COMPARISON OF PROCESS‐BASED AND ARTIFICIAL NEURAL NETWORK APPROACHES FOR STREAMFLOW MODELING IN AN AGRICULTURAL WATERSHED1

ABSTRACT: The performance of the Soil and Water Assessment Tool (SWAT) and artificial neural network (ANN) models in simulating hydrologic response was assessed in an agricultural watershed in southeastern Pennsylvania. All of the performance evaluation measures including Nash‐Sutcliffe coefficient of efficiency (E) and coefficient of determination (R²) suggest that the ANN monthly predictions were closer to the observed flows than the monthly predictions from the SWAT model. More specifically, monthly streamflow E and R² were 0.54 and 0.57, respectively, for the SWAT model calibration period, and 0.71 and 0.75, respectively, for the ANN model training period. For the validation period, these values were ?0.17 and 0.34 for the SWAT and 0.43 and 0.45 for the ANN model. SWAT model performance was affected by snowmelt events during winter months and by the model's inability to adequately simulate base flows. Even though this and other studies using ANN models suggest that these models provide a viable alternative approach for hydrologic and water quality modeling, ANN models in their current form are not spatially distributed watershed modeling systems. However, considering the promising performance of the simple ANN model, this study suggests that the ANN approach warrants further development to explicitly address the spatial distribution of hydrologic/water quality processes within watersheds.     

Evaluation of Existing and Modified Wetland Equations in the SWAT Model

The ability to accurately simulate flow and nutrient removal in treatment wetlands within an agricultural, watershed‐scale model is needed to develop effective plans for meeting nutrient reduction goals associated with protection of drinking water supplies and reduction of the Gulf of Mexico hypoxic zone. The objectives of this study were to incorporate new equations for wetland hydrology and nutrient removal in Soil and Water Assessment Tool (SWAT), compare model performance using original and improved equations, and evaluate the ramifications of errors in watershed and tile drain simulation on prediction of NO₃‐N dynamics in wetlands. The modified equations produced Nash‐Sutcliffe Efficiency values of 0.88 to 0.99 for daily NO₃‐N load predictions, and percent bias values generally less than 6%. However, statistical improvement over the original equations was marginal and both old and new equations provided accurate simulations. The new equations reduce the model's dependence on detailed monitoring data and hydrologic calibration. Additionally, the modified equations increase SWAT's versatility by incorporating a weir equation and an irreducible nutrient concentration and temperature coefficient. Model improvements enhance the utility of SWAT for simulating flow and nutrients in wetlands and other impoundments, although performance is limited by the accuracy of inflow and NO₃‐N predictions from the contributing watershed. Editor's note : This paper is part of the featured series on SWAT Applications for Emerging Hydrologic and Water Quality Challenges. See the February 2017 issue for the introduction and background to the series.     

HYDROLOGIC MODELING OF TWO GLACIATED WATERSHEDS IN NORTHEAST PENNSYLVANIA1

ABSTRACT: A hydrologic modeling study, using the Hydrologic Simulation Program - FORTRAN (HSPF), was conducted in two glaciated watersheds, Purdy Creek and Ariel Creek in northeastern Pennsylvania. Both watersheds have wetlands and poorly drained soils due to low hydraulic conductivity and presence of fragipans. The HSPF model was calibrated in the Purdy Creek watershed and verified in the Ariel Creek watershed for June 1992 to December 1993 period. In Purdy Creek, the total volume of observed stream-flow during the entire simulation period was 13.36 × 10⁶ m³ and the simulated streamflow volume was 13.82 × 10⁶ m³ (5 percent difference). For the verification simulation in Ariel Creek, the difference between the total observed and simulated flow volumes was 17 percent. Simulated peak flow discharges were within two hours of the observed for 30 of 46 peak flow events (discharge greater than 0.1 m³/sec) in Purdy Creek and 27 of 53 events in Ariel Creek. For 22 of the 46 events in Purdy Creek and 24 of 53 in Ariel Creek, the differences between the observed and simulated peak discharge rates were less than 30 percent. These 22 events accounted for 63 percent of total volume of streamflow observed during the selected 46 peak flow events in Purdy Creek. In Ariel Creek, these 24 peak flow events accounted for 62 percent of the total flow observed during all peak flow events. Differences in observed and simulated peak flow rates and volumes (on a percent basis) were greater during the snowmelt runoff events and summer periods than for other times.     

