River Meanders -

Theory of
Minimum Variance
By WALTER B. LANGBEIN and LUNA B. LEOPOLD

PHYSIOGRAPHIC AND HYDRAULIC STUDIES OF RIVERS

GEOLOGICAL SURVEY PROFESSIONAL PAPER 422-H

The geometry of a meander is that of a random

waZh whose most freguent form minimizes the
sum of the s p a r e s of the changes in direction
in each unit length. Changes in direction
close& approximate& a sine function of channel
distance. Depth, veZocity, and sZope are
aqusted so as t o decrease the variance of
shear and the friction factor in a meander over
that in an otherwise comparabZe straght reach
of the same river r

UNITED STATES GOVERNMENT P R I N T I N G OFFICE, WASHINGTON : 1966

 UNITED STATES DEPARTMENT OF THE INTERIOR

STEWART L. UDALL, Secretary

GEOLOGICAL SURVEY

William T. Pecora, Director

For sale by the Superintendent of Documents, U.S. Government Printing Office

Washington, D.C. 20402 - Price 20 cents
 d

CONTENTS

Page
i18
9
0
11
13
15

ILLUSTRATIONS
Page
FIGURE1. Most frequent random w a l k s _ - _ _ - _ - - _ _ _ - _ _ - - - - - _ - - _ - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -133
2. Plan view of channel and graph of direction angle #I as a function of distance along a meander- - - - - - - - - - - - - - - - - 3
3. Maps of meander reaches of several rivers and graphs showing relation between channel direction and chaniiel
distance-_---------------------------------------------------------------------------------------- 6
4. Map of laboratory meander and plot of thalweg direction against distance __________________________________ 8
5. Comparison of four symmetrical sinuous curves having equal wavelength and sinuosity--- - - - - - - - - - - - - - - - - - - - 9 _
6. Plan and profiles of straight and curved reaches, Baldwin Creek near Lander, Wyo- - _ _ _ _ _ _ _ _ - _ _ _ _ _ _ _ _ _ _ _ _ _ _ _10
7. Plan and profile of curved and straight reaches of Pop0 Agie River, yZ mile below Hudson, Wyo- - - - - - - - - - - - - - - 11
8. Plan and pro6le of straight and curved reaches on Pole Creek a t Clark’s Ranch near Pinedale, Wyo - - - - - - - - - - - 12
9. Relation of sinuosity to variances of shear and friction factor_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 14
_____------
10. Relations between velocity, depth, and slope in straight and curved reaches, Pole Creek, near Pinedale, Wvo----- 14

TABLES
Page
€I 13
14
14
 SYMBOLS
ratio of variances (table 3)
a coefficient (equation 1)
sinuosity, ratio of path distance to downvalley distance
radius of curvature of a bend
coefficient of correlation (ttible 3)
unit distance along path
a fitctor or variable
P probability of a particular direction (cqiiat,ion 1 )
U variance of a particulnr ftictor, T
TI water depth
S energy slope
M total path distance in a single wavelength
N number of measurements
V velocity
a a constant of integration (equation 2)
Y unit weight of water
x wavelength
?r pi, ratio of circumference to radius
P radius of curvature of path
U standard deviation
4J angle that path at a given point makes with mean downpath direction
w maximum angle a path makes with mean downpath direction
IV
 PHYSIOGRAPHIC AND HYDRAULIC STUDIES OF RIVERS
~~

RIVER MEANDERS-THEORY OF MINIMUM VARIANCE

B.~ LANGBEIN
By W A I ~ T E Z and LUNAB. LEOPOLD

ABSTRACT
and Wolman, 1960, p. 774). This leads t o visual
Meandcrs arc the result of erosion-deposition processes tending similarity regardless of scale. When one looks a t a
toward the most stable form in which t h e variability of certain
stream on a planimetric map without first glancing a t
csscntial properties is minimized. This minimization involves
the adjustmcnt of thc planimetric gcometry and thc hydraulic inap scale, it is not immediately obvious whether i t is a
factors of depth, velocity, and local slope. large river or a small stream.
The planimctric geomctry of a meandcr is that of a random Explanation of river meanders have been varied. It
walk whose most frequent form minimizcs the sum of thc squarcs has been suggested that meanders are caused by such
of the changes in direction in each successive unit length. The processes as-
direction anglcs arc then sine functions of channel distancc.
This yields a meander shape typically present in meandering regular erosion and deposition;
rivers and has the characteristic that the ratio of meander length secondary circulation in the cross-sectional plane;
to average radius of curvature in t h e bend is 4.7. seiche effect analogous to lake seiches.
Depth, velocity, and slope are shown by field observations to Attempts, however, to utilize these theories to calcu-
be adjusted so as to decrease the variance of shear and the friction late the forms of meanders have failed.
factor in a meander cwve over that in a n otherwise comparable
straight rcach of the same river. Although various phenomena-including some of
Since theory and observation indicate meandcrs achieve the those mentioned above, such as cross circulation-are
minimum variance postulated, i t follows that for channels in intimately involved in tlhe deviation of rivers from a
which alternating pools and riffles occur, meandering is the most straight course, the development of meanders is pnt-
probable form of channel geometry and thus is more stable ently related t o the superposition of many diverse
geometry than a straight or nonmcaiidcring alincment .
effects. Although each of these individual effects is
INTRODUCTION in itself completely deterministic, so many of them
So ubiquitous are curves in rivers and so common are occur that they cannot be followed in detail. As
smooth and regular meander forms, they have attracted postulated in this paper, such effects can be treated as
the interest of investigators from many disciplines. if they were stochastic-that is, as if they occurred-in
The widespread geographic distribution of such forms a random fashion.
and their occurrence in various settings had hardly been This paper examines the consequences of this pos-
appreciated before air travel became common. More- tulate in relation to (1) the planimetric geometry of
over, air travel has emphasized how commonly the form meanders, and (2) the variations in such hydraulic
of valleys-not merely the river channels within them- properties as depth, velocity, and slope in meanders as
assumes a regular sinuosity which is comparable to contrasted with straight reaches.
meanders in river channels. Also, investigations of the The second problem required new datn. These were
physical characteristics of glaciers and oceans as well as obtained during the snowmelt season of high discharge
landforms led to the recognition that analogous forms in the years 1959-64 by the junior author with the
occur in melt-water channels developed on glaciers and invaluable and untiring assistance of William W.
even in the currents of the Gulf Stream. Emmett and Robert M . Myrick. Leon W. Wiard was
The striking similarity in physical form of curves in with us for some of the work. T o each of them, the
these various settings is the result of certain geometric authors are very grateful for their important contri-
proportions apparently common to all, that is, a nearly bution, measured in part by the internal consistency
constant ratio of radius of curvature to meander length of the various field data, and the satisfactory closures
and of radius of curvature to channel width (Leopold of surveys made under difficult conditions.
€I 1
H2 PHYSIOGRAPHIC AND HYDRAULIC STUDIES OF RIVERS

