
ing ground elevations using a topographic map may be prudent to adjust the pipe diameters by limiting roughness
adequate for the latter, whereas the former may require coefficients to certain predetermined values. For example,
 CALIBRATION OF HYDRAULIC NETWORK MODELS 315

Table 1. Typical Hazen–William Pipe Roughness Factors

Pipe Size
Pipe Material Pipe Age (Inches) C Factor

Cast Iron New All Sizes 130

5 years old 12 and Over 120
8 119
4 118
10 years old 24 and Over 113
12 111
4 107
20 years old 24 and Over 100
12 96
4 89
30 years old 30 and Over 90
16 87
4 75
40 years old 30 and Over 83
16 80
4 64
50 years old 40 and Over 77
24 74
4 55
Welded Steel Values of C the same as for cast-iron
pipes, 5 years older
Riveted Steel Values of C the same as for cast-iron
pipes, 10 years older
Wood Stave Average value, 120
regardless of age
Concrete or Concrete Large sizes, good workmanship, steel 140
Lined forms
Large sizes, good workmanship, 120
wooden forms
Centrifugally spun 135
Vitrified In good condition 110
Plastic or Drawn Tubing 150

1 inch = 25.4 mm.

if the computed Hazen–William roughness coefficient for MODEL STUDIES TO DETERMINE CALIBRATION DATA
a pipeline is 40, one could limit the coefficient to about SOURCES
90 and compute the pipe diameter that would produce the
measured pressure drop. After model parameters have been estimated, the accuracy
of the model parameters can be assessed, which is done
by executing the computer model using the estimated
Spatial and Temporal Distribution of Nodal Demands parameter values and observed boundary conditions and
then comparing the model results with the results
The second major parameter determined in calibration from actual field observations. Collecting accurate and
analysis is the average (steady-state analysis) or tem- appropriate field data is crucial for proper calibration
porally varying (extended-period analysis) demand to be of a hydraulic network model. One should use at least
assigned to each junction node. Initial average estimates 10 to 20 pressure readings measured at locations evenly
of nodal demands can be obtained by identifying a region of spread over the water distribution system for steady-state
influence associated with each junction node, identifying calibration. These pressure readings should ideally be
the types of demand units in the service area, and multi- collected when the network is under ‘‘stressed’’ conditions,
plying the number of each type by an associated demand i.e., during a peak-hour demand condition, during filling
factor. Alternatively, the estimate can be obtained by first cycle of a ground-level pumped storage tank, or during
identifying the area associated with each type of land use hydrant flow tests. Under these conditions, the increased
in the service area and then multiplying the area of each flowrates make the pressure readings more sensitive to
type by an associated demand factor. In either case, the small changes in pipe roughness values. The hydrant
sum of these products is an estimate of the demand at flows that produce more than 15 psi drop from static
the junction node. Although in theory the first approach pressures are ideal for calibration studies. In most
should be more accurate, the latter approach is more expe- distribution systems, the peak-hour demands may not
dient. Estimates of unit demand factors are available from provide 15 psi drop at sufficient number of locations.
water resource handbooks and textbooks (3). Ground-level pumped storage tanks are not common in
316 CALIBRATION OF HYDRAULIC NETWORK MODELS

most distribution systems. Therefore, hydrant flow tests performing a fire flow test are illustrated by McEnroe
provide a logical alternative for pressure reading in most et al. (4).
systems. However, not every hydrant flow will result in In order to obtain sufficient data for an adequate model
an adequate pressure drop. Hydrants located at or near calibration, it is important that data from several fire
large pressure mains, supply sources, and elevated storage flow tests be collected. Before conducting each test, it
tanks tend to provide less than the desired pressure drop. is also important that the associated system boundary
On the other hand, certain hydrants might yield very condition data be collected, which includes information
little flow while showing large pressure drops (e.g., 80 psi on tank levels, pump status, etc. It is a common practice
of pressure drop while yielding only 200 gpm) because of for the local fire departments to conduct hydrant flow
localized losses such as constricted flow areas. Although tests and record the time of day and corresponding flows
a network model that is yet to be calibrated may not be and pressures. However, in most cases, such records do
meaningful for planning and operational studies, it can not include the boundary conditions associated with each
provide some insight into the locations for hydrant flow hydrant flow test, as the main purpose for their tests is
tests that provide reasonable flows and adequate pressure to rate the fire hydrant and not necessarily for hydraulic
drops. The following procedure is recommended to locate calibration. Therefore, care must be taken to avoid hydrant
hydrants for calibration field studies. flow data that does not include the associated boundary
conditions data. Table 2 may be used as a template when
• Perform hydrant flow analysis on all hydrants collecting calibration data using hydrant flow tests.
in the distribution system. Hydrant flow analysis
appears to be a standard feature in most commercial Telemetry Data
modeling packages and provides information on
In addition to static test data, data collected over an
flowrate that a hydrant can deliver while maintaining
extended period of time (typically 24 hours) can be very
a minimum pressure of 20 psi throughout the
useful for use in calibrating network models. The most
distribution system. One could use the nearest
common type of data will include flowrate data, tank
junction nodes in place of hydrants if the model does
water level data, and pressure data. Depending on the
not contain the hydrant information.
level of instrumentation and telemetry associated with
• Compute the flowrate per unit pressure drop the system, much of the data may be already collected as
(difference between static and residual pressure) for part of the normal operations. For example, most systems
each hydrant. collect and record tank levels and average pump station
• Rank order the hydrants from lowest to highest of discharges on an hourly basis. These data are especially
flowrate per unit pressure drop. useful in verifying the distribution of demands among
• Select about 15 to 25 hydrants from the top of the list the various junction nodes. If such data are available,
that are evenly scattered throughout the distribution the data should first be checked for accuracy before use
system. Care must be taken to avoid hydrants with in the calibration effort. If such data are not readily
unusually low flowrates to pressure drop ratios. A available, the modeler may have to install temporary
closer look at these hydrants and related flows might pressure gauges or flowmeters in order to obtain the data.
reveal problems with the network model. The authors In the absence of flowmeters in lines to tanks, inflow
have also found that unusually high flowrate to or discharge flowrates can be inferred from incremental
pressure drop ratios could be because of network readings of the tank level.
anomalies that can be corrected by a close observation In systems with ground-level pumped storage tanks
of network model parameters. (Fig. 1), the tanks are filled during off-peak demand
periods using the high system pressures. Pressures in
the distribution system drop dramatically during tank-
DATA COLLECTION filling operation, thereby providing an opportunity to use
the associated data for model calibration. Tank telemetry
The next step is to acquire calibration data from field data, along with a few static pressure readings within
studies. This section describes the general procedures for the distribution system where a significant drop occurs,
collecting and recording field data. could be used for calibrating the network model in lieu of
hydrant flow test data.
Fire Flow Tests
Fire flow tests are useful for collecting both discharge and EVALUATE MODEL RESULTS
pressure data for use in calibrating hydraulic network
models. Such tests are normally conducted using both Using fire flow data, the model is used to simulate the
a normal pressure gauge (for measuring both static and discharge from one or more fire hydrants by assigning
dynamic heads) and a Pitot gauge (for use in calculating the observed hydrant flows as nodal demands within the
discharge). In performing a fire flow test, at least two model. The flows and pressures predicted by the model are
separate hydrants are first selected for use in the data then compared with the corresponding observed values in
collection effort. One hydrant is identified as the pressure an attempt to assess model accuracy. In using telemetry
or residual hydrant, whereas the remaining hydrant is data, the model is used to simulate the variation of
identified as the flow hydrant. The general steps for tank water levels and system pressures by simulating
 Table 2. Sample Hydrant Flow Data Collection Table for Calibration Studies
Flowing Hydrant Residual Hydrant Tank Levels Pump Status Remarks

Static Discharge Static Residual

Test Date Time Name Elevation Pressure Pressure Flowrate Name Elevation Pressure Pressure A B . . .. A B . . ..

1
2
3
4
5
6
7
317

8
9
10
11
12
13
14
15
16
17
18
318 CALIBRATION OF HYDRAULIC NETWORK MODELS

microcalibration. Each of these steps is described in the

following sections.

Storage tank MACROCALIBRATION

Elevation In the event that one or more of the measured variable

values are different from the modeled values by an amount
that is deemed to be excessive (i.e., greater than 30%), it is
Filling cycle (valve B closed) likely that the cause for the difference may extend beyond
A errors in the estimates for either the pipe roughness values
Connection to
distribution or the nodal demands. Possible causes for such differences
B
system are many, but may include
Pumping cycle (valve A closed)
• valves that are open in the system but modeled as
closed or partly closed
Plan view • valves that are closed or partly closed in the system
but not modeled correctly
Figure 1. Schematic of a ground-level pumped storage facility. • inaccurate pump curves or pumps defined by useful
power rather than head/flow data
• incorrect pump (on/off) status data or pump operating
the operating conditions for the day over which the field
speeds for variable speed pumps
data was collected. The predicted tank water levels are
then compared with the observed values in an attempt to • inaccurate tank telemetry data
assess model accuracy. • inaccurate flow or pressure gauges
Model accuracy may be evaluated using various • incorrect pipe sizes and lengths (typographi-
criteria. The most common criteria are absolute pressure cal errors)
difference (normally measured in psi) or relative pressure • inaccurate pressure regulating valve settings
difference (measured as the ratio of the absolute pressure • incorrect network geometry
difference to the average pressure difference across the
• inaccurate junction or hydrant elevations
system). In most cases, a relative pressure difference
criteria is normally preferred. For extended period • hydrants located on wrong pipes near multiple pipe
simulations, comparisons are normally made between the junctions (Fig. 2)
predicted and observed tank water levels. To a certain • incorrect pressure zone boundaries
extent, the desired level of model calibration will be related • missing or incorrect demand data—especially per-
to the intended use of the model. For example, a higher taining to large users (e.g., hospitals, schools, etc.)
level of model calibration will normally be required for
water quality analysis or an operational study as opposed The best way to adequately address such errors is
to use of the model in a general planning study. For most to systematically review the data associated with the
applications, a maximum state variable (i.e., pressure model in order to ensure its accuracy (5). In most cases,
grade, water level, flowrate) deviation of less than 5–10% some data will be less reliable than other data. This
is generally regarded as satisfactory (1). observation provides a logical place to start in an attempt
Deviations between results of the model application to identify the problem. Model sensitivity analysis provides
and the field observations may be caused by several another means of identifying the source of discrepancy.
factors, including: (a) erroneous model parameters (i.e., For example, if it is suspected that a valve is closed, this
pipe roughness values and nodal demand distribution), assumption can be modeled by simply closing the line in
(b) erroneous network data (i.e., pipe diameters, lengths, the model and evaluate the resulting pressures. Potential
etc.), (c) incorrect network geometry (i.e., pipes connected errors in pump curves can sometimes be minimized by
to the wrong nodes, etc.), (d) incorrect pressure zone simulating the pumps with negative inflows set equal to
boundary definitions, (e) errors in boundary conditions observed pumps discharges (6).
(i.e., incorrect PRV value settings, tank water levels, Plotting pressure contours pertaining to the simulation
pump curves, etc.), (f) errors in historical operating that produced undue low pressure (compared with the
records (i.e., pumps starting and stopping at incorrect measured value) might provide an insight into certain
times), (g) measurement equipment errors (i.e., pressure problem areas. For example, a significant flow constriction
gauges not properly calibrated, etc.), and (h) measurement (wrong pipe diameter or partially closed valve) in the
error (i.e., reading the wrong values from measurement model would produce rapid pressure drops that could be
instruments). The last two sources of errors can hopefully easily traced using the pressure contour plots. Probing
be eliminated or at least minimized by developing and into headloss/1000 values in the vicinity of the hydrant
implementing a careful data collection effort. Elimination under question might help trace the problem pipes. In
of the remaining errors will frequently require the general, it would be relatively easier to trace the problems
iterative application of the last three steps of the model if the simulated pressure reading is significantly lower
calibration process—macrocalibration, sensitivity, and than the measured value. Only after the model results
 CALIBRATION OF HYDRAULIC NETWORK MODELS 319

In performing a steady-state calibration, the model

parameters are adjusted to match pressures and flowrates
associated with static observations. The normal source
for such data is from fire flow tests. In an extended-
100 mm
period calibration, the model parameters are adjusted
to match time-varying pressures and flows as well as
250 mm 250 mm tank water level trajectories. In most cases, the steady-
state calibration will be more sensitive to changes in pipe
Location of hydrant in the system
roughness, whereas the extended-period calibration will be
more sensitive to changes in the distribution of demands.
As a result, one potential calibration strategy would be to
first fine tune the pipe roughness parameter values using
100 mm the results from fire flow tests and then try to fine tune
the distribution of demands using the flow/pressure/water
level telemetry data.
250 mm 250 mm Historically, most attempts at microcalibration have
typically employed an empirical or trial and error
Correct location of hydrant in the model
approach. Such an approach can prove to be extremely
time-consuming and frustrating when dealing with most
typical water systems. The level of frustration will, of
course, depend somewhat on the expertise of the modeler,
100 mm the size of the system, and the quantity and quality of
the field data. Several commercial software packages offer
an optimized calibration feature that would ease some of
250 mm 250 mm the frustration associated with model calibration. Genetic
algorithm (GA)-based techniques for model calibration
Hydrant placed on a wrong pipe in the model appear to be popular with the commercial software
(due to automated assignment of hydrant to the nearest pipe) packages. The optimized calibration techniques obtain
an optimal set of decision variables (e.g., pipe roughness
Figure 2. Example depicting misplaced hydrant.
values and nodal demands) that minimize the average
deviation between measured and calculated pressures and
and the observed conditions are within some reasonable flows while satisfying all system and bound constraints (7).
degree of correlation (usually less than 20% error) should Although one would expect this to be a one-time process,
the final step of micro-level calibration be attempted. several factors force repeated use of this approach until
a more satisfactory set of decision variables are obtained.
SENSITIVITY ANALYSIS One of the reasons for repeated use of optimized calibration
runs is the way the pipelines are grouped. GA-based
Before attempting a micro-level calibration, it is helpful to approaches solve for optimal group roughness values
perform a sensitivity analysis of the model in order to help instead of roughness values for individual pipelines.
identify the most likely source of model error. This analysis Alternatively, they can solve for an optimal multiplication
can be accomplished by varying the different model factor for the preassigned roughness values for all the
parameters by different amounts and then calculating pipes in a group. Pipes of similar roughness characteristics
the associated effect. For example, many current network (estimated based on material, age, and diameters of the
models have as an analysis option the capability to make pipes) are treated as one group. However, this is a highly
multiple simulations in which global adjustment factors subjective process and therefore offers opportunities for
can be applied to pipe roughness values or nodal demand regrouping and reapplying the optimized calibration
values. By examining such results, the user can begin technique. Similarly, the explicit bound constraints on
to identify which parameters have the most significant the pipe group roughness values offer opportunities for
impact on the model results and thereby identify potential relaxing/tightening of the bounds and reapplying the
parameters for subsequent fine tuning through micro-level optimized calibration technique. The iterative procedure
calibration. for model calibration using optimized calibration approach
is summarized in the following.
MICROCALIBRATION
• Setup the model to include calibration data (observed
After the model results and the field observations are pressures and flow readings).
in reasonable agreement, a micro-level model calibration
should be performed. As discussed previously, the • Group the pipes based on similar roughness charac-
two parameters adjusted during this final calibration teristics.
phase will normally include pipe roughness and nodal • Estimate the roughness values for each group
demands. In many cases, it may be useful to break of pipes.
the microcalibration into two separate steps: (a) steady- • Set up an upper and lower bound for each group of
state calibration, and (b) extended-period calibration. pipes based on estimated group roughness values. As
320 NUMERICAL MODELING OF CURRENTS