WATERSHED SCALE MODELING OF CRITICAL SOURCE AREAS OF RUNOFF GENERATION AND PHOSPHORUS TRANSPORT1

ABSTRACT: A curve number based model, Soil and Water Assessment Tool (SWAT), and a physically based model, Soil Moisture Distribution and Routing (SMDR), were applied in a headwater watershed in Pennsylvania to identify runoff generation areas, as runoff areas have been shown to be critical for phosphorus management. SWAT performed better than SMDR in simulating daily streamflows over the four‐year simulation period (Nash‐Sutcliffe coefficient: SWAT, 0.62; SMDR, 0.33). Both models varied streamflow simulations seasonally as precipitation and watershed conditions varied. However, levels of agreement between simulated and observed flows were not consistent over seasons. SMDR, a variable source area based model, needs further improvement in model formulations to simulate large peak flows as observed. SWAT simulations matched the majority of observed peak flow events. SMDR overpredicted annual flow volumes, while SWAT underpredicted the same. Neither model routes runoff over the landscape to water bodies, which is critical to surface transport of phosphorus. SMDR representation of the watershed as grids may allow targeted management of phosphorus sources. SWAT representation of fields as hydrologic response units (HRUs) does not allow such targeted management.     

Hydrologic Calibration and Validation of SWAT in a Snow‐Dominated Rocky Mountain Watershed,Montana, U.S.A.1

Abstract: The Soil and Water Assessment Tool (SWAT) has been applied successfully in temperate environments but little is known about its performance in the snow‐dominated, forested, mountainous watersheds that provide much of the water supply in western North America. To address this knowledge gap, we configured SWAT to simulate the streamflow of Tenderfoot Creek (TCSWAT). Located in central Montana, TCSWAT represents a high‐elevation watershed with ～85% coniferous forest cover where more than 70% of the annual precipitation falls as snow, and runoff comes primarily from spring snowmelt. Model calibration using four years of measured daily streamflow, temperature, and precipitation data resulted in a relative error (RE) of 2% for annual water yield estimates, and mean paired deviations (Dv) of 36 and 31% and Nash‐Sutcliffe (NS) efficiencies of 0.90 and 0.86 for monthly and daily streamflow, respectively. Model validation was conducted using an additional four years of data and the performance was similar to the calibration period, with RE of 4% for annual water yields, Dv of 43% and 32%, and NS efficiencies of 0.90 and 0.76 for monthly and daily streamflow, respectively. An objective, regression‐based model invalidation procedure also indicated that the model was validated for the overall simulation period. Seasonally, SWAT performed well during the spring and early summer snowmelt runoff period, but was a poor predictor of late summer and winter base flow. The calibrated model was most sensitive to snowmelt parameters, followed in decreasing order of influence by the surface runoff lag, ground water, soil, and SCS Curve Number parameter sets. Model sensitivity to the surface runoff lag parameter reflected the influence of frozen soils on runoff processes. Results indicated that SWAT can provide reasonable predictions of annual, monthly, and daily streamflow from forested montane watersheds, but further model refinements could improve representation of snowmelt runoff processes and performance during the base flow period in this environment.     

Modeling the Spatially Varying Water Balance Processes in a Semiarid Mountainous Watershed of Idaho1