MEANDER GEOMETRY PATH O F GREATEST PROBABILITY BETWEEN TWO

POINTS 1
Meandering in rivers can be considered in two con-
texts. the first involves the whole profile from any The niodel is a simple one; a river has a finite prob-
headwater tributary downstxeam through the main ability p to deviate by an angle d4 from its previous
trunk-that is, the longitudinal profile of the river sys- direction in progressing an eIementa1 distance ds along
tem. The second contest includes a meandering reach itrs pat'h. The probability distribut<ion as a function
of a river in which the channel in its lateral migrations of deviation angle may be assumed to be normal
niay take various planimetric forms or paths between (Gaussian) which would be described by
two points in the valley. d.p =c exp 51 (
&Ids 7)
The river network as a whole is an open system
tending toward a steady state and within which several
where SI is the standard deviation and where c is
hydraulically related factors are mutually interacting
defined by the condition that f dp= 1.O.
and adjusting-specifically, velocity, depth, width,
The actual meander path then corresponds to the
hydraulic resistance, and slope. These adjust to
most probable river path proceeding between two
accommodate the discharge and load contributed by
points A and B, if the direction of flow a t the point A
the drainage basin. The adjustment takes place not
and the length of the path between A and B is fixed
only by erosion and deposition but also by variation
and the probability of a change in direction per unit
in bed forms which affect hydraulic resistance and thus
river length is given by the probability distribution
local competence to carry debris. Previous applica-
above. This formulation of the problem of river
tion of the theory of minimum variance to the whole
meanders is identical to that of a class of random-walk
river system shows that downstream change of slope
problems which have been studied by Von Schelling
(long profile) is intimately related t o downstream
(1951, 1964). The Von Schelling solution demon-
changes in other hydraulic variables. Because dis-
strated that the arc length s is defined by the following
charge increasas in the downstream direction, minimum
elliptic integral:
variance of power expenditure per unit length leads
toward greater concavity in the longitudinal profile. s=-
1 J d2(a--cos
d4J
I$)
(2)
However, greater concavity is opposed by the tendency @

for uniform power expenditures on each unit of bed where 4 is the direction angle measured from the line
area. A B and CY is a constant of integration. It is con-
I n the context of the whole river system, a meandering venient to set
a=cos w (3 1
segment, often but not always concentrated in down-
stream rather than upstream portions of the system, in which w becomes the maximum angle the path makes
tends to provide greater concavity by lengthening the from the origin with the mean direction. Curves for
downstream portion of the profile. By increasing the w=4Oo, goo, l l O o , which all show patterns character-
concavity of the profile, the product of discharge and istically seen in river meanders, are given in figure 1
slope, or power per unit length becomes more uniform (after Von Schelling, 1964, p. 8). Von Schelling (1951)
along a stream that increases in flow downstream. showed that a general condition for the most frequent
Thus the meander decreases the variance of power per path for a continuous curve of given length between
unit length. Though the occurrence of meanders affects the two points, A and B, is
the stream length and thus the river profile, channel
slope in the context of the whole river is one of the -=a
As miiiinium,
(4)
P2
ndjiistable and dependent parameters, being determined
by miitual interaction with the other dependent factors. where As is t i unit distance along the path and p is the
I n the second context of any reach or segment of radius of curvature of the path in that unit distance.
river length, then, average slope is given, and local But since
changes of channel plan must maintain this average p=(%) (5)
slope. Between any two points on the valley floor,
however, a variety of paths are possible, any of which where A+ is the angle b y which direction is changed
would maintain the same slope and thus the same in distance As, therefore,
length. A thesis of the present paper is that the
meander form occupies a most probable among all == As
arninimum.
possible paths. This path is defined by a random-walk I Acknowled@nents me due A. E. SCheideggrr for his assistsncc in clsrifying the
model. iiiathciiiatical relationships of this section.
 RIVER MEANDERS-THEORY O F MINIMUM VARIANCE €13
Since the sum of all the directional changes is zero, or
CA+=O. (7)

equation G is the most probable condition in which trhe

variance (mmn square of deviations in direction) is
ininitnum.

w= 90“ w= 110“ W

l.-Exa~nplcs of most frequent randonl-walk paths of given lengths between

FIGURE E DISTANCE ALONG CHANNEL
two specifled points on n plane. Adapted from H. von Schelling.
5.-Theoretical ineander In plan view (A) and a plot 01 its direction angle
FIGURE +
THE BINE- GENERATED CURVE as B function of distance along the channel path ( B ) .

