Distributed Watershed Model Research Articles

Integrated environmental modeling for efficient aquifer vulnerability assessment using machine learning

Nitrate contamination in groundwater was evaluated using the concept of integrated aquifer assessment by combining groundwater characterization and risk analysis with tiered approaches for land and surface runoff contamination by soil chemicals and leaching of contamination to groundwater in the Upper White River Watershed (UWRW) in Indiana. Integrated aquifer vulnerability assessment was conducted using an integration of a distributed watershed model (Soil and Water Assessment Tool [SWAT]) and a machine learning technique (Geospatial-Artificial Neural Network [Geo-ANN]). The results indicate that integrated aquifer vulnerability assessment performed well based on the model performance (NSE/R2/PBIAS = 0.66/0.70/0.07). Thus, the overall assessment of aquifer vulnerability can be performed using the integrated aquifer vulnerability assessment technique provided in this study. Moreover, this approach provides an efficient guide for managing groundwater resources for policy makers and groundwater-related researchers.

A simple temperature-based method to estimate heterogeneous frozen ground within a distributed watershed model

Abstract. Frozen ground can be important to flood production and is often heterogeneous within a watershed due to spatial variations in the available energy, insulation by snowpack and ground cover, and the thermal and moisture properties of the soil. The widely used continuous frozen ground index (CFGI) model is a degree-day approach and identifies frozen ground using a simple frost index, which varies mainly with elevation through an elevation–temperature relationship. Similarly, snow depth and its insulating effect are also estimated based on elevation. The objective of this paper is to develop a model for frozen ground that (1) captures the spatial variations of frozen ground within a watershed, (2) allows the frozen ground model to be incorporated into a variety of watershed models, and (3) allows application in data sparse environments. To do this, we modify the existing CFGI method within the gridded surface subsurface hydrologic analysis watershed model. Among the modifications, the snowpack and frost indices are simulated by replacing air temperature (a surrogate for the available energy) with a radiation-derived temperature that aims to better represent spatial variations in available energy. Ground cover is also included as an additional insulator of the soil. Furthermore, the modified Berggren equation, which accounts for soil thermal conductivity and soil moisture, is used to convert the frost index into frost depth. The modified CFGI model is tested by application at six test sites within the Sleepers River experimental watershed in Vermont. Compared to the CFGI model, the modified CFGI model more accurately captures the variations in frozen ground between the sites, inter-annual variations in frozen ground depths at a given site, and the occurrence of frozen ground.

Modeling the impacts of changing climatic extremes on streamflow and sediment yield in a northeastern US watershed

Study regionWe investigate the impacts of local temperature and precipitation trends on discharge and sediment loading by applying the model to a watershed in the northeastern US, where trends in increasing precipitation exceed those of other regions in North America. Study focusIn this study we simulate the response of watershed sediment loading to changing frequencies and magnitudes of extreme precipitation events using a coupled model that explicitly simulates streambank erosion and failure within a distributed watershed model. To drive the model, we use meteorological inputs from general circulation models (GCMs) as well as from a statistical weather generator (WG). New hydrological insights for the regionChanges in the timing and magnitude of snow melt and spring flows, as well as associated sediment mobilization, resulted from increases in temperature. Increases in discharge and sediment load resulted from increases in precipitation events exceeding the 95th percentile. In runs driven by WG weather data, positive trends were evident in peak (as well as annual) discharge and suspended sediment yields over the years modeled. No clear trends were seen in GCM-driven runs, which do not capture historically-observed trends in extreme precipitation. This work is consistent with other studies in that it shows important changes in discharge and sediment yields from a watershed resulting from ongoing changes in climate.

