SWAT: MODEL USE, CALIBRATION,

AND VALIDATION
J. G. Arnold, D. N. Moriasi, P. W. Gassman, K. C. Abbaspour, M. J. White,
R. Srinivasan, C. Santhi, R. D. Harmel, A. van Griensven,
M. W. Van Liew, N. Kannan, M. K. Jha

ABSTRACT. SWAT (Soil and Water Assessment Tool) is a comprehensive, semi-distributed river basin model that requires a
large number of input parameters, which complicates model parameterization and calibration. Several calibration tech-
niques have been developed for SWAT, including manual calibration procedures and automated procedures using the shuf-
fled complex evolution method and other common methods. In addition, SWAT-CUP was recently developed and provides
a decision-making framework that incorporates a semi-automated approach (SUFI2) using both manual and automated
calibration and incorporating sensitivity and uncertainty analysis. In SWAT-CUP, users can manually adjust parameters
and ranges iteratively between autocalibration runs. Parameter sensitivity analysis helps focus the calibration and uncer-
tainty analysis and is used to provide statistics for goodness-of-fit. The user interaction or manual component of the
SWAT-CUP calibration forces the user to obtain a better understanding of the overall hydrologic processes (e.g., baseflow
ratios, ET, sediment sources and sinks, crop yields, and nutrient balances) and of parameter sensitivity. It is important for
future calibration developments to spatially account for hydrologic processes; improve model run time efficiency; include
the impact of uncertainty in the conceptual model, model parameters, and measured variables used in calibration; and as-
sist users in checking for model errors. When calibrating a physically based model like SWAT, it is important to remember
that all model input parameters must be kept within a realistic uncertainty range and that no automatic procedure can
substitute for actual physical knowledge of the watershed.
Keywords. Autocalibration, Hydrologic model, SWAT, Validation.

T
he SWAT (Soil and Water Assessment Tool) mod- decisions on water resources and nonpoint-source pollution
el is a continuous-time, semi-distributed, process- in large river basins. The first version of SWAT was devel-
based river basin model. It was developed to oped in the early 1990s and released as version 94.2. Engel
evaluate the effects of alternative management et al. (1993) reported the first application of SWAT in the
peer-reviewed literature; Srinivasan and Arnold (1994) and
Arnold et al. (1998) later published the first peer-reviewed
Submitted for review in November 2011 as manuscript number description of a geographic information system (GIS) inter-
SW 9517; approved for publication by the Soil & Water Division of face for SWAT and overview describing the key compo-
ASABE in May 2012.
The authors are Jeffrey G. Arnold, ASABE Fellow, Agricultural En- nents of SWAT, respectively. Arnold and Forher (2005) de-
gineer, USDA-ARS Grassland, Soil and Water Research Laboratory, Tem- scribed the expanding global use of SWAT as well as
ple, Texas; Daniel N. Moriasi, ASABE Member, Research Hydrologist, several subsequent releases of the model: versions 96.1,
USDA-ARS Great Plains Agroclimate and Natural Resources Research
Unit, El Reno, Oklahoma; Philip W. Gassman, ASABE Member, Asso-
98.2, 99.2, and 2000. Gassman et al. (2007) provided fur-
ciate Scientist, Center for Agricultural and Rural Development, Iowa State ther description of SWAT, including SWAT version 2005,
University, Ames, Iowa; Karim C. Abbaspour, Senior Scientist, Eawag: and also presented an in-depth overview of over 250
Swiss Federal Institute for Aquatic Science and Technology, Dübendorf, SWAT-related applications that were performed worldwide.
Switzerland; Michael J. White, ASABE Member, Agricultural Engineer,
USDA-ARS Grassland, Soil and Water Research Laboratory, Temple, Krysanova and Arnold (2008), Douglas-Mankin et al.
Texas; Raghavan Srinivasan, ASABE Member, Director, Spatial Sci- (2010), and Tuppad et al. (2011) provide further updates on
ences Laboratory, Texas A&M University, College Station, Texas; Chin- SWAT application and development trends, and the latter
nasamy Santhi, Associate Research Scientist, Texas Research and Exten- two articles provide further description of SWAT version
sion Center, Temple, Texas; Daren Harmel, ASABE Member, Research
Leader, USDA-ARS Grassland, Soil and Water Research Laboratory, 2009, the latest release of the model.
Temple, Texas; Ann van Griensven, Associate Professor, UNESCO-IHE The development of SWAT is a continuation of USDA
Institute for Water Education, Delft, The Netherlands; Michael W. Van Agricultural Research Service (ARS) modeling experience
Liew, Hydrologic Computer Simulation Modeler, Department of Biologi-
cal Systems Engineering, University of Nebraska, Lincoln, Nebraska; Na-
that spans a period of over 30 years (Gassman et al., 2007;
rayanan Kannan, ASABE Member, Assistant Research Scientist, Texas Williams et al., 2008). The current SWAT model includes
Research and Extension Center, Temple, Texas; and Manoj K. Jha, key components contributed from USDA-ARS models as
ASABE Member, Assistant Professor, College of Engineering, North well as from other models (fig. 1). Core pesticide transport,
Carolina A&T, Greensboro, North Carolina. Corresponding author: Jef-
frey G. Arnold, USDA-ARS Grassland, Soil and Water Research Labora- hydrology, and crop growth models that have been incorpo-
tory, 808 East Blackland Road, Temple, TX 76502; phone: 254-770-6502; rated into SWAT can be traced to earlier USDA-ARS field-
e-mail: [email protected].

Transactions of the ASABE

Vol. 55(4): 1491-1508 2012 American Society of Agricultural and Biological Engineers ISSN 2151-0032 1491
 Figure 1. Schematic of SWAT development history and model adaptations (adapted from Gassman et al., 2007).

scale models (fig. 1): the Groundwater Loading Effects of components shown in figure 1 as well as other routines, and
Agricultural Management Systems (GLEAMS) model also features several pre- and post-processing software
(Leonard et al., 1987), the Chemicals, Runoff, and Erosion tools, including the widely used ArcGIS SWAT (ArcSWAT)
from Agricultural Management Systems (CREAMS) model GIS interface (Olivera et al., 2006). Extensive SWAT2009
(Knisel, 1980), and the Environmental Policy Integrated documentation can also be accessed at the SWAT website
Climate model (Williams et al., 2008; Wang et al., 2011). (http://swatmodel.tamu.edu), including theoretical docu-
These components were first grafted into the Simulator for mentation describing all equations, a user’s manual describ-
Water Resources in Rural Basins (SWRRB) model (fig. 1; ing model inputs and outputs, ArcSWAT and Map Window
Arnold and Williams, 1987), along with other key compo- interface manuals, and a developer’s manual. In addition to
nents including a weather generator, sediment routing rou- the model documentation, access is also provided at the
tine, and groundwater submodel (Arnold and Allen, 1999). website to all supporting software, selected journal articles
The initial version of SWAT was then created by interfac- and other publications, a SWAT literature database, previ-
ing SWRRB with the routing structure in the Routing Out- ous and forthcoming conference information, forthcoming
puts to Outlet (ROTO) model (fig. 1; Arnold et al., 1995b). workshops, SWAT-related job openings, and an email
Expanded routing and pollutant transport capabilities have newsletter called SWATbytes. The core SWAT development
since been incorporated into the model (fig. 1), including and user support team is located at the USDA-ARS Grass-
reservoir, pond, wetland, point source, and septic tank ef- land, Soil and Water Research Laboratory and the Texas
fects as well as enhanced sediment routing routines (Arnold AgriLife Blackland Research Center in Temple, Texas.
et al., 2010b) and in-stream kinetic routines from the SWAT development is also occurring at other research sites
QUAL2E model (Brown and Barnwell, 1987). Additional in North America and in other regions (Gassman et al.,
modifications that have been incorporated into SWAT 2010), and multiple user groups have developed world-
(fig. 1) include an improved carbon cycling routine based wide, including SWAT, ArcSWAT, VizSWAT, SWAT-CUP,
on the CFARM model (Kemanian, 2011), alternative daily Latin American, southeast Asia, Africa, Iran, and Brazil us-
and subdaily hydrology routines including the Green-Ampt er groups.
infiltration method (Green and Ampt, 1911), temporal ac- Many of the previous studies published in the extensive
counting of management practice and land use changes and body of peer-reviewed and other SWAT literature describe
enhanced subsurface tile drainage, filter strips, grassed wa- calibration and validation approaches used for verifying the
terways, irrigation, and other improved representations of accuracy of the model for the simulated conditions. These
conservation and management practices (fig. 1; Arnold et testing procedures have been reported at varying levels of
al., 2010b). The temporal accounting routine allows users detail for a wide range of watershed scales, environmental
to introduce the adoption of different selected management conditions, and application goals worldwide (e.g., Gassman
practices or account for changes in land use part way et al., 2007, 2010). More in-depth procedures have also
through a SWAT simulation run, such as the hydrologic and been reported for specific aspects of the calibration and val-
pollutant impacts simulated by Chiang et al. (2010) in re- idation process, such as the guidelines proposed by Moriasi
sponse to temporal changes in pasture use for a 32 km2 wa- et al. (2007) regarding specific statistical criteria to judge
tershed in northwest Arkansas. the success of SWAT (and other model) testing results.
The current SWAT2009 code incorporates all of the However, a comprehensive overview of all key facets re-