an initial guess, these bounds should be within 20% NUMERICAL MODELING OF CURRENTS
to 30% of the estimated values.
• Perform optimized calibration. RUDI RAJAR
University of Ljubljana
• Compare measured and calibrated pressure readings.
Ljubljana, Slovenia
• If not satisfactory, or if any of the calibrated
group roughness values reach the bounds, relax the
bounds by a small percentage and rerun optimized INTRODUCTION
calibration.
• Repeat the trials until satisfactory results are Description of Currents
obtained.
Currents—water movement—appear in all surface
• If the bounds had to be relaxed beyond a reasonable waters: rivers, lakes, coastal seas, and oceans. This flow of
value for a group of pipes, it would imply a problem water is almost always turbulent, which means with many
with network data and not necessarily the roughness irregular eddies of different size. Currents are sometimes
values. Go back to macrocalibration probing the near to steady movement (not changing with time), but
network in and around the area of concern, and mostly the movement is unsteady, significantly changing
repeat macrocalibration, sensitivity analysis, and with time.
microcalibration steps. In rivers the currents are more aligned nearly in one
dimension (1D), although in larger rivers they can be quite
Although this procedure outlines the iterative method
complex and three-dimensional (3D). In shallower rivers
for obtaining optimal set of group roughness values, it can
or coastal seas the modeling of currents is sometimes
be easily modified when calibrating the model for other
two-dimensional (2D), but in reality all currents are
parameters such as nodal demands and their spatial and
three-dimensional (3D); that is, the water velocities are
temporal distribution (7)
in all three directions, although often they are an order of
magnitude greater in the two horizontal directions than
SUMMARY AND CONCLUSION
in the vertical direction.
Network model calibration should always be performed In surface waters the current circulation has different
before any network analysis planning and design study. orders of magnitude.
A seven-step methodology for network model calibration Global currents are large-scale currents, typical for
has been proposed. Historically, one of the most difficult every world ocean. They are often quasisteady in long-term
steps in the process has been the final adjustment of average, and with their enormous transport of heat from
pipe roughness values and nodal demands through the the equator to the poles, and vice versa, they significantly
process of micro-level calibration. With the advent of affect climate and weather on the continents. The main
recent computer technology, it is now possible to achieve driving forces are the temperature gradients between the
good model calibration with a reasonable level of success. poles and the equator and the wind. Due to the earth’s
As a result, little justification remains for failing to develop rotation, the main pattern of global currents is typically
good calibrated network models before conducting network from the poles toward the equator on the west side of
analysis. It is expected that future developments and continents (cold currents), and from the equator to the
applications of both GIS and SCADA technology will lead poles on the east side of continents (warm currents).
to even more efficient tools. Some of these global currents are the cold Peru (or
Humboldt) current along the west coast of South America
BIBLIOGRAPHY (toward the equator), the warm Kuroshio current from
the equator along the east coast of Asia, and the warm
1. Walski, T.M. (1995). Standards for model calibration. Proc. Gulf Stream along the east coast of North America. The
AWWA Computer Conf., Norfolk, VA, pp. 55–64. last is deflected by the topography of the coastline toward
2. Pipe2000. (2004). Users Guide, Pipe2000 Series Models northern Europe, causing a relatively mild climate along
GUI—KYPIPE Hydraulic Model. KYPipe LLC, Lexington, KY. the Scandinavian coastline.
3. Cesario, L. (1995). Modeling, Analysis and Design of Water Local currents (mesoscale and microscale currents) are
Distribution Systems. American Water Works Association, currents in smaller regions, as in coastal (or shelf) seas,
Denver, CO. which are defined as shallower than about 200 m, to
4. McEnroe, B., Chase, D., and Sharp, W. (1989). Field testing distinguish them from the oceans. Of course, local currents
water mains to determine carrying capacity. Technical Paper at a smaller scale appear also in lakes and rivers.
EL-89, Environmental Laboratory of the Army Corps of
Engineers Waterways Experiment Station, Vicksburg, MS.
Importance of Current Circulation
5. Walski, T.M. (1990). Sherlock Holmes meets Hardy Cross, or
model calibration in Austin, Texas. J. AWWA. 82(3): 34. Numerical modeling of currents usually includes model-
6. Cruickshank, J.R. and Long, S.J. (1992). Calibrating Com- ing of additional parameters that are important in water
puter Model of Distribution Systems. Proc. AWWA Computer ecology. When solving practical problems, the follow-
Conf., Nashville, TN. ing parameters must be determined: (1) water current
7. Lingireddy, S. and Ormsbee, L.E. (2002). Hydraulic network (circulation), (2) transport of sediments with currents,
calibration using genetic optimization. Civil Environ. Engi- (3) transport of any contaminants, and (4) biochemical
neer. Syst. 19(1): 13–39. processes of contaminants.
 NUMERICAL MODELING OF CURRENTS 321

Water current circulation is always the basis for calcu- parameters were then measured and used to model nature,
lation of the transport of any contaminant. But in many using physical laws of similitude. However, there are
cases, knowledge of the current velocities is used directly, several difficulties in applying this methodology: physical
for example, in designing river or marine coastal pro- models are expensive and time consuming to build and
tection, where the computation of current patterns helps operate; it is practically impossible to simulate very large
determine bank and riverbed erosion and the optimum regions; and it is almost impossible to reliably simulate
form of the protection structures. Transport of sediment is biochemical processes.
also important from two points of view. In river reservoirs, Field measurements are becoming very important.
lakes, or coastal bays, sedimentation is often a big problem. As numerical models almost always demand calibration
Over the years, river reservoirs upstream of dams are filled and verification of the simulated results, reliable data for
with sediments and thus lose their useful storage volume. comparison can only be obtained by field measurements.
Even more important is the fact that most contaminants This technology is developing very fast. Remote sensing
are bound to suspended sediments and often over 90% of (e.g., satellite images) can provide the pattern of
the contaminant transport is the result of sediment trans- sea surface temperature, or chlorophyll concentration
port with less than 10% in dissolved form. Determining over very large regions. As these patterns depend on
the transport by currents of any contaminants in surface transport by currents, this methodology can be used for
waters is very important. These can be nutrients, affecting verification and calibration of numerical models. In situ
the growth of algae and possibly causing eutrophication, measurements (directly by ships, buoys) are also necessary
or any toxic chemical. Coastal seas often receive large in combination with the remote sensing technique.
amounts of contaminants from rivers. Many coastal settle- In fact, reliable simulation of currents and contaminant
ments discharge their wastewater (sometimes untreated) transport can only be attained by the combined application
into coastal seas. Determination of currents and contam- of numerical models and field measurements.
inant transport—most often by numerical modeling—is
essential to determine consequences of pollution and to
plan the optimum location of wastewater outlets. A spe- NUMERICAL MODELS
cial problem is oil spill, where numerical modeling of the
spreading oil slick can help in clean up or can help to find What Is a Numerical Model?
the culprit of the pollution. Heat can also be considered as
A simplified definition of numerical model is the following:
a kind of contaminant. Coastal thermal or nuclear power
plants dispose of their waste heat into the sea. As the a computer code that can describe or ‘‘simulate’’ a certain
water temperature should not increase over a certain limit physical phenomenon, in this case current circulation
(usually 3 ◦ C over the ‘‘natural’’ temperature), these power and contaminant transport in surface water bodies.
plants should be located at sites where strong currents The simulation is never perfect, as the real physical
transport away and mix the waste heat. Because contami- phenomenon is too complex, and too many parameters
nants are not only transported and spread (dispersed) over and processes are involved. However, most often, the
the water body but are also prone to different transforma- simulations are accurate enough to essentially help at
tions due to biochemical processes, another quite difficult solving different scientific and engineering problems.
task of numerical models is to simulate these processes. Every numerical model basically consists of three parts:
The importance of global currents is crucial. As already (1) basic equations, describing the physical phenomena;
stated, global currents significantly affect climate and (2) all the necessary input data (e.g., description of
weather on the continents. The El Niño phenomenon is topography, data on physical properties of water and
well known to cause catastrophic changes of weather contaminants); (3) ‘‘forcing factors’’—the data about forces
over large parts of continents. Global currents are also that cause water and sediment movement and sources
important carriers of nutrients and contaminants of of contamination. The modelers call these ‘‘boundary
all types, and they influence the fish migration. Many conditions,’’ because these data determine the behavior
oceanographic studies are carried out to determine details of the model and essentially influence the results.
of global currents, their changes (e.g., prediction of the A numerical model can be compared to a physical model:
effect of future global warming), and their important in both cases the form (topography) of the water body
influence on the environment. must be given (for numerical model in digital form),
as well as the physical properties of the water and
Methods of Determination of Currents boundary conditions. The final results are the same in
both cases. The result of a numerical model can even be
To determine currents in surface waters, there are three presented as an animation, showing, for example, water
methods: numerical modeling, physical modeling, and circulation—similar to a movie taken on a physical model.
field measurements. Numerical modeling is the most
appropriate methodology. However, most often it should Dimensions of Numerical Models of Currents
be combined with field measurements and sometimes with
physical modeling. One-dimensional models simulate the flow in one dimen-
Physical models were widely used before the era of fast sion only—for example flow along a river. Topography
computers. Studied regions were built at a reduced scale is given with cross sections; the distance between them
in large hydraulic laboratories, and water was pumped depends on how big the river is and on the required accu-
to the models. Current velocities and other important racy. For unsteady flow, the result is time-dependent depth
322 NUMERICAL MODELING OF CURRENTS

and velocity in each cross section. Contaminant concen- Numerical Methods

tration can also be calculated in every cross-section and at
The water movement or current is described by a system
every time step.
of differential equations, basically a scalar continuity
Two-dimensional models are usually used in shallow
equation (mass conservation) and a vector momentum
coastal seas or lakes. The computational region (in
equation, which has as many scalar components as the
the horizontal plane) is divided into control volumes
dimension of the model. Because these equations are
of rectangular or other polygonal form (possibly also
nonlinear, the solution is very complex. Basically two
curvilinear) and bottom and bank topography is given
groups of numerical methods are used for their solution:
as input data. There is only one control volume from
finite difference methods and finite element methods. With
surface to the bottom. The result of simulation is depth
the first group of methods, the computational domain is
and two components of horizontal velocity and possibly
divided into control volumes of polygonal forms (possibly
contaminant concentration in every control volume and at
with curvilinear sides), while with finite element methods,
every time step.
the domain is divided into different unequal geometric
Three-dimensional models are similar to 2D models:
bodies. The finite difference methods are a bit simpler
the difference is that the computational region is
at transforming the differential equations into finite
additionally divided into layers along the depth. In this
difference forms, while the finite element method can
way, the whole computational domain is divided into
better adjust the control volumes to irregularly shaped
‘‘bricks’’ in all three dimensions. Besides the parameters
boundaries. In the simulation of currents, the finite
mentioned with 2D models, the vertical velocity component
difference models are more often used.
is also computed.
The number of control volumes into which the
Three-dimensional models are the most sophisticated
computational domain is divided depends on the accuracy
and basically the most accurate. However, the real
of the input data and on the required accuracy of the
accuracy and reliability of the models depend on the
results. Typically, the domain is divided into 100–150 cells
accuracy of the input data and on the calibration,
in each horizontal direction, and into 15–30 layers along
verification, and validation of the models.
the depth. Computing time is long and fast computers are
needed for the computation of currents.
Forcing Factors
The water movement or, current is caused and influenced Verification, Calibration, and Validation of Models
by four ‘‘forcing factors.’’ For each simulated case, the applied model has to be
Wind has an important effect on the water surface. checked for reliability.
With its shear stress it causes the surface layer of water Verification of a model is the first test, which
to move basically in the direction of the blowing wind. shows whether the phenomenon is correctly simulated.
Due to the law of continuity, this movement also causes Then the model should be calibrated. Most often there
movement of other water layers. If the phenomenon is are a few parameters in the model simulations for
taking place in a closed basin, layers near the bottom must which the proper values are not accurately known, so
move in the opposite direction to the wind. In general, the a series of simulations are carried out with variable
movement is complex and depends on the bottom and values of these parameters. The simulation results are
bank topography. In larger seas and oceans, the earth’s compared with results of measurements (or possibly
rotation essentially influences the current pattern (1). results of another model). In this way the proper values
A good numerical model must take into account all of the parameters are determined. (Example parameters
these influences. are diffusion coefficients for turbulence or dispersion
Tide influences the current pattern. When simulating of contaminants.) When the simulated results are in
currents inside a semienclosed bay or coastal sea, the acceptable agreement with measurements, the model is
boundary condition for the model is represented by the ‘‘verified’’ and ‘‘calibrated.’’ It is then used to simulate
time variation of tidal level at the open boundaries. other similar cases, for which the model parameters
Inflow of rivers also affects the currents, with its inflow were not calibrated. If the results are in agreement with
of mass and momentum. measurements, the model is ‘‘validated.’’
Density gradients—variations of density with space
and/or time—also cause water circulation. The density
changes are most often caused by gradients of temperature CASE STUDIES
and salinity; therefore, this type of forcing is also called
A: Transport of Radioactive Contaminants in Japan Sea
‘‘thermohaline forcing.’’
The relative importance of the four forcing factors can The Japan Sea is a large, almost enclosed sea between
vary with different cases. Wind and river inflow are usually the Japanese islands, Sakhalin Island, the Siberian coast,
the most important forcing factors in lakes and smaller and Korea. Its length is about 1800 km, its width 900 km
coastal seas, while tide is the most important factor in (see Fig. 1), and the maximum depth is about 3900 m. The
small and middle-sized coastal areas. Density gradients relatively weak exchange of water with the Pacific Ocean
and winds are very important for ocean circulation. A good is achieved through four straits (Fig. 1), the widest being
numerical model should include all four forcing factors, Tsushima Strait between Korea and the Japanese islands.
although in many cases some less important ones can be A large amount of radioactive (RA) waste has been
neglected entirely. deposited in the Japan Sea, off Vladivostok, at a depth
 NUMERICAL MODELING OF CURRENTS 323