Evaluating the risk of phosphorus loss with a distributed watershed model featuring zero-order mobilization and first-order delivery

Many semi-distributed models that simulate pollutants' losses from watersheds do not handle well detailed spatially distributed and temporal data with which to identify accurate and cost-effective strategies for controlling pollutants issuing from non-point sources. Such models commonly overlook the flow pathways of pollutants across the landscape. This work aims at closing such knowledge gap by developing a Spatially and Temporally Distributed Empirical model for Phosphorus Management (STEM-P) that simulates the daily phosphorus loss from source areas to receiving waters on a spatially-distributed grid-cell basis. STEM-P bypasses the use of complex mechanistic algorithms by representing the phosphorus mobilization and delivery processes with zero-order mobilization and first-order delivery, respectively. STEM-P was applied to a 217km2 watershed with mixed forest and agricultural land uses situated in southwestern China. The STEM-P simulation of phosphorus concentration at the watershed outlet approximated the observed data closely: the percent bias (Pbias) was −7.1%, with a Nash-Sutcliffe coefficient (ENS) of 0.80 on a monthly scale for the calibration period. The Pbias was 18.1%, with a monthly ENS equal to 0.72 for validation. The simulation results showed that 76% of the phosphorus load was transported with surface runoff, 25.2% of which came from 3.4% of the watershed area (classified as standard A critical source areas), and 55.3% of which originated from 17.1% of the watershed area (classified as standard B critical source areas). The standard A critical source areas were composed of 51% residences, 27% orchards, 18% dry fields, and 4% paddy fields. The standard B critical source areas were mainly paddy fields (81%). The calculated spatial and temporal patterns of phosphorus loss and recorded flow pathways identified with the STEM-P simulations revealed the field-scale critical source areas and guides the design and placement of effective practices for non-point source pollution control and water quality conservation.

융설과 토양의 동결-융해 과정을 고려한 겨울철 토양온도의 시공간 분포 모의

Soil temperature is one of the most important environmental factors that govern hydrological and biogeochemical processes related to diffuse pollution. In this study, considering the snowmelting and the soil freezing-thawing processes, a set of computer codes to estimate winter soil temperature has been developed for CAMEL(Chemicals, Agricultural Management and Erosion Losses), a distributed watershed model. The model was calibrated and validated against the field measurements for three months at 4 sites across the study catchment in a rural area of Yeoju, Korea. The degree of agreement between the simulated and the observed soil temperature is good for the soil surface(R2 0.71∼0.95, RMSE 0.89～1.49℃). As for the subsurface soils, however, the simulation results are not as good as for the soil surface(R2 0.51～0.97, RMSE 0.51～5.08℃) which is considered resulting from vertically-homogeneous soil textures assumed in the model. The model well simulates the blanket effect of snowpack and the latent heat flux in the soil freezing-thawing processes. Although there is some discrepancy between the simulated and the observed soil temperature due to limitations of the model structure and the lack of data, the model reasonably well simulates the temporal and spatial distributions of the soil temperature and the snow water equivalent in accordance with the land uses and the topography of the study catchment.

A mixed discrete-continuous variable multiobjective genetic algorithm for targeted implementation of nonpoint source pollution control practices

[1] Implementation of nonpoint source (NPS) pollution control strategies at the watershed scale hinges on abating pollutant movement from the landscape to water bodies at minimum cost. This paper presents an integrated simulation-optimization approach for targeted implementation of agricultural conservation practices at the watershed scale. A multiobjective genetic algorithm (NSGA-II) with mixed discrete-continuous decision variables was coupled with a distributed watershed model, Soil and Water Assessment Tool (SWAT), to identify optimal types and locations of conservation practices for nutrient and pesticide control at the watershed scale. Previous optimization studies have used binary representation of nonpoint source pollution controls, even though many could be better characterized as continuous variables. In this study, a novel discrete-continuous decision variable, also known as mixed-variable, representation was used to enhance the versatility of the approach by evaluating more options during the search process. Application of the proposed framework in the Eagle Creek Watershed, Indiana, indicated that the optimal suite of conservation practices from the mixed-variable NSGA-II was more effective in meeting water quality targets at lower costs than the solution from binary-variable optimization. However, the mixed-variable approach was considerably more computationally demanding for assessing tradeoffs between environmental and economic factors. A method for hybridization of binary and mixed-variable NSGA-II methods in the context of nonpoint source pollution control practices was developed to enhance the computational efficiency of the optimization procedure. As a result, the number of model simulations required for convergence to the Pareto-optimal solutions was reduced by 96%. The conceptual complexity and computational requirements of optimization-based approaches are impediments to their wider application for targeted implementation of NPS pollution control strategies. The methods and finding of this study address these issues and could result in a more effective implementation of management strategies at the watershed scale.