1492 TRANSACTIONS OF THE ASABE

quired for an ideal SWAT calibration and validation process er and differentiates between annual and perennial plants.
is currently lacking in the literature. Thus, the objectives of The plant growth model is used to assess removal of water
this study are as follows: (1) to provide a brief description and nutrients from the root zone, transpiration, and bio-
of the key SWAT components, (2) present a general over- mass/yield production. SWAT uses the Modified Universal
view of a logical calibration and validation sequence, Soil Loss Equation (MUSLE) (Williams and Berndt, 1977)
(3) describe calibration options and parameters in more de- to predict sediment yield from the landscape. In addition,
tail, (4) show how the calibration and validation process is SWAT models the movement and transformation of several
applied for two case studies, and (5) discuss weaknesses forms of nitrogen and phosphorus, pesticides, and sediment
and future research needs regarding calibration and valida- in the watershed. SWAT allows the user to define manage-
tion approaches with SWAT. ment practices taking place in every HRU.
Once the loadings of water, sediment, nutrients, and pes-
ticides from the land phase to the main channel have been
MODEL DESCRIPTION determined, the loadings are routed through the streams and
reservoirs within the watershed. The water balance for res-
SWAT operates on a daily time step and is designed to
ervoirs includes inflow, outflow, rainfall on the surface,
predict the impact of land use and management on water,
evaporation, seepage from the reservoir bottom, and diver-
sediment, and agricultural chemical yields in ungauged wa-
sions.
tersheds. The model is process based, computationally effi-
Model equations are given in the SWAT theoretical doc-
cient, and capable of continuous simulation over long time
umentation (http://swatmodel.tamu.edu) and in Arnold et
periods. Major model components include weather, hydrol-
al. (1998). Gassman et al. (2007) presents an overview of:
ogy, soil temperature and properties, plant growth, nutri-
(1) climatic inputs and HRU hydrologic balance; (2) crop-
ents, pesticides, bacteria and pathogens, and land manage-
ping, management inputs, and HRU-level pollutant losses;
ment. In SWAT, a watershed is divided into multiple
and (3) flow and pollutant routing. Arnold et al. (2010b)
subwatersheds, which are then further subdivided into hy-
describe current research on enhancements to SWAT to
drologic response units (HRUs) that consist of homogene-
route water across discretized landscape units that simulate
ous land use, management, topographical, and soil charac-
the impacts of spatial land use changes and land manage-
teristics. The HRUs are represented as a percentage of the
ment on the hillslope-valley continuum.
subwatershed area and may not be contiguous or spatially
identified within a SWAT simulation. Alternatively, a wa-
tershed can be subdivided into only subwatersheds that are
characterized by dominant land use, soil type, and man- SWAT CALIBRATION AND VALIDATION
agement. SWAT input parameters are process based and must be
Water balance is the driving force behind all the pro- held within a realistic uncertainty range. The first step in
cesses in SWAT because it impacts plant growth and the the calibration and validation process in SWAT is the de-
movement of sediments, nutrients, pesticides, and patho- termination of the most sensitive parameters for a given
gens. Simulation of watershed hydrology is separated into watershed or subwatershed. The user determines which var-
the land phase, which controls the amount of water, sedi- iables to adjust based on expert judgment or on sensitivity
ment, nutrient, and pesticide loadings to the main channel analysis. Sensitivity analysis is the process of determining
in each subbasin, and the in-stream or routing phase, which the rate of change in model output with respect to changes
is the movement of water, sediments, etc., through the in model inputs (parameters). It is necessary to identify key
channel network of the watershed to the outlet. Below is a parameters and the parameter precision required for cali-
brief description of the processes simulated by SWAT. De- bration (Ma et al., 2000). In a practical sense, this first step
tails of these processes are given in the SWAT theoretical helps determine the predominant processes for the compo-
documentation (http://swatmodel.tamu.edu). nent of interest. Two types of sensitivity analysis are gener-
The hydrologic cycle is climate driven and provides ally performed: local, by changing values one at a time, and
moisture and energy inputs, such as daily precipitation, global, by allowing all parameter values to change. The two
maximum/minimum air temperature, solar radiation, wind analyses, however, may yield different results. Sensitivity
speed, and relative humidity, that control the water balance. of one parameter often depends on the value of other relat-
SWAT can read these observed data directly from files or ed parameters; hence, the problem with one-at-a-time anal-
generate simulated data at runtime from observed monthly ysis is that the correct values of other parameters that are
statistics. Snow is computed when temperatures are below fixed are never known. The disadvantage of the global sen-
freezing, and soil temperature is computed because it im- sitivity analysis is that it needs a large number of simula-
pacts water movement and the decay rate of residue in the tions. Both procedures, however, provide insight into the
soil. Hydrologic processes simulated by SWAT include sensitivity of the parameters and are necessary steps in
canopy storage, surface runoff, infiltration, evapotranspira- model calibration.
tion, lateral flow, tile drainage, redistribution of water with- The second step is the calibration process. Calibration is
in the soil profile, consumptive use through pumping an effort to better parameterize a model to a given set of lo-
(if any), return flow, and recharge by seepage from surface cal conditions, thereby reducing the prediction uncertainty.
water bodies, ponds, and tributary channels. SWAT uses a Model calibration is performed by carefully selecting val-
single plant growth model to simulate all types of land cov- ues for model input parameters (within their respective un-

55(4): 1491-1508 1493

certainty ranges) by comparing model predictions (output) of SWAT crop growth and/or yield calibrations are reported
for a given set of assumed conditions with observed data by Hu et al. (2007), Ng et al. (2010a), Andersson et al.
for the same conditions. The final step is validation for the (2011), and Nair et al. (2011). Several studies (e.g., Santhi
component of interest (streamflow, sediment yields, etc.). et al., 2001; Engel et al., 2007) also recommend that
Model validation is the process of demonstrating that a giv- streamflow, sediment, and nutrient transport be calibrated
en site-specific model is capable of making sufficiently ac- sequentially (in that order) because of interdependencies
curate simulations, although “sufficiently accurate” can between constituents due to shared transport processes.
vary based on project goals (Refsgaard, 1997). Validation Even though a complete set of hydrologic and water quality
involves running a model using parameters that were de- data are rarely available, all available data should be con-
termined during the calibration process, and comparing the sidered. We recommend that baseflow and surface runoff
predictions to observed data not used in the calibration. In be separated from the observed total daily streamflow using
general, a good model calibration and validation should in- a baseflow filter. The baseflow filter developed by Arnold
volve: (1) observed data that include wet, average, and dry et al. (1995a) and modified by Arnold and Allen (1999) is
years (Gan et al., 1997); (2) multiple evaluation techniques available at http://swatmodel.tamu.edu/software/baseflow-
(ASCE, 1993; Legates and McCabe, 1999; Boyle et al., filter-program. Baseflow and recharge data from this pro-
2000); (3) calibrating all constituents to be evaluated; and cedure have shown good correlation with those produced
(4) verification that other important model outputs are rea- by SWAT (Arnold et al., 2000). To help with the recom-
sonable. In general, graphical and statistical methods with mended SWAT calibration and validation process, a pro-
some form of objective statistical criteria are used to de- gram was recently developed by White et al. (2012). This
termine when the model has been calibrated and validated. program warns users if selected model outputs vary outside
Calibration can be accomplished manually or using auto- typical ranges to ensure that processes are realistically sim-
calibration tools in SWAT (van Griensven and Bauwens, ulated (http://swatmodel.tamu.edu/software/swat-check).
2003; Van Liew et al. (2005) or SWAT-CUP (Abbaspour et Following these simple recommendations can avoid major
al., 2007). errors in scenario analysis, which is the primary objective
Ideally, calibration and validation should be process and of most SWAT-related projects.
spatially based, while taking into account input, model, and Calibration and validation are typically performed by
parameter uncertainties. A good example of process-based splitting the available observed data into two datasets: one
calibration involves streamflow. Streamflow processes are for calibration, and another for validation. Data are most
comprised of the water balance in the land phase of the hy- frequently split by time periods, carefully ensuring that the
drology, including ET, lateral flow, surface runoff, return climate data used for both calibration and validation are not
flow, tile flow (if present), channel transmission losses, and substantially different, i.e., wet, moderate, and dry years
deep aquifer recharge. If data are available for each of these occur in both periods (Gan et al., 1997). Data may also be
processes, they should be calibrated individually. For sedi- split spatially, with all available data at a given monitoring
ments, nutrients, pesticides, and bacteria, sources and sinks location assigned to the calibration phase and correspond-
should be considered. If a longer time period is available ingly performing the validation at one or more other gauges
for hydrology than water quality data, it is important to use within the watershed. This approach can be necessary when
all the hydrology data available for calibration and valida- users are faced with data-limited situations that preclude
tion to capture long-term trends. This process-based cali- performing a split-time calibration and validation using a
bration should be done at the subwatershed or landscape single gauge. SWAT users have also used calibrated param-
level to ensure that variability in the predominant processes eters from a watershed with approximately similar climatic,
for each of the subwatersheds is captured instead of deter- soils, and land use conditions for validation in their study
mining global (watershed-wide) processes. There are, how- watershed, or vice versa. Split-location calibration and val-
ever, generally insufficient observed data to enable a full idation approaches have been performed in some previous
spatial calibration and validation at the watershed scale. SWAT studies (e.g., Arnold et al., 2001; Van Liew and Gar-
The metrics and methods used to compare observed data to brecht, 2003; Cao et al., 2006; Parajuli et al., 2009).
model predictions are also important. Multiple graphical Although these are the recommended calibration and
and statistical methods could be used, such as time-series validation approaches, they are not enforced, and thus there
plots, Nash-Sutcliffe efficiency (NSE; Nash and Sutcliffe, are several ways in which SWAT has been calibrated and
1970), and percent bias. validated. Most published SWAT applications report both
A general calibration flowchart for flow, sediment, and graphical and statistical hydrologic calibration results, es-
nutrients is shown in figure 2 to aid with the manual model pecially for streamflow, and hydrologic validation results
calibration process. Users should check the water balance are also reported for a large percentage of the studies. Simi-
components (ET, surface/baseflow ratios, tile flow propor- lar pollutant testing results are also reported for many
tions, plant yield, and biomass) during the calibration pro- SWAT studies, although not nearly as many as are reported
cess to make sure the predictions are reasonable for the for streamflow results. An extensive array of statistical
study region or watershed. Because plant growth and bio- techniques can be used to evaluate SWAT hydrologic and
mass production can have an effect on the water balance, pollutant predictions; for example, Coffey et al. (2004) de-
erosion, and nutrient yields, reasonable local/regional plant scribe nearly 20 potential statistical tests that can be used to
growth days and biomass production should be verified judge SWAT predictions, including coefficient of determi-
during model calibration to the extent possible. Examples nation (r2), NSE, root mean square error (RMSE), nonpar-

1494 TRANSACTIONS OF THE ASABE

 Figure 2. Example SWAT manual calibration flowchart (from Engel et al., 2007; adapted from Santhi et al., 2001).

ametric tests, t-test, objective functions, autocorrelation, served data is indicated by an NSE value of 1. NSE values
and cross-correlation. By far, the most widely used statis- ≤0 indicate that the observed data mean is a more accurate
tics reported for calibration and validation are r2 and NSE. predictor than the simulated output. Both NSE and r2 are
The r2 statistic can range from 0 to 1, where 0 indicates no biased toward high flows. To minimize this bias, some re-
correlation and 1 represents perfect correlation, and it pro- searchers have taken the log of flows for statistical compar-
vides an estimate of how well the variance of observed val- ison or have developed statistics for low and high flow sea-
ues are replicated by the model predictions(Krause et al., sons (Krause et al., 2005). Krause et al. (2005) provide
2005). A perfect fit also requires that the regression slope further discussion regarding the strengths and weaknesses
and intercept are equal to 1 and 0, respectively; however, of using r2, NSE, and other efficiency criteria measures.
the slope and intercept have typically not been reported in An extensive list of r2 and NSE calibration and/or vali-
published SWAT studies. If r2 is the primary statistical dation statistics is provided by Gassman et al. (2007) for
measure, it should always be used with slope and intercept 115 SWAT studies that reported hydrologic results as well
to ensure that means are reasonable (slope = 1) and bias is as 37 SWAT studies that reported pollutant results. Similar
low. NSE values can range between -∞ to 1 and provide a r2 and NSE statistical compilations for an additional 20
measure how well the simulated output matches the ob- SWAT studies are reported by Douglas-Mankin et al.
served data along a 1:1 line (regression line with slope (2010), and 23 SWAT studies are reviewed by Tuppad et al.
equal to 1). A perfect fit between the simulated and ob- (2011). These statistics provides valuable insight regarding