Stratton, Benjamin T., Venakataramana Sridhar, Molly M. Gribb, James P. McNamara, and Balaji Narasimhan, 2009. Modeling the Spatially Varying Water Balance Processes in a Semiarid Mountainous Watershed of Idaho. Journal of the American Water Resources Association (JAWRA) 45(6):1390‐1408. Abstract: The distributed Soil Water Assessment Tool (SWAT) hydrologic model was applied to a research watershed, the Dry Creek Experimental Watershed, near Boise Idaho to investigate its water balance components both temporally and spatially. Calibrating and validating SWAT is necessary to enable our understanding of the water balance components in this semiarid watershed. Daily streamflow data from four streamflow gages were used for calibration and validation of the model. Monthly estimates of streamflow during the calibration phase by SWAT produced satisfactory results with a Nash Sutcliffe coefficient of model efficiency 0.79. Since it is a continuous simulation model, as opposed to an event‐based model, it demonstrated the limited ability in capturing both streamflow and soil moisture for selected rain‐on‐snow (ROS) events during the validation period between 2005 and 2007. Especially, soil moisture was generally underestimated compared with observations from two monitoring pits. However, our implementation of SWAT showed that seasonal and annual water balance partitioning of precipitation into evapotranspiration, streamflow, soil moisture, and drainage was not only possible but closely followed the trends of a typical semiarid watershed in the intermountain west. This study highlights the necessity for better techniques to precisely identify and drive the model with commonly observed climatic inversion‐related snowmelt or ROS weather events. Estimation of key parameters pertaining to soil (e.g., available water content and saturated hydraulic conductivity), snow (e.g., lapse rates, melting), and vegetation (e.g., leaf area index and maximum canopy index) using additional field observations in the watershed is critical for better prediction.     

Modeling sediment and nitrogen export from a rural watershed in eastern Canada using the soil and water assessment tool

Watershed simulation models can be used to assess agricultural nonpoint-source pollution and for environmental planning and improvement projects. However, before application of any process-based watershed model, the model performance and reliability must be tested with measured data. The Soil and Water Assessment Tool version 2005 (SWAT2005) was used to model sediment and nitrogen loads from the Thomas Brook Watershed, which drains a 7.84 km rural landscape in the Annapolis Valley of Nova Scotia, Canada. The Thomas Brook SWAT model was comprised of 28 subbasins and 265 hydrologic response units, most of them containing agricultural land use, which is the main nonpoint nitrogen source in the watershed. Crop rotation schedules were incorporated into the model using field data collected within Agriculture and Agri-Food Canada's Watershed Evaluation of Beneficial Management Practices program. Model calibration (2004-2006) and validation (2007-2008) were performed on a monthly basis using continuous stream flow, sediment, and nitrogen export measurements. Model performance was evaluated using the coefficient of determination, Nash-Sutcliff efficiency (NSE), and percent bias (PBIAS) statistics. Study results show that the model performance was satisfactory (NSE > 0.4; > 0.5) for stream flow, sediment, nitrate-nitrogen, and total nitrogen simulations. Annual corn, barley, and wheat yields were also simulated well, with PBIAS values ranging from 0.3 to 7.2%. This evaluation of SWAT demonstrated that the model has the potential to be used as a decision support tool for agricultural watershed management in Nova Scotia.     

Effects of calibration on L-THIA GIS runoff and pollutant estimation

Urbanization can result in alteration of a watershed's hydrologic response and water quality. To simulate hydrologic and water quality impacts of land use changes, the Long-Term Hydrologic Impact Assessment (L-THIA) system has been used. The L-THIA system estimates pollutant loading based on direct runoff quantity and land use based pollutant coefficients. The accurate estimation of direct runoff is important in assessing water quality impacts of land use changes. An automated program was developed to calibrate the L-THIA model using the millions of curve number (CN) combinations associated with land uses and hydrologic soil groups. L-THIA calibration for the Little Eagle Creek (LEC) watershed near Indianapolis, Indiana was performed using land use data for 1991 and daily rainfall data for six months of 1991 (January 1-June 30) to minimize errors associated with use of different temporal land use data and rainfall data. For the calibration period, the Nash-Sutcliffe coefficient was 0.60 for estimated and observed direct runoff. The calibrated CN values were used for validation of the model for the same year (July 1-December 31), and the Nash-Sutcliffe coefficient was 0.60 for estimated and observed direct runoff. The Nash-Sutcliffe coefficient was 0.52 for January 1, 1991 to December 31, 1991 using uncalibrated CN values. As shown in this study, the use of better input parameters for the L-THIA model can improve accuracy. The effects on direct runoff and pollutant estimation of the calibrated CN values in the L-THIA model were investigated for the LEC. Following calibration, the estimated average annual direct runoff for the LEC watershed increased by 34%, total nitrogen by 24%, total phosphorus by 22%, and total lead by 43%. This study demonstrates that the L-THIA model should be calibrated and validated prior to application in a particular watershed to more accurately assess the effects of land use changes on hydrology and water quality.     

