In Soil and Water Assessment Tool (SWAT), oftentimes, Critical Source Area (CSA), the minimum upstream drainage area that is required to initiate a stream, is used to subdivide a watershed. In the current literature, CSA has been used as a trial and error process to define the subwatershed levels. On the other hand, the ongoing collaboration of the United States Environmental Protection Agency Office of Water and the United States Geological Survey has promoted a national level predefined catchments and flowlines called National Hydrography Dataset (NHD) Plus to ease watershed modeling in the United States. The introduction of NHDPlus can eliminate the uncertain nature in defining the number of subwatersheds required to model the hydrologic system. This study demonstrates an integrated modeling environment with SWAT and NHDPlus spatial datasets. A spatial tool that was developed in a Geographical Information System (GIS) environment to by-pass the default watershed delineation in ArcSWAT, the GIS interface to SWAT, with the introduction of NHDPlus catchments and flowlines, was used in this study. This study investigates the effect of the spatial size (catchment area) of the NHDPlus and the input data resolution (cell/pixel size) within NHDPlus catchments on SWAT streamflow and sediment prediction. In addition, an entropy based watershed subdivision scheme is presented by using the landuse and soil spatial datasets with the conventional CSA approach to investigate if one of the CSAs can be considered to produce the best SWAT prediction on streamflow. Two watersheds (Kings Creek, Texas and Sugar Creek, Indiana) were used in this study. The study shows that there exists a subwatershed map that does not belong to one of the subwatershed maps produced through conventional CSA approach, to produce a better result on uncalibrated monthly SWAT streamflow prediction. Beyond the critical threshold, the CSA threshold which gives the best uncalibrated monthly streamflow prediction among a given set of CSAs, the SWAT performance can be improved further by subdividing some of the subwatersheds at this critical threshold. The study also shows that the input data resolution (within each NHDPlus catchments) does not have an influence on SWAT streamflow prediction for the selected watersheds. However, there is a change on streamflow prediction as the area of the NHDPlus catchment changes. Beyond a certain catchment size (8-9% of the watershed area), as the input data resolution becomes finer, the total sediment increases whereas the sediment prediction in high flow regime decreases. As the NHDPlus catchment size changes, the stream power has an influence on total sediment prediction. However, as the input data resolution changes, but keeping the NHDPlus catchment size constant, the Modified Universal Soil Loss Equation topographic factor has an influence on total sediment prediction.

Elevated phosphorus (P) loading of freshwater lakes and reservoirs often results in poor water quality and negative ecological effects. Critical source areas (CSA) of P in the watershed can be difficult to identify and control. A useful concept for identification of a CSA is the P risk index (P Index) that evaluates the P risk associated with distinct source and transport pathways. The objectives of this study were to create a GIS model that adapts the Minnesota (MN) P Index for use at the watershed scale in a mixed-use, mountain environment, and to evaluate its effectiveness relative to field-based assessment. A GIS-based model of the MN P Index, adapted for montane environments and relying primarily on publicly available geospatial data, was created and applied in the Wallsburg watershed, located in the mountains of Central Utah. One necessary data input, P found in plant residue of common Utah ecosystems, was found lacking after literature review. We experimentally determined a range of observed values from multiple ecosystems to adapt and validate the GIS model. The GIS P Index was evaluated against the results of 58 field scale applications of the MN P Index conducted throughout the watershed. The field-scale analysis resulted in about 14% of the sites sampled being identified as high or very high risk for P transport to surface water. Spatially, these high risk areas were determined to be a geographic cluster of fields near the lower middle agricultural section of the watershed. The GIS model visually and spatially identified the same cluster of fields as high risk areas. Various soil test P scenarios were explored and compared to the known 58 site values. Soil test phosphorus had little effect on the GIS model's ability to accurately predict P risk in this watershed suggesting that high volume soil sampling is not always necessary to identify CSAs of P. Variable hypothetical livestock density scenarios were also simulated. The GIS model proved sensitive to variable P inputs and highlighted the necessity of accurate applied P source data. On average the model under-predicted the known field-site values by a risk score of 1.3, which suggests reasonable success in P assessment based on the categorical risk scores of the MN P Index and some potential for improvement. The GIS model has great potential to give land managers the ability to quickly locate potential CSAs and prioritizing remediation efforts to sites with greatest risk.