55(4): 1491-1508 1495

the hydrologic performance of the model across a wide yield and streamflow components of interest in the region.
spectrum of conditions. To date, no absolute criteria for
judging model performance have been firmly established in ACCOUNTING FOR UNCERTAINTY BANDS
the literature, and for good reason: the criteria for judgment The above statistical indices only apply to the compari-
of model performance should be tied to the intended use of son of two signals and are not adequate when outputs are
the model (Engel et al., 2007). SWAT has been used for a expressed as uncertainty bands. In this case, as the simula-
variety of applications, ranging from simple hydrologic as- tion results are usually expressed by the 95% prediction
sessments for watershed planning to the assignment of uncertainties (95PPU), they cannot be compared with the
blame and damages in a court of law (White et al., 2011). observation signals using the traditional r2 and NSE statis-
The risk of adverse impacts arising from model prediction tics. For this reason, Abbaspour et al. (2004, 2007) suggest
uncertainty or error for a particular application should be a using two measures, referred to as the P-factor and the R-
consideration during the calibration. However, for a more factor. The P-factor is the percentage of the measured data
typical application, Moriasi et al. (2007) proposed that NSE bracketed by the 95PPU. This index provides a measure of
values should exceed 0.5 in order for model results to be the model’s ability to capture uncertainties. As all the “true”
judged satisfactory for hydrologic and pollutant loss evalu- processes are reflected in the measurements, the degree to
ations performed on a monthly time step (and that appro- which the 95PPU does not bracket the measured data indi-
priate relaxing and tightening of the standard be performed cates the prediction error. Ideally, the P-factor should have
for daily and annual time step evaluations, respectively). a value of 1, indicating 100% bracketing of the measured
Assuming this criterion for both the NSE and r2 values at data, hence capturing or accounting for all the correct pro-
all time steps, the majority of the calibration and validation cesses. The R-factor, on the other hand, is a measure of the
statistics listed by Gassman et al. (2007) were judged satis- quality of the calibration and indicates the thickness of the
factory, or adequately replicating observed streamflows and 95PPU. Its value should ideally be near zero, hence coin-
other hydrologic indicators. However, it is clear that poor ciding with the measured data. The combination of P-factor
test statistics occurred for parts or all of some studies. The and R-factor together indicate the strength of the model cal-
poorest results generally occurred for daily predictions, alt- ibration and uncertainty assessment, as these are intimately
hough this was not universal (e.g., Grizzetti et al., 2005). linked.
Interestingly, Douglas-Mankin et al. (2010) found that all
of the daily flow calibration statistics reported among the CALIBRATION APPROACHES
20 SWAT studies they reviewed were satisfactory or better, Conventionally, calibration is performed manually and
based on the 0.5 criterion described above. Tuppad et al. consists of changing model input parameter values to pro-
(2011) further found that 85% of the daily flow statistics duce simulated values that are within a certain range of the
compiled from their review of 23 SWAT studies also met measured data (Balascio et al., 1998). However, when the
this criterion. When they combined their overall compiled number of parameters used in the manual calibration is
statistics with those reported by Gassman et al. (2007) and large, especially for complex hydrologic models, manual
Douglas-Mankin et al. (2010), Tuppad et al. (2011) also re- calibration can become labor-intensive (Balascio et al.,
ported that 72% of 134 calibration results and 58% of 113 1998) and automated calibration methods are preferred.
validation results were rated as satisfactory or better, and Both manual algorithms and automated methods have been
21% of calibration results and 12% of validation results developed for calibration of SWAT simulations. An itera-
were rated as very good, where “very good” meant NSE tive approach is usually used for manual calibration involv-
(and r2) >0.75, again based on monthly flow criteria pro- ing the following steps: (1) perform the simulation;
posed by Moriasi et al., 2007. (2) compare measured and simulated values; (3) assess if
Some of the poorer testing results reported in previous reasonable results have been obtained; (4) if not, adjust in-
SWAT studies can be partially attributed to inadequate spa- put parameters based on expert judgment and other guid-
tial coverage of precipitation inputs, which can occur be- ance within reasonable parameter value ranges; and (5) re-
cause of an inadequate number of rain gauges in the simu- peat the process until it is determined that the best results
lated watershed or coarse subwatershed configurations that have been obtained. Several studies present systematic
failed to capture the spatial detail of available rainfall data strategies for performing streamflow and/or pollutant cali-
(e.g., Cao et al., 2006; Conan et al., 2003; Bouraoui et al., bration and validation. Coffey et al. (2004) recommend us-
2002, 2005). Inadequate model calibration (Bosch et al., ing NSE and r2 for analyzing monthly output and median
2004), measurement uncertainty in streamflow data (Har- objective functions, sign test, autocorrelation, and cross-
mel et al., 2006a, 2009), and short streamflow records (Mu- correlation for assessing daily output based on comparisons
leta and Nicklow, 2005) can also result in weak SWAT hy- of SWAT results with measured streamflow. Santhi et al.
drologic predictions. Most reported SWAT studies contain (2001) propose a manual calibration approach (including a
both calibration and validation, while others performed on- flowchart) that was a function of sensitive input parameters
ly calibration due to a lack of observed data. In a few cases, (15 were selected), realistic uncertainty ranges, and satis-
calibration of SWAT was not performed. For example, factory r2 and NSE statistical results. Cao et al. (2006) also
Srinivasan et al. (2010) describe an uncalibrated application present a flowchart of their SWAT manual calibration ap-
of SWAT for the Upper Mississippi River basin in the proach, which was based on multiple hydrologic outputs
north-central U.S., which was conducted with the goal of and multiple gauge sites. Nair et al. (2011) present another
determining how the default parameters represented crop

1496 TRANSACTIONS OF THE ASABE

systematic SWAT calibration approach, including a sche- (2009) used SWAT-CUP to calibrate a hydrologic model of
matic of the methodology that incorporates crop the Thukela River basin in South Africa. In the above ap-
growth/yield calibration and validation along with stream- plications, the goodness of fit criteria is provided by P-
flow and pollutant loss calibration and validation. Many factor and R-factor. For the objective function, however, a
studies also report the use of automated techniques for cali- weighted version of r2 (Krause et al., 2005) was selected as
brating SWAT, which typically rely on Monte Carlo or oth- the efficiency criterion:
 b R2
er sampling schemes to estimate the best choice of values

Φ=
for multiple input parameters, without violating practical or if b ≤ 1

 b R 2
theoretical boundaries for each specific input parameter. (1)
−1
The input values are usually determined over the course of if b > 1
iterative SWAT simulations, which sometimes number in
where b is the slope of the regression line between meas-
the thousands. Several optimization schemes have been
ured and simulated signals. A major advantage of this effi-
used in SWAT autocalibration applications, including gen-
ciency criterion is that it ranges from 0 to 1, which com-
eralized likelihood uncertainty estimation (GLUE), shuffled
complex evolution (SCE), and the Parameter Estimation pared to NSE with a range of -∞ to 1, ensures that in a
(PEST) program (Doherty, 2004). Govender and Everson multisite calibration the objective function is not governed
(2005) and Wang and Melesse (2005) used PEST to cali- by a single or a few badly simulated stations.
brate key hydrology-related parameters for SWAT applica-
tions in South Africa and northwest Minnesota, respective- CALIBRATION PARAMETERS
ly. Wang and Melesse (2005) also found that manual Numerous studies have reported input parameters used
calibration resulted in more accurate predictions than the in SWAT model calibration. Table 1 summarizes the pa-
automated PEST approach. Ng et al. (2010b) described ad- rameters used in 64 studies and in studies previously sum-
vantages and disadvantages of using PEST versus the marized by Douglas-Mankin et al. (2010) and Tuppad et al.
GLUE method for calibrating SWAT for a watershed in (2011). All of these studies include detailed reporting of
central Illinois. Setegn et al. (2009) and Razavi et al. (2010) model parameterization and calibration procedures, includ-
described additional SWAT calibration studies that relied ing tables with parameter ranges and/or final values. Many
partially on GLUE methodology for watersheds located in publications in the literature (https://www.card.iastate.edu/
Ethiopia and south-central New York. To determine opti- swat_articles) do not adequately report changes in parame-
mum input parameters based on the global objective criteria ters. Model parameters used in calibration studies and even
for a simulation of a river basin in Belgium, van Griensven in the selected publications exhibited gaps. Tuppad et al.
and Bauwens (2003, 2005) incorporated an SCE module di- (2011) re-emphasizes an important point made by Douglas-
rectly into the SWAT code. Calibration parameters, corre- Mankin et al. (2010): “Improved reporting of calibration
sponding parameter ranges, and measured daily streamflow and validation procedures and results, perhaps guided by a
and pollutant data were input for the application, which re- set of standard reporting guidelines, is essential for ade-
quired several thousand SWAT simulations for completion. quate interpretation of each study and comparison among
Eckhardt and Arnold (2001) and Eckhardt et al. (2005) used studies in the future. This increased information would also
similar SCE-based automatic calibration methods for form the basis for assigning typical parameters and ranges
SWAT simulations of German watersheds. Other SCE- for use in both manual or automatic calibration and uncer-
based SWAT automatic calibration applications are reported tainty processes.”
by Di Luzio and Arnold (2004) and Van Liew et al. (2005, Table 1 categorizes parameters by process. Since SWAT
2007). is a comprehensive model that simulates process interac-
Automatic calibration and uncertainty analysis capabil- tions, many parameters will impact multiple processes. For
ity is now directly incorporated in SWAT2009 (Gassman et example, CN directly impacts surface runoff; however, as
al., 2010) via the SWAT-CUP software developed by Ea- surface runoff changes, all components of hydrology bal-
wag (2009). A number of previous SWAT application pro- ance change. Soil erosion and nutrient transport are also di-
jects report automated calibration/validation and uncertain- rectly impacted by surface runoff, as are plant growth and
ty analysis using SWAT-CUP. Abbaspour et al. (2007) nutrient cycling. This is the primary reason why most man-
performed a multi-objective calibration and validation of ual calibration methods start with the hydrology balance
the Thur watershed in Switzerland using discharge, sedi- and streamflow, then move to sediment, and finally cali-
ment, nitrate, and phosphate in the objective function with brate nutrients and pesticides, as shown in figure 2 (Santhi
uncertainty analysis. Schuol et al. (2008a, 2008b) calibrated et al., 2001). It is evident from table 1 that hydrology is cal-
with uncertainty analysis and validated models of west Af- ibrated in most studies, with CN2, AWC, ESCO, and
rica and the entire continent of Africa. Yang et al. (2008) SURLAG used routinely. The baseflow process is also of-
compared five different optimization algorithms in SWAT- ten calibrated with the baseflow recession parameters used
CUP and calibrated a watershed in China (2007) using the in many studies.
MCMC algorithm. Faramarzi et al. (2009) used SWAT to Parameters for sediment calibration are used less often
build a hydrologic model of Iran and calibrated and vali- due to inadequate reporting or studies that did not focus on
dated it with the SUFI2 algorithm accounting for prediction sediment. It is interesting to note that of the studies in
uncertainty. Akhavan et al. (2010) calibrated a model of ni- which sediment was calibrated, channel parameters were
trate leaching for a watershed in Iran, and Andersson et al. used more often than parameters affecting sediment

55(4): 1491-1508 1497

Table 1. Calibration parameters reported in 64 selected SWAT watershed studies.[a] Numbers in parentheses are the number of times the param-
eter was used in calibration. Definitions of variables are found in the SWAT user manual (http://swatmodel.tamu.edu/documentation).
Process Input Parameters
Surface CN2 AWC ESCO EPCO SURLAG OV_N
runoff (36) (28) (23) (10) (22) (8)
Baseflow GW_ALPHA GW_REVAP GW_DELAP GW_QWN REVAPMN RCHARG_DP
(28) (18) (21) (12) (13) (14)
Snow SFTMP SMFMN SMFMX SMTMP TIMP SNO50COV SNOCOVMX
(11) (14) (18) (13) (7) (4) (3)
Sediment from PRF APM SPEXP SPCON CH_EROD CH_COV
channels (10) (7) (10) (11) (6) (7)
Sediment from USLE_P USLE_C USLE_K LAT_SED SLSOIL SLOPE
landscape (7) (7) (7) (1) (2) (8)
N from RCN UBN GWNO3 ERORGN NPERCO ANION_EXCL
landscape (1) (3) (2) (5) (11) (2)
P from PSP PHOSKD UBP PPERCO GWQMINP ERORGP
landscape (5) (6) (5) (8) (1) (5)
Pesticides KOC HL_SOIL HL_FOL WSOL WOFFW
(1) (1) (1) (1) (2)
Subsurface TDRAIN GDRAIN DEP_IMP
tile (1) (2) (1)
N and P BC1 BC2 BC3 BC4 RS4 RS5
from channels (2) (2) (2) (2) (2) (1)
Plant growth GSI HI BLAI PHU CN_YLD
(3) (1) (3) (1) (1)
Bacteria BACTRDQ BACTMIX BCNST CFRT_KG WDPRCH WDPQ
(1) (1) (1) (1) (1) (1)
Other BIOMIX SOL_ROCK MSK_COL MSK_CO2 CBNINT SOL_BD ALPHA_BNR EVRCH SOL_ALB LAT_TTIME
(4) (1) (1) (2) (1) (3) (1) (1) (2) (1)
[a]
Abbaspour et al. (2007), Ahl et al. (2008), Alibuyog et al. (2009), Behera and Panda (2006), Bekele and Nicklow (2007), Benaman and Shoemaker
(2004), Benaman et al. (2005) , Bekele and Knapp (2010), Cheng et al. (2007), Chin et al. (2009), Chu et al. (2004), Coffey et al. (2010), Debele et
al.(2008), Di Luzio and Arnold (2004), Douglas-Mankin et al.(2010), Du et al.(2006), Easton et al.(2008), Eckhardt et al.(2002), Eckhardt et al.
(2005), Engel et al. (2007), Ghaffari et al. (2010), Gikas et al. (2006), Gitau et al. (2008), Green et al. (2006, 2007), Green and van Griensven (2008),
Heuvelmans et al. (2004, 2006), Hu et al. (2007), Inamdar and Naumov (2006), Jha et al. (2010), Lemonds and McCray (2007), Maski et al. (2008),
Meng et al. (2010), Mukundan et al. (2010), Muleta and Nicklow (2005), Narasimhan et al. (2010), Santhi et al. (2008), Shoemaker et al. (2007),
Starks and Moriasi (2009), Sui and Frankenberger (2008), Sudheer et al. (2007), Tuppad et al. (2011),van Griensven et al. (2008), Van Liew et al.
(2007), White and Chubey (2005), Zhang et al. (2008a), Zhang et al. (2008b).