A Simple Process‐Based Snowmelt Routine to Model Spatially Distributed Snow Depth and Snowmelt in the SWAT Model1

Abstract: We present a method to integrate a process‐based (PB) snowmelt model that requires only daily temperature and elevation information into the Soil and Water Assessment Tool (SWAT) model. The model predicts the spatiotemporal snowpack distribution without adding additional complexity, and in fact reduces the number of calibrated parameters. To demonstrate the utility of the PB model, we calibrate the PB and temperature‐index (TI) SWAT models to optimize agreement with stream discharge on a 46‐km² watershed in northwestern Idaho, United States, for 10 individual years and use the calibrated parameters for the year with the best agreement to run the model for 15 remaining years. Stream discharge predictions by the PB and TI model were similar, although the PB model simulated snowmelt more accurately than the TI model for the remaining 15‐year period. Spatial snow distributions predicted by the PB model better matched observations from LandSat imagery and a SNOTEL station. Results for this watershed show that including PB snowmelt in watershed models is feasible, and calibration of TI‐based watershed models against discharge can incorrectly predict snow cover.     

SENSITIVITY ANALYSIS,CALIBRATION, AND VALIDATIONS FOR A MULTISITE AND MULTIVARIABLE SWAT MODEL1

The ability of a watershed model to mimic specified watershed processes is assessed through the calibration and validation process. The Soil and Water Assessment Tool (SWAT) watershed model was implemented in the Beaver Reservoir Watershed of Northwest Arkansas. The objectives were to: (1) provide detailed information on calibrating and applying a multisite and multivariable SWAT model; (2) conduct sensitivity analysis; and (3) perform calibration and validation at three different sites for flow, sediment, total phosphorus (TP), and nitrate‐nitrogen (NO₃‐N) plus nitrite‐nitrogen (NO₂‐N). Relative sensitivity analysis was conducted to identify parameters that most influenced predicted flow, sediment, and nutrient model outputs. A multi objective function was defined that consisted of optimizing three statistics: percent relative error (RE), Nash‐Sutcliffe Coefficient (R_NS²), and coefficient of determination (R²). This function was used to successfully calibrate and validate a SWAT model of Beaver Reservoir Watershed at multi‐sites while considering multivariables. Calibration and validation of the model is a key factor in reducing uncertainty and increasing user confidence in its predictive abilities, which makes the application of the model effective. Information on calibration and validation of multisite, multivariable SWAT models has been provided to assist watershed modelers in developing their models to achieve watershed management goals.     

Hydrologic and atrazine simulation of the Cedar Creek Watershed using the SWAT model

One of the major factors contributing to surface water contamination in agricultural areas is the use of pesticides. The Soil and Water Assessment Tool (SWAT) is a hydrologic model capable of simulating the fate and transport of pesticides in an agricultural watershed. The SWAT model was used in this study to estimate stream flow and atrazine (2-chloro-4-(ethylamino)-6-(isopropylamino)-s-triazine) losses to surface water in the Cedar Creek Watershed (CCW) within the St. Joseph River Basin in northeastern Indiana. Model calibration and validation periods consisted of five and two year periods, respectively. The National Agricultural Statistics Survey (NASS) 2001 land cover classification and the Soil Survey Geographic (SSURGO) database were used as model input data layers. Data from the St. Joseph River Watershed Initiative and the Soil and Water Conservation Districts of Allen, Dekalb, and Noble counties were used to represent agricultural practices in the watershed which included the type of crops grown, tillage practices, fertilizer, and pesticide application rates. Model results were evaluated based on efficiency coefficient values, standard statistical measures, and visual inspection of the measured and simulated hydrographs. The Nash and Sutcliffe model efficiency coefficients (E(NS)) for monthly and daily stream flow calibration and validation ranged from 0.51 to 0.66. The E(NS) values for atrazine calibration and validation ranged from 0.43 to 0.59. All E(NS) values were within the range of acceptable model performance standards. The results of this study indicate that the model is an effective tool in capturing the dynamics of stream flow and atrazine concentrations on a large-scale agricultural watershed in the midwestern USA.     