L-THIA를 이용한 낙동강수계 임하댐유역 비점오염원의 공간적 분포해석

Long-Term Hydrologic Impact Assessment (L-THIA) model which is a distributed watershed model was applied to analyze the spatial distribution of surface runoff and nonpoint source pollutant loading from Imha watershed during 2001~2010. L-THIA CN Calibration Tool linked with SCE-UA was developed to calibrate surface runoff automatically. Calibration (2001~2005) and validation (2006~2010) of monthly surface runoff were represented as 'very good' model performance showing 0.91 for calibration and 0.89 for validation as Nash-Sutcliffe (NS) values. Average annual surface runoff from Imha watershed was 218.4 mm and Banbyun subwatershed was much more than other watersheds due to poor hydrologic condition. Average annual nonpoint source pollutant loading from Imha wateshed were 2,295 ton/year for <TEX>$BOD_5$</TEX>, 14,752 ton/year for SS, 358 ton/year for T-N, and 79 ton/year for T-P. Amount of pollutant loading and pollutant loading rates from Banbyun watershed were much higher than other watersheds. As results of analysis of loading rate from grid size (<TEX>$30m{\times}30m$</TEX>), most of high 10 % of loading rate were generated from upland. Therefore, major hot spot area to manage nonpoint source pollution in Imha watershed is the combination of upland and Banbyun subwatershed. L-THIA model is easy to use and prepare input file and useful tool to manage nonpoint source pollution at screening level.

On the Downscaling of Meteorological Fields Using Recurrent Networks for Modelling the Water Balance in a Meso-Scale Catchment Area of Saxony, Germany

In this study, recurrent networks to downscale meteorological fields of the ERA-40 re-analysis dataset with focus on the meso-scale water balance were investigated. Therefore two types of recurrent neural networks were used. The first approach is a coupling between a recurrent neural network and a distributed watershed model and the second a nonlinear autoregressive with exogenous inputs (NARX) network, which directly predicted the component of the water balance. The approaches were deployed for a meso-scale catchment area in the Free State of Saxony, Germany. The results show that the coupled approach did not perform as well as the NARX network. But the meteorological output of the coupled approach already reaches an adequate quality. However the coupled model generates as input for the watershed model insufficient daily precipitation sums and not enough wet days were predicted. Hence the long-term annual cycle of the water balance could not be preserved with acceptable quality in contrary to the NARX approach. The residual storage change term indicates physical restrictions of the plausibility of the neural networks, whereas the physically based correlations among the components of the water balance were preserved more accurately by the coupled approach.

Predicting event response in a nested catchment with generalized linear models and a distributed watershed model

AbstractThis study focuses on the prediction of event‐based runoff coefficients (an important descriptor of flood events) for nested catchments up to an area of 50 km² in the Eastern Ore Mountains. The four main objectives of the study are (i) the prediction of runoff coefficients with the statistical method of generalized linear models, (ii) the comparison of the results of the linear models with estimates of a distributed conceptual model, (iii) the comparison of the dynamics of observed soil moisture and simulated saturation deficit of the hydrological model and (iv) the analysis of the relationship between runoff coefficient and observed and simulated wetness.Different predictor variables were selected to describe the runoff coefficient and were differentiated into variables describing the catchment's antecedent wetness and meteorological forcing. The best statistical model was estimated in a stepwise approach on the basis of hierarchical partitioning, an exhaustive search algorithm and model validation with jackknifing. We then applied the rainfall runoff model WaSiM ETH to predict the runoff processes for the two larger catchments. Locally measured small‐scale soil moisture (acquired at a scale of four to five magnitudes smaller than the catchment) was identified as one of the key predictor variables for the estimation of the runoff coefficient with the general linear model. It was found that the relationship betweenobserved and simulated (using WaSiM ETH) wetness is strongly hysteretic. The runoff coefficients derived from the rainfall runoff simulations systematically underestimate the observed values. Copyright © 2012 John Wiley &amp; Sons, Ltd.