transport from the landscape. A potential explanation is that CASE STUDIES

there is more uncertainty in the channel sediment routing Two case studies were chosen as examples of SWAT
parameters, and thus users feel more comfortable using validation. The first study by Van Liew et al. (2005) high-
them in calibration. These parameters can also be very sen- lights the advantages of manual and automated calibration
sitive, making adjustment very effective during the calibra- techniques, and the second study by Rouholahnejad et al.
tion process. However, it is critical to ensure that sources (2012a) uses the semi-automated SUFI2 program. Both
and sinks of sediment and that the ratio of upland sources case studies emphasize that no automatic calibration proce-
versus channel sources and deposition are realistic, even dure can substitute for actual physical knowledge of water-
though measured data are often relatively scarce. Sediment shed processes.
measurement is very difficult and involves considerable er-
ror. Furthermore, the modified Universal Soil Loss Equa-
MANUAL AND AUTOMATED CALIBRATION
tion in SWAT is inadequate in many cases, such as in ac-
The calibration study by Van Liew et al. (2005) com-
counting for the “second storm effect” reported by
pared and discussed both manual and automated calibration
Abbaspour et al. (2007). Therefore, adjustment of the pa-
techniques for five watersheds at the ARS Little River ex-
rameters is actually compensating for the lack of precision
perimental watersheds at Tifton, Georgia (fig. 3) and at the
in the measurement or errors in the conceptual model.
Little Washita experimental watersheds operated by ARS
Based on the literature review, it is also evident that
scientists at El Reno, Oklahoma (fig. 4). The locations rep-
many processes are not as rigorously calibrated as hydrolo-
resent a wide range of climate, soils, and land use (table 2).
gy and streamflow. Only a few studies adequately reported
calibration parameters for N, P, pesticides, bacteria, tile Manual Calibration
flow, and plant growth. Standard reporting guidelines SWAT was calibrated manually by following a multistep
would help form the basis for assigning parameters and procedure recommended by Neitsch et al. (2002). For the
ranges for these processes. Little Washita, the upper watershed (subwatershed 526)
was calibrated first, and the parameters in that subwater-
shed were then held constant while the larger watershed
(subwatershed 550) was calibrated on that portion of the
subwatershed below the outlet of 526. Although it was rec-
ognized that computational differences between measured
and simulated streamflow at the outlet of subwatershed 526

1498 TRANSACTIONS OF THE ASABE

 watershed B. Manual calibration attempted to minimize to-
tal flow (minimized average annual percent bias), accom-
panied by visual inspection of daily hydrographs and dura-
tion of daily flow curves. The sum of squares of residuals
objective function could have been used in the manual cali-
bration, but preliminary testing showed that the total mass
balance method in combination with the inspection of dura-
tion of daily flow curves gave a better representation of the
range of simulated flows (Van Liew et al., 2005). Detailed
calibration of SWAT on the Little Washita was previously
reported by Van Liew and Garbrecht (2003). A preliminary
calibration was conducted on a monthly basis to identify
the order of magnitude of all parameters to reproduce prop-
er runoff volumes and seasonal characteristics. The runoff
curve number (CN2) that governs the surface runoff re-
sponse was first calibrated. Second, the groundwater
“revap” coefficient (GW_REVAP), the threshold depth of
water in the shallow aquifer required for return flow
(REVAPMN), and the deep aquifer percolation fraction
(RCHRG_DP), which governs the fraction of percolation
from the root zone to the deep aquifer, were calibrated.
Third, the baseflow recession factor (ALPHA_BF) and the
groundwater delay (GW_DELAY) parameters were cali-
brated so that the monthly measured versus simulated hy-
drographs agreed well (Van Liew et al., 2005). This prelim-
inary calibration was followed by a fine-tuning at the daily
Figure 3. Location of the Little River Experimental Watershed in time scale so that the predicted versus measured peak flows
Georgia (from Van Liew et al., 2005). and recession curves on a daily time step matched as close-
ly as possible. This same approach was taken in the manual
were passed on to subwatershed 550, this approach to cali- calibration of the Little River.
bration was the most reasonable option that could be exer-
cised, based on the availability and quality of existing da- Autocalibration
tasets in the watershed (Van Liew et al., 2005). A similar The autocalibration procedure described by Van Liew et
approach was taken in calibrating the Little River. The up- al. (2005) was developed by van Griensven and Bauwens
per portion of the watershed (subwatershed F) was calibrat- (2003) and is based on the shuffled complex evolution al-
ed first, which was then followed by a calibration of sub- gorithm (SCE-UA; Duan et al., 1992) that allows for the

Figure 4. Location of the Little Washita River Experimental Watershed in Oklahoma (from Van Liew et al., 2005).

55(4): 1491-1508 1499

Table 2. Number of subbasins, number of hydrologic response units, drainage areas, and land use types for the two USDA ARS experimental
watersheds.
No. of No. of Drainage Area Land Use Type (%)
Watershed Subbasins HRUs (km2) Range/Pasture Crop Forest Wetland Misc.
Little River F 12 51 114 19 45 26 9 1
Little River B 40 161 330 10 42 45 2 1
Little Washita 526 22 138 160 59 28 6 0 7
Little Washita 550 73 486 600 66 19 9 0 6
Little Washita 522 66 413 538 66 18 9 0 7

calibration of model parameters based on a single objective sary after autocalibration. Caution must also be exercised in
function. The SCE-UA has been widely used in watershed using the autocalibration tool so that the selection of initial
model calibration and other areas of hydrology, such as soil lower and upper ranges in the parameters results in cali-
erosion, subsurface hydrology, remote sensing, and land brated values that are representative of watershed condi-
surface modeling, and has generally been found to be ro- tions (Van Liew et al., 2005).
bust, effective, and efficient (Duan, 2003).
Parameters in SWAT were calibrated at the daily time SEMI-AUTOMATED SUFI2
scale in a distributed fashion using the automated calibra- For the second case study, an example calibration of the
tion procedure, in which observed and simulated outputs Danube project (Rouholahnejad et al., 2012b) was selected
were compared at the same outlet points as the manual cal- using SWAT-CUP. Rouholahnejad et al. (2012a) referred to
ibration. With the completion of a given optimization, two the process of parameter assignment as parameterization.
sets of calibrated parameters were computed for the Little Correct parameterization is an important step in model cal-
River that corresponded to subwatersheds F and B, and two ibration and must be based on the knowledge of the hydro-
sets were computed for the Little Washita that correspond- logic processes and variability in soil, land use, slope, and
ed to subwatersheds 526 and 550. Default values suggested location as defined by the subbasin number. Parameteriza-
by van Griensven (2002) were selected as initial upper and tion, therefore, could be defined as “the process of impart-
lower ranges for the respective model parameters. Minimiz- ing the analyst’s knowledge of the physical processes of the
ing the sum of squares residuals was used as the objective watershed to the model.” No automatic calibration proce-
function in the autocalibration procedure. dure can substitute for actual physical knowledge of the
The Van Liew et al. (2005) study highlighted an im- watershed, which can translate into correct parameter rang-
portant difference that must be realized in comparing the es for different parts of the watershed. These ranges can ef-
manual versus autocalibration approaches. The autocalibra- fectively guide the optimization routine. Hence, correct pa-
tion approach was strictly a quantitative comparison that rameterization can result in faster and more accurate model
involved minimizing the difference between measured and calibration with smaller prediction uncertainty. SWAT-CUP
simulated values. The manual approach involved both includes automated as well as semi-automatic procedures
quantitative and qualitative comparisons, since it involved for model calibration. The following steps are suggested in
using the total mass controller in conjunction with graph- a calibration exercise with the semi-automated program
ical comparisons of monthly and daily hydrographs and du- SUFI2:
ration of daily flow curves to calibrate the model against 1. Develop initial or default SWAT input parameters
measured data. Use of the manual calibration accentuates (as created by ArcSWAT or other GIS interfaces)
the tradeoffs that exist in achieving total mass balance, rea- and prepare the input files for SWAT-CUP.
sonable hydrograph responses, and adequate representation 2. Run the model with initial parameters and plot the
of the range in flows. Van Liew et al. (2005) suggest that simulated and observed variables at each gauging
the strengths of both the manual and autocalibration ap- station for the entire period of record.
proaches can be used to facilitate the calibration process. 3. Based on step 2, divide the entire period into cali-
With proper selection of the upper and lower ranges for bration and validation periods while attempting to
model parameter values, autocalibration can provide an ini- ensure that both periods have a similar number of
tial parameter set with minimal labor on the part of the user. wet and dry years and similar average water bal-
Depending on the particular modeling needs, a manual ap- ances.
proach can then be taken to refine the calibration, so that an 4. Determine the most sensitive parameters for the
appropriate balance is achieved regarding the amount, tim- observed values of interest. This information can
ing, and distribution of the output variable. usually be deduced from the literature (see ta-
Results of the Van Liew et al. (2005) suggest that the au- ble 1).
tocalibration option in SWAT provides a powerful, labor- 5. Assign an initial uncertain range (typically 20% to
saving tool that can be used to substantially reduce the frus- 30%) to each parameter globally, meaning scaling
tration and subjectivity that often characterize manual cali- the parameters identically for each HRU.
brations. If used in combination with a manual approach, 6. Run the SWAT-CUP-SUFI2 model 300 to
the autocalibration tool shows promising results in provid- 500 times and view the results for each gauged
ing initial estimates for model parameters. To maintain outlet, as shown in figure 5.
mass balance and adequately represent the range in magni- 7. Perform the global sensitivity analysis and view
tude of output variables, manual adjustments may be neces- the results. At this stage, the P-factor and p-value

1500 TRANSACTIONS OF THE ASABE

 Figure 5. Observed flow, 95% model uncertainty, and best estimation at gauging station 209 before calibration.

t-statistic can be used to eliminate non-sensitive realistic ranges, as influenced by the uncertainty in defining
parameters from the calibration process. the parameter.
8. After observing model performance in step 6, re- In the manner described above, the parameters of each
gionalize the respective parameters. For example, observational gauge can be used to spatially calibrate the
as shown in figure 5, the model systematically un- model in the drainage area between the gauges. At this
derestimated baseflow at outlet q_209 (in subbasin point, the analyst’s knowledge of processes in the water-
209), and there is an early shift in the flow peak. shed could also be implemented in the optimization.
To increase baseflow, decrease deep percolation Figure 6 shows the results after implementing the above
(GWQMN), decrease the groundwater revap coef- changes in the parameters and running the model, where
ficient (GW_REVAP), and increase the threshold NS increased from -1.5 to 0.2. Additional iterations can fur-
depth of water in shallow aquifer (REVAPMN). To ther improve the results. Details on parameterization and
correct the early shift, decrease the slope results can be found in Rouholahnejad et al. (2012a,
(HRU_SLP), increase Manning’s roughness coeffi- 2012b). After calibration, the model should be run for the
cient (OV_N), increase the value of overland flow validation period to assess its performance.
rate (SLSUBBSN), and increase snow melt param-
eters (SMTMP).
To increase baseflow and delay peaks, identify the sub- DISCUSSION
basins that contribute to the outlet at subbasin 209 and im-
Gassman et al. (2010) discussed trends in SWAT use and
plement the changes in the respective parameters. For ex-
the technical and networking factors that are regarded as
ample, make changes in the parameters only in subbasins
strengths of the model, which include: web-based docu-
draining into 209, and set new ranges for the parameters us-
mentation, user support groups, SWAT literature database,
ing one-at-a-time sensitivity analysis implemented in
GIS interface tools, pre- and post-processing tools, open
SWAT-CUP. The parameters must always be kept within
source code, regional and international conferences, and

Figure 6. Observed flow, 95% model uncertainty, and best estimation at gauging station 209 after calibration.