Modeling of Sediment Yield From Anjeni-Gauged Watershed,Ethiopia Using SWAT Model1

Setegn, Shimelis G., Bijan Dargahi, Ragahavan Srinivasan, and Assefa M. Melesse, 2010. Modeling of Sediment Yield From Anjeni-Gauged Watershed, Ethiopia Using SWAT Model. Journal of the American Water Resources Association (JAWRA) 46(3):514-526. DOI: 10.1111/j.1752-1688.2010.00431.x Abstract: The Soil and Water Assessment Tool (SWAT) was tested for prediction of sediment yield in Anjeni-gauged watershed, Ethiopia. Soil erosion and land degradation is a major problem on the Ethiopian highlands. The objectives of this study were to evaluate the performance and applicability of SWAT model in predicting monthly sediment yield and assess the impacts of subbasin delineation and slope discretization on the prediction of sediment yield. Ten years monthly meteorological, flow and sediment data were used for model calibration and validation. The annual average measured sediment yield was 24.6 tonnes/ha. The annual average simulated sediment yield was 27.8 and 29.5 tones/ha for calibration and validation periods, respectively. The study found that the observed values showed good agreement with the simulated sediment yield with Nash-Sutcliffe efficiency (NSE) = 0.81, percent bias (PBIAS) = 28%, RMSE-observations standard deviation ratio (RSR) = 0.23, and coefficient of determination (R²) = 0.86 for calibration and NSE = 0.79, PBIAS = 30%, RSR = 0.29, and R² = 0.84 for validation periods. The model can be used for further analysis of different management scenarios that could help different stakeholders to plan and implement appropriate soil and water conservation strategies.     

APPLICATION OF THE DR3M WATERSHED MODEL ON A SMALL URBAN BASIN1

ABSTRACT: Data collected at a 79-acre urban watershed in Albuquerque, New Mexico, were used to calibrate and verify the Distributed Routing Rainfall-Runoff Model, a parametric watershed model. Standard errors of estimate for the 38 calibration storms were 33 percent and 38 percent, respectively, for volumes and peaks; and for the 46 verification storms were 29 percent and 37 percent, respectively, for volumes and peaks. Correlation coefficients for peaks were 0.8 and 0.95, respectively, for calibration and verification storms.     

HYDROLOGICAL MODELING OF THE IROQUOIS RIVER WATERSHED USING HSPF AND SWAT1

ABSTRACT: The performance of two popular watershed scale simulation models — HSPF and SWAT — were evaluated for simulating the hydrology of the 5,568 km² Iroquois River watershed in Illinois and Indiana. This large, tile drained agricultural watershed provides distinctly different conditions for model comparison in contrast to previous studies. Both models were calibrated for a nine‐year period (1987 through 1995) and verified using an independent 15‐year period (1972 through 1986) by comparing simulated and observed daily, monthly, and annual streamflow. The characteristics of simulated flows from both models are mostly similar to each other and to observed flows, particularly for the calibration results. SWAT predicts flows slightly better than HSPF for the verification period, with the primary advantage being better simulation of low flows. A noticeable difference in the models' hydrologic simulation relates to the estimation of potential evapotranspiration (PET). Comparatively low PET values provided as input to HSPF from the BASINS 3.0 database may be a factor in HSPF's overestimation of low flows. Another factor affecting baseflow simulation is the presence of tile drains in the watershed. HSPF parameters can be adjusted to indirectly account for the faster subsurface flow associated with tile drains, but there is no specific tile drainage component in HSPF as there is in SWAT. Continued comparative studies such as this, under a variety of hydrologic conditions and watershed scales, provide needed guidance to potential users in model selection and application.     