For graphing meanders it is easier t o make use of the Inasmuch as the ratio w/1/2(1-cos~) is nearly con-
approximation that. stant (=1.05), over the range of possible values of w,
meander path lengths, 44, are inversely proportional t o
the standard deviation U ; thus, 34=6.6/u.
Equation 10 defines a curve in which the direction C$
is a close approximation for is a sine function of the distance s along the meander.
Figure 2 shows a theoretical meander developed from R
sine function. The meander thus is not a sine curve,
but will be referred to as a “sine-generated curve.”
where angles w and 4 are deviations from the central Thus meander loops are generated by the sine curve
axis with the downstream direction as zero. With defined by equation 10, and the amount of horseshoe
this simplification looping dependsron the value of w . This means that a t
+=u sin
O J 3 ( 1 -cos w )
w
relative distance a equal to
S
and 1, 4 has a value o l
zero or the channel is locally directed in the mean
S
direction. At distance T~equal to )$ and :i the d u e
of 4 has its largest value w , as c m be seen in figure 2.
wlirre A4 is tlie tot,iil p& d i s t u c e along LL meander, The graph on figure 2A has been const,ructed for L:
and where w is the maximum angle the path makes value of w= l l O o , and corresponds almost exactly to
from the mean downvalley direction. The maximum the 110’ curve of figure 1 calculated by the exact
possible value of w is 2.2 radians, for at this angle equations of Von Schelling.
meander limbs intersect. When w equals 2.2 radians,
It will be noted that the plan view of the channel
the pattern is a closed figure “8.”
(fig. 2.4) is not sinusoidal; only the channel direction
Comparison of equations 9 and 10 indicates that changes as sinusoidal function of distance (fig. 2B).
The meander itself is more rounded and has a relatively
uniform radius of curvature in the bend portion. This
can be noted in the fact that a sine curve has quite a
all angles being in radians, as before. straight segment as it crosses tho x -axis.
H4 PHYSIOGRAPHIC AND HYDRAULIC STUDIES O F RIVE,RS

The tangent to the sine function in figure 2B a t any Figures 3 0 and 3P are among the best known incised
point is A4JAs which is the reciprocal of the local meanders in the western United States, where the
radius of curvature of the meander. The sine curve is Colorado River and some of its tributaries are a thou-
nearly straight as it crosses the zero axis. Therefore, sand or more feet below the lowest of the surrounding
the radius of curvature is nearly constant in a meander benches, the cliffs rising a t very steep angles up from
bend over two portions covering fully’a third of the the river’s very edge. Even these provide a plot of
length of each loop. direction versus length which is closely approximated
There are, of course, plenty of irregular meanders by a sine curve.
bul those which are fairly uniform display a close ac- By far the most syriiinetrical arid uniform riieander
cordance to a true sine curve when channel direction reach we have ever seen in the field is the Pop0 Agie
is plotted against distance along the channel, as will be River near Hudson, Wyo., for which flow data will be
shown. discussed later. The relation of channel direction to
distance is compared with a sine curve in the lower part
ANALYSIS O F SOME FIELD EXAMPLES
of figure 3F.
Some field examples presented in figure 3 will now be The laboratory experiments conducted by Friedkin
discussed briefly with a view to demonstrating that (1945) provide an example of near-ideal meanders.
atiiong a variety of meander shapes the sine-generated The meander shown in figure 4 is one of those that
curve fits the actual shape quite well and better than developed greatest sinuosity. The channel shown on
a1t ernatives. figure .$Awas developed in Mississippi River sand on :t
Figure 3 4 upper part, shows the famous Greenville “ v d e y ” slope of 0.006, by rates of flow that T-ariecl
bends a t Greenville, Miss., before the artificial cutoffs from 0.05 to 0.24 cfs over an 1s-hour period on a
changed the pattern. The crosses in the lower diagram schedule that simulated the fluctuations of discharge
represent values of channel direction, +-relative to a on the Mississippi River. The sand bed was homo-
chosen zero azimuth, which for convenience is the mean geneous and the meander mapped by Friedkin is more
downstream direction-plotted against channel dis- regular than those usually observed in natural streams.
tance. The full curve in the lower diagram is a sine The direction angles of the thalweg were measured a t
curve chosen in wavelength and amplitude t o approx- intervals of 2.5 feet. These direction angles are
imate the river data. A sine curve that was fitted to plotted on the graph, figure 4B. Since each meander
tlie crosses was used to generate a plot of channel had a thalweg length of 50 feet, the data for the two
direction against distance which has been superimposed meanders are shown on the same graph. Where the
on the map of the river as the dashed line in the upper two bends had the same direction only one point is
diagram. It can be seen that there is reasonable agree- plotted. The data correspond closely to the sine
ment. curve shown.
The same technique has been used in the other ex-
amples in figure 3 chosen to cover a range in shape from COMPARISON O F VARIANCES O F DIFFERENT
MEANDER CURVES
oxbows t o flat sinuous form, and including a variety of
sizes. Figure 3B shows a stream that not only is not As a close approximation t o the theoretical minimum,
as oxbowed as the Greenville bends but also represents when a meander is such that the direction, 4, in a g k e n
a river of much smaller size, the Mississippi River unit length As is a sine function of the distance along
being about three-quarters of a mile wide and Blackrock the curve, then the sum of squares of changes in direction
Creek about 50 feet wide. from the mean direction is less than for any other
Figure 3C shows the famous Paw Paw, W. Vtt., bends common curve. I n figure 5, four curves are presented.
of the Potomac River, similar to and in the same area Or~econsists of two portions of circles which have
as the great meanders of the tributary Shenandoah been joined. Another is a pure sine curve. Thc
River. The curves in these rivers are characterized by third is a parabola. A fourth is sine-generated in
being exceptionally elongated in that the amplitude is that, as has been explained, the change of direction
unusually large for the wavelength and both are large bears R sinusoidal relation to distnnce along the curve.
for the river width, characteristics believed to be in- All four have the same waveleiigth and sinuosity-
fluenced by elongation nlong the direction of a fracture the ratio of curve length to straight line distance.
system in the rock (Hack and Young, 1959). Despite However, the sums of squares of the changes in direction
these peculiarities the sine-generated curve fits the as measiired in degrees over 10 equally spaced lengths
river well. along the curves differ greatly as follows:
 R N E R MEANDERS-THEORY OF MINIMUM VARIANCE H5
Curve z (A+)2 has the dimension of L-I or the reciprocal of length.
Parabola- - - _- - - - - - - - - - - - - - - - - - - - - - - - - - 5210 As previously shown, u is inversely proportional to
Sine c u r v e _ _ _ - - - - - _ _ - - - - - - - - - - - - - - - - - - - 5200 meander length.
Circular curve--- - - - - - - - - _ _ - - - - - - - - - - - - - 4840 Bend radius is related to wave length and is vir-
Sine-generated curve.. - - - - - - - - - - - - - - - - - - - 3940 tually independent of sinuosity. Defined as before,
The “sine-generated” curve has the least sum of squares. as the average over the yS of channel length for which I$
The theoretical minimum curve is identical within is nearly linearly related to channel distance, bend
practical limits of drawing. radius R is
MEANDER LENGTH, SINUOSITY, AND BEND RADIUS
The planimetric geometry of river meanders has been
Since 4 ranges from S 0 . 5 t~o - 0 . 5 ~ over this near-
defined as follows
+=w sin -& 21r
(10)
linear range, A#=@. Substituting for w its algebraic
equivalent in terms of sinuosity,
where 4 equals the direction a t location s, w is the max-
imum angle the meander takes relative to the general
w=2.2 -J”l-
ci
direction of the river, and M is the channel length of a and since M= kX, bend radius equals
meander. The values of w and M can be defined further
as follows:
The angle u is a unique function of the sinuosity and
is independent of the meander length. It ranges from Some typical values for bend radius in terms of
zero for a straight line path of zero sinuosity t o a max- sinuosity are
imum of 125’ for “gooseneck” meanders a t point of in-
Bend radius waoelength
cipient crossing. The sinuosity, k, equals the average k (R) Ratio bend radius
of the values of cos 4 over the range from 1$=0 t o 4 = w . 1. 25 _ _ . _ _ _ - _ 0. 215X 4.6_
_ _ _ _ _ _ _ _
Thus a relationship can be defined between k and w . An 1. 5 _ _ _ _ _ _ _ _ - _ _ _ _ _ _ _ _ _ .20X 5. 0
approximate algebraic expression is 2. 0 _ _ _ - _ - _ _ - _ - _ _ _ _.22X ___ 4. 5
2. 5 . _ _ _ _ _ _ _ .24X 4. 2 _ _
_ _ _ _ _ _ _ _