132° 136° 139°

Japan sea
50

48°
ri a

nt Soya Scale
rre
be

strait
cu
Si

15 cm/s
an
Lim

Latitude north
44° 300
0 45 45 km
Vladivostok
M1
NO.9. Tsugaru
N KCC

M3M2 strait
40°
br. 0
ore 100 1000
40
fsh
Of
50
00

0
100
0
200
1000
36°
NSB
n
pa
EKWC

0
100
Ja

35 Figure 1. Measured surface currents in

Tsushima S.
the Japan Sea (left) and surface currents
130 135 140 simulated by the PCFLOW3D model
Longitude east (right). (From Ref. 3.).

of about 3500 m. Concern existed in Japan that leaking fishing is possible) after 3 years; the amount would
containers with RA waste could cause contamination increase in time and would attain a maximum after about
of fish and other marine organisms in the Japan Sea. 30 years. But the maximum concentrations would be on
A research model should answer the questions: (1) In the order 1E-2 Bq/m3 , which is far below the natural
what concentrations and after what time would the RA background values and hence would not represent any
contaminants reach the surface layers? (2) Where might danger for the environment.
they be transported by currents and dispersion? In the
simulations, the possible leaking of the radionuclides was B: Simulation of Mercury Cycling in the Mediterranean Sea
supposed to be at location No. 9 (Fig.1). Mercury is a highly toxic metal, especially in its
The first task was to determine the current cir- organic form, monomethylmercury. An international
culation (4). A 3D numerical model PCFLOW3D was project carried out in 2002–2004 had the goal to determine
used (2,3). The 3D circulation was driven by three forcing the natural and anthropogenic sources of mercury in the
factors: thermohaline forcing, wind, and the inflow/outflow Mediterranean region and to determine the distribution
surface currents. The computations were carried out of the most important mercury forms by the use of a
taking into account the measured seasonally averaged three-dimensional numerical model (5).
temperature and salinity distribution and seasonal winds. Again, the current circulation had to be determined
A uniform wind field was accounted for (9.5 m/s in win- first. It was calculated for four annual seasons. Forcing
ter, 0 m/s in summer). The inflow current through the factors were the following: (1) thermohaline forcing, that
Tsushima Strait (a branch of the global Kuroshio current) is, density gradients due to temperature and salinity
was taken into account. distribution (obtained by measurements) in the whole
After some calibration, the results of the model were computational domain and also along the depth; (2) wind,
compared with the observed pattern of the surface currents and (3) inflow from rivers and exchange of water through
(Fig.1). It can be seen that agreement is good; the the Gibraltar and Bosporus. Verification and calibration
simulated spirals have the same direction and strength as of the model were carried out on the basis of numerous
the measured ones. Some measurements also confirmed measurements. Figure 2 shows a comparison of simulated
proper simulation of bottom currents. and measured currents for the summer season.
On the basis of the computed winter and summer Transport and dispersion of mercury were then simu-
hydrodynamic circulation, transport and dispersion of lated, sources arriving from rivers, through straits from
radioactive contaminants were further simulated. The adjacent seas, and from some known point sources of
calculated amount of radioactive contaminants released in contamination. Verification and calibration of the trans-
90 d was 1 TBq (1012 becquerels), from dumping location port–dispersion part of the model were carried out on
No. 9 (Fig. 1). the basis of numerous measurements of relevant param-
The final results of the simulations have shown that eters. Three cruises by a large research ship Urania over
the RA contaminants would reach the surface mainly in the whole Mediterranean Sea provided measurements.
the northern part of the Japan Sea. The first contaminants Figure 3 shows the distribution of mercury, simulated
would reach the surface layers (above 180 m depth, where over one year of inflow from the sources of contamination.
 44

42

40

38

36

34

32

−4 −2 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36

Scale:

10 cm/s

100 km

Figure 2. (Top) Measured average

currents in the Mediterranean in the
summer season. (Bottom). Simulated
currents for the same season.

Total Hg, t = 1 yr, kpart = 0.1 l /mg

45 +3 Surface layer: 0−10 m

+5
+1
40

35 +2

30

25

20
+10
15 +6

10 +8
+9
+4
+7
5

Figure 3. Concentration of total mercury in the 10 20 30 40 50 60 70 80

surface layer of the Mediterranean Sea, after ng/L
1 year of simulation by PCFLOW3D model.
Sources of contamination are from rivers and
point sources, marked by numbers. land .0001 .001 .01 .1 1 10 100

324
 UNCERTAINTY ANALYSIS IN WATERSHED MODELING 325

One of the goals of the project was also to couple modeling of a watershed as described above. The main
the water model with the atmospheric model (which uncertainties are listed and described here. The first
also simulates circulation and mercury transport) into a uncertainty present when modeling watershed systems
general integrated model. As there is permanent exchange originates from the modelers’ limited understanding of the
of mercury between the water surface and the atmosphere, real world (i.e., uncertainty in the perceptual model used).
the integrated model gives a more realistic result about
global mercury cycling and contamination in the region. Perceptual Model (PM) Uncertainty. A PM (1) is a
The general model was finally used as a tool to predict conceptual representation of the watershed that is
possible measures to diminish concentration of toxic subsequently translated into mathematical (numerical)
mercury in the Mediterranean region. form in the model. The PM is therefore based on our
understanding of the real-world watershed system. This
understanding might be low, particularly with respect to
BIBLIOGRAPHY
subsurface system characteristics and water flowpaths,
1. Duxbury, A.C. and Duxbury A.B. (1989). An Introduction to
and therefore our PM might be uncertain.
the World’s Oceans. Wm. C. Brown Publishers, Dubuque, IA.
The next group of uncertainties is related to the
2. Rajar, R. and C̆etina, M. (1997). Hydrodynamic and water
quality modelling: an experience. Ecol. Model. 101: 195–207. mathematical/numerical model and its components. These
components and the related uncertainties are:
3. C̆etina, M., Rajar, R., and Povinec, P. (2000). Modelling of
circulation and dispersion of radioactive pollutants in the
Japan Sea. Ocean. Acta 23(7): 819–836. Model Structure (Equation) Uncertainty. The ambigu-
4. Yoon, J.H. (1995). Numerical modelling of the ocean miniature
ity present in the PM usually means that a variety of
‘‘Japan Sea.’’ In: Proceedings of the International Workshop on model structures seem to be feasible representations of
Numerical Prediction of Ocean Variations, Tokyo, March. the watershed at hand prior to any modeling being under-
5. Rajar, R., C̆etina, M., Z̆agar, D., and Petkovs̆ek, G. (2004). taken. Unfortunately, analysis of the performance of the
Three-dimensional simulation of mercury cycling in the selected model structures in reproducing observed water-
Mediterranean Sea. In: International Conference on Mercury shed behavior is often insufficient to distinguish between
as a Global Pollutant, Ljubljana, June, Part 2, pp. 1324–1327. them and some ambiguity about an appropriate structural
system representation remains.
Model Parameter Uncertainty. Models of watershed
UNCERTAINTY ANALYSIS IN WATERSHED hydrology necessarily aggregate the natural heterogeneity
MODELING of watershed characteristics and processes into individual
model components. These components are characterized
THORSTEN WAGENER by model parameters that describe, for example, storage
The Pennsylvania State size, water residence time, or flowpath. The scale at which
University these parameters can be measured is commonly different
University Park, Pennsylvania from the scale at which they are applied within the model.
At least some of the model parameters therefore have to
be estimated through a process of parameter adjustment
INTRODUCTION (manually or automatically), where the parameters are
varied until observed and modeled system behavior agree
All watershed modeling studies unavoidably contain as much as possible. Most watershed models contain a
uncertain elements. This uncertainty is propagated into large number of nonlinearly interacting parameters, which
the model predictions and impacts subsequent decision- makes this a very difficult task, and some uncertainty
making. If not taken into account properly, this can commonly remains with respect to the question of whether
lead to model predictions that are assumed to be more a unique best parameter set has been found.
reliable than they actually are. The usefulness of these Model State Uncertainty. With respect to state uncer-
predictions in any decision-making context therefore tainty one has to distinguish between initial state uncer-
becomes questionable. tainty and state uncertainty after a certain time series
A watershed model is the numerical representation of data has been processed by the model. The first uncer-
of our perception of the hydrologic system at hand. It tainty can usually be taken care of by including a warm-up
consists of a model structure (i.e., underlying equations), period that is excluded from the performance analysis.
model parameters (i.e., system invariants), and model Alternatively, the initial state can be optimized, similarly
states. The model is driven by measurements of variables to the parameter adjustment process described above.
entering the watershed (e.g., precipitation and energy) State uncertainty during model simulation stems from
and compared to observations of the system response (e.g., biases in the model predictions. It can be reduced using
streamflow or groundwater heads). data assimilation techniques, that is, approaches that
recursively assimilate measurements of the model output
SOURCES OF UNCERTAINTY variable from previous time steps to adjust the model
states to reduce predictive uncertainty.
Sources of uncertainty in this context are manifold and Variable Uncertainty. This is another, less often
basically include every element that is needed for the considered uncertainty, originating from the problem
326 UNCERTAINTY ANALYSIS IN WATERSHED MODELING

of simulated and observed variables not being exactly parameter sets within a certain model structure
the same quantity (2). For example, predictions and that provide behavioral predictions. This search is
measurements of soil moisture might be at different scales commonly performed using various types of random
and therefore a perfect match should not be expected or is sampling (e.g., uniform random sampling or types
not even something the modeler should strive for (3). of stratified sampling) or using population evolution
Input (Forcing) and System Response Data Uncertainty. based techniques. The use of Monte Carlo Markov
There are also uncertainties that are related to the chains can be very efficient in this context if the
measurements of forcing and response variables. These assumptions that have to be made about the shape
stem, for example, from an inability to capture the spatial of the response surface are justified.
distribution of rainfall or from problems in capturing the The second part of the search is related to the
true streamflow discharge. These uncertainties can only problem that potentially different model structures
be reduced outside the modeling process, for example, can provide behavioral predictions for a certain case.
through better measurement techniques. This is an even more difficult task since the model
space is not just infinite, but also unbounded. No
Generally, no satisfactory means of considering all systematic approach exists to explore this search
these uncertainties simultaneously in a watershed model- space. Approaches suggested include the use of
ing exercise have yet been developed (3). Additionally, the ensembles of model structure (e.g., in a Bayesian
modeler is faced with the problem that, in many cases, framework) (5) or the inclusion of a more complex
the distance between the (erroneous) observation and the error model that takes the imperfection of the
uncertain prediction of streamflow is the only measure of selected model structure into account (6).
deviation (error) at each time step, providing very little 3. Propagate the Predictions of All Identified Models
information for model evaluation. into the Output Space While Considering Other
Uncertainties (e.g., in the Forcing Data). Three
basic approaches are applied for the propagation
APPROACHES TO UNCERTAINTY ANALYSIS stage, first-order, statistical, and set-theoretic. The
first approach only propagates the mean and
A wide variety of techniques for uncertainty analysis in the variance of the predicted variable, based on
watershed hydrology are available and often applied (3,4). means and variances of inputs and parameters. An
These methods differ, for example, in underlying philoso- overview of the most popular first-order approaches
phy and assumptions and in sampling strategies applied. is provided by Melching (7). Statistical approaches
Little guidance is available to help modelers decide what sample from the posterior parameter distributions
approach might be the most suitable one for their situ- derived in step 2 and usually make assumptions
ation (1). While the differences are large, there is some about the distributions of input uncertainties to
overlap between all the approaches. In principle, all derive ensembles of predictions. Ensembles of inputs
techniques for uncertainty analysis can be reduced to and/or parameter sets, not necessarily sampled from
three basic steps. The main potential approaches that are strict statistical distributions, are also propagated
applied for uncertainty analysis in watershed hydrology in the set-theoretic approach.
are described based on this three-step approach.
CONCLUSION
1. Define What Constitutes a Behavioral Model. The
definition of what constitutes a behavioral model Uncertainty is an unavoidable element in any watershed
can roughly be divided into two types of definitions, modeling exercise. It has to be considered if the model
those that follow a strict statistical approach, and simulations should be used in a decision-making context.
those using set-theoretic measures not consistent Main uncertainties include those stemming from a lack
with traditional statistics. The former are based on of understanding of the hydrologic system, the problem of
assumptions about the structure of the model resid- identifying appropriate model structure and parameters,
uals, that is, the differences between observed and and uncertainties in the measured data. While many
simulated system response. Typical assumptions are approaches exist that can be used for uncertainty analysis,
that the error is Gaussian and has a constant vari- little guidance is available to choose the best approach for
ance in time (3). The second type is less bound by a a certain study. A three-step structure is common to all
formal statistical framework, but this is commonly approaches and potential implementations are discussed
defended with the statement that the structure of above. Future developments in this area will focus
the uncertainty is unknown in most modeling stud- on a more realistic representation of the uncertainties
ies and that restricting assumptions are therefore present, techniques that allow for the explicit and
not justifiable. Examples of the second type include simultaneous consideration of all uncertainties, and better
the generalized likelihood uncertainty estimation sampling strategies.
(GLUE) (2) framework and fuzzy measures.
2. Find All Models that Comply With This Definition BIBLIOGRAPHY
in the Feasible Parameter Space. The search for
all models that are consistent with the definition 1. Beven, K. (2004). Does an interagency meeting in Washington
consists of two parts. First is the search for all imply uncertainty? Hydrol. Proc. 18: 1747–1750.
 WATERSHED MODELING 327