EVALUATION OF THE APPLICABILITY OF THE SWAT MODEL FOR COASTAL WATERSHEDS IN SOUTHEASTERN LOUISIANA1

ABSTRACT: The Soil and Water Assessment Tool (SWAT) has been used for hydrologic analyses at various watershed scales. However, little is known about the model's performance in coastal watersheds. In this study SWAT was evaluated for its applicability in three Louisiana coastal watersheds: the Amite, Tickfaw, and Tangipahoa River watersheds. The model was calibrated with daily discharge from 1976 to 1977 and validated from 1979 to 1999 for the Amite and Tangipahoa and with daily discharge from 1979 to 1989 for the Tickfaw. Deviation of mean discharge and the Nash‐Sutcliffe model efficiency were used to evaluate model behavior. The study found that Manning's roughness coefficient for the main channel, SCS curve number, and soil evaporation compensation factor were the most sensitive parameters for these coastal watersheds. The Manning's roughness coefficient showed the greatest effect on the response time of surface runoff, suggesting the critical role of channel routing in hydrologic modeling for lowland watersheds. The SWAT model demonstrated an excellent performance, with Nash‐Sutcliffe efficiencies of 0.935, 0.940, and 0.960 for calibrations of the Amite, Tickfaw, and Tangipahoa watersheds, respectively, and of 0.851, 0.811, and 0.867 for validations. The modeling results demonstrate that SWAT is capable of simulating hydrologic processes for medium scale to large scale coastal lowland watersheds in Louisiana.     

WATER QUALITY MODEL OUTPUT UNCERTAINTY AS AFFECTED BY SPATIAL RESOLUTION OF INPUT DATA1

ABSTRACT: Resolution of the input GIS data used to parameterize distributed‐parameter hydrologic/water quality models may affect uncertainty in model outputs and impact the subsequent application of model results in watershed management. In this study we evaluated the impact of varying spatial resolutions of DEM, land use, and soil data (30 × 30 m, 100 × 100 m, 150 × 150 m, 200 × 200 m, 300 × 300 m, 500 × 500 m, and 1,000 × 1,000 m) on the uncertainty of SWAT predicted flow, sediment, NO₃‐N, and TP transport. Inputs included measured hydrologic, meteorological, and watershed characteristics as well as water quality data from the Moores Creek watershed in Washington County, Arkansas. The SWAT model output was most affected by input DEM data resolution. A coarser DEM data resolution resulted in decreased representation of watershed area and slope and increased slope length. Distribution of pasture, forest, and urban areas within the watershed was significantly affected at coarser resolution of land use and resulted in significant uncertainty in predicted sediment, NO₃‐N, and TP output. Soils data resolution had no significant effect on flow and NO₃‐N predictions; however, sediment was overpredicted by 26 percent, and TP was underpredicted by 26 percent at 1,000 m resolution. This may be due to change in relative distribution of various hydrologic soils groups (HSGs) in the watershed. Minimum resolution for input GIS data to achieve less than 10 percent model output error depended upon the output variable of interest. For flow, sediment, NO₃‐N, and TP predictions, minimum DEM data resolution should range from 30 to 300 m, whereas minimum land use and soils data resolution should range from 300 to 500 m.     

Spatiotemporal Variability of Snow Depletion Curves Derived from SNODAS for the Conterminous United States, 2004‐2013

Assessment of water resources at a national scale is critical for understanding their vulnerability to future change in policy and climate. Representation of the spatiotemporal variability in snowmelt processes in continental‐scale hydrologic models is critical for assessment of water resource response to continued climate change. Continental‐extent hydrologic models such as the U.S. Geological Survey National Hydrologic Model (NHM) represent snowmelt processes through the application of snow depletion curves (SDCs). SDCs relate normalized snow water equivalent (SWE) to normalized snow covered area (SCA) over a snowmelt season for a given modeling unit. SDCs were derived using output from the operational Snow Data Assimilation System (SNODAS) snow model as daily 1‐km gridded SWE over the conterminous United States. Daily SNODAS output were aggregated to a predefined watershed‐scale geospatial fabric and used to also calculate SCA from October 1, 2004 to September 30, 2013. The spatiotemporal variability in SNODAS output at the watershed scale was evaluated through the spatial distribution of the median and standard deviation for the time period. Representative SDCs for each watershed‐scale modeling unit over the conterminous United States (n = 54,104) were selected using a consistent methodology and used to create categories of snowmelt based on SDC shape. The relation of SDC categories to the topographic and climatic variables allow for national‐scale categorization of snowmelt processes.     