55(4): 1491-1508 1501

model training workshops. The fundamental strengths of tion process. Fitting rainfall data, however, should be exer-
SWAT are flexibility in combining upland and channel pro- cised with caution, as rainfall is a driving variable and fit-
cesses and simulation of land management. As noted by ting it may mask the importance of other parameters.
Gassman et al. (2007), each process is a simplification of
reality and could be improved. Gassman et al. (2007) also
discussed several weaknesses that include: simplified rep- FUTURE DEVELOPMENTS
resentation of HRUs, simulation of certain management
The international modeling community has made signif-
practices, pathogen fate and transport, in-stream sediment
icant strides over the last decade on model parameteriza-
routing and kinetic functions, static soil carbon, subsurface
tion, calibration, and uncertainty analysis. As each of the
tile flow and nitrate losses, and routines for automated sen-
remaining issues is addressed, users will build greater con-
sitivity, calibration, and input uncertainty analysis. Consid-
fidence in model results and improve conservation and en-
erable progress has been made on many of these weakness-
vironmental policy development.
es since Gassman et al. (2007); however, some processes
are difficult to characterize accurately due to insufficient
IMPROVED ACCOUNTING FOR HYDROLOGIC PROCESSES
monitoring data, inadequate data to parameterize inputs, or
Current autocalibration tools optimize the accuracy at
insufficient understanding of the processes themselves.
one or more stream gauges, without regard for predictions
Arnold et al. (2010a) developed routines to route flow
at locations without measured data. Sites lacking measured
across the landscape between HRUs, which allows for
data have no weight in the autocalibration procedure. It is
more process-based simulation of riparian and floodplain
important that models accurately predict hydrographs at se-
processes. Documentation and interfaces are being devel-
lected points in a watershed; however, it is equally im-
oped to guide users in parameterizing management scenari-
portant that the model simulates the processes realistically
os. White and Arnold (2009) developed improved routines
at all locations. For example, if surface runoff is overesti-
for vegetative filter strips, and Arabi et al. (2006) suggest
mated, it is likely that ET and/or subsurface and tile flow
appropriate input parameterization for several structural
are underestimated, resulting in overestimation of sediment
management practices. A web-based tool for spatial man-
yields and underestimation of subsurface nitrate yields.
agement scenario parameterization for SWAT has been de-
This will cause errors when parameterizing variables relat-
veloped within the eRAMS (Environmental Risk Assess-
ed to sediment and nutrient transport and result in unrealis-
ment and Management System) interface (www.eramsinfo.
tic policy recommendations when running scenarios that
com/erams_beta). Progress has also been made on a dy-
target erosion and fertilizer management. Similarly, it is
namic soil carbon model (Kemanian et al., 2011) and on
important to realistically simulate sediment sources and
improving the tile flow and nitrate submodels (Moriasi et
sinks within a watershed in addition to sediment loads at a
al., 2011, 2012). Several other new model components are
gauge. If upland erosion is overpredicted, and thus channel
in final development, such as modeling different types of
erosion is underpredicted to match measured gauge loads,
septic systems (Jeong et al., 2011), simulation of urban
then management practices designed to reduce erosion
processes at shorter time intervals, and urban best man-
from the landscape may show significant impact on total
agement practices. These new developments will comple-
sediment yields, while in reality the practices would have
ment SWAT modeling processes and calibration in urban
little impact at the basin outlet. Before calibrating time-
and septic system dominant watersheds.
series of nutrient loadings (N and P) at gauging stations, the
As previously noted, SWAT is a comprehensive, semi-
overall nutrient balance of the watershed should be exam-
distributed model that uses readily available inputs. The
ined. This step will ensure that proper processes and
weakness in a comprehensive watershed model is the high
sources are realistically simulated, such as amount of ferti-
number of parameters, which complicates model parame-
lizer applied, nutrient uptake by plants, denitrification, fixa-
terization and calibration. Van Griensven and Bauwens
tion, volatilization, nitrification, and organic versus soluble
(2003) overcame some of these problems by developing an
nutrient loadings. Nutrient calibration should focus on cali-
autocalibration method that reduced multiple objective
brating the major constituents rather than calibrating total N
functions into a single global criterion in an objective way,
and total P. SWAT has a dynamic nutrient simulation rou-
thus solving the weighting problem. Abbaspour et al.
tine that considers transformation and movement of all con-
(2007) developed autocalibration and uncertainty software
stituents at multiple levels. Major constituents for N load-
for SWAT, called SWAT-CUP, which includes the method
ings are organic N and mineral N. Similarly, organic P and
of van Griensven and Bauwens (2003) and other methods,
mineral P constitute the total P. If monitoring data are not
including a multi-site, semi-automated inverse modeling
directly available for nutrient constituents, their proportions
routine (SUFI-2) for calibration and uncertainty analysis.
should at least be verified. For example, total N may be cal-
Schuol et al. (2008b) and Abbaspour et al. (2009) applied
ibrated to match the observed data, but the relative contri-
SWAT-CUP for a blue-green water analysis of the continent
bution of organic and mineral N should also be checked for
of Africa and the country of Iran. In SWAT-CUP, all SWAT
the specific region. In addition, if plant growth is not
parameters can be included in the calibration process, in-
properly simulated, the model may not be properly parame-
cluding all water quality parameters, crop parameters, crop
terized or calibrated, which may result in errors with crop-
rotation and management parameters, and weather genera-
ping systems and fertilizer management scenarios. Nair et
tor parameters. Furthermore, rainfall and temperature can
al. (2011) suggest that crop yield comparison be added to
also be treated as random variables and fitted in the calibra-

1502 TRANSACTIONS OF THE ASABE

the calibration procedure. Compared to traditional ap- uncertainty or error in the measured input data (e.g., rainfall
proaches that do not include crop yield calibration, Nair et and temperature), (2) the uncertainty or error in the meas-
al. (2011) produced improved prediction efficiencies, espe- ured data used in model calibration (e.g., river discharges
cially for the nutrient balance. Faramarzi et al. (2009) and sediment load), and (3) the uncertainty or error in the
found that inclusion of irrigation made a significant differ- conceptual model and model parameters (e.g., hydrologic
ence in the simulation of hydrology and calculation of cor- processes). Abbaspour (2005) states that there is an inti-
rect evapotranspiration. mate relationship between calibration and uncertainty anal-
ysis and that they must be performed simultaneously. In
SPATIAL CALIBRATION other words, calibration must always be accompanied by an
Even when models are calibrated at multiple gauge sites assessment of the goodness of the calibration, taking into
within a watershed, further spatial calibration would im- account all modeling errors.
prove accuracy. In large river basins, rainfall, runoff, and The uncertainties in the conceptual model and model pa-
water yield can vary widely across the basin. When spatial rameters, as well as the uncertainty in measured data used
data are available for runoff, water yield, or ET, spatial cal- in calibration, all affect simulation quality and appropriate-
ibration at the subbasin level can be used to calibrate the ness; therefore, Harmel et al. (2010) developed a simple
local water balance better, which significantly improves the model evaluation matrix to incorporate data and simulation
temporal (time series) calibration of streamflow at the uncertainty in model evaluation and reporting. In addition,
gauges in the basin (Santhi et al., 2008). When such data the modified goodness-of-fit indicator calculations of Har-
are not available, calibration at multiple gauges can be used mel et al. (2010), which are based on Haan et al. (1995), are
to capture the spatial variation in flow, as reported by sev- currently being incorporated into the Abasspour et al.
eral authors (White and Chaubey, 2005; Qi and Grunwald, (2007) SWAT-CUP software.
2005; Santhi et al., 2001). Remotely sensed estimates of
ET, leaf area index, residue cover, and soil moisture have GUIDELINES FOR CALIBRATION AND
the potential to improve spatial calibration. Remotely VALIDATION PERIODS
sensed data could also be used for calibrating landscape Since it is impossible to replicate watersheds and river
processes. Land use export coefficients and point-source basins, common practice in hydrologic studies is to divide
loads should also be verified. Data are usually limited or the measured data either temporally or spatially for calibra-
unavailable; however, databases from research plots and tion and validation (Engel et al., 2007). One view suggests
small watersheds (e.g., MANAGE database; Harmel et al., that both wet and dry periods be included in both the cali-
2006b, 2008) have been assembled and are useful to ensure bration and validation periods (Gan et al., 1997), ensuring
reasonable load estimates from different land uses (HRUs) that both periods reflect the range of conditions under
within a watershed. which a model is expected to perform. This is often not fea-
sible due to limitations in the length of monitoring data
RUN TIME EFFICIENCY available for calibration and validation. Previous studies
Many of the autocalibration techniques require hundreds (Kannan et al., 2007; Van Liew and Garbrecht, 2003) rec-
or thousands of simulations to find the optimal solution. ommend including a wet period with high runoff events in
SWAT-CUP, through a parallel processing scheme devel- the calibration period. A contrasting view from Reckhow
oped for the Windows platform, allows individual runs to (1994) contends that validation conditions should be differ-
be sent to different processors, thus taking advantage of ent in the sense that the important processes and forcing
multiprocessor PCs, supercomputers, and clusters (Whit- functions or responses differ from the calibrated conditions,
taker et al., 2004). Yalew et al. (2010) split individual as the purpose of validation is to provide an independent
SWAT simulations into several submodels, ran the submod- assessment of model performance. There remains some
els in parallel, collected the subbasin outputs at a central confusion in the literature about what validation is and
computer, and then performed the routing. This technique what it means to validate a model (Rykiel 1996), and there
allows individual simulations to be parallelized and re- are currently no guidelines for separating measured data for
quires minor modification to the source code. Parallelizing calibration and validation. Research to determine the im-
the SWAT code by sending sections of code to different pact of selection of calibration and validation periods on
processors is also being examined. model parameterization would benefit the modeling com-
munity and advance the science of modeling. Guidelines
IMPACT OF UNCERTAINTY ON CALIBRATION for selection of the periods should consider recommending
AND DECISION MAKING a minimum length of period required for calibration and
Because models are used to develop and evaluate water validation. Such guidelines could be developed by conduct-
resource policy, several recent pleas have been made to ing additional model runs and analysis in watersheds with
consider inherent uncertainties in model development and extensive observed data.
application (e.g., Beven, 2006; Bende-Michl et al., 2011).
Definition and quantification of calibration uncertainty in AUTOMATED ERROR CHECKING
distributed hydrological modeling has become the subject Model interfaces and automated calibration routines
of much research in recent years (Abbaspour, 2005). Three have simplified SWAT calibration and validation such that
sources of uncertainty or error must be considered: (1) the the effort required is a fraction of that needed a decade ago.
These advances allow SWAT application by less-