2. Beven, K.J. (2000). Rainfall–Runoff Modeling: The Primer. land-use variability and change (8,9), calculation of design
John Wiley, Chichester, UK. floods (10,11), as load models linked to water quality
3. Gupta, H.V., Beven, K.J., and Wagener, T. (2004). Model investigations (12), real-time flood forecasting (13–15),
calibration and uncertainty estimation. In: Encyclopedia and the provision of boundary conditions for atmospheric
of Hydrological Sciences. M. Anderson (Ed.). John Wiley, circulation models (16,17). This multitude of, often very
Chichester, UK. specific, tasks and circumstances has contributed to the
4. Duan, Q., Sorooshian, S., Gupta, H.V., Rousseau, A., and Tur- development of a vast number of watershed models
cotte, R. (Eds.). (2003). Advances in Calibration of Watershed starting in the early 1960s (see Todini (18), for a historical
Models. AGU Monograph, pp. 29–47.
review). These models are usually a mixture of linear and
5. Newman, S.P. (2002). Accounting for conceptual model uncer- nonlinear functions, combined to represent those processes
tainty via maximum likelihood Bayesian model averaging.
occurring in the specific watershed and important for the
Acta Univ. Carolinae Geol. 46(2/3): 529–534.
study objectives at hand. Watershed models are also used
6. Kennedy, M.C. and O’Hagan, A. (2001). Bayesian calibration
in water quality studies, but the focus here is simulation
of computer models. J. R. Statist. Soc. 63(3): 425–464.
of the quantitative response and general features of
7. Melching, C.S. (1995). Reliability estimation. In: Computer these models.
Models of Watershed Hydrology. V.P. Singh (Ed.). Water
Resources Publishers, Washington, DC, pp. 69–118.
WATERSHED PROCESSES

WATERSHED MODELING The hydrologic cycle is a closed system in the sense that
no water is lost. Only part of the processes that constitute
THORSTEN WAGENER the hydrologic cycle are relevant for watershed modeling.
The Pennsylvania State A watershed is therefore an open system with respect
University to inputs (mainly precipitation and energy) and outputs
University Park, Pennsylvania (mainly streamflow, but also surface and subsurface
losses). A schematic description of the main processes
INTRODUCTION in a typical watershed system is shown in Fig. 1. Models
of watershed hydrology represent these processes using
Models of watershed hydrology are important tools in different mathematical approaches and different levels
hydrological investigations for operational and research of spatial detail. It is important to consider that the
purposes. Areas of application are the extension of part of the watershed system reproduced in the model
streamflow time-series in space (1–3) and time (4,5), usually depends on the prevailing processes and also on
evaluation of a watershed response to climate (6,9) and/or the modeling purpose and available data (3). For example,

Evapotranspiration Precipitation
(rain, snow)

Snow accumulation Channel

(snow water storage) precipitation
SURFACE

snowmelt
VERTICAL FLUXES

Interception and surface moistening

Throughfall and stemflow

Overland
flow Channel
Soil surface storage
storage and
(depression storage...)
routing
Infiltration
UNSATURATED

Soil moisture recharge

(capillary water)

Gravity water storage and flow Interflow Basin

including macropore flow discharge
Capillary rise Percolation
LATERAL FLUXES

Groundwater in upper horizons Baseflow

LEGEND
SATURATED

(shallow, perched)
Systems
Deep percolation (storages)
Delayed
Groundwater in deeper horizons baseflow
(large scale aquifers...) Fluxes
(flows) Figure 1. Schematic of watershed
processes (after Reference 19).
328 WATERSHED MODELING

routing processes are usually not required if the modeling that large parts of the (heterogeneous) watershed are
purpose is simulation of the watershed outflow at a integrated in a single (homogeneous) element. The con-
monthly time-step. sequence of this process is that the model parameters
lose some of their physical meaning and cannot be mea-
sured on the required scale. Rather, the modeler has
WATERSHED MODEL CLASSIFICATION
to use observations of the watershed response to find
appropriate values for the model parameters (see next
It is common to classify watershed models into three
section). Some models are based on a semidistributed
distinct types (20; Fig. 2): (1) empirical (also called data-
approach, in which different models represent individ-
based, metric, or black-box), (2) conceptual (also called
ual subwatersheds. This approach is finding increased
parametric, explicit soil moisture accounting, or gray-box)
attention in the literature (28). Conceptual models form
and (3) physically based (also called mechanistic or white-
the large majority of models used in operational hydrol-
box) model structures. There are, however, models that
ogy. Their dependence on measurements of the watershed
have characteristics of more than one group.
response (mainly streamflow) limits their use in ungauged
Empirical models use available time series of input and
watersheds. However, research is ongoing to resolve this
output variables (precipitation, streamflow, temperature,
problem (3,29).
etc.) to derive both the model structure and the
Physical models (30,31) are based on conservation of
corresponding parameter values. They are therefore
mass, momentum, and energy. The spatial discretization
purely based on the information retrieved from the
applied is usually based on grids, but sometimes also
data and generally do not include prior knowledge
on some type of hydrologic response unit or triangular
about catchment behavior and flow processes, hence, the
irregular networks (32). Physical models are therefore
name black box. Popular examples of empirical models
particularly attractive for studies that require a high
are artificial neural networks (ANN; 21,22) and transfer
level of spatial detail, such as studies on soil erosion or
functions (TF; 23). Empirical models are usually spatially
diffuse-source pollution (33). The initial idea underlying
lumped; they treat the catchment as a single unit.
these models was that the degree of physical realism on
Variants on purely empirical TF models are data-based
which these models are based would be sufficient to relate
mechanistic models. They constrain the degrees of freedom
their parameters, such as soil moisture characteristics
to those structures that are physically interpretable,
and unsaturated zone hydraulic conductivity functions for
thereby using the hydrologist’s understanding of the
subsurface flow or friction coefficients for surface flow, to
natural system (24–26).
physical characteristics of the catchment (18), thus elimi-
Conceptual models (3,27) are built from storage ele-
nating the need for observed system response to condition
ments that represent parts of the watershed where water
the parameters of the model. However, the currently avail-
is temporarily stored, such as soil, aquifers, or streams.
able physical models do not fulfill this ideal. They suffer
These elements are filled through fluxes such as rainfall,
from extreme data demand, scale-related problems (e.g.,
infiltration, or percolation, and empty through evapotran-
the measurement scales are different from the process and
spiration, runoff, drainage, etc. Based on a conceptualiza-
model (parameter) scales), and overparameterization (34).
tion of the real-world watershed, the modeler, specifies the
A number of key parameters—applied to a large number of
structure of these models a priori, for example, the num-
elements—still has to be estimated from measurements
ber of and connections between storage elements. These
to capture the uniqueness of a specific watershed (35).
models are also usually spatially lumped. This means
The expectation that these models could be applied to
ungauged catchments has therefore not been fulfilled (36).
Beven (34) argues that this type of model is applied in
(a) (b)
a way similar to lumped conceptual models, though on a
different scale.
A general characteristic of all current models is that
they do not provide reliable predictions in ungauged
basins (3). The general behavior of a watershed is
reasonably constrained by the model structure (i.e.,
the constituent equations), but some adjustment of
model parameters is required to capture the unique
(c) characteristics of the system under investigation. A large
number of popular models of watershed hydrology can be
found in Singh (37) and Singh and Frevert (38,39).

WATERSHED MODELING PROCEDURES

Many authors have suggested procedures for using water-

shed models. Examples are given in Beven (32) and
Figure 2. Classification of watershed models according to in Anderson and Burt (40). They vary slightly in the
spatial discretization used: (a) lumped, (b) semidistributed, and steps included and definitions used. A procedure using
(c) distributed. commonly found elements is shown in Fig. 3 (after
 WATERSHED MODELING 329

knowledge-based approaches to support the decision-

Model structure making process of the hydrologist for selecting an
selection
appropriate model structure.
Sensitivity analysis tests the sensitivity of the model
Sensitivity output to changes in the model parameters (or even
analysis to uncertainty in the model input or in other model
elements). This can be done before parameter estimation
to find parameters that are particularly important for
Parameter
estimation
model performance—the estimation process can then be
focused on these, or it can be done after the calibration as
part of the uncertainty analysis—to identify parameters
that are particularly uncertain. The simplest approach is
Generation
a perturbation technique where one parameter is varied
over its feasible range while the remaining parameters are
kept at reasonable values, although more sophisticated
Prediction
(incl. uncertainty) approaches exist (32).
As stated before, the current generation of watershed
Figure 3. Typical elements of a watershed modeling procedure. models generally requires some sort of parameter estima-
tion to enable the model to match observed and calculated
system behavior, for example, to give reliable predic-
Reference 3). The elements considered are model struc-
tions. This process is sometimes referred to as calibration.
ture selection, sensitivity analysis, parameter estimation,
Within this step, the performance of selected parameter
generalization, and prediction (including uncertainty esti-
sets in reproducing the observed system behavior is eval-
mation).
uated (Fig. 4). either by visual inspection or by calculating
The first step for the modeler is to select (or
some function F that summarizes how well simulated and
develop) a model structure, a combination of mathematical
observed system behavior match:
equations, that describes the watershed system as a
function of the system characteristics (e.g., saturation F = f (e) (1)
or infiltration excess dominated runoff production), the
modeling objective (e.g., do the effects of routing have to where f is some aggregation function applied to e, the
be considered if only the monthly response is wanted?), vector of residuals calculated as
and the available data (e.g., a simple evapotranspiration et = zt − yt (2)
function has to be used if measurements of variables
such as humidity or wind speed are not available). where z is the observed system response, for example,
Personal preferences and experience in using a particular a time series of streamflow; y is the calculated system
model structure might also play an important role in response; and t is the time step. The modeler attempts
the selection process. Woolhiser and Brakensiek (41, to make the function F as small (or large depending on
p. 15) conclude that ‘‘objective methods of choosing the the chosen formulation) as possible, either by manually
best model (structure) have not yet been developed, adjusting the parameters or through some automatic
so this choice remains part of the art of hydrological procedure. See Duan et al. (42) for a state-of-the-art review
modelling.’’ However, future modeling tools might include of current techniques.

True True Observed

input Real world response response
Amplitude

Simulated
Observed Model (q) Time
response
input

Prior
q
info
Adjustment

Measurement

Figure 4. Calibration of a watershed model. The symbol θ represents a parameter vector, y is

the system output (usually streamflow), and t denotes time.
330 WATERSHED MODELING