Automatic Calibration of Hydrologic Models With Multi‐Objective Evolutionary Algorithm and Pareto Optimization1

Abstract: In optimization problems with at least two conflicting objectives, a set of solutions rather than a unique one exists because of the trade‐offs between these objectives. A Pareto optimal solution set is achieved when a solution cannot be improved upon without degrading at least one of its objective criteria. This study investigated the application of multi‐objective evolutionary algorithm (MOEA) and Pareto ordering optimization in the automatic calibration of the Soil and Water Assessment Tool (SWAT), a process‐based, semi‐distributed, and continuous hydrologic model. The nondominated sorting genetic algorithm II (NSGA‐II), a fast and recent MOEA, and SWAT were called in FORTRAN from a parallel genetic algorithm library (PGAPACK) to determine the Pareto optimal set. A total of 139 parameter values were simultaneously and explicitly optimized in the calibration. The calibrated SWAT model simulated well the daily streamflow of the Calapooia watershed for a 3‐year period. The daily Nash‐Sutcliffe coefficients were 0.86 at calibration and 0.81 at validation. Automatic multi‐objective calibration of a complex watershed model was successfully implemented using Pareto ordering and MOEA. Future studies include simultaneous automatic calibration of water quality and quantity parameters and the application of Pareto optimization in decision and policy‐making problems related to conflicting objectives of economics and environmental quality.     

Modeling Hydrology in a Small Rocky Mountain Watershed Serving Large Urban Populations1

Abstract: This article describes the development of a calibrated hydrologic model for the Blue River watershed (867 km²) in Summit County, Colorado. This watershed provides drinking water to over a third of Colorado’s population. However, more research on model calibration and development for small mountain watersheds is needed. This work required integration of subsurface and surface hydrology using GIS data, and included aspects unique to mountain watersheds such as snow hydrology, high ground‐water gradients, and large differences in climate between the headwaters and outlet. Given the importance of this particular watershed as a major urban drinking‐water source, the rapid development occurring in small mountain watersheds, and the importance of Rocky Mountain water in the arid and semiarid West, it is useful to describe calibrated watershed modeling efforts in this watershed. The model used was Soil and Water Assessment Tool (SWAT). An accurate model of the hydrologic cycle required incorporation of mountain hydrology‐specific processes. Snowmelt and snow formation parameters, as well as several ground‐water parameters, were the most important calibration factors. Comparison of simulated and observed streamflow hydrographs at two U.S. Geological Survey gaging stations resulted in good fits to average monthly values (0.71 Nash‐Sutcliffe coefficient). With this capability, future assessments of point‐source and nonpoint‐source pollutant transport are possible.     

Effects of soil data resolution on SWAT model stream flow and water quality predictions

The prediction accuracy of agricultural nonpoint source pollution models such as Soil and Water Assessment Tool (SWAT) depends on how well model input spatial parameters describe the characteristics of the watershed. The objective of this study was to assess the effects of different soil data resolutions on stream flow, sediment and nutrient predictions when used as input for SWAT. SWAT model predictions were compared for the two US Department of Agriculture soil databases with different resolution, namely the State Soil Geographic database (STATSGO) and the Soil Survey Geographic database (SSURGO). Same number of sub-basins was used in the watershed delineation. However, the number of HRUs generated when STATSGO and SSURGO soil data were used is 261 and 1301, respectively. SSURGO, with the highest spatial resolution, has 51 unique soil types in the watershed distributed in 1301 HRUs, while STATSGO has only three distributed in 261 HRUS. As a result of low resolution STATSGO assigns a single classification to areas that may have different soil types if SSURGO were used. SSURGO included Hydrologic Response Units (HRUs) with soil types that were generalized to one soil group in STATSGO. The difference in the number and size of HRUs also has an effect on sediment yield parameters (slope and slope length). Thus, as a result of the discrepancies in soil type and size of HRUs stream flow predicted was higher when SSURGO was used compared to STATSGO. SSURGO predicted less stream loading than STATSGO in terms of sediment and sediment-attached nutrients components, and vice versa for dissolved nutrients. When compared to mean daily measured flow, STATSGO performed better relative to SSURGO before calibration. SSURGO provided better results after calibration as evaluated by R(2) value (0.74 compared to 0.61 for STATSGO) and the Nash-Sutcliffe coefficient of Efficiency (NSE) values (0.70 and 0.61 for SSURGO and STATSGO, respectively) although both are in the same satisfactory range. Modelers need to weigh the benefits before selecting the type of data resolution they are going to use depending on the watershed size and level of accuracy required because more effort is required to prepare and calibrate the model when a fine resolution soil data is used.