o (radians) =2.2@
These values agree very closely with those found
or by measurement of actual meanders. Leopold and
w=125O
4-.7 Wolman (1960) measured the meander characteristics
of 50 samples, including rivers from various parts of
Sinuosity as measured by parameter IC, as explained, is the United States, that range in width from a few
thought to be a consequence of the profile development feet to a mile. For this group, in which sinuosities
which is only secondarily influenced by reaction from were dominantly between 1.1 and 2.0, the average
the meander development. ratio of meander length to radius of curvature was
I n the random-walk model, the standard deviation 4.7, which is equivalent to .213X. The agreement
of changes in direction per unit of distance is u which with values listed above from the theory is satisfactory.
s
3 Relation of o to sinuosity, k.
The curve defined by 4=w sin 360° is believed

kx-
M t o underlie the stable form of meanders. That actual
2 cos +As meanders are often irregular is well known, but as
where
+=m sin 2 2s.
observed above, those meanders that are regular in
M geometry conform to this equation. Deviations (or
With an assumed value of W , values of + atj24 equally fspaced~intervalsof
3/M were computed. The reciprocal of the average value of cos q5 equals ‘hoise”), it is surmised, are due t o two principal
the sinuosity, k. causes: (1) shifts from unstable to a stable form caused
Values of o and k so computed were as follows:
by random actions and varying flow, and (2) non-
O
(radians) k homogeneities such as rock outcrops, differences in
alluvium, or even trees.
The irregularities are more to be observed in “free”
T h e values of L approximate the following
(that is, “living”) meanders than in incised meanders.
k=4.84/(4.84-d) During incision, irregularities apparently tend to be
or

a See footnote 2, this page.