The problem with this step of conditioning the 11. Lamb, R. (2000). An approach to the calibration of a concep-
parameters on observations is that it can result in tual rainfall-runoff model for flood frequency estimation by
parameter values that depend on the specific time series continuous simulation. Water Resour. Res. 35: 3103–3114.
used (e.g., a particular climate period). Some modelers 12. Mroczkowski, M., Raper, G.P., and Kuczera, G. (1997). The
include a generalization stage to avoid this. During this quest for more powerful validation of conceptual catchment
stage, the model (including the parameter set found most models. Water Resour. Res. 33(10): 2325–2335.
representative of the system at hand) is applied to another 13. Kitanidis, P.K. and Bras, R.L. (1980). Real-time forecasting
time series of measured system response, not used during with a conceptual hydrologic model 1. Analysis of uncertainty.
parameter estimation, but usually of the same variable. A Water Resour. Res. 16(6): 1025–1033.
model is assumed generally applicable if its performance 14. Kitanidis, P.K. and Bras, R.L. (1980). Real-time forecasting
does not degrade significantly during this process (43). with a conceptual hydrologic model 2. Applications and
The model can then be used for prediction. However, results. Water Resour. Res. 16(6): 1034–1044.
there is a wide variety of factors that introduce uncertainty 15. Moore, R.J. (1999). Real-time flood forecasting systems:
into the modeling process. These are for example, the Perspectives and prospects. In: R. Casale, and C. Margottini
inability of the modeler to find a ‘‘best’’ parameter (Eds.), Floods and Landslides: Integrated Risk Assessment,
set, uncertainties in measuring input and output data, Springer-Verlag, Berlin, pp. 147–189.
and problems in finding an appropriate model structure 16. Dümenil, L. and Todini, E. (1992). A rainfall-runoff scheme
to represent the system under investigation. These for use in the Hamburg climate model. In: J.P. O’Kane (Ed.),
uncertainties have to be considered and incorporated in Advances in Theoretical Hydrology, Elsevier, Amsterdam,
the model predictions. It is therefore sensible to state these pp. 129–157.
uncertainties explicitly and include confidence limits that 17. Wood, E.F., Lettenmaier, D.P. and Zatarin, V.G. (1992).
denote the ranges in which the predictions of the model A land-surface hydrology parameterisation with sub-grid
are most likely to fall (44,45). variability for general circulation models. J. Geophys. Res.
97(D3): 2717–2728.
BIBLIOGRAPHY 18. Todini, E. (1988). Rainfall-runoff modelling—past, present
and future. J. Hydrol. 100: 341–352.
1. Manley, R.E. (1978). Simulation of flows in ungauged basins.
19. Becker, A. (1992). Criteria for a hydrologically sound
Hydrological Sci. Bull. 23(1): 85–101.
structuring of large scale land surface process models.
2. Sefton, C.E.M. and Howarth, S.M. (1998). Relationships In: J.P. O’Kane, (Ed.) Advances in Theoretical Hydrology,
between dynamic response characteristics and physical
Elsevier, Amsterdam, pp. 97–111.
descriptors of catchments in England and Wales. J. Hydrol.
211: 1–16. 20. Wheater, H.S., Jakeman, A.J., and Beven, K.J. (1993).
Progress and directions in rainfall-runoff modelling. In:
3. Wagener, T., Wheater, H.S., and Gupta, H.V. (in press).
A.J. Jakeman, M.B. Beck, and M.J. McAleer (Eds.), Mod-
Rainfall-Runoff Modelling in Gauged and Ungauged Catch-
elling Change in Environmental Systems, John Wiley, Chich-
ments. Imperial College Press/World Scientific, London/River
Edge, NJ, UK/USA. ester pp. 101–132.
4. Lees, M.J. and Wagener, T. (2000). Review of HYCOM 21. Hsu, K.-L., Gupta, H.V., and Sorooshian, S. (1995). Artificial
rainfall-runoff modeling undertaken for the review of neural network modelling of the rainfall-runoff process. Water
Southern Water’s source yields (1997). Unpublished Report to Resour. Res. 31(10): 2517–2530.
Southern Water, Imperial College Consultants, London, UK. 22. Dawson, C. and Wilby, R. (1998). An artificial neural network
5. Lees, M.J. and Wagener, T. (2000). HYCOM Rainfall-Runoff approach to rainfall-runoff modelling. Hydrological Sci. J. 43:
Modelling of Bewl-Darwell Reservoir Source Rivers. Unpub- 47–66.
lished Report to Southern Water, Imperial College Consul- 23. Young, P.C. (1984). Recursive Estimation and Time-Series
tants, London, UK. Analysis: An Introduction. Springer-Verlag, Berlin, Germany.
6. Jakeman, A.J., Chen, T.H., Post, D.A., Hornberger, G.M., 24. Young, P.C. (1992). Parallel processes in hydrology and water
Littlewood, I.G., and Whitehead, P.G. (1993). Assessing quality: A unified time-series approach. J. IWEM 6: 598–612.
uncertainties in hydrological response to climate at large
25. Jakeman, A.J. and Hornberger, G.M. (1993). How much
scale. Macroscale modelling of the hydrosphere. Proc.
complexity is warranted in a rainfall-runoff model? Water
Yokohama Symp., IAHS Publ. No. 214, pp. 37–47.
Resour. Res. 29(8): 2637–2649.
7. Cameron, D., Beven, K.J., and Naden, P. (2001). Flood fre-
26. Young, P.C. and Beven, K.J. (1994). Data-based mechanistic
quency estimation under climate change (with uncertainty).
modelling and the rainfall-flow nonlinearity. Environmetrics
Hydrol. Earth Syst. Sci. 4(3): 393–405.
5: 335–363.
8. Binley, A.M., Beven, K.J., Calver, A., and Watts, L.G. (1991).
Changing responses in hydrology: Assessing the uncertainty 27. O’Connell, P.E. (1991). A historical perspective. In:
in physically based model predictions. Water Resour. Res. D.S. Bowles, and P.E. O’Connell (Eds.) Recent Advances in
27(6): 1253–1261. the Modelling of Hydrologic Systems, Kluwer Academic,
Dordrecht, pp. 3–30.
9. Parkin, G. , O’Donnell, D., Ewen, S., Bathurst, S.C.,
O’Connell, P.E., and Lavabre, S. (1996). Validation of 28. Boyle, D.P., Gupta, H.V., Sorooshian, S., Koren, V., Zhang, Z.,
Catchment Models for Predicting Land-use and Climate and Smith, M. (2001). To wards improved streamflow fore-
Change Impacts 2. Case Study for a Mediterranean casts: The value of semi-distributed modelling. Water Resour.
Catchment. J. Hydrol. 175, pp. 595–613. Res. 37(11): 2739–2759.
10. Calver, A., Lamb, R., and Morris, S.E. (1999). River flood 29. Abdullah, F.A. and Lettenmaier, D.P. (1997). Development
frequency estimation using continuous runoff modelling. Proc. of regional parameter estimation equations for a macroscale
Inst. Civ. Eng. Water Maritime Energy 136: 225–234. hydrologic model, J. Hydrol. 197: 230–257.
 MODELING OF WATER QUALITY IN SEWERS 331

30. Freeze, R.A. and Harlan, R.L. (1969). Blueprint for a from sewer overflows. In addition, sewage under anaerobic
physically-based digitally simulated hydrologic response conditions may cause corrosion of pipes, malodors, and
model. J. Hydrol. 9: 237–258. pose health risks for sewage workers. Hence, it is of utmost
31. Abbott, M.B., Bathurst, J.C., Cunge, J.A., O’Connell, P.E., importance to analyze and understand the transport
and Rasmussen, J. (1986). An introduction to the European and transformation of wastewater in sewers. Ideally,
Hydrological System—Système Hydrologique Europeén,
modeling of sewer systems and wastewater treatment
SHE. 1. History and philosophy of a physically-based, dis-
plants should be based on the same general principles
tributed modelling system. J. Hydrol. 87: 45–59.
and done in integrated models to pursue an integrated
32. Beven, K.J. (2000). Rainfall-Runoff Modeling: The Primer.
understanding and management of sewers, wastewater
John Wiley, Chichester, U.K.
treatment plants, and receiving waters. However, it must
33. Refsgaard, J.C. and Abbott, M.B. (1996). The role of dis-
be stressed that the transformation of wastewater in
tributed hydrological modelling in water resource man-
agement. In: M.B. Abbott, and J.C. Refsgaard (Eds.), Dis-
sewers is not identical to the processes in activated sludge
tributed Hydrological Modelling, Kluwer Academic, Dor- and biofilm systems—so even though several similarities
drecht, pp. 1–16. exist between sewers and wastewater treatment plants,
34. Beven, K.J. (1989). Changing ideas in hydrology—The case modeling of water quality in sewers is a discipline of it
of physically-based models. J. Hydrol. 105: 157–172. own. This article focuses on how to model water quality
35. Calver, A. (1988). Calibration, sensitivity and validation of in sewers.
a physically based rainfall-runoff model. J. Hydrol. 103:
102–115. THE SEWER AS A PHYSICAL, BIOLOGICAL AND
36. Refsgaard, J.C. and Knudsen, J. (1996). Operational valida- CHEMICAL REACTOR1
tion and intercomparison of different types of hydrological
models. Water Resour. Res. 32(7): 2189–2202. To model water quality in sewers, it is necessary to
37. Singh, V.P. (Ed.). (1995). Computer Models of Watershed know the various types of sewers and their water
Hydrology. Water Resources Publishers, Highlands Ranch. quality characteristics. For a detailed description of the
38. Singh, V.P. and Frevert, D. (Eds.). (2002). Mathematical processes in sewer, refer to specialized literature (1).
Models of Large Watershed Hydrology, Vol. 1. Water Around the world, all sewers can basically be divided
Resources Publishers, Highlands Ranch. into three categories:
39. Singh, V.P. and Frevert, D. (Eds.). (2002). Mathematical
Models of Small Watershed Hydrology, Vol. 2. Water 1. storm sewers
Resources Publishers, Highlands Ranch.
2. wastewater sewers
40. Anderson, M.G. and Burt, T.P. (1985). Modelling strategies.
In: M.G. Anderson, and T.P. Burt (Eds.), Hydrological 3. combined sewers.
Forecasting, John Wiley & Sons, New York, pp. 1–13.
41. Woolhiser, D.A. and Brakensiek, D.L. (1982). Hydrologic Storm Sewers. Storm sewers dominantly transport
system synthesis. In: C.T. Haan, H.P. Johnson, and rain water, mainly runoff, during rain from city areas.
D.L. Brakensiek (Eds.), Hydrologic Modelling of Small Typically, storm water runoff has low concentrations
Watersheds. ASAE Monograph, Washington, DC, pp. 3–16. of pollution and the concentration time for the storm
42. Duan, Q., Gupta, H.V., Sorooshian, S., Rousseau, A.N., and sewer system is usually rather short—of the order of
Turcotte, R. (Eds.). (2003). Calibration of watershed models. minutes to a few hours. If the concentration time is short
AGU Water Science and Application, 6, Washington, DC. and the concentrations are low, then aerobic conditions
43. Klemeš, V. (1986). Operational testing of hydrological simu- will dominate, and there will be only a low level of
lation models. Hydrological Sci. J. 31(1): 13–24. transformation and biological activity in the storm water.
44. Beven, K.J. and Binley, A.M. (1992). The future of distributed
Models: model calibration and uncertainty prediction. Hydrol. Sanitary Sewers. Sanitary sewers transport wastewater
Process. 6: 279–298. from residences, commercial areas, and industries.
45. Gupta, H.V., Beven, K., and Wagener, T. (2004). Model Depending on the local layout of the sewer system, the
calibration and uncertainty estimation. In: M.G. Anderson wastewater may be transported by gravity or in pumped
(Ed.), Encyclopedia of Hydrological Sciences, John Wiley, lines (in areas with insufficient ground slope available).
Chichester, UK. The concentration time for sanitary sewers may be up to
1–2 days. Wastewater typically has high concentrations
of pollution and biodegradable material, so wastewater
MODELING OF WATER QUALITY IN SEWERS sewers are potentially a place of high biological activity.
OLE MARK
Combined Sewers. Combined sewers transport just
Asian Institute of Technology
wastewater in dry weather and both wastewater and
Hørsholm, Denmark
1
The terminology ‘‘The sewer as a physical, chemical and
biological reactor’’ has been largely promoted through the work of
Modeling of water quality in sewers emerged during the the Sewer Systems & Processes Working Group under IWA and in
late 1990s. The water quality and amount of water in particular promoted by Prof. Thorkild-Hvitved-Jacobsen, Aalborg
sewers are dominant factors for designing wastewater University, Denmark, and Prof. Richard Ashley, University of
treatment plants and for analyzing water quality impacts Bradford, UK.
332 MODELING OF WATER QUALITY IN SEWERS

runoff from city areas during rain. The concentration time Modeling of Quasi-Conservative Pollution
in a combined sewer system, may be rather, short or it
The term Quasi-Conservative Pollution covers pollution
may rise to the order of 1–2 days. During dry weather
that is not significantly transformed while it is in the sewer
flow, combined sewers operate like sanitary sewers, but
system. Ammonia and phosphorous are water quality
during rain, rainwater that has a lower concentration
parameters, which certainly are not 100% conservative,
of pollutants is added, and the transport time of the
but which in most cases can be modeled with sufficient
water to the wastewater treatment plant is significantly
accuracy assuming that they are conservative (6,7).
reduced. When a combined sewer system is overloaded, the
Dissolved conservative pollution can typically be
excess water is discharged into receiving waters (a river,
modeled by assigning a measured concentration to the
lake, or a coastal area), and this combined sewer overflow
wastewater and to the rain. The resulting concentration in
(CSO) may seriously deteriorate the water quality in the
the sewers is then effectively a dilution of the wastewater
receiving waters.
based on the land use and local industrial load to the sewer
The impact on recipients from overflows from sewer
system. At this point, it is important that the model applied
systems can be divided into three typical problems:
for pollution modeling represents the transport of pollution
inside each pipe. The model must have computational grid
1. toxic impact due to un-ionized ammonia discharged points within each pipe otherwise there will be too much
mainly from CSOs dispersion in long pipes, resulting in lowered peaks of
2. oxygen depletion concentration.
3. hygienic problems caused by fecal coliform.
Modeling of Nonconservative Pollution

Hence, the selection of an appropriate modeling approach The term Nonconservative Pollution covers situations
for water quality in sewers depends both on the type of where biological and chemical processes play a significant
sewer system and the nature of the problem. Un-ionized role during the time the water passes through the
ammonia in a combined sewer system may be accurately sewer/drainage system. In this case, modeling the
modeled by applying an advective dispersion model, pollution by an advective–dispersion model will be
considering that ammonia is conservative. However, it is a insufficient and will give misleading results. Some
must to include biological and chemical processes to model examples of Nonconservative Pollution are BOD, COD,
hydrogen sulfide in a pumped wastewater pipe whose and hydrogen sulfide. If the sewer system has a very short
retention time is more than 3–4 hours. The following concentration time, even BOD and COD may be modeled
sections focus on modeling various types of water quality by ignoring the water quality processes, but a contribution
problems in sewers. from erosion of sediment deposits in the sewers must still
be included.
The advective–dispersion model basically has only
DETERMINISTIC MODELING OF WATER QUALITY one calibration parameter, the dispersion coefficient,
because advection of the pollution depends fully on the
hydrodynamics. To determine the dispersion coefficient,
The strength of deterministic models for water quality
a tracer measurement can be done with salt, rhodamine,
in sewers is increased understanding of the individual
uranine, or a similar substance, or the model can be
processes and their interaction. Deterministic modeling of
calibrated based on measurements during rainfall of a
water quality in sewers and urban drainage can be divided
quasi-conservative pollutant such as ammonia. A tracer
into two groups:
study is highly recommended because it verifies that the
model accurately describes the transport of a conservative
1. quasi-conservative pollution substance under known flow conditions. In addition to
2. nonconservative pollution. finding the dispersion coefficient, the tracer measurement
program serves three purposes:
Modeling water quality in sewers is a challenging task,
which assumes that the model is built and verified step by • verification of the assumptions about infiltration to
step. Note that a small error in the hydrodynamic model the sewer system
is carried 100% into the water quality model. A volume • the possibility of fine-tuning the Manning numbers
error of 10% more water in the hydrodynamic model will along the sewers in the tracer measurement
directly imply that the concentration of pollution in the • validation of the capability of the modeling sys-
water quality model will be too small. At present, modeling tem to describe the physical transport of the dis-
of pollutant transport may be calculated by using an solved pollutants.
advective–dispersion model such as MOUSE TRAP (3).
Models transporting the pollution by pure advection (no In conclusion, a tracer study serves well as a verification
dispersion) are InfoWorks (4) and the SWMM models (EPA process of a hydrodynamic model. If the model fails to
SWMM, MIKE SWMM, and XP SWMM) (5). As software reproduce the tracer results, then there is no reason to
for modeling sewers develops rapidly, it is suggested that continue the study to model anything related to water
the reader always seek the latest information from the quality. The final set of input data to the advection-
software suppliers of the models. dispersion model consists of
 MODELING OF WATER QUALITY IN SEWERS 333

my-g/L
Measured and simulated tracer concentration
1000.000

500.000

0.000
12:30 12:45 13:00 13:15 13:30 13:45 14:00 14:15 14:30
1997 7/4

Measured and simulated discharge

m3/s
0.400
0.300
0.200
0.100
0.000
12:30 12:45 13:00 13:15 13:30 13:45 14:00 14:15 14:30
1997 7/4 Figure 1. Example of tracer study results, showing
Simulated: Measured: that the hydrodynamic model needs to be recalibrated.