55(4): 1491-1508 1503

experienced users and those without sufficient background tion techniques. However, the models are only tools to aid
in hydrology, sedimentology, soil science, and nutrient dy- in the decision process and are never a substitute for user
namics. In particular, the use of automated calibration understanding of the processes and management practices
software may produce simulated values that appear appro- occurring in the watershed to guide calibration. We recom-
priate because they adequately mimic the measured data mend that users study the watershed thoroughly, understand
used in calibration and validation, but the model may con- the processes involved, identify the specific project needs
tain input data errors and/or inappropriate parameter ad- and scenarios to be analyzed, parameterize SWAT for the
justments not readily identified by the user or the autocali- watershed, compare the model prediction with observed da-
bration software. SWAT Check (White et al., 2012) is a ta, and then develop and implement a calibration plan. The
stand-alone program that examines model output relative to SWAT-CUP case study shows the potential to combine spa-
typical ranges, creates process-based figures for visualiza- tial and process data along with user understanding of the
tion of output values, and detects common model applica- watershed into the autocalibration process. We recommend
tion problems. The program examines 56 model outputs continued addition of spatial process information to SWAT-
and summaries and prompts users if unusual values are en- CUP, and incorporating all the checks on processes and er-
countered. SWAT Check is currently in beta release, with rors from White et al. (2012).
updates pending based on user feedback. This software In many SWAT applications, additional fine-tuning of
should assist the SWAT community (especially new users) the calibration may be needed after scenario analysis is
in developing better model applications. completed. It is extremely important to remember that alt-
hough some SWAT input parameters are empirical, they are
MANUAL AND AUTOMATED CALIBRATION all physically based and must be kept within realistic rang-
Manual calibration of distributed watershed models like es as influenced by the uncertainty in defining the parame-
SWAT is difficult and almost infeasible in many large-scale ter.
applications. However, manual calibration forces the user
to better understand the model, the important processes in
the watershed, and parameter sensitivity. Tools like REFERENCES
VizSWAT (http://swatmodel.tamu.edu/software/vizswat. Abbaspour K. C. 2005. Calibration of hydrologic models: When is
aspx) can be used to visualize complex spatial data from a model calibrated? In Proc. Intl. Congress on Modelling and
HRUs, subbasins, and reaches. Van Liew et al. (2005) sug- Simulation (MODSIM’05), 2449-2455. A. Zerger and R. M.
gested that autocalibration be attempted first, followed by Argent, eds. Melbourne, Australia: Modelling and Simulation
manual calibration, to ensure that average annual means Society of Australia and New Zealand.
and the general balances are correct. Another approach is to Abbaspour, K. C., A. Johnson, and M. Th. van Genuchten. 2004.
perform manual calibration first on the average annual hy- Estimating uncertain flow and transport parameters using a
drologic balance and average annual loads (minimizing sequential uncertainty fitting procedure. Vadose Zone J. 3(4):
1340-1352.
percent bias). This approach was used by Jeong et al.
Abbaspour K. C., M. Vejdani, and S. Haghighat. 2007. SWAT-
(2010) for the calibration of flow using the subdaily rainfall CUP calibration and uncertainty programs for SWAT. In Proc.
and runoff version of SWAT. Autocalibration with a narrow Intl. Congress on Modelling and Simulation (MODSIM’07),
window of parameter ranges can then be performed to fine- 1603-1609. L. Oxley and D. Kulasiri, eds. Melbourne,
tune daily and subdaily statistics. Australia: Modelling and Simulation Society of Australia and
SWAT-CUP provides a decision-making framework that New Zealand.
incorporates a semi-automated approach (SUFI2) using Abbaspour, K. C., M. Faramarzi, S. S. Ghasemi, and H. Yang.
both manual and automated calibration incorporating sensi- 2009. Assessing the impact of climate change on water
tivity and uncertainty analysis. Users can manually adjust resources in Iran. Water Resour. Res. 45: W10434, doi:
10.1029/2008WR007615.
parameters and ranges iteratively between autocalibration
Ahl, R. S., S. W. Woods, and H. R. Zuuring. 2008. Hydrologic
runs. Users can also use output from sensitivity and uncer- calibration and validation of SWAT in a snow-dominated
tainty analysis as they iteratively move between manual Rocky Mountain watershed, Montana, U.S.A. J. American
and autocalibration. Parameter sensitivity analysis helps fo- Water Resour. Assoc. 44(6): 1411-1430.
cus the calibration, and uncertainty analysis is used to pro- Akhavan, S., J. Abedi-Koupai, S. F. Mousavi, M. Afyuni, S. S.
vide statistics for goodness-of-fit. The user interaction or Eslamian, and K. C. Abbaspour. 2010. Application of SWAT
manual component of the calibration forces the user to ob- model to investigate nitrate leaching in Hamadan-Bahar
tain a better understanding of the overall hydrologic pro- watershed, Iran. J. Agric. Ecosystem and Environ. 139(4): 675-
cesses (baseflow ratios, ET, sediment sources and sinks, 688.
Alibuyog, N. R., V. B. Ella, M. R. Reyes, R. Srinivasan , C.
crop yields, and nutrient balances) and of parameter sensi-
Heatwole, and T. Dillaha. 2009. Predicting the effects of land
tivity. By integrating these tools in the calibration process- use change on runoff and sediment yield in Manupali River
es, SWAT-CUP provides a powerful approach to watershed subwatersheds using the SWAT model. Intl. Agric. Eng. J.
calibration. 18(1-2): 15-25.
Andersson, J. C. M., A. J. B. Zehnder, G. P. W. Jewitt, and H.
WATERSHED CALIBRATION Yang. 2009. Water availability, demand, and reliability of in
The ultimate goal of calibrating a watershed model situ water harvesting in smallholder rainfed agriculture in the
should be to incorporate spatial processes into the calibra- Thukela River basin, South Africa. Hydrol. Earth System Sci.
13(12): 2329-2347.

1504 TRANSACTIONS OF THE ASABE

Andersson, J. C. M., A. J. B. Zehnder, J. Rockstrom, and H. Yang. 10(5): 363-374.
2011. Potential impacts of water harvesting and ecological Bende-Michl, U., M. Volk, R. D. Harmel, L. Newham, and T.
sanitation on crop yield, evaporation, and river flow regimes in Dalgaard. 2011. Monitoring strategies and scale-appropriate
the Thukela River basin, South Africa. Agric. Water Mgmt. hydrologic and biogeochemical modelling for natural resource
98(7): 1113-1124. management: Conclusions and recommendations from a
Arabi, M., R. S. Govindaraju, M. M. Hantush, and B. A. Engel. session held at the iEMSs 2008. Environ. Modelling Software
2006. Role of watershed subdivision on modeling the 26(4): 538-542.
effectiveness of best management practices with SWAT. J. Beven, K. 2006. On undermining the science? Hydrol. Proc.
American Water Resour. Assoc. 42(2): 513-528. 20(14): 3141-3146.
Arnold, J. G., and P. M. Allen. 1999. Automated methods for Bosch, D. D., J. M. Sheridan, H. L. Batten, and J. G. Arnold.
estimating baseflow and groundwater recharge from 2004. Evaluation of the SWAT model on a coastal plain
streamflow records. J. American Water Resour. Assoc. 35(2): agricultural watershed. Trans. ASAE 47(5): 1493-1506.
411-424. Bouraoui, F., L. Galbiati, and G. Bidoglio. 2002. Climate change
Arnold, J. G., and N. Fohrer. 2005. SWAT2000: Current impacts on nutrient loads in the Yorkshire Ouse catchment
capabilities and research opportunities in applied watershed (U.K.). Hydrol. Earth Syst. Sci. 6(2): 197-209.
modeling. Hydrol. Proc. 19(3): 563-572. Bouraoui, F., S. Benabdallah, A. Jrad, and G. Bidoglio. 2005.
Arnold, J. G., and J. R. Williams. 1987. Validation of SWRRB: Application of the SWAT model on the Medjerda River basin
Simulator for water resources in rural basins. J. Water Resour. (Tunisia). Physics Chem. Earth. 30(8-10): 497-507.
Plan. Manage. ASCE 113(2): 243-256. Boyle, D. P., H. V. Gupta, and S. Sorooshian. 2000. Toward
Arnold, J. G., P. M. Allen, R. Muttiah, and G. Bernhardt. 1995a. improved calibration of hydrologic models: Combining the
Automated baseflow separation and recession analysis strengths of manual and automatic methods. Water Resour.
techniques. Ground Water 33(6): 1010-1018. Res. 36(12): 3663-3674.
Arnold, J. G., J. R. Williams, and D. R. Maidment. 1995b. Brown, L. C., and T. O. Barnwell Jr. 1987. The enhanced stream
Continuous-time water and sediment-routing model for large water quality models QUAL2E and QUAL2E-UNCAS:
basins. J. Hydraul. Eng. ASCE 121(2): 171-183. Documentation and user manual. EPA/600/3-87/007. Athens,
Arnold, J. G., R. Srinivasan, R. S. Muttiah, and J. R. Williams. Ga.: U.S. EPA, Environmental Research Laboratory.
1998. Large-area hydrologic modeling and assessment: Part I. Cao, W., W. B. Bowden, T. Davie, and A. Fenemor. 2006. Multi-
Model development. J. American Water Resour. Assoc. 34(1): variable and multi-site calibration and validation of SWAT in a
73-89. large mountainous catchment with high spatial variability.
Arnold, J. G., R. S. Muttiah, R. Srinivasan, and P. M. Allen. 2000. Hydrol. Proc. 20(5): 1057-1073.
Regional estimation of baseflow and groundwater recharge in Cheng, H., W. Ouyang, F. Hao, X. Ren, and S. Yang. 2007. The
the Upper Mississippi River basin. J. Hydrol. 227(1-4): 21-40. nonpoint-source pollution in livestock-breeding areas of the
Arnold, J. G., P. M. Allen, and D. S. Morgan. 2001. Hydrologic Heihe River basin in Yellow River. Stochastic Environ. Res.
model for design and constructed wetlands. Wetlands 21(2): Risk Assess. 21(3): 213-221.
167-178. Chiang, L., I. Chaubey, M. W. Gitau, and J. G. Arnold. 2010.
Arnold, J. G., P. M. Allen, M. Volk, J. R. Williams, and D. D. Differentiating impacts of land use changes from pasture
Bosch. 2010a. Assessment of different representations of management in a CEAP watershed using the SWAT model.
spatial variability on SWAT model performance. Trans. ASABE Trans. ASABE 53(5): 1569-1584.
53(5): 1433-1443. Chin, D. A., D. Sakura-Lemmessy, D. D. Bosch, and P. A. Gay.
Arnold, J. G., P. W. Gassman, and M. J. White. 2010b. New 2009. Watershed-scale fate and transport of bacteria. Trans.
developments in the SWAT ecohydrology model. In Proc. 21st ASABE 52(1): 145-154.
Watershed Technology Conf.: Improving Water Quality and Chu, T. W., A. Shirmohammadi, H. Montas, and A. Sadeghi. 2004.
Environment. ASABE Publication No. 701P0210cd. St. Evaluation of the SWAT model’s sediment and nutrient
Joseph, Mich.: ASABE. components in the Piedmont physiographic region of
ASCE. 1993. Criteria for evaluation of watershed models. J. Irrig. Maryland. Trans. ASAE 47(5): 1523-1538.
Drainage Eng. 119(3): 429-442. Coffey, M. E., S. R. Workman, J. L. Taraba, and A. W. Fogle.
Balascio, C. C., D. J. Palmeri, and H. Gao. 1998. Use of a genetic 2004. Statistical procedures for evaluating daily and monthly
algorithm and multi-objective programming for calibration of hydrologic model predictions. Trans. ASAE 47(1): 59-68.
a hydrologic model. Trans. ASAE 41(3): 615-619. Coffey, R., E. Cummins, N. Bhreathnach, V. O. Flaherty, and M.
Behera, S., and R. K. Panda. 2006. Evaluation of management Cormican. 2010. Development of a pathogen transport model
alternatives for an agricultural watershed in a subhumid for Irish catchments using SWAT. Agric. Water Mgmt. 97(1):
subtropical region using a physical process-based model. 101-111.
Agric. Ecosystems Environ. 113(1-4): 62-72. Conan, C., G. de Marsily, F. Bouraoui, and G. Bidoglio. 2003. A
Bekele, E. G., and H. V. Knapp. 2010. Watershed modeling to long-term hydrological modelling of the Upper Guadiana
assessing impacts of potential climate change on water supply River basin (Spain). Physics Chem. Earth 28(4-5): 193-200.
availability. Water Resour. Mgmt. 24(13): 3299-3320. Debele, B., R. Srinivasan, and J. Y. Parlange. 2008. Coupling
Bekele, E. G., and J. W. Nicklow. 2007. Multi-objective automatic upland watershed and downstream waterbody hydrodynamic
calibration of SWAT using NSGA-II. J. Hydrol. 341(3-4): 165- and water quality models (SWAT and CE-QUAL-W2) for
176. better water resources management in complex river basins.
Benaman, J., and C. A. Shoemaker. 2004. Methodology for Environ. Modeling Assess. 13(1): 135-153.
analyzing ranges of uncertain model parameters and their Di Luzio, M., and J. G. Arnold. 2004. Formulation of a hybrid
impact on total maximum daily load process. Reston, Va.: calibration approach for a physically based distributed model
ASCE. with NEXRAD data input. J. Hydrol. 298(1-4): 136-154.
Benaman, J., C. A. Shoemaker, and. D. A. Haith. 2005. Doherty, J. 2004. PEST: Model-Independent Parameter
Calibration and validation of Soil and Water Assessment Tool Estimation User Manual. 5th ed. Brisbane, Australia:
on an agricultural watershed in upstate New York. J. Hydrol. Watermark Numerical Computing. Available at: www.pest