Model calibration and uncertainty analysis for runoff in the Chao River Basin using sequential uncertainty fitting

Abstract The Chao River is one of the most important surface water sources for drinking water in Beijing. Due to the impacts of human activities and climate change, the Chao River basin is facing water scarcity. Therefore, it is very important to effectively manage water resources, while the distributed watershed model is the useful and effective tool. Soil and Water Assessment Tool (SWAT) was selected to set up hydrological model in the Chao River basin. Model calibration and uncertainty analysis were performed with Sequential Uncertainty Fitting (SUFI-2), which is one of the programs integrated with SWAT in the package SWAT-CUP (SWAT Calibration and Uncertainty Programs). Results showed that the p-factor was 0.85 and the r-factor was 1.12 in calibration period (1995-1999) while the p-factor was 0.83 and the r-factor was 2.15 in validation period (2000-2002). When values of p-factor and r-factor are accepted, further goodness of fit can be quantified by the coefficient of determination (R2) and Nash–Sutcliffe coefficient (NS) between the observed and the final best simulated data. The results indicated that R2 was 0.90 and NS was 0.88 in calibration period, while R2 was 0.77 and NS was 0.74 in validation period. The results of calibration and uncertainty analysis were satisfactory.

일강우 내삽을 이용한 일유량 시뮬레이션 및 앙상블 유량 발생

입력자료의 불확실성은 강우-유출 모의에서 중요한 불확실성 요소 중의 하나이다. 본 연구에서는 먼저 세 가지의 서로 다른 내삽 기법을 통해 계산된 강수 입력 자료 (관측값을 각 소유역의 중심점으로 내삽하여 추정한 입력자료임)들이 강우-유출 모형에 미치는 영향을 분포형 수문모형 (PRMS)을 이용하여 분석하였으며, 내삽오차를 바탕으로 발생한 입력자료를 앙상블 유량 예측에 이용하는 과정을 수문학적으로 서로 다른 두개 하천 유역에 적용하였다. 또한 Monte Carlo기법을 이용하여 수문 모형의 매개변수가 서로 다른 입력자료의 특성에 따라 변화하는 양상을 구분하여 보았다. 본 연구에서 제시된 앙상블 유량 예측방법은 기상 예측 및 기상 모형의 결과물 등의 입력자료를 이용함으로써 중/장기 유량 예측에 활용될 수 있을 것으로 판단된다. Input uncertainty is one of the major sources of uncertainty in hydrologic modeling. In this paper, first, three alternate rainfall inputs generated by different interpolation schemes were used to see the impact on a distributed watershed model. Later, the residuals of precipitation interpolations were tested as a source of ensemble streamflow generation in two river basins in the U.S. Using the Monte Carlo parameter search, the relationship between input and parameter uncertainty was also categorized to see sensitivity of the parameters to input differences. This analysis is useful not only to find the parameters that need more attention but also to transfer parameters calibrated for station measurement to the simulation using different inputs such as downscaled data from weather generator outputs. Input ensembles that preserves local statistical characteristics are used to generate streamflow ensembles hindcast, and showed that the ensemble sets are capturing the observed steamflow properly. This procedure is especially important to consider input uncertainties in the simulation of streamflow forecast.

Multi-Objective Optimization for Distributed Watershed Model Parameters Using Real-coded GA

Multi-Objective Optimization for Distributed Watershed Model Parameters Using Real-coded GA

Simulation of land use–soil interactive effects on water and sediment yields at watershed scale

Influences of vegetation management on soil erosion have been extensively studied. However, interactive effects between land use and soil are poorly documented in literature. Given the importance of understanding such effects for successful watershed management, the objective of this study was to examine the land use–soil interactive effects on water and sediment yields for the 117,845-ha drainage area upstream of the U.S. Geological Survey flow gauging station 08101000 in the Cowhouse Creek watershed located in north central Texas. The examination was implemented using the Soil and Water Assessment Tool (SWAT), a distributed watershed model that has been widely used to tackle problems relevant to nonpoint source pollution. A SWAT model was calibrated and validated in accordance with the observed daily streamflows at this gauging station. Subsequently, the calibrated model was used to examine changes of water and sediment yields as a result of the conversion of range brush to range grasses on an individual soil basis. The results indicated that for the study area, the removal of range brush would result in an annual water yield increase of 24 mm ha −1 treated area. However, the removal on an upland soil with a moderately high permeability was predicted to increase the annual water yield by 80 mm ha −1 treated area, while it would result in a small increase of annual sediment loading (4.2 t ha −1 treated area) and a minimal alteration to the existing spatial patterns of sediment sources. The increase of water yield would be larger for the removal of range brush on a soil that is adjacent to the stream channels. For a given soil, the predicted water yield increase was greater for the wetter hydrologic condition than that for the drier one. A reasonable generalization of this study was that the development of best management practices for watershed health and sustainability may need to take into account land use–soil interactive effects on an individual soil basis.