1. diurnal variation of dry weather flow the advective dispersion mode, preferably based on a
2. diurnal variation of concentration hydrodynamic model. At this time, only the MOUSE TRAP
3. infiltration model has such capabilities (3). The SWMM models (EPA
SWMM, MIKE SWMM and XP SWMM) and InfoWorks
4. rainfall runoff
(4) have only an advection model.
5. major industrial loads in terms of flow and The transformation of pollution may either be modeled
concentration. by applying a model based on decay rates or by using
a model based on the physical, biological, and chemical
The discrepancies between a tracer study and advective processes, such as the WATS model (1). The WATS model
dispersion, shown in Fig. 1, are for the city of Helsingborg, has thus been successfully applied to wastewater modeling
Sweden (7). in the Estoril sewer system in Portugal (10). Good case
During the measurements, the tracer dose was added studies demonstrating the applicability of water quality
as a half-minute long pulse. From the figure, it can be seen process models during rain are still lacking at this time.
that there was a very good match between the measured To calibrate a water quality process model, it is of
and simulated flow, but the timing of the measured and utmost importance to have a minimum of two stations
the simulated peaks was out of phase for the tracers. It was measuring hydraulic and water quality parameters in the
found that this occurred because the Manning numbers important sewers in the catchment because the water
were never calibrated properly and only a flow gauge and quality parameters will determine a process rate, which
not a water level gauge was available at this location in can be found only by comparing the measurements from
the sewers. Simulation results by applying an advective two or more measurement stations. Some typical water
dispersion model for modeling various water quality quality processes under aerobic conditions can be seen in
parameters can be seen in Fig. 2 for the Helsingborg sewer Fig. 3.
system in Sweden.

Sediment Attached Pollution MODELING HYDROGEN SULFIDE

Pollution may be attached to sediments, especially BOD, Understanding and modeling hydrogen sulfide is impor-
COD and phosphorous (2,8). When suspended sediment tant because it introduces problems such as corrosion
or sediment deposits exist in sewers, it is of utmost of pipes and other metal constructions in sewers. It is
importance to include a model that can also simulate toxic, it poses a serious health risk to the sewer main-
sediment transport processes in the sewers—including tenance staff, and it releases noxious odors to the city
erosion and deposition of sediments. Principles for environment. Further, it may inhibit sewage treatment
modeling sediment transport in sewers can be found in processes (11). A high sulfide content may be toxic to fish
Mark et al. (9) and IWA (2). in streams affected by sewer overflow (12). Hydrogen sul-
fide develops under anaerobic conditions in sewers, and
Modeling Nonconservative Pollution by Application of a
it interacts with anaerobic water quality processes (1,13).
Water Quality Model
Hydrogen sulfide has traditionally been modeled by using
Nonconservative pollution in sewers must be modeled empirical equations such as: Thistlethwayte (14), Boon
by describing the water quality processes along with and Lister (15), Pommery & Parkhurst (16), and Nielsen
334 MODELING OF WATER QUALITY IN SEWERS

Inflow WWTP Öresundsverket

[mg/l]
Ammonia - N, mg/l Flow, m3/s
35.0 [m3/s] Time series of DISCHARGE BRANCHES (1oktm.prf)

[m] 30.0
2.0
25.0
12400.0 20.0 1.5
15.0 1.0
12200.0
10.0 0.5
12000.0 5.0 0.0
0: 0 6: 0 12: 0 18: 0 0: 0 6: 0 12: 0 18: 0 0: 0
11800.0 00:00 06:00 12:00 18:00 00:00 06:00 12:00 18:00 00:00
30-9-1997 1-10-1997 2-10-1997
30-9-1997 1-10-1997 2-10-1997
11600.0
[mg/l] Ammonia - N, mg/l
11400.0 35.0
30.0
11200.0
Measured, analysed 25.0
11000.0 20.0
Mouse HD/AD simulated
15.0
10800.0
10.0
10600.0 5.0
6: 0 12: 0 18: 0 0: 0 6: 0 12: 0 18: 0 0: 0
10400.0 30-9-1997 1-10-1997 2-10-1997
10200.0 [mg/l] COD, mg/l
10000.0 700.0
9800.0 600.0
500.0
9600.0 400.0
9400.0 300.0
200.0
9200.0 [mg/l]
Ammonia - N, mg/l
6: 0 12: 0 18: 0 0: 0 6: 0 12: 0 18: 0
9000.0 35.0
30-9-1997 1-10-1997
30.0
8800.0
25.0 [mg/l] Total - P, mg/l
11.0
20.0
10500.0 10000.0 9500.0 9000.0 8500.0 8000.0 7500.0 7000.0 10.0
15.0 9.0
8.0
10.0 7.0
6.0
0: 0 6: 0 12: 0 18: 0 0: 0 6: 0 12: 0 18: 0 0: 0 5.0
30-9-1997 1-10-1997 2-10-1997 4.0
3.0
2.0
0: 0 6: 0 12: 0 18: 0 0: 0 6: 0 12: 0 18: 0 0: 0
30-9-1997 1-10-1997 2-10-1997

Figure 2. Wet weather sampling. Examples of measured flow and wastewater analyses
compared to preliminary MOUSE hydrodynamic and advective dispersion simulation results.

and Hvitved-Jacobsen (17). Note that the four empirical for bacterial decay are salinity, temperature, and light
models in these references were developed by using data intensity. At present, limited knowledge is available
from pressure mains, so they are not generally applicable concerning the transport of coliform in sewers; it is
to all sewer systems. At present, the only hydrogen sul- modeling fecal coliform in sewers by using a first-order
fide model in commercial software is the MOUSE TRAP decay rate is suggested (21); it is conservative when the
model (18). This model includes an advective dispersion concentration time in the sewer system is short—up to a
model and four empirical equations for sulfide buildup. few hours (22).
In addition, oxygen demand and extra Michaelis–Menten
expressions are included in the model, so it approaches a A PROCEDURE FOR MODELING WATER QUALITY IN
more deterministic description of sulfide production. SEWERS
A more recent and generally applicable model for
hydrogen sulfide in sewers is the WATS model (1). The In most cases, sediment affects pollution significantly, and
WATS model applies an integrated aerobic and anaerobic when the concentration time of the sewer system is long (as
concept to the transformation of organic matter and sulfur a rule of thumb, 3–4 hours), the interaction of BOD/COD
in wastewater (1). and oxygen must be taken into account in the model
to produce reliable results. When the oxygen level plays a
significant role, for instance, in modeling hydrogen sulfide,
MODELING FECAL COLIFORM IN SEWERS a water quality process model is a must. A consistent
procedure for water quality modeling is outlined here.
Experience modeling fecal coliforms has shown that Each step in the modeling procedure must be completed
coliform cannot be considered conservative or without and verified before continuing to the next step. If the model
natural decay in the receiving waters (20), for example, results at a step are unsatisfactory, then it is not feasible
in the sea, where the main controlling parameters to continue because the outcome of each step depends on
 MODELING OF WATER QUALITY IN SEWERS 335

Reaeration

Oxygen

Decay Heterotrophic
growth
BOD
Decay suspended
BOD
dissolved Hydrolysis
Biofilm Depo-
Erosion Erosion sition

Sediment BOD BOD

oxygen interstitial sediment
demand liquid

Sediment

Figure 3. Example of water quality processes in wastewater under aerobic conditions.

the success of the previous steps. The modeling procedure 7. Model total COD and BOD as the outcome of
is as follows: the modeling of the dissolved COD/BOD and the
sediment transport of particulate COD and BOD.
1. Make a tracer verification of the hydrodynamic
model, and apply a pure advection and/or advec-
tive–dispersion model. LONG-TERM SIMULATION OF POLLUTION
2. Model a real quasi-conservative pollutant, for
example, ammonia for both dry weather flow
and rainstorms. High computational speed has made it feasible to carry
out hydrodynamic and water quality simulations applying
3. Analyze the ratio of dissolved and particulate COD
deterministic models. These models allow computing
and BOD, and check if a ‘‘first flush’’ exists.
annual loads of pollutants to a WWTP or annual loads
4. If the water quality parameters in the study are discharged from CSOs. In addition, extreme events can be
attached to sediment, then model the sediment modeled, such as those giving rise to high CSO discharges
transport of the suspended solids. of water and pollutants that can cause fish kills. Through
5. To determine the water quality process rates, obtain long term simulation, extreme events can be assigned
measurements of flow and pollution for relevant return periods, providing information concerning how
periods at upstream and a downstream locations in often various events occur. In Fig.4, the result from
the sewer system. a 3-month simulation is seen. Long-term simulations
6. Model dissolved COD and BOD using the process generally require data spanning three to five times the
rates from step 5. (return) period of interest.
336 MODELING OF WATER QUALITY IN SEWERS

800 processes in the sewers play a key role in this analysis,

700 COD-tot and, as more information becomes available concerning
modeling of water quality in sewers, more solutions,
600
and hopefully, better and sustainable solutions may be
500 developed for handling the water in sewer systems.
400
300 BIBLIOGRAPHY
200
100 1. Hvitved-Jacobsen, T. (2002). Sewer Processes—Microbial and
Chemical Process Engineering of Sewer Networks. CRC Press,
0 Boca Raton, FL.
20/08/98 09/09/98 29/09/98 19/10/98 08/11/98
2. IWA/IAHR Joint Committee on Urban Drainage. (2003).
50 Solids in Sewers. IWA Publishing.
45 N-tot 3. Garsdal, H., Mark, O., Dørge, J., and Jepsen, S.-E. (1994).
40 MOUSE TRAP: Modelling of Water Quality Processes and
35 the Interaction of Sediments and Pollutants in Sewers.
Specialized International Conference: The Sewer as a
30
Physical, Chemical & Biological Reactor. Aalborg, Denmark.
25
20 4. Bouteligier, R., Vaes, G., and Berlamont, J. (2001). Dissolved
Pollutant Modelling in Urban Drainage Systems—A Case
15
Study of the Wezembeek Oppem Catchment on the Transport
10
of Ammonium using INFO WORKS CS. MTERM Workshop
5 on Modelling of Water Quality in Urban Drainage Systems,
0 Asian Institute of Technology, December.
20/08/98 09/09/98 29/09/98 19/10/98 08/11/98
5. Huber, W.C. and Dickinson, R.E. (1988). Storm Water Man-
9.0 agement Model—SWMM, Version 4 User’s Manual. US Envi-
P-tot ronmental Protection Agency, Athens, GA.
8.0
6. Mark, O., Appelgren, C., and Kosir, M. (1996). Water Quality
7.0
Modelling for the Ljubljana Master Plan. 7th International
6.0 Conference on Urban Storm Drainage, Hanover, Germany.
5.0 7. Mark, O., Hernebring, C., and Magnusson, P. (1998). Optimi-
4.0 sation and control of the inflow to a wastewater treatment
3.0 plant using integrated modelling tools. Water Sci. Technol.
37(1): 347–354.
2.0
8. Crabtree, R.W. (1989). Sediments in Sewers. J. Inst. Water
1.0
Environ Manage. 3.
0.0
9. Mark, O., Appelgren, C., and Larsen, T. (1995). Principles
20/08/98 09/09/98 29/09/98 19/10/98 08/11/98
and approaches for numerical modelling of sediment trans-
port in sewers. Water Sci. Technol. 31(7): 107–115.
Model
10. Mourato, S., Almeida, M., Matos, J., and Hvitved-Jacob-
Measured daily average sen, T. (2002). Modelling in-sewer pollutant degradation
processes in the Costa do Estoril sewer system. Sewer
Figure 4. Simulated ammonium concentration in the inflow Processes and Networks. France—Accepted for publication
to the Øresundsverket, WWTP, Helsingborg, compared with in Water Science and Technology in 2003.
analyzed 24 hourly values (each workday). 11. Bentzen, G. et al. (1995). Controlled dosing of nitrate for
prevention of H2S in a sewer network and the effects on
the subsequent treatment process. Water Sci. Technol. 31(7):
FUTURE PERSPECTIVES AND CHALLENGES IN RELATION 293–302.
TO MODELING OF WATER QUALITY IN SEWERS
12. Nielsen, P.H., Raunkjær, K., Norsker, N.H., Jensen, N.A.,
and Hvitved-Jacobsen, T. (1992). Transformation of wastew-
At present, it is feasible to model water quality and ater in sewer systems—a review paper. Water Sci. Technol.
pollution in sewers. However, for successful modeling, it 25(6): 17–31.
is of great importance to understand the hydrodynamics, 13. Tanaka, N., Hvitved-Jacobsen, T., and Horie, T. (1998).
the transport of pollution, and the potential associated Transformation of wastewater organic matter in sewers under
biological and chemical processes. Further, a model changing aerobic/anaerobic conditions. Water Sci. Technol.
describing the dominant processes is essential. And, 72(6): 651–666.
finally, a systematic procedure must be applied whereby 14. Thistlethwayte, D.K.B. (Ed.). (1972). The Control of Sulfides
the performance of the water quality model is verified step in Sewage Systems. Butterworth, Sydney, Australia, p. 173.
by step. 15. Boon, A.G. and Lister, A.R. (1975). Formation of sulphide in
Holistic analysis of the sewer system, the wastewater rising main sewers and its prevention by injection of oxygen.
treatment plant, and receiving waters is still an emerging Prog. Water Technol. 7: 289–300.
area that has potential for integrated analyses of overall 16. Pommery, R.D. and Parkhurst, J.D. (1976). The Forecasting
performance. The sewer system and the water quality of Sulphide Build-up Rates in Sewers. Proceeding of the
 MODELING OF URBAN DRAINAGE AND STORMWATER 337

