Watershed-scale nonpoint source (NPS) pollution models have become important tools to understand, evaluate, and predict the negative impacts of NPS pollution on water quality. Today, there are many NPS models available for users. However, different types of models possess different form and structure as well as complexity of computation. It is difficult for users to select an appropriate model for a specific application without a clear understanding of the limitations or strengths for each model or tool. This review evaluates 14 more commonly used watershed-scale NPS pollution models to explain how and when the application of these different models are appropriate for a given effort. The models that are assessed have a wide range of capacities that include simple models used as rapid screening tools (e.g., Long-Term Hydrologic Impact Assessment (L-THIA) and Nonpoint Source Pollution and Erosion Comparison Tool (N-SPECT/OpenNSPECT)), medium-complexity models that require detail data input and limited calibration (e.g., Generalized Watershed Loading Function (GWLF), Loading Simulation Program C (LSPC), Source Loading and Management Model (SLAMM), and Watershed Analysis Risk Management Frame (WARMF)), complex models that provide sophisticated simulation for NPS pollution processes with intensive data and rigorous calibration (e.g., Agricultural Nonpoint Source pollution model (AGNPS/AnnAGNPS), Soil and Water Assessment Tool (SWAT), Stormwater Management Model (SWMM), and Hydrologic Simulation Program Fortran (HSPF)), and modeling systems that integrate various sub-models and tools, and contain the highest complexity to solve all phases of hydrologic, hydraulic, and chemical dynamic processes (e.g., Automated Geospatial Watershed Assessment Tool (AGWA), Better Assessment Science Integrating Point and Nonpoint Sources (BASINS) and Watershed Modeling System (WMS)). This assessment includes model intended use, components or capabilities, suitable land-use type, input parameter type, spatial and temporal scale, simulated pollutants, strengths and limitations, and software availability. Understanding the strengths and weaknesses of each watershed-scale NPS model will lead to better model selection for suitability and help to avoid misinterpretation or misapplication in practice. The article further explains the crucial criteria for model selection, including spatial and temporal considerations, calibration and validation, uncertainty analysis, and future research direction of NPS pollution models. The goal of this work is to provide accurate and concise insight for watershed managers and planners to select the best-suited model to reduce the harm of NPS pollution to watershed ecosystems.

Soil erosion and lake sediment loading are primary concerns of watershed managers around the world. In the Xinjiang River Basin of China, severe soil erosion occurs primarily during monsoon periods, resulting in sediment flow into Poyang Lake and subsequently causing lake water quality deterioration. Here, we identified high-risk soil erosion areas and conditions that drive sediment yield in a watershed system with limited available data to guide localized soil erosion control measures intended to support reduced sediment load into Poyang Lake. We used the Soil and Water Assessment Tool (SWAT) model to simulate monthly and annual sediment yield based on a calibrated SWAT streamflow model, identified where sediment originated, and determined what geographic factors drove the loading within the watershed. We applied monthly and daily streamflow discharge (1985–2009) and monthly suspended sediment load data (1985–2001) to Meigang station to conduct parameter sensitivity analysis, calibration, validation, and uncertainty analysis of the model. The coefficient of determination (R²), Nash–Sutcliffe efficiency (NSE), percent bias (PBIAS), and RMSE -observation’s standard deviation ratio (RSR) values of the monthly sediment load were 0.63, 0.62, 3.8%, and 0.61 during calibration, respectively. Spatially, the annual sediment yield rate ranged from 3 ton ha⁻¹year⁻¹ on riparian lowlands of the Xinjiang main channel to 33 ton ha⁻¹year⁻¹ on mountain highlands, with a basin-wide mean of 19 ton ha⁻¹year⁻¹. The study showed that 99.9% of the total land area suffered soil loss (greater than 5 ton ha⁻¹year⁻¹). More sediment originated from the southern mountain highlands than from the northern mountain highlands of the Xinjiang river channel. These results suggest that specific land use types and geographic conditions can be identified as hotspots of sediment source with relatively scarce data; in this case, orchards, barren lands, and mountain highlands with slopes greater than 25° were the primary sediment source areas. This study developed a reliable, physically-based streamflow model and illustrates critical source areas and conditions that influence sediment yield.

Accurate streamflow prediction plays a pivotal role in hydraulic project design, nonpoint source pollution estimation, and water resources planning and management. However, the highly non-linear relationship between rainfall and runoff makes prediction difficult with desirable accuracy. To improve the accuracy of monthly streamflow prediction, a seasonal Support Vector Regression (SVR) model coupled to the Soil and Water Assessment Tool (SWAT) model was developed for 13 subwatersheds in the Illinois River watershed (IRW), U.S. Terrain, precipitation, soil, land use and land cover, and monthly streamflow data were used to build the SWAT model. SWAT Streamflow output and the upstream drainage area were used as two input variables into SVR to build the hybrid SWAT-SVR model. The Calibration Uncertainty Procedure (SWAT-CUP) and Sequential Uncertainty Fitting-2 (SUFI-2) algorithms were applied to compare the model performance against SWAT-SVR. The spatial calibration and leave-one-out sampling methods were used to calibrate and validate the hybrid SWAT-SVR model. The results showed that the SWAT-SVR model had less deviation and better performance than SWAT-CUP simulations. SWAT-SVR predicted streamflow more accurately during the wet season than the dry season. The model worked well when it was applied to simulate medium flows with discharge between 5 m³ s^-1 and 30 m³ s^-1, and its applicable spatial scale fell between 500 to 3000 km². The overall performance of the model on yearly time series is “Satisfactory”. This new SWAT-SVR model has not only the ability to capture intrinsic non-linear behaviors between rainfall and runoff while considering the mechanism of runoff generation but also can serve as a reliable regional tool for an ungauged or limited data watershed that has similar hydrologic characteristics with the IRW.