55(4): 1491-1508 1505

 homepage.org/Model-independence.php. Accessed 25 June BMPs at the farm-scale level. Trans. ASABE 51(6): 1925-
2012. 1936.
Douglas-Mankin, K. R., R. Srinivasan, and J. G. Arnold. 2010. Govender, M., and C. S. Everson. 2005. Modelling streamflow
Soil and Water Assessment Tool (SWAT) model: Current from two small South African experimental catchments using
development and applications. Trans. ASABE 53(5): 1423- the SWAT model. Hydrol. Proc. 19(3): 683-692.
1431. Green, C. H., and A. van Griensven. 2008. Autocalibration in
Du, B., A. Saleh, D. B. Jaynes, and J. G. Arnold. 2006. Evaluation hydrologic modeling: Using SWAT2005 in small-scale
of SWAT in simulating nitrate nitrogen and atrazine fates in a watersheds. Environ. Modelling Software 23(4): 422-434.
watershed with tiles and potholes. Trans. ASABE 49(4): 949- Green, C. H., M. D. Tomer, M. Di Luzio, and J. G. Arnold. 2006.
959. Hydrologic evaluation of the Soil and Water Assessment Tool
Duan, Q. D. 2003. Global optimization for watershed model for a large tile-drained watershed in Iowa. Trans. ASABE
calibration. In Calibration of Watershed Models, 89-104. 49(2): 413-422.
Water Science and Application Series, Vol. 6. Washington, Green, C. H., J. G, Arnold, J. R. Williams, R. Haney, and R. D.
D.C.: American Geophysical Union. Harmel. 2007. Soil and Water Assessment Tool hydrologic and
Duan, Q. D., V. K. Gupta, and S. Sorooshian. 1992. Effective and water quality evaluation of poultry litter application to small-
efficient global optimization for conceptual rainfall-runoff scale subwatersheds in Texas. Trans. ASABE 50(4): 1199-
models. Water Resour. Res. 28(4): 1015-1031. 1209.
Easton, Z. M., D. R. Fuka, M. T. Walter, D. M. Cowan, E. M. Green, W. H., and G. A. Ampt. 1911. Studies on soil physics: 1.
Schneiderman, and T. S. Steenhuis. 2008. Re-conceptualizing The flow of air and water through soils. J. Agric. Sci. 4(1): 11-
the Soil and Water Assessment Tool (SWAT) model to predict 24.
runoff from variable source areas. J. Hydrol. 348(3-4): 279- Grizzetti, B., F. Bouraoui, and G. De Marsily. 2005. Modelling
291. nitrogen pressure in river basins: A comparison between a
Eawag. 2009. SWAT-CUP. Dübendorf, Switzerland: Swiss statistical approach and the physically based SWAT model.
Federal Institute of Aquatic Science and Technology. Available Physics Chem. Earth 30(8-10): 508-517.
at: www.eawag.ch/organisation/abteilungen/siam/software/ Haan, C. T., B. Allred, D. E. Storm, G. J. Sabbagh, and S. Prahhu.
swat/index_EN. 1995. Statistical procedure for evaluating hydrologic/water
Eckhardt, K., and J. G. Arnold. 2001. Automatic calibration of a quality models. Trans. ASAE 38(3): 725-733.
distributed catchment model. J. Hydrol. 251(1-2): 103-109. Harmel, R. D., R. J. Cooper, R. M. Slade, R. L. Haney, and J. G.
Eckhardt, K., S. Haverkamp, N. Fohrer, and H. G. Frede. 2002. Arnold. 2006a. Cumulative uncertainty in measured
SWAT-G, a version of SWAT99.2 modified for application to streamflow and water quality data for small watersheds. Trans.
low mountain range catchments. Physics Chem. Earth 27(9- ASABE 49(3): 681-701.
10): 641-644. Harmel, R. D., S. Potter, P. Casebolt, K. Reckhow, C. Green, and
Eckhardt, K., N. Fohrer, and H. G. Frede. 2005. Automatic model R. Haney. 2006b. Compilation of measured nutrient load data
calibration. Hydrol. Proc. 19(3): 651-658. for agricultural land uses in the United States. J. American
Engel, B. A., R. Srinivasan, J. G. Arnold, C. Rewerts, and S. J. Water Resour. Assoc. 42(5): 1163-1178.
Brown. 1993. Nonpoint-source (NPS) pollution modeling Harmel, R. D., S. Qian, K. Reckhow, and P. Casebolt. 2008. The
using models integrated with geographic information systems MANAGE database: Nutrient load and site characteristics
(GIS). Water Sci. Tech. 28(3-5): 685-690. updates and runoff concentration data. J. Environ. Qual. 37(6):
Engel, B., D. Storm, M. White, J. Arnold, and M. Arabi. 2007. A 2403-2406.
hydrologic/water quality model application protocol. J. Harmel, R. D., D. R. Smith, K. W. King, and R. M. Slade. 2009.
American Water Resour. Assoc. 43(5): 1223-1236. Estimating storm discharge and water quality data uncertainty:
Faramarzi, M., K. C. Abbaspour, R. Schulin, and H. Yang. 2009. A software tool for monitoring and modeling applications.
Modeling blue and green water availability in Iran. Hydrol. Environ. Modelling Software 24(7): 832-842.
Proc. 23(3): 486-501. Harmel, R. D., P. K. Smith, and K. L. Migliaccio. 2010.
Gan, T. Y., E. M. Dlamini, and G. F. Biftu. 1997. Effects of model Modifying goodness-of-fit indicators to incorporate both
complexity and structure, data quality, and objective functions measurement and model uncertainty in model calibration and
on hydrologic modeling. J. Hydrol. 192(1): 81-103. validation. Trans. ASABE 53(1): 55-63.
Gassman, P. W., M. Reyes, C. H. Green, and J. G. Arnold. 2007. Heuvelmans, G., B. Muys, and J. Feyen. 2004. Evaluation of
The Soil and Water Assessment Tool: Historical development, hydrological model parameter transferability for simulating the
applications, and future directions. Trans. ASABE 50(4): 1211- impact of land use on catchment hydrology. Physics Chem.
1250. Earth 29(11-12): 739-747.
Gassman, P. W., J. G. Arnold, R. Srinivasan, and M. Reyes. 2010. Heuvelmans, G., B. Muys, and J. Feyen. 2006. Regionalisation of
The worldwide use of the SWAT model: Technological drivers, the parameters of a hydrological model: Comparison of linear
networking impacts, and simulation trends. In Proc. 21st regression models with artificial neural nets. J. Hydrol. 319(1-
Century Watershed Technology: Improving Water Quality and 4): 245-265.
Environment. ASABE Publication No. 701P0210cd. St. Hu, X., G. F. McIsaac, M. B. David, and C. A. L. Louwers. 2007.
Joseph, Mich.: ASABE. Modeling riverine nitrate export from an east-central Illinois
Ghaffari, G., S. Keesstra, J. Ghodousi, and H. Ahmadi. 2010. watershed using SWAT. J. Environ. Qual. 36(4): 996-1005.
SWAT-simulated hydrological impact of land-use change in Inamdar, S. and A. Naumov. 2006. Assessment of sediment yields
the Zanjanrood basin, northwest Iran. Hydrol. Proc. 24(7): for a mixed land use Great Lakes watershed: Lessons from
892-903. field measurements and modeling. J. Great Lakes Res. 32(3):
Gikas, G., T. Yiannakopoulou, and V. Tsihrintzis. 2006. Modeling 471-488.
of nonpoint-source pollution in a Mediterranean drainage Jeong, J., N. Kannan, J. G. Arnold, R. Glick, L. Gosselink, and R.
basin. Environ. Modeling Assess. 11(3): 219-233. Srninvasan. 2010. Development and integration of subhourly
Gitau, M. W., W. J. Gburek, and P. L. Bishop. 2008. Use of the rainfall-runoff modeling capability within a watershed model.
SWAT model to quantify water quality effects of agricultural Water Resour. Mgmt. 24(15): 4505-4527.