8th International Conference on Water Pollution Research, Today’s advances in computer technology allow many
Sydney. cities all over the world to report overflows from their
17. Nielsen, P.H. and Hvitved-Jacobsen, T. (1988). Effect of sewers yearly using a combination of measurements
sulfate and organic matter on the hydrogen sulfide formation and modeling, and in many places, they manage local
in biofilms of filled sanitary sewers. J. Water Pollut. Control and minor flooding problems using computer-based
Fed. 60: 627–634.
solutions. This involves developing computer models
18. Erichsen, A.C., Andersen, K.K., and Mark, O. (1999). Sul- of the drainage/sewer system, for instance, by using
phide Occurrence in Sewer Networks—a New Topic in
software such as InfoWorks (1), MIKE URBAN (2),
MOUSE TRAP. 3rd DHI Software Conference, Helsingør,
Denmark.
MOUSE (3), PCSWMM (4), and SWMM (4). All are
hydrodynamic modeling packages that solve the one-
19. Van Assel, J. and Carrette, R. (2002). A practical application
of Urban Pollution Modelling in Flanders (Belgium): The
dimensional equations of St. Venant.
catchment of Tielt. Water Sci. Technol. 45(3): 125–129.
20. Evison, L.M. (1986). Comparative studies on the survival of TYPES OF MODELS USED FOR URBAN DRAINAGE AND
indicator organisms and pathogens in fresh and sea water. STORMWATER SYSTEMS
Water Sci. Technol. 20: 1051–1054.
21. Rauch, W. et al. (2002). Mathematical modeling of integrated
Any model is a simplified mathematical representation of
urban drainage systems. Water Sci. Technol. 45(3): 81–94.
a physical system. This representation may be based on a
22. Tomicic, B., Lutzen, A., and Mark, O. (2001). Integrated
deterministic approach (i.e., a fixed relationship between
Modeling of the Sewer System and the Receiving Waters for
the Island of Ischia. Conference: UDM. May, Orlando, FL.
physical disturbance and its effect, such as model results)
or it could be stochastic, involving terms of probability
in the model inputs and the interpretation of model
MODELING OF URBAN DRAINAGE AND results. Models for urban drainage and stormwater are
predominantly deterministic models, and the text below
STORMWATER refers only to this class.
OLE MARK Furthermore, deterministic models applied to urban
drainage can be roughly classified as physical models
Asian Institute of Technology
Hørsholm, Denmark and conceptual models. Placing a model in one of these
classes relies on the level of mathematical sophistication
SUTAT WEESAKUL of the treatment of underlying physical processes in the
DHI Water & Environment
model. Practically, models are classified according to the
Bangkok, Thailand importance of empirical parameters for the models’ ability
to describe these processes accurately. The reliance on
empirical parameters classifies the model as conceptual
INTRODUCTION and emphasizes the need for validation against field
measurements. In this context, various hydrologic models
The construction and operation of urban drainage and would belong to the class of conceptual models; a
stormwater systems have historically been driven by the hydrodynamic network model is an example of a physical
objectives of maintaining public hygiene and preventing deterministic model.
flooding. All over the world, cities have grown rapidly
during the last century, and city authorities have built and
continuously expanded storm and wastewater drainage BRIEF HISTORY OF URBAN DRAINAGE MODELING
facilities. Many old cities have developed according to their
varying historical needs and visions. Hence, the layout and Modeling of urban sewer and drainage systems dates
design of the infrastructure have gradually developed into back to the late 1960s when the U.S. EPA funded the
rather complex systems, which require tools to handle development of the first version of the EPA Stormwater
such situations. Management Model (SWMM). The first version 1.0
An understanding of the physical system and its of SWMM was released in 1971. Around this time,
interaction with the environment is a prerequisite for Wallingford (UK) and DHI Water and Environment (DK)
effective planning and management of urban drainage developed their first one-dimensional flow models, but it
and stormwater systems. In this respect, computer models was not until 1982–86, with the release of the commercial
provide the opportunity for well-structured analyses of software packages, MicroWASSP (1982), MOUSE (3), and
the current performance of a system and the effect of PCSWMM (4), that they were no longer ‘labeled’ as ‘in-
potential changes in the system; they offer a sound house software.’ These modeling packages came very soon
scientific framework for coordinated management and after the first PCs hit the market. Since the release of
planning. Apart from assessing different scenarios, the personal computers (PC), modeling software for urban
models also help to improve process understanding. drainage has closely followed and fully used the full
Urban drainage models are used to understand the powers of the PC. As PCs have become more powerful,
often rather complex interaction between rainfall and the modeling software has increased in complexity, and
overflow/flooding. Once the existing conditions have been the size of the urban drainage model has gone from a
analyzed and understood, alleviating schemes can be maximum of around 250–500 pipes in the mid-1980s
evaluated, and the optimal scheme implemented. to modeling the drainage systems of mega cities, (e.g.,
338 MODELING OF URBAN DRAINAGE AND STORMWATER

models with thousands of pipes). Computer modeling are obtained by manual or automated sampling, followed
of urban drainage has not added much new within the by laboratory analyses.
basic theory of urban hydrology and hydraulics, but it
brought fast and user-friendly computations of the St.
APPLICATIONS OF MODELS TO URBAN DRAINAGE AND
Venant equations to the broad engineering community.
STORMWATER SYSTEMS
During the 1990s, modeling of urban drainage broke
new ground by developing both theoretical and modeling
Urban drainage and stormwater models are usually set up
approaches for sediment transport and water quality
and applied for specific analyses, and a model of the same
in sewers, processes which today are still not fully
part of a city may be differently schematized and calibrated
understood. Lately, major modeling packages have been
for different objectives, based on the application type.
fully embedded in Geographical Information Systems
The next sections briefly describe common applications of
(GIS) to ease model building based on the registrations in
urban drainage and stormwater modeling:
the databases of the local authorities and to present model
output in conjunction with the cities’ infrastructural data.
• modeling of water quality
• modeling of urban flooding
CALIBRATION AND VERIFICATION OF MODELS: FIELD • modeling the interaction between urban drainage
MEASUREMENTS systems, receiving waters, and wastewater treat-
ment plants.
Urban drainage models are expected to reproduce the
behavior of the modeled system with a high level of MODELING OF WATER QUALITY
accuracy. This is usually ensured through parameter
calibrations and the verification of model results against Modeling of wastewater is important for predicting the
the measured performance of the real system. Calibration inflow of both water quantity and quality to wastewater
involves adjusting the model’s key parameters, to treatment plants and for precise descriptions of overflow
minimize the differences between the model results and from sewer systems (overflow of a mix of wastewater
the field measurements (e.g., water levels). A continuous and rainwater from combined sewer systems)—combined
period or a set of intermittent events used for calibration sewer overflows (CSO) and overflow from sanitary sewer
should include a full range of expected operational systems (SSO). Traditionally, overflows are modeled by
conditions in the system. Verification provides proof that using a rainfall runoff model as input to a hydrodynamic
the model generates results within the acceptable error pipe flow model in conjunction with a diurnal description
range. Reliable verification should be carried out for of the wastewater produced by the people living in the city.
a simulation period or intermittent event(s), which are In the past, the performance of a sewer system was
different and independent of those used for the model characterized by most Environmental Protection Agencies
calibration. (EPAs) by the number of overflow events per year. If
All models must, if possible, be calibrated before a sewer system had less than, e.g., five overflows/year,
application. Conceptual models require more attention it was considered satisfactory performance. This was
in this respect compared to fully deterministic, physical a performance criterion which was feasible to check
models because the assumptions of the conceptual models and implement, but it could not account for the actual
must also be proven valid. Thus, in urban drainage volumes spilled. It was realized that five overflows/year
modeling, the focus of model calibration is usually on could be spilling one thimble of water five times, or it
hydrologic models (conceptual), whereas the deterministic could be huge volumes of water spilled five times. When
hydrodynamic models of the drainage network typically modeling was introduced, it was possible to compute
require only minor adjustments for accurate performance. volumes from all overflows in a drainage system. In
However, independently of the applied model type, good the past, they all had to be measured to get a picture
modeling practice requires thorough model verification of the complete overflow from the system. Relatively
before application. cheap and reliable application of models for calculating
Reliable field measurements in sewers and drains overflow volumes and pollution emissions for the entire
and the results of laboratory analysis (in the case system changed the EPA focus from a frequency/year to
of water quality modeling) are essential for successful a consideration of the yearly overflow volumes, whenever
model verification (i.e., for the model application as a that was more appropriate. For instance, for lakes, the
whole). The harsh environment in sewers and urban accumulated overflow of nitrogen and phosphorus from
drains (poisonous gases and pathogenic microorganisms the drainage system is most important, whereas the
in confined conduits, aggressive fluids, floating debris), number of spills is less important. For smaller streams, the
as well as particular hydraulic conditions (alternating loads from extreme events causing high ammonia or low
free-surface and pressurized flow, rapid flow surges, oxygen concentrations may be the dominant processes.
etc.) requires special monitoring equipment and highly The approach to sewer rehabilitation/optimization was
trained specialists for good results. Routinely measured changed from a reduction of the number of spills/year to
variables include water level, flow velocity, conductivity, a reduction of the yearly overflow volumes together with
temperature, pH, etc. The chemical and biological consideration of the load to receiving waters. In other
properties of the water, as well as sediment characteristics, words, ‘‘The focus went downstream.’’
 MODELING OF URBAN DRAINAGE AND STORMWATER 339

The load to streams and rivers would typically be

computed based on a conceptual hydraulic model such
as the SAMBA model (5). This kind of model is simplified
so that it includes only computation of a time–area runoff
hydrograph together with the most significant elements
in the pipe/drainage system, weirs, pumps, locations with
divergent flow, and outlets. Due to the simplifications, the
calculations are very fast and suitable for scanning a large
number of alternative layouts of drainage systems. In the
early days of computer modeling, this kind of modeling
was the only way to calculate annual loads and extreme
statistics of discharges of water and pollutants based on
historical time series of rain events. The model focused on
the requirements from the EPA.
The introduction of water quality modeling of urban
drainage such as MOSQITO (6) and MOUSE TRAP (7)
in the middle of the 1990s, extended the possibility
of hydrodynamic modeling of ‘‘clean’’ water to include Figure 1.
modeling of the transport of pollution and sediments in
the sewer and drainage systems. This further enhanced
the description of the modeling of wastewater and the cities in developed countries typically have smaller scale
pollution loads to the receiving water because this kind of problems, due to locally insufficient sewer capacity. Other
water quality modeling provides an accurate description cities, for instance, in Asia-Pacific, can have more severe
of the dilution process when rainwater mixes with sewage, problems because there is insufficient drainage and much
together with an estimate of the water quality processes in heavier local rainfall. The situation is further aggravated
the system. The outcome was more accurate computations because cities in developing countries grow rapidly these
of concentrations of conservative pollutants in the CSO days, but do not have the funds necessary to extend and
and, for example, ammonia can be modeled accurately in rehabilitate their existing drainage systems. For more
most cases. For more information concerning modeling information concerning urban flooding, refer to the section
water quality, refer to the section in the Encyclopedia of in the Encyclopedia of Water, Urban Flooding.
Water, Modeling of Water Quality in Sewers. Modeling urban flooding differs significantly from
During the 1990s, long-term simulations of drainage modeling pipe flow, which has a confined geometry
network hydraulics and water quality gradually became can be measured accurately and easily. The flow is
standard practice, facilitating simulation of long historical basically one-dimensional and accurately described by
time series of rain and their impact in terms of overflow or the St. Venant equations. When the water leaves the
load on the wastewater treatment plant. This combination drainage/stormwater system, it flows onto a three-
of the simulation of wet weather conditions by using the dimensional ground surface where the flow routes and
dynamic pipe flow model and a simple hydrologic model paths are not well defined. The surface must be described
to simulate the dry weather conditions makes it possible by a digital elevation model containing roads and houses
to simulate many years (>20 years, but feasible length in the flooded areas. Modeling surface flooding takes
of simulation period depends on model size, computer on additional data in an environment where the local
speed, and rainfall data duration) continuously within a geometry is difficult to describe accurately and the flow
reasonable computational time on a PC. The results can be pattern may be one- or two-dimensional. Recently, a
presented based on the statistical occurrence of impacts, model was developed that combined a one-dimensional
for example, in terms of overflow volume. Understanding pipe flow model and a two-dimensional surface flood
the behavior of the system and its interaction with model (8). This approach allows exchanging surcharge
the environment, gained from long-term simulation, is water and flow in the pipe network The results can be
superior to simulation of selected single rains because the used to identify hazardous flood zones, where high velocity
uncertainty from selecting a few representative rains is occurs (9). Another two-dimensional model using a shallow
removed. Furthermore, based on a long-term simulation, water equation has been developed to study the flow
the statistics are made on the computed output (the impact characteristic of street flooding when water escapes from
of the rain) in terms of CSO, SSO, or flooding. In the past, manholes (10). Modeling urban flooding is an emerging
the statistics were made during the selection of the rains area where applications have started to appear during
because as they would be input for the simulation. the last decade. For further information, refer to Mark
et al. (11).
MODELING OF URBAN FLOODING Examples of cities that experience urban flooding
problems on a larger scale are Bangkok, Thailand; Dhaka,
The problems arising from urban flooding range from Bangladesh; Ho Chi Min City, Vietnam; Mumbai, India;
minor ones, such as water getting into the basements and Phnom Penh, Cambodia. In the year 2000, nearly
of a few houses, to major incidents where large parts of 17,000 telephone lines in Mumbai went out of service after
cities are inundated for several days (Fig. 1). Most modern flooding occurred, and the electricity supply was switched
340 MODELING OF URBAN DRAINAGE AND STORMWATER

off for a long time in most areas for safety reasons. The Furthermore, there is a general lack of methodology
floods reached 1.5 m deep in the locations that were most and technology available for integrated planning and
severely inundated. Fifteen lives were lost in the Mumbai management. Finally, there is simply a lack of a regulatory
flooding (Times of India, July 25, 2000). In Dhaka, even a framework for ‘‘integrated thinking.’’ An exception to this
little rain can inundate parts of the city for several days. ‘‘rule’’ is the UPM procedure (14) in the United Kingdom,
The situation was highlighted in September 1996, when which is highly recommended reading. Today’s challenge
the residents of Dhaka experienced ankle to knee-deep is to move the legislation from individual consideration of
water in the streets. Heavy traffic jams occurred because each subsystem’s performance to integrated management
of the stagnant water on the streets, and daily activities in of the urban wastewater system.
parts of the city were almost paralyzed. In 1983, Bangkok The quality of an integrated model for an urban
was flood inundated for six months, which caused direct drainage system depends strongly on the quality of
damage worth approximately 6,600 million baht ($149 the WQ sewer model, which today is the weakest
million) (12). link in integrated modeling. Despite these uncertainties,
integrated modeling is still very beneficial, and the future
MODELING THE INTERACTIONS AMONG URBAN will surely be dominated by integrated modeling.
DRAINAGE SYSTEMS, RECEIVING WATERS, AND
WASTEWATER TREATMENT PLANTS REAL-TIME MODELING