789-282-66-2
HG PHYSIOGRAPHIC A N D HYDRAULIC S T U D I E S OF RIVEhRS

0
- 1 2 MILES
0
u
50 100 FEET

DISTANCE ALONG CHANNEL, I N FEET

DISTANCE ALONG CHANNEL, IN MILES

G
X D INCISED COLORADO RIVER AT THE LOOP,UTAH
1 2MILES <
150-
_I

WEST VIRGINIA

U
z
4

b - 0
=E
E
CI 100 i
-
DISTANCE ALONG CHANiCEL I N THOUSANDS OF FEET
5
DISTANCE ALONG CHANNEL IN rHOlJSANDb OF K E l 0
FIGUBE
3.-MAPS OF RIVER CIIANNELB A N D
A Crpper. Map o f channel compared l o sine-generated curve (dashrd); lower, plot of actual channel dircction against distance (crosses) and a sine curve (full line).
B Upper, Mapof channel eouipared to sine-generated curve (dashed); lower, plot of actual channel direction against distance (crosses) and a sine curve (full lhe).
C, Upper, Ma polchannel L . v ~ ~ > "t To sine-gonerntcd
c~ curvc (dashed); lower, plot of actual channel directionagainst distance (crosses) and a slne curve (full line).
D , Upper. Planimetric mal) of rivcr; lower, plot of actual chanuel direction against distance (crosses) and a sine curve (full line).
 RIVER MEANDERS-THEORY OF MINIMUM VARIANCE H7

10

E . l N C l S E D SAN JUAN RIVER

AT THE GOOSENECKS, UTAH

channel

-I-

I I I I
5+00 10+00 15+00 20 t o o
I
25+00
I
30+00 3 00 +
40 00
STATIONING ALONG CHANNEL, I N FEET

PLOTS OF CHANNEL DIRECTION AQAINST DISTANCE

E , Right, Planimetricmap ofriver; Zeft, plot of actual channel direction against distance (crosses) and a sine curve (fullline). F, Upper, Planimetric nlap of river;
lower, plot of actual channel direction against distance (crosses) and a sine curve (full Itne).
H8 PHYSIOGRAPHIC AND HYDRAULIC STUDIES O F RIVERS

into the Missouri and Colorado River systems from

the Wind River Range in central Wyoming, where good
weather for field work generally exists a t the base of
the mountains during the week or two in late spring
or early summer when snowmelt runoff reaches its peak.
Some of the smaller streams presented the difficulty
that the area of study lay 15 to 20 miles from the far-
in feet
thest watershed divide-a distance sufficient to delay
I
the peak discharge by 9 t o 12 hours. Maximum snow-
melt in midafternoon caused a diurnal peak flow which
0 =65" Sin (S/50x360") reached the study area shortly after midnight and
meant surveying and stream gaging by flashlight and
lantern illumination of level rod and notebook.
The plan of field work involved the selection of one
E
8 or more meanders of relatively uniform shape which
-100 I I I I occurred in the same vicinity with a straight reach of
0 10. 20 30 40 50 river a t least long enough to include two d e s separated
B DISTANCE ALONG CHANNEL, IN FEET by a pool section. No tributaries would enter between
FIGURE
4.-Mapoflaboratorymeanderand plot ofthalweg direction against distance.
the curved and straight reaches so that a t any time
A, Plan view; B, Comparison of direction angles with sine curve. discharge in the two would be identical. It was
assumed and checked visually, that the bed and bank
averaged out and only the regular form is preserved, materials were identical in the two reaches as well.
and where these intrench homogeneous rocks, the Later bed sampling showed that there were some
result is often a beautiful series of meanders as in differences in bed-material size which had not been
the San Juan River or the Colorado River. apparent to the eye.
A curve introduces an additional form of energy dis-
MEA NDERS COMPARED WITH STRAIGHT REACHES sipation not present in a comparable straight reach,
Meanders are common features of rivers, whereas an energy loss due to change of flow direction (Leopold
straight reaches of any length are rare. It may be and others, 1960). This additional loss is concentrated
inferred. therefore, that straightness is a temporary in the zone of greatest curvature which occurs midway
state. As the usual and more stable form, according between the riffle bars. The additional loss is a t a
to the thesis of this report, the meander should be location where, in a straight reach, energy dissipation
characterized by lower variances of the hydraulic fac- is smaller than the average for the whole reach.
tors, a property showfi by Maddock (Written commun., The introduction of river bends, then, tends to equal-
1965) to prevail in self-formed channels. ize the energy dissipation through each unit length,
The review of the characteristics of meanders and but does so a t the expense of greater contrasts in bed
the theories which had been advanced to explain them elevation. Moreover, curvature introduces a certain
showed that quantitative data were nearly nonexistent additional organization into the distribution of channel-
on many aspects of the hydraulics of flow during those bed features. The field measurements were devised to
high stages or discharges a t which channel adjustment measure and examine the variations in these quantities.
actively takes place. Though more will be said about individual study
Because there are certain similarities and certain reaches when the data are discussed, a salient feature
differences between straight segments or reaches of was apparent which had dot previously been noticed
river and meandering reaches, it also seemed logical to when the first set of such field data was assembled; this
devote special attention to comparison of straight and feature can be seen in figure 6. Where the profiles of
curved segments during channel-altering flow. An bed and water surface are plotted on the same graph
initial attempt to obtain such data was made during for comparison, it is apparent that a t moderately high
1954-58 on a small river in Maryland, near enough to stage-from three-fourths bankfull to bankfull-the
home and ofice to be reached on short notice during larger-scale undulation of the riverbed caused by pool
the storm flows of winter. Though something was and shallow has been drowned out, as the streamgager
learned, it became clear that rivers which rise and fall would say, or no longer causes an undulation in the
in stage quickly do not allow a two-man team with water-surface profile of the meander. I n contrast, a t
modest equipment t o make the desired observations. the same stage in the straight reach the flat and steep
Attention was then devoted to the rivers which drain alternations caused by pool and riffle are still discernible.
 RIVER MEANDERS-THEORY O F MINIMUM VARIANCE €19
On the other hand, the profile revealed that the undula- curve does not have a high probability of occurrence,
tion of the bed along the stream length is of larger and to find such a combination, airplane photographic
magnitude in the curved than in the straight reach. traverses a t about 1,000 feet above the ground were
Considering the features exemplified in figure 6, one flown for some distance along the east front of the
could reason as follows. In a straight reach of channel Wind River Range.* Even after reaches were chosen
Ilic dunes, bars, pools, and rimes form more or less from the photographs, ground inspection resulted in
independently of the ch:tnnel pat tern owing to grain discard of significant proportions of possible sites.
interaction. T o the occurrence of pool and riffle the When a satisfactory site was located, a continuous
channel must adjust. The riffle cnuscs a zone of greater- water-stage recorder was established; benchmarks and
thm-average stcepness which is also a zone of greater- staff gage were installed and the curved and straight
than-average energy cspeaditure. The pool, on the reaches mapped by plane table. Cross sections were
other hand, being relatively free of large gravel on the staked a t such a spacing that about seven would be
bed and of larger depth relative to bed roughness ele- included in a length equal to one pool-and-riffle
ments, offers less resistance to flow and there the energy sequence.
expenditure is less than average. The result is a Water-surface profiles were run by leveling a t one or
stcpped water-surface profile-flat over pool and steep more stages of flow. Distances between water-surface
over rime. shots were equal for a given stream; IO-foot distances
Once pools and riffles form with their consequent were used on small streams having widths of 10 to 20
vnriations in depth, width remaining rclatively uniform, feet and 25-foot spacing used for streams 50 to 100
then slopes niust vary. If slopes vary so that bed shear feet wide. Usually two rods, one on each streambank,
rcmaiiis constant, then slope would vary inversely with accompanied the instrument. Shots were taken op-
depth. However, this result would require that the posite one another along the two banks. Water-
friction factor vary directly as the square of the depth. surface elevations were read to 0.01 foot on larger
If, on the other hand, the friction factor remained uni- rivers, and to 0.001 foot on small ones. For the small
form and the slope varied accordingly, then bed shear rivers an attachment was used on a surveying rod
would vary inversely as the square of the depth. Uni- which in construction resembled a standard point gage
formity in bed shear and the friction factor is incompat- used in laboratory hydraulic practice for measurement
ible, and the slope adjusts so as to accommodate both of wa ter-surf ace elevation.
a1)out equably and minimize their total variance. Velocity measurements by current meter were
usually made a t various points across the stream a t
FIELD MEASUREMENTS each cross section or a t alternate ones. I n the small
The general program of field measurements has been streams these current-meter measurements were made
described. The main difficulty experienced was that by wading, those in the larger rivers from a canoe. For
the combination of conditions desired occurs rarely. canoe measurements a tagline had to be stretched
A straight reach in proximity to a regular meander across the river a t the staked cross sections, an opera-
tion fraught with some difficulty where brush lined the
banks and when the velocity was high a t near-bankfull
stage.
Figures 6, 7, and 8 show examples of the planimetric
maps and bed and water-surface profiles. The sinus-
oidal change of channel direction along the stream
length for figure 7 was presented in figure 3.
REDUCTION OF DATA
Mean depths and mean velocities were calculated
for each cross section. The average slopes of the water
surface between the cross sections were also computed
from the longitudinal profile. T o reduce these quan-
tities to nondimcnsional form, they were expressed in
ratio to the respective means over each reach. The
variances of these ratios were computed by the usual
formula
~ ~~