1506 TRANSACTIONS OF THE ASABE

Jeong, J., C. Santhi, J. G. Arnold, R. Srinivasan, S. Pradhan, and Fausey. 2011. Importance of crop yield in calibrating
K. Flynn. 2011. Development of algorithms for modeling watershed water quality simulation models. J. American Water
onsite wastewater systems within SWAT. Trans. ASABE 54(5): Res. Assoc. 47(6): 1285-1297.
1693-1704. Narasimhan, B., R. Srinivasan, S. T. Bednarz, M. R. Ernst, and P.
Jha, M. K., K. E. Schilling, P. W. Gassman, and C. F. Wolter. M. Allen. 2010. A comprehensive modeling approach for
2010. Targeting land-use change for nitrate-nitrogen load reservoir water quality assessment and management due to
reductions in an agricultural watershed. J. Soil Water Cons. point and nonpoint source pollution. Trans. ASABE 53(5):
65(6): 342-352. 1605-1617.
Kannan, N., S. M. White, F. Worrall, and M. J. Whelan. 2007. Nash, J. E., and J. E. Sutcliffe. 1970. River flow forecasting
Sensitivity analysis and identification of the best through conceptual models: Part I. A discussion of principles.
evapotranspiration and runoff options for hydrological J. Hydrol. 10(3): 282-290.
modeling in SWAT-2000. J. Hydrol. 332(3-4): 456-466. Neitsch, S. L., J. G. Arnold, J. R. Kiniry, R. Srinivasan, and J. R.
Kemanian, A. R., S. Julich, V. S. Manoranjan, and J. G. Arnold. Williams. 2002. Soil and Water Assessment Tool, User
2011. Integrating soil carbon cycling with that of nitrogen and Manual, Version 2000. Temple, Tex.: Grassland, Soil and
phosphorus in the watershed model SWAT: Theory and model Water Research Laboratory.
testing. Ecol. Modelling 222(12): 1913-1921. Ng, T. L., J. W. Eheart, X. Cai, and F. Miguez. 2010a. Modeling
Knisel, W. G. 1980. CREAMS: A field-scale model for chemicals, Miscanthus in the Soil and Water Assessment Tool (SWAT) to
runoff, and erosion from agricultural management systems. simulate its water quality effects as a bioenergy crop. Environ.
Conservation Research Report No. 26. Washington, D.C.: Sci. Tech. 44(18): 7138-7144.
USDA National Resources Conservation Service. Ng, T. L., J. W. Eheart, and X. M. Cai. 2010b. Comparative
Krause, P., D. P. Boyle, and F. Base. 2005. Comparison of calibration of a complex hydrologic model by stochastic
different efficiency criteria for hydrological model assessment. methods GLUE and PEST. Trans. ASABE 53(6): 1773-1786.
Adv. Geosci. 5: 89-97. Olivera, F., M. Valenzuela, R. Srinivasn, J. Choi, H. Cho, S. Koka,
Krysanova, V., and J. G. Arnold. 2008. Advances in and A. Agrawal. 2006. ArcGIS-SWAT: A geodata model and
ecohydrological modeling with SWAT: A review. Hydrol. Sci. GIS interface for SWAT. J. American Water Resources Assoc.
J. 53(5): 939-947. 42(2): 295-309.
Legates, D. R., and G. J. McCabe. 1999. Evaluating the use of Parajuli, P. B., N. O. Nelson, L. D. Frees, and K. R. Mankin.
“goodness-of-fit” measures in hydrologic and hydroclimatic 2009. Comparison of AnnAGNPS and SWAT model
model validation. Water Resour. Res. 35(1): 233-241. simulation results in USDA-CEAP agricultural watersheds in
Lemonds, P. J., and J. E. McCray. 2007. Modeling hydrology in a south-central Kansas. Hydrol. Proc. 23(5): 748-763.
small rocky mountain watershed serving large urban Qi, C., and S. Grunwald. 2005. GIS-based hydrologic modeling in
populations. J. American Water Resour. Assoc. 43(4): 875-887. the Sandusky watershed using SWAT. Trans. ASAE 48(1): 169-
Leonard, R. A., W. G. Knisel, and D. A. Still. 1987. GLEAMS: 180.
Groundwater loading effects on agricultural management Razavi, S., B. A. Tolson, L. S. Matott, N. R. Thomson, A.
systems. Trans. ASAE 30(5): 1403-1418. MacLean, and F. R. Seglenieks. 2010. Reducing the
Ma, L., J. C. Ascough II, L. R. Ahuja, M. J. Shaffer, J. D. Hanson, computational cost of automatic calibration through model
and K. W. Rojas. 2000. Root Zone Water Quality Model preemption. Water Resour. Res. 46: W11523, doi:
sensitivity analysis using Monte Carlo simulation. Trans. 10.1029/2009WR008957.
ASAE 43(4): 883-895. Reckhow, K. H. 1994. Water quality simulation modeling and
Maski, D., K. R. Mankin, K. A. Janssen, P. Tuppad, and G. M. uncertainty analysis for risk assessment and decision making.
Pierzynski. 2008. Modeling runoff and sediment yields from Ecol. Model. 72(1-2): 1-20.
combined in-field crop practices using the Soil and Water Refsgaard, J. C. 1997. Parameterisation, calibration, and
Assessment Tool. J. Soil Water Cons. 63(4): 193-203. validation of distributed hydrological models. J. Hydrol.
Meng, H., A. M. Sexton, M. C. Maddox, A. Sood, C. W. Brown, 198(1): 69-97.
R. R. Ferraro, and R. Murtugudde. 2010. Modeling Rouholahnejad, E., K. C. Abbaspour, M. Vejdani, R. Srinivasan,
Rappahannock River basin using SWAT-Pilot for Chesapeake R. Schulin, and A. Lehmann. 2012a. A parallelization
Bay watershed. Trans. ASAE 26(5): 795-805. framework for calibration of hydrological models. Environ.
Moriasi, D. N., J. G. Arnold, M. W. Van Liew, R. L. Binger, R. D. Modelling Software 31: 28-36.
Harmel, and T. Veith. 2007. Model evaluation guidelines for Rouholahnejad, E., K. C. Abbaspour, R. Srinivasan, R. Schulin,
systematic quantification of accuracy in watershed and A. Lehmann. 2012b. Using a hydrological model of the
simulations. Trans. ASABE 50(3): 885-900. Danube River basin to calculate nitrate and phosphate loads. J.
Moriasi, D. N., J. G. Arnold, G. G. Vasquez-Amabile, B. A. Hydrol. (in preparation).
Engel, and C. G. Rossi. 2011. Shallow water table depth Rykiel, E. J. 1996. Testing ecological models: The meaning of
algorithm in SWAT: Recent developments. Trans. ASABE validation. Ecol. Model. 90(3): 229-244.
54(5): 1705-1711. Santhi, C., J. G. Arnold, J. R. Williams, W. A. Dugas, R.
Moriasi, D. N., C. G. Rossi, J. G. Arnold, and M. D. Tomer. 2012. Srinivasan, and L. M. Hauck. 2001. Validation of the SWAT
Evaluating hydrology of SWAT with new tile drain equations. model on a large river basin with point and nonpoint sources.
J. Soil Water Cons. (accepted). J. American Water Resour. Assoc. 37(5): 1169-1188.
Mukundan, R., D. E. Radcliffe, and L. M. Risse. 2010. Spatial Santhi, C., N. Kannan, J. G. Arnold, and M. Di Luzio. 2008.
resolution of soil data and channel erosion effects on SWAT Spatial calibration and temporal validation of flow for
model predictions of flow and sediment. J. Soil Water Cons. regional-scale hydrologic modeling. J. American Water
65(2): 92-104. Resour. Assoc. 44(4): 829-846.
Muleta, M. K., and J. W. Nicklow. 2005. Sensitivity and Schuol, J., K. C. Abbaspour, R. Srinivasan, and H. Yang. 2008a.
uncertainty analysis coupled with automatic calibration for a Estimation of freshwater availability in the west African
distributed watershed model. J. Hydrol. 306(1-4): 127-145. subcontinent using the SWAT hydrologic model. J. Hydrol.
Nair, S. S., K. W. King, J. D. Witter, B. L. Sohngen, and N. R. 352(1-2): 30-49.

55(4): 1491-1508 1507

Schuol, J., K. C. Abbaspour, H. Yang, R. Srinivasan, and A. J. B. area data: Can it make a difference in calibrating the
Zhender. 2008b. Modeling blue and green water availability in hydrologic response for watershed-scale modeling? J. Soil
Africa. Water Resour. Res. 44: W07406, doi: Water Cons. 62(3): 162-170.
10.1029/2007WR006609. Wang, X., and A. M. Melesse. 2005. Evaluation of the SWAT
Setegn, S. G., R. Srinivasan, and B. Dargahi. 2009. Hydrological model’s snowmelt hydrology in a northwestern Minnesota
modelling in the Lake Tana basin, Ethiopia, using SWAT watershed. Trans. ASAE 48(4): 1359-1376.
model. Open Hydrol. J. 2: 49-62. Wang, X., N. Kannan, C. Santhi, S. R. Potter, J. R. Williams, and
Shoemaker, C. A., R. G. Regis, and R. C. Fleming. 2007. J. G. Arnold. 2011. Integrating APEX output for cultivated
Watershed calibration using multistart local optimization and cropland with SWAT simulation for regional modeling. Trans.
evolutionary optimization with radial basis function ASABE 54(4): 1281-1298.
approximation. Hydrol. Sci. 52(3): 450-465. White, K. L., and I. Chaubey. 2005. Sensitivity analysis,
Srinivasan, R., and J. G. Arnold. 1994. Integration of a basin-scale calibration, and validations for a multisite and multivariable
water quality model with GIS. Water Resour. Bull. 30(3): 453- SWAT model. J. American Water Resour. Assoc. 41(5): 1077-
462. 1089.
Srinivasan, R., X. Zhang, and J. G. Arnold. 2010. SWAT White, M. J., and J. G. Arnold. 2009. Development of a simplistic
ungauged: Hydrological budget and crop yield predictions in vegetative filter strip model for sediment and nutrient retention
the Upper Mississippi River basin. Trans. ASABE 53(5): 1533- at the field scale. Hydrol. Proc. 23(11): 1602-1616.
1546. White, M. J., D. E. Storm, P. R. Busteed, M. D. Matlock, and R.
Starks, P. J., and D. N. Moriasi. 2009. Spatial resolution effect of R. West. 2011. Evaluating potential phosphorus management
precipitation data on SWAT calibration and performance: impacts in the Lake Eucha basin using SWAT. Trans. ASABE
Implications for CEAP. Trans. ASABE 52(4): 171-1180. 54(3): 827-835.
Sudheer, K. P., I. Chaubey, V. Garg, and K. W. Migliaccio. 2007. White, M. J., R. D. Harmel, J. G. Arnold, and J. R. Williams.
Impact of time scale of the calibration objective function on 2012. SWAT Check: A screening tool to assist users in the
the performance of watershed models. Hydrol. Proc. 21(25): identification of potential model application problems. J.
3409-3419. Environ. Quality (in press). Available at: http://swatmodel.
Sui, Y., and J. R. Frankenberger. 2008. Nitrate loss from tamu.edu/software/swat-check.
subsurface drains in an agricultural watershed using Whittaker, G. R. 2004. Use of a Beowulf cluster for estimation of
SWAT2005. Trans. ASABE 51(4): 1263-1272. risk using SWAT. Agron. J. 95(5): 1495-1497.
Tuppad, P., K. R. Douglas-Mankin, T. Lee, R. Srinivasan, and J. Williams, J. R., and H. D. Berndt. 1977. Sediment yield prediction
G. Arnold. 2011. Soil and Water Assessment Tool (SWAT) based on watershed hydrology. Trans. ASAE 20(6): 1100-1104.
hydrologic/water quality model: Extended capability and wider Williams, J. R., J. G. Arnold, J. R. Kiniry, P. W. Gassman, and C.
adoption. Trans. ASABE 54(5): 1677-1684. H. Green. 2008. History of model development at Temple,
van Griensven, A. 2002. Developments towards integrated water Texas. Hydrol. Sci. 53(5): 948-960.
quality modeling for river basins. Publication No. 40. Brussels, Yalew, S. G., A. van Griensven, and L. Kokoszkiewcz. 2010.
Belgium: Vrije Universiteit, Department of Hydrology and Parallel computing of a large-scale spatially distributed model
Hydraulic Engineering. using the Soil and Water Assessment Tool (SWAT). In Proc.
van Griensven, A., and W. Bauwens. 2003. Multiobjective 2010 Intl. Congress on Environ. Modeling and Software:
autocalibration for semidistributed water quality models. Water Modeling for Environment’s Sake. Manno, Switzerland:
Resour. Res. 39(12): 1348-1356. International Environmental Modeling and Software Society.
van Griensven, A., and W. Bauwens. 2005. Application and Available at: www.iemss.org/iemss2010/index.php?n=Main.
evaluation of ESWAT on the Dender basin and Wister Lake Proceedings. Accessed 1 October 2011.
basin. Hydrol. Proc. 19(3): 827-838. Yang, J., K. C. Abbaspour, P. Reichert, and H. Yang. 2008.
van Griensven, A., T. Meixner, R. Srinivasan, and S. Grunwald. Comparing uncertainty analysis techniques for a SWAT
2008. Fit-for-purpose analysis of uncertainty using split- application to Chaohe basin in China. J. Hydrol. 358(1-2): 1-
sampling evaluations. Hydrol. Sci. 53(5): 1090-1103. 23.
Van Liew, M. W., and J. Garbrecht. 2003. Hydrologic simulation Zhang, X., R. Srinivasan, B. Debele, and F. Hao. 2008a. Runoff
of the Little Washita River experimental watershed using simulation of the headwaters of the Yellow River using the
SWAT. J. American Water Resour. Assoc. 39(2): 413-426. SWAT model with three snowmelt algorithms. J. American
Van Liew, M. W., J. G. Arnold, and D. D. Bosch. 2005. Problems Water Resour. Assoc. 44(1): 48-61.
and potential of autocalibrating a hydrologic model. Trans. Zhang, X., R. Srinivasan, and M. Van Liew. 2008b. Multisite
ASAE 48(3): 1025-1040. calibration of the SWAT model for hydrologic modeling.
Van Liew, M. W., C. H. Green, and P. J. Starks. 2007. Unit source Trans. ASABE 51(6): 2039-2049.

1508 TRANSACTIONS OF THE ASABE