‘‘Integrated modeling’’ in urban drainage is a term At present, hydrologists working in urban areas are
describing ‘‘Modeling of the interaction between urban facing many new challenges imposed by the ever-changing
drainage systems, the receiving waters, and wastewater hydrologic environment in cities. In such conditions,
treatment plants.’’ The term, ‘‘receiving waters’’ is used emphasis is focused on managing the urban systems as
here for groundwater, streams, rivers, lakes, the sea, etc. efficiently as possible within the existing infrastructure by
Most urban areas have these three components. Few who applying currently available information and technology,
are involved in managing these systems would disagree where real-time modeling plays an increasingly important
with the statement: ‘‘It is logical and beneficial to plan, role. It occurs in two types of applications:
design and operate these components in an integrated
manner’’ (13). • modeling real-time control and active real-time
Integrated modeling is a complex exercise due to the control of urban drainage systems
sheer size of the model problem and also due to the
• modeling as a real-time decision support tool.
different modeling approaches for each subsystem that
has developed during modeling history. The hydrodynam-
ics in flow and water levels can be directly transferred MODELING REAL-TIME CONTROL AND ACTIVE
between subsystems, but presently water quality modeling REAL-TIME CONTROL OF URBAN DRAINAGE SYSTEMS
is handled completely differently in channels/rivers and at
the wastewater treatment plant. This causes problems In the early 1990s real-time control of sewers (RTC) was
in the interface and data transfer between subsystems, recognized as a big leap forward to achieving superior
which again impacts accurate modeling of WQ processes. operation of many sewer systems. Since then, RTC has
The logic behind integrated management is old, certainly been proven an eminent solution for some sewer
and a number of separate tools exist. The advent of systems to reduce the risk of flooding and the amount of
computer models has been the biggest single factor in the CSO at the same time. But RTC has not been implemented
development of integrated analyses. Because of technology to the extent it was foreseen in the early 1990s, and
limitations in the past, historical reality has been removed it seems as if RTC never became ‘‘the big thing’’ that
from integration, and independent pragmatic criteria for was envisaged. Some of the reasons behind the slow
each component have been the norm (13). These are implementation of RTC in the real world are due to the
some examples. risks of applying RTC, and the safety precautions which
have to be taken to ensure smooth and secure operation
• Sewers are typically designed to contain storms of a of the RTC system in case of a power failure or other
specific return frequency without surcharging, rather abnormal incidents.
than protecting against flooding. The aim of RTC modeling of an urban drainage/
• Combined sewer overflows are engineered to pass stormwater system is to develop the strategies applied
forward flow dictated by downstream sewer or WWTP in real life for pumps, weirs, and gates and to evaluate the
capacity rather than the ability of the receiving water impact and risk from malfunctions, in a system controlled
to accept the overflows. by RTC. For example, what is the flood risk if a weir or a
• Treatment plants themselves are often designed to gate is stuck during a rain?
cope with an arbitrary multiple of dry weather flow, Today RTC is applicable for
rather than treating all flow from the sewer system.
• reducing the flooding risk by using the storage
In most parts of the world, there have been fragmented capacity within the sewer system,
planning and management of the sewer system, the • reducing pollution spills from sewer overflows by
wastewater treatment plant, and receiving waters. retaining more stormwater and sewage within the
 MODELING OF URBAN DRAINAGE AND STORMWATER 341

sewer system as a result of attenuated flows FUTURE PERSPECTIVES AND CHALLENGES IN MODELING
and storage, URBAN DRAINAGE
• reducing capital costs by minimizing the storage Calibration Support for Urban Drainage and Stormwater
and flow carrying capacity requirements within Systems
the system,
Today, urban drainage and stormwater models are cali-
• reducing operating costs by optimizing pumping
brated to reproduce measurements of flow, water levels,
costs and providing the information necessary to
and velocity through a manual calibration procedure.
implement effective maintenance procedures, and
The reliability of the modeling is highly dependent on
• enhancing wastewater treatment plant performance the adequacy of the calibration procedure. Tradition-
by balancing inflow loads and allowing the plant to ally, the model is calibrated manually using a trial-
operate closer to its design capacity, thereby reducing and-error parameter adjustment procedure. At present,
the variability of the final effluent. however, automatic calibration support is being widely
discussed (16) as a replacement for the manual calibration
When the feasibility of RTC is evaluated for a sewer procedure, which is subjective and can be time-consuming.
system, it is important to be aware of the changes in Automatic methods for model calibration take advantage
flow pattern introduced by RTC. If RTC is applied during of the speed and power of computers, and are objective
dry weather flow, for example, to level out the flow to and relatively easy to implement. Automatic calibra-
a wastewater treatment plant, it must be investigated if tion supports are still emerging. So they need to be
the new flow conditions introduced by using RTC cause checked, and the calibration must be thoroughly evalu-
sediment deposits to develop in new and undesired places. ated by an experienced engineer because the automatic
If this is the case, then sediment deposits may increase the calibration procedure may stop at a local minimum of
risk of flooding, or the entire saving gained by applying the objective function, or it may not converge sufficiently
RTC may be spent on increased cleaning and maintenance. toward the chosen objective due to deficiencies in the
Neither would be a good outcome. measured data.

WATER QUALITY MODELING

MODELING AS A REAL-TIME DECISION SUPPORT TOOL
Modeling water quality in sewers is still an emerging area
Applying a model in conjunction with real-time data (e.g., because water quality and sediment transport processes
rain data) provides means for improving the information are difficult to describe precisely and the theory is still
about the current status of the urban drainage system. developing. At present, a number of practical operational
In general, rain gauges provide information only at problems in sewers push the research toward sustainable
specific locations. The application of a model provides urban drainage systems and practical solutions to handle
the possibility of filling information gaps between the significant problems such as hydrogen sulfide production.
rain gauges and computing the future impact of the rain. Hydrogen sulfide causes a noxious smell in the public
Models can be automatically executed in real time, for environment; it corrodes pipes and pumps in sewer
example, when the rainfall at specified locations exceeds networks; and poses a health risk to people working
preset threshold values. Based on the forecast by a inside sewers, for example, during maintenance. When
hydrologic model or an urban drainage model, the level of a sewer pipe collapses due to corrosion from hydrogen
information about the hydrologic system can be extended sulfide, it imposes huge costs on society to restore the
into the future. Real-time hydrologic information makes it pipe. Therefore, there is a need to understand and
possible to detect potential hazards shortly before or just manage the phenomenon of hydrogen sulfide formation,
after they happen, for example, heavy rainfall recorded and the emerging development of modeling tools for
in certain parts of the catchment that causes flooding, this problem makes it possible to assess causes and
or overflows, and actions can be taken to mitigate the alleviating measures.
problems in the drainage system. Another example is that
local traffic control can redirect traffic to roads not flooded MODELING THE INTERACTION BETWEEN URBAN
or provide decision support to the operators of pumps and DRAINAGE SYSTEMS, RECEIVING WATERS, AND
weirs in the system (15). WASTEWATER TREATMENT PLANTS
The application of real-time hydrologic information is
not the full solution to overflow and flooding problems, The future challenge within integrated modeling is to
but it may mitigate flood problems and in this way develop a consistent modeling formulation for water
reduce the economic losses and the stress on people quality in sewers, the wastewater treatment plant, and
and the environment. Finally, the information generated receiving waters, which can be carried across the three
by a real-time hydrologic information system can be subsystems. Furthermore, water quality processes in
applied by using the historical data for design and sewers are at present still an emerging area of research,
maintenance analyses to achieve better functionality of and need more research. Finally, legislation must be
the urban hydrologic system before the next heavy rain revised nearly all over the world to deal with permits
arrives. In this fashion, the system can also be used to to systems which are optimized from a holistic point of
train operators. view. Today, around the world, most legislation focuses
 Next Page

342 MODELING UNGAUGED WATERSHEDS

only on one single sub-system, independently of the MODELING UNGAUGED WATERSHEDS

rest. This will open optimization of the total urban
drainage system rather than optimizing each subsystem THORSTEN WAGENER
individually. The Pennsylvania State
University
BIBLIOGRAPHY University Park, Pennsylvania

1. Andrews, T. (2002). Building better integrated network data.

World Water Environ. Eng. July: 26.
INTRODUCTION
2. Andersen, H.S., Larsen, L.C., Lindberg, S., Tamasauskas, H.,
and Mark, O. (2004). MIKE URBAN—Integration of Mod-
Models of watershed hydrology are irreplaceable tools in
elling and GIS. 6th Int. Conf. Urban Drainage Mod-
elling—UDM’04, Dresden, Germany.
today’s research and practice. Their areas of application
are wide reaching, from water management and flood
3. Lindberg, S, Nielsen, J.B., and Carr, R. (1989). An inte-
grated PC-modelling system for hydraulic analysis of
forecasting, to their use as load models for water quality
drainage systems. 1st Aust. Conf. Tech. Comput. Water Ind.: studies. A vast number of these models have been
Watercomp ’89, Melbourne, Australia. developed since the 1960s and they differ in temporal
4. James, W. and Robinson, M.A. (1984). PCSWMM3, Version and spatial discretization, processes described, and in
3 of the runoff, executive and extended transport blocks the constituent equations used. However, there are also
adapted for the IBM-PC, Proc. Stormwater Water Qual. similarities that have consequences for the way these
Modelling Conf., sponsored by USEPA and Ontario Ministry models are commonly applied. All models of watershed
of Environment, Burlington, Ontario, Sept. 6–7, published hydrology aggregate the real hydrologic system on a
by CHI Hamilton, pp. 39–52. particular element scale in space and time. The spatial
5. Johansen, N.B., Linde-Jensen, J.J., and Harremoës, P. scale might vary from grid cells of tens of meters to
(1984). Computing combined system overflow based on models that treat the whole catchment as a single
historical rain series. Chalmers University of Technology, unit, and the temporal scale might vary, for example,
3rd Int. Conf. Urban Storm Drainage, Göteborg, Sweden.
from 15-minute intervals to monthly time steps. It is
6. Blanc, D., Kellagher, R., Phan, L., and Price, R. (1995). common to assume homogeneity of processes or watershed
FLUPOL-MOSQITO, models, simulations, critical analysis
characteristics on scales smaller than the one applied, that
and development. Water Sci. Technol. 32(1): 185–192.
is, parameters are assumed to be effective on a certain
7. Garsdal, H., Mark, O., Dørge, J., and Jepsen, S.-E. (1994).
scale (Fig. 1), although the heterogeneity of the real world
MOUSE TRAP: Modelling of water quality processes and
the interaction of sediments and pollutants in sewers.
on smaller scales is sometimes described by distribution
Specialized Int. Conf.: Sewer Phys. Chem. Biol. Reactor, functions (2). The characteristics of each model element,
Aalborg, Denmark. for example, storage or infiltration capacity, are described
8. Alam, J.M. (2003). Two-dimensional Urban Flood Modeling by parameters within the model. A common problem is that
for Real Time Flood Forecasting for Dhaka city, Bangladesh. the scale on which these characteristics can be measured
AIT Thesis no. WM-02-06. Asian Institute of Technology, are usually different, mostly smaller, than the scale of
Thailand. the model element. The effect of this difference between
9. Nania L., Gomez M., and Dolz, J. (2002). Analysis of risk model and measurement scale is that one has to revert
associated with urban runoff, case study: City of Mendoza, to alternative methods to estimate model parameters.
Argentina. 9th Int. Conf. Urban Drainage, Sept. 8–13, The usual approach is to observe the responses of the
Portland, OR. real hydrologic system, for example, streamflow, and
10. Schmitt T.G., Schilling, W., Saegrov, S., and Nieschulz, K.P. compare them to predictions of the model. The modeler
(2002). Flood risk management for urban drainage system then adjusts the model parameters, in a process usually
by simulation and optimisation. Global Solutions for Urban referred to as calibration, until model predictions and
Drainage, 9th Int. Conf. Urban Drainage, Sept. 8–13,
observations are as close as possible. Calibration can be
Portland, OR.
performed using manual or automatic techniques and the
11. Mark, O., Weesakul, S., Apirumanekul, C., Boonya Aroon-
available literature on this subject is immense. Between
net, S., and Djordjević, S. (2004). Potential and limitations of
3 to 10 years of observations are required for calibration,
1-D modelling of urban flooding. Accepted for J. Hydrol.
depending on the model complexity and the informational
12. Boonya-aroonnet, S., Weesakul, S., and Mark, O. (2002).
Modelling of urban flooding in Bangkok. 9th Int. Conf. Urban
content of the data (3). Shorter periods can suffice when
Drainage, Portland, OR. the data sufficiently trigger the response modes of the
13. Rauch, W. et al. (2002). Mathematical modeling of integrated
model (4).
urban drainage systems. Water Sci. Technol. 45(3): 81–94. Alternatives to model calibration have to be found
14. Foundation for Water Research. (1998). Urban Pollution
when no or insufficient time series of the variable under
Management (UPM) Manual, 2nd Edn. FR/CL 0009. investigation are available for this process. This is a
15. Mark, O. et al. (2002). A real-time hydrological information
common problem, even in countries that have extensive
system for Bangkok. Int. Conf. Urban Hydrol. 21st Century, measuring networks such as the United Kingdom that has
Oct. 14–18, Kuala Lumpur, Malaysia. more than 1400 gauging stations (5). It is also possible
16. Wangwongwiroj, N., Schlütter, F., and Mark, O. (2003). that sufficiently long streamflow time series are available
Principles and practical aspects of an automatic calibration but that the modeling objective is the prediction of a
procedure for urban rainfall-runoff models. Accepted for the different variable, for example, groundwater levels. Then,
J. Urban Water. a calibration with respect to the variable under study