6.-Coniparison of four synimet,rical silluous curvcs IiavitIg equal wive-

FIGURE The authors acknowledge the assistance of Herbert E. Skibitzke’and David
4 E.
length and sinuosity. Jones in the photo reconnaissance.
Ill0 PHYSIOGRAPHIC AND HYDRAULIC STUDIES OF RIVE,RS

-(+y (11) As shown by Maddock (1965), the behavior of many

movable-bed streams can be explained by a tendency
where u22 is the variance of the quantity X and N is toward least variation in bed shear and in the friction
number of measurements in each reach. factor. Accordingly, the variances of the bed shew

100 0 100 200 300FEEl

M

i
1

-
0 50 100FEEl

0 1 CO 200 300 400

DISTANCE ALONG CHANNEL. IN FEET
H
FIGTJHE R.--I’Ian and profiles o f straight, and curved reaches, Baldwill Creek m a r Lmdcr. Wyo. A . Planimetric map showing location of rcilchcs;
B, Profiles of centerline of bed and average of water-surfacc elevations ob the two sidcs of the strcam; curved rcneh is dashed line and straigb~rcuch
is full line.
 RIVER MEANDERS-THEORY OF MINIMUM VARIANCE H11

15

o Gravel b a r s a s of s t a g e = 3 6 5

and of the friction factor-again considered as ratios to Width is not included in the analysis because it is
their respective means-were also computed. Bed shear is relatively uniform, and has no characteristic that dis-
equal to the product rDS, where y is the unit weight tinguishes it between curved and straight reaches.
of water, D is the depth, and S is the slope of the energy Table 1 shows a sample computation for Pole
profile. The unit weight of water is a constant that Creek, and table 2 lists a summary for the five streams
may be neglected in this analysis of variances. I n this surveyed.
study water-surface slopes will be used in lieu of slopes
COMPARISON OF RESULTS
of the energy profile. This involves an assumption
that the velocity head is small relative to the accuracy The results are listed in table 2. It will be noticed
of measurement of water surface. I n any case, where that the vmiance of slope in each reach is larger than
the velocity-head corrections were applied, they were that of depth or velocity. The contrast, however, is
either small or illogical, and so were neglected. less in the meandering than in the straight reach; this
Similarly, the variance of the quantity DS/vz, which reduction may be noted by the figures in the column
is proportional to the Darcy-Weisbach friction factor, headed a, which is the ratio of the variance of slope to
was computed. the sum of those of depth and velocity. Among the
H12 PHYSIOGRAPHIC AND HYDRAULIC STUDIES OF RIVERS