Resample or not?! Effects of resolution of DEMs in watershed modeling

AbstractResolution of data is a sensitive issue in environmental modeling. Professionals from diverse background use Geographic Information Systems (GIS) for environmental modeling. Availability of GIS software and publicly available digital data make modeling more accessible to anyone, but without proper understanding of the resampling and resolution of the digital database, model results are of little practical significance. It is imperative that the government agencies and their contractors responsible for using GIS‐based modeling to implement regulatory goals must have strong foundational knowledge of resolution of data and their implications. This study utilizes the SWAT (Soil and Water Assessment Tool) model integrated with ArcView to examine how sensitive the SWAT model was to the resolution of Digital Elevation Models (DEMs) while predicting the streamflow. This research uses a case study to demonstrate to the modeling community that resolution of data matters when predicting flow. If there were no effect of resolution on the modeled results then the original 90 m DEM and the original 30 m DEM resampled to 90 m would show similar trends. Initial input layers into SWAT were: DEMs, soils, landuse (LU) and meteorological data. The model‐predicted streamflow was validated against USGS (US Geological Survey) stream gauge data. DEMs are available at 30, 90 and 300 m resolution originally; therefore, this study analysed the sensitivity of the predicted streamflow at 30, 90 and 300 m resolution. Results indicated that SWAT is indeed sensitive to the resolution of the DEMs: original 90 and 30 m DEM resampled to 90 m did not show the same trend. Therefore, the effects of resolution cannot be ignored and resampling may not be adequate in modeling stream flows using a distributed watershed model. Copyright © 2009 John Wiley &amp; Sons, Ltd.

A distributed hydro-environmental watershed model with three-zoned cell profiling

A cell-based distributed watershed model is developed which enables us to simulate the hydrological and hydraulic aspects of the watershed in a refined fashion. With three-zoned cell profiling, the model is composed of three sub-models; tank model for a surface water zone, soil moisture model for a surface soil zone, and unconfined shallow groundwater flow model for a subsurface zone. Inclusion of the soil moisture sub-model modified to reroute the infiltration, routed from the tank sub-model, into the return flow and the groundwater recharge features the model. The groundwater flow sub-model, numerically approximated by use of the finite volume method and the implicit time-marching scheme, considers a network of on-farm drainage canals as internal boundaries, which is an essential need for modeling the watershed including farmlands. Cascade-linking of the three sub-models in a cell and assembling of all the cells over the entire watershed domain provides the global equations system to be solved. Applicability of the model is demonstrated with its practical application to a real watershed in that paddy and upland crop fields take great part of the land-use practice. It is then indicated in a quantified manner that rice farming significantly contribute as a major groundwater recharger in an irrigation period to fostering and conservation of regional water resources. Along with appropriately profiling a cell, the model is so versatile and tough that it can be applied without difficulty to a watershed of diverse terrains and land-uses and the computations can stably be carried out. It is thus concluded that the model presently developed could be a powerful “watershed simulator” to investigate and assess the time-varying hydro-environmental properties of a watershed while separating and integrating the hydrological and hydraulic components of particular importance.

Multiple-criteria calibration of a distributed watershed model using spatial regularization and response signatures

Summary This paper explores the use of a semi-automated multiple-criteria calibration approach for estimating the parameters of the spatially distributed HL-DHM model to the Blue River basin, Oklahoma. The study was performed in the context of Phase 2 of the DMIP project organized by the Hydrology Lab of the NWS. To deal with the problem of ill conditioning, we employ a regularization approach that constrains the search space using information contained in a priori estimates of the spatially distributed parameter fields developed from soils and other geo-spatial datasets. Unlike the commonly used spatial-multiplier method, our more general approach allows the parameters to depart non-uniformly (to some degree) from the a priori spatial pattern. The approach reduces the number of unknowns to be estimated using historical input–output data from 860 to 35. Two commonly used summary statistics of the model residuals, MSE and MSEL, are used to optimize fitting of the model to both the peaks and the recession periods of the time series data. A signature measure approach is used to select parameter sets that are close to Pareto-optimal in terms of MSE and MSEL, but which provide more consistent representation of the hydrologic behavior of the watershed as summarized by measures derived from the flow duration curve. While the results support the methods used in this analysis and show considerable improvement over the a priori parameter estimates, we find that the basin has some peculiar behaviors (including time non-stationarity) that the HL-DHM model as implemented is not set up to reproduce.