A
From planetable map by Leopold and Emmett

, n\o+oo
-
0 100 200 300FEET

18+00\ ‘
>
Lz B
a
Em a I I I I I I I I I I I I I I I
E 4.00 5+00 10+00 Stationing along straight reach
a
z+ 6 -
6
D
&-4
W
w
LL

z 2
z -
2
5 0
2
1 otoo 5tOO 10+00 Stationing along curved reach 15+00
w
DISTANCE ALONG CHANNEL. IN FEET

8.-Plan and profile of straight and curved reaches on Pole Creek at Clark’s Ranch near Pinedale, Wyo. A , Planimetricmapshowing location of
FIGURE
reaches; B, Profiles of centerline of bed and average of water-surface elevations on the two sides of the stream; curved reach is dashed line and straight
reach is a full line.

five rivers studied, this ratio is uniformly less in the This improved correlation is shown in the following
meandering reaches than in the straight reaches. tabulation (table 3) of the respective coefficients of
The data listed in table 2 also show that the variance determination (= squareof the coefficient of correlation.)
in bed shear and in the frict.ion factor are uniformly Figure 10 shows graphical plots for the Pole Creek
lower in the curved than in the straight reach. The data listed in table 2. The graph shows t,l.iat in the
data on table 2 suggest that the decrease in the variance meanders, depths, velocities, and slopes are more
of bed shear and the friction factor is related to the systematically organized than in straight reaches.
sinuosity. As shown on figure 9, the greater the To summarize, the detailed comparisons of meander-
sinuosity the less the average of these two variances. ing and straight reaches confirm the hypothesis t,hat in
When one considers, for example, that the variances a meander curve the hydraulic parameters are so ad-
of depth and velocity are greater in the curved reach, justed that greater uniformity (less variability) is estab-
and even that the variance of slope may be greater (as lished among them. If a river channel is considered in
for example Pole Creek), the fact that the product DS a steady state, then the form achieved should be such
and the ratio DS/V2have consistently lower variances as to avoid concentrating variability in one aspect a t the
must reflect a higher correlation between these vari- expense of another. Particularly a steady-state form
ables in tho meandered than in the straight reach. minimizes the ~ariationsin forces on the boundary-
 RIVER MEANDERS-THEORY O F M I N I M U M VARIANCE H13

Depth Velocity ( u ) Mean dcpth Mean vclocity Slope (S) in reach

(1)) in reach (u) in reach - - 1)s (ratio

1
DS/u’(ratio
Station (bet) (ratio to (ratio to to mean) to mean)
Ratio to Feet per Ratio to mean) mean) Feet per Ratio to
sec. mean 1,000 ft. mean

Straight reach

1
___
0. 79 3. 67
’’ 30 0. 98 1. 21
I 1. 67
I 1. 20
I 1. 18 0. 81
1. 17
1. 00
3. 26
2. 44
1. 12
. *9 1 1. 08 1. 01 2. 14 1. 55 1. 67 1. 64

1. 05
.93
2. 44
3. 52
. 87
1. 25
1. 02
.99
.90
no

1.06
1. 01
1 1 1.80
2. 24
l:A:
oe

.58
1. 46
1. 10
.52
1. 43
.88 2. 12 .76
1. 04 .3s
1. 19 2. 26 . 80
-- _-__
2. 81 - _ _ - _ _ _ _ . _ _ _ _ _ _ _ _ . -- --- - --
.042 __________ .21

Curved reach

2. 68 1. 23 2. 24 0. 82
1. 06 0. 88 1. 03 0. 50 0. 53 0. 69
1. 96 .90 2. 45 .90
. 78 1. 08 2. 20 1. 07 .84 . 72
1. 44 .66 3. 46 1. 27
.88 1. 08 2. 00 .98 . 86 .74
2. 37 1. 09 2. 45 .90
1. 42 .75 . 70 .34 . 48 .85
3. 77 1. 74 1. 63 .60
1. 32 . 89 1. 14 . 56 .74 1. 22
1. 95 . 90 2. 62 . 96
.74 1. 23 3. 94 1. 92 1. 42 .94
1. 25 .58 4. 10 1. 50
.58 1. 49 3. 97 1. 94 1. 13 .51
1. 25 .58 4. 02 1. 48
.83 1. 15 3. 17 1. 55 1. 29 . 98
2. 34 1. 08 2. 20 .81
1. 16 .77 .27 .13 . 15 .25
2. 67 1. 23 1. 98 .73
2. 17 I 2. 05
------.

that is, the steady-state form tends toward more uni- Subsequent work resulted in the postulate that the
form distribution of both bed stress and the friction behavior of rivers is such as to minimize the variations
f RC tor. in their several properties (Leopold and Langbein,
1963), and the present work shows how this postulate
INTERPRETIVE DISCUSSION
applies to river meanders as had been prognosticated.
After a review of data and theories concerning river Total variability cannot be zero. A reduction in the
meanders Leopold and Wolman (1960) concluded that, variability in one factor is usually accompanied by an
meander geometry is “related in some unknown manner increase in that of another. For example, in the mean-
to a more general mechanical principle” (p. 774). The der, the sine-generated curve has greater variability in
status of knowledge suggested that the basic reason for changes in direction than the circle. This greater vari-
meandering is related t o energy expenditure, and they ability in changes in direction is, however, such as to
concluded (p. 788) as follows. “Perhaps abrupt dis- decrease the total angular change. The sine-generated
continuities in the rate of energy expenditure in a reach curve, as an approximation to the theoretical curve,
of channel are less compatible with conditions of equilib- minimizes the total effect.
rium than is a more or less continuous or uniform ratc The meandering river has greater changes in bed
of energy loss.” v contours than a straight reach of a river. However,
H14 PHYSIOGRAPHIC AND HYDRAULIC STUDIES O F RIVElRS

TABLE2.-Sinuosity and variances for straight and curved reaches on Jive streams
Sinuosity: ratio of patli distance to downvalley distance. Variances: u
:
, depth; a:, velocity; uz, slope; &, bod shear; uj, Darcy-Weisbach friction facton

Stream Sinuosity (k) / ? I a t l u i l a

Straight reaches

Pole Creek near l'inedalc, Wyo..-- - _ _ - - - - _ _ - - - 1. 05 0.019 0. 042

0. 26 4.2 0. 25 0. 21
Wind River near Dubois, Wyo _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 1. 12 .03 . 017
.16 3. 4 .I4 * 11
Baldwin Crcck near Lander, Wyo _ _ _ _ _ - - - - -.- - 1. 0 .065 .058
.94 7. 8 .55 . 40
Pop0 Agie River near Hudson, Wyo _ _ _ _ _ _ _ _ _ _ 1. 02 . 026 . 70
.017 16. .46 .30
Mill Creek near Ethetc, Wyo _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 1. 02 . 053 ______.2.8 ___2. 6 .32 __________
Curved reaches

Pole Creek near Pinedalc, Wyo _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 2. 1 0.067 0.088 0. 41 2. G 0. 15 0. 07

Wind River near DuBois, Wyo _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 2. 9 .074 .032 .
13 1. 2 .
06 .Oi3
Baldwin Creek near Lander, Wyo _ _ _ _ _ - - - - - - - - 2. 0 .048 .017 .20 3. 1 .15 . 14
Pop0 Agie River near Hudson, Wyo _ _ _ _ _ _ _ _ _ _ 1. 65 . 044 .029 .38 5. 2 .28 . 12
Mill Creek ncar Ethete, Wyo _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 1. 65 .145 _ _ _ _ _ _ _ _ _ _ .64 2. 2 .21 ____--.-

3.0 n
z
0
8 4
v,
oc Curved reach
W
a X
2 X
t;
U
W
> X r =0.95
r, r =0.52
ffl z 2
2 2.0 i
i- Straight
z
ffl
o
0
reach
d 1
> 1 2 0 1 2 3
WATER-SURFACE SL.OPE, IN FEET PER HUNDRED FEET

Zq
JStraight reaches

i
% , "
I " I 3
1.o I I ..
0.1 0.2 0.3 0.4 0.5
VARIANCE OF SHEAR AND FRICTION FACTOR t,
f 2
PICURE
9.-Rrlation of inrnndcr sinuosity to arcrage of vnri:?ncm of friction hctor
z r = -0.21
and shear. r
k
w 1
n
reach
3.-Comparison
TABLE of coeficients of determination
0 4
1 2
Lris, coefflclent of linear detorrnination betmeen dopth and slolm; &, coefficient of
linear determination between velocity and slope] WATER-SURFACE SLOPE, IN FEET PER HUNDRED FEET
~ _ _ _ _ _ _ _
1 Straight reach I Curved reach Fiaum lO.-Relntions between velocity, depth, and slope, in straight and curved
roaches, Pole Creek, near Pincdale, Wyo. r=correlntlon coefficient.

0.72 0.90
.56 .42 these changes in depth produce changes in the slope of
.42
.58
.73
.62
the water surface and so reduce the variability in bed
.88 - - - -....- shear and in the friction factor, as well as to lessen the
contrastabetween the variances of depth and slope.
 RIVER MEANDERS-THEORY OF MINIMUM VARIANCE €115
These considerations lead to the inference that the REFERENCES
meandering pattern is more stable than a straight reach
Ilaclc, J. T., and Yoiing, R. S., 1959, Intrenched ineanrlers of the
in streams otherwise comparable. The meanders them- North Fork of the Shenandoah River, Vn.: U.S. Gcol.
selves shift continuously; the meandering behavior is Survey Prof. Paper % P A , p. 1-10.
stable through time. Langbein, W. B., 1964, Geometry of river channels: Jour.
This discussion concerns the ideal case of uniform Hydro. Div., Am. Soc. Civil Engineers, p. 301-312.
lithology. Nature is never so uniform and there are Leopold, L. B., Bagnold, R. A., Wolman, M. G., and Brush,
changes in rock hardness, structural controls, and other L. M.,1960, Flow resistance in sinuous or irregular chan-
heterogeneities related to earth history. Yet, despite nels: U.S. Geol. Survey Prof. Paper 282-D, p. 111-134.
these natural inhomogeneities, the theoretical forms Leopold, L. B., and Langbein, W. B., 1962, The concept of
entropy in landscape evolution: U.S. Gcol. Survey Prof.
show clearly. Paper 500-A, 20 p. se 75
This does not imply that a change from meandering Lcopold, L. B., and Wolinan, M.G., 1957, ltiver channel pat-
to straight course does not occur in nature. The in- terns-braided, meandering, and straight: U.S. Geol. Sur-
ference is that such reversals are likely to be less com- vey Prof. Paper 282-B, 84 p.
mon than t.he maintenance of a meandering pattern. 1960, River meanders: Bull. Geol. Soc. Am., v. 71, p.
The adjustment toward this stable pattern is, as in 769-794.
other geomorphic processes, made by the mechanical Von Schelling, IJermann, 1951, Most frequent particle paths in
effects of erosion and deposition. The theory of mini- a plane: Am. Geophys. Union Trans., v. 32, p. 222-226.
mum variance adjustment describes the net river be- 1964, Most frequent random walks: Gen. Elec. Co.
havior, not the processes. Rept. 64GL92, Schenectady, N.Y.

U . S . GOVClllVMENT P R I N T I N G 0 F F I C E : l Q S S