Hypsometric control on surface and subsurface runoff

A fundamental problem in hydrology is relating the basin hydrological response to the geomorphologic properties of a catchment. In this technical note, we show that the hypsometric distribution exerts control on surface and subsurface runoff partitioning by isolating its effect with respect to other basin characteristics. We conduct simulations using a distributed watershed model for hypsometric realizations developed by modifying the contour line values of a real basin. Results indicate that the runoff components are a function of the basin hypsometric form. In general, a relatively less eroded (convex) basin exhibits higher total runoff that is more dominated by subsurface processes, while a relatively more eroded (concave) basin shows less total runoff with a higher fraction of surface response. Hypsometric differences are also observed in the relations between base flow discharge and the mean groundwater depth and the variable source area. Furthermore, the hypsometric form reveals clear signatures on the spatial distribution of soil moisture and runoff response mechanisms.

Combined Monitoring and Modeling Indicate the Most Effective Agricultural Best Management Practices

Although water quality problems associated with agricultural nonpoint source (NPS) pollution have prompted the rapid and widespread adoption of a variety of so called "best management practices" (BMPs), there have been few realistic efforts to assess their combined effectiveness in reducing NPS pollution. This study used the Variable Source Loading Function (VSLF) model, a distributed watershed model, to simulate phosphorus (P) loading from an upstate New York dairy farm before and after the implementation of a suite of BMPs. With minimal calibration, the model calculates the dissolved P (DP) losses from impervious surfaces (e.g., barnyards), the plant/soil complex, field-applied manure, and loads associated with baseflow conditions. The simulated DP loads agreed well with measured loads for both the pre-BMP and post-BMP periods. More importantly, results showed that BMPs reduced DP loads by 35%, which is over half of the expected reduction if all manure was removed from the watershed, i.e., approximately 50% reduction. The model results indicate that had no BMPs been installed DP loads would be approximately 37% greater than observed at the watershed outlet. The most effective BMPs were those that disassociated pollutant loading areas from areas prone to generating runoff, i.e., hydrologically sensitive areas. By contrast, attempts to reduce P content in manure were somewhat less effective. This study demonstrates that a combination of distributed, mechanistic modeling and long-term monitoring provides better insights into the effectiveness of water quality protection efforts than either individually.

Rainfall characteristics define the value of streamflow observations for distributed watershed model identification

We present evidence that the characteristics of rainfall events strongly control the value of streamflow observations for the identification of distributed watershed models. A series of synthetic rainfall events with different spatio‐temporal extents and dynamics are used to investigate spatially‐distributed global parameter sensitivities for a typical watershed model. The model's parametric sensitivities vary greatly with rainfall distribution characteristics, location of the model cell in relation to the watershed's gauged outlet, and, to a lesser degree, the initial model states. This study demonstrates that the information content of streamflow is a dynamic property and that distributed model identification methodologies should consider the impact of spatio‐temporal rainfall dynamics.

Peak Flow Forecasting with Radar Precipitation and the Distributed Model CASC2D

The processing speed of computers and availability of spatial hydrologic data make distributed watershed models a viable approach for many applications, including peak flow forecasting. This study demonstrates the feasibility of using radar precipitation data with a distributed watershed model to obtain increased lead-time for peak flow forecasting. The CASC2D watershed model is applied to the Hassyampa River watershed in central Arizona using radar-based rainfall estimates from the National Weather Service WSR-88D weather radar. An application of radar rainfall data as input to the CASC2D model is then presented by which precipitation forecasts are generated by extrapolation of precipitation patterns from radar images. The calibrated model is run for two rainfall events: with and without linear precipitation forecasts. Results of this study confirm that the precipitation forecasts based on extrapolation of rainfall patterns from radar can produce increases in the forecast lead-time. For the two events presented, the forecast lead-time was increased by four to five hours, which can be significant when issuing flood warnings.