Few elements in surface waters are monitored as closely as phosphorus (P) due to its role in the eutrophication and degradation of surface waters. Limiting P mobilization from source areas is, therefore, a central goal of water quality protection plans. But the work of locating sources in mixed-use watersheds is challenged by the spatial and temporal variability of critical source areas (CSAs) of P. Synoptic sampling is a proven method for capturing the spatial variation of water quality parameters in surface waters, though it's not often used to track temporal dynamics across the same study area. Phosphorus fractionation is an analytical method that divides the total P (TP) in water into fractions, which for this study included total dissolved P (TDP), particulate P (PP), dissolved reactive P (DRP), and dissolved organic P (DOP). The objective of this study was to demonstrate the utility of combining temporally repeated synoptic sampling with simple P fractionation as a unique strategy for locating and characterizing CSAs of P. Seven synoptic sampling campaigns were conducted over a two-year period (March 2015 – July 2016) in a rural, montane watershed in north central Utah, USA. In each campaign, we sampled 18 sites across three tributaries (Main Creek, Spring Creek, and Little Hobble Creek) during three distinct, annual hydrologic periods (rising flow, peak flow, and baseflow). Temporal repetition clearly identified the rising flow period as the period with greatest P loading in the watershed. Combining repeated synoptic sampling and P fractionation successfully identified CSAs of P and most probable transfer pathways. Specifically, stream segments along lower Spring Creek and Main Creek were associated with the greatest increases of PP loads during periods of rising flow and peak flow. In the same time periods, the greatest DOP loads stemmed from forested areas as well as areas in the lower watershed associated with winter grazing of cattle. The watershed exhibited a significant background concentration of DRP from groundwater-driven subsurface sources in the lower half of the watershed that persisted year-round. These assessments can be used to develop management practices that limit various P loads from these respective critical source areas. The characterization of CSAs could not have been made using only a traditional synoptic sampling approach. This study demonstrated that the combination of repeated synoptic sampling and P fractionation can be an effective technique for locating and characterizing critical P source areas in order to guide best management practices that improve surface water quality.

A distributed approach can improve functionality of H/WQ (Hydrology and Water Quality) modeling by facilitating a way to explicitly incorporate spatial characteristics of a watershed into the model. The time-area approach, with its intuitive and inherently distributed concept, provides a simple method to simulate runoff mechanisms. This study developed a distributed model based on the time-area approach with the goal of improved utility and efficiency in H/WQ modeling. Uncertainty is always introduced into watershed modeling because of imperfect knowledge and scale dependant spatial heterogeneity and temporal variability. Uncertainty analysis can provide a modeler, policy maker, and stakeholder with reliability information, better understanding, and better communication about the modeling results. This study quantified uncertainty of the model parameter and output through uncertainty analysis in order to assess risk in watershed management. The main goal of this study was to develop a hydrology and sediment transport model capable of routing overland flow using a time-area concept and providing reliability of the modeling results in a probabilistic manner through uncertainty analysis. The HYSTAR (HYdrology and Sediment transport simulation using Time-ARea method) model incorporates a modified Curve Number (CN) method and the newly devised time-area routing method to estimate runoff. HYSTAR is capable of simulating direct runoff, base flow, soil moisture, and sediment load in a distributed manner and in an hourly time step. In the model, the modified CN and a continuity equation are used to calculate infiltration of the routed runoff as well as rainfall on every overland cell. The effective direct runoff volume is distributed over downstream areas using the newly developed routing method. A direct runoff hydrograph is constructed directly through the discrete convolution of the time-area histogram and the effective direct runoff volume map without employing a unit hydrograph. In addition, sediment transport is simulated using the routing method and the sediment transport capacity approach without using a delivery ratio. The sensitivity analysis found that the CN and root zone depth were the most critical parameters for runoff simulation with HYSTAR. The model provided acceptable performance in predicting runoff and sediment load of a subwatershed of the Owl Run Watershed (ORD) with the Nash-Sutcliffe efficiency coefficient and coefficient of determination greater than 0.5. However, it failed to reproduce runoff for a subwatershed of Polecat Creek Watershed (PCA), where data show that runoff is not immediately responsive to rainfall. Uncertainty analysis revealed that the confidence intervals of the simulated monthly runoff and sediment load corresponded to 9.7 % and 10.2 % of their averages, respectively, at a significance level of 0.05. In addition, the average ranges of variation created by the Digital Elevation Model (DEM) and National Land Cover Data (NLCD) errors in the simulated monthly runoff and sediment load were equivalent to 7.5 % and 15.9 % of the average of their calibrated values, respectively. Based on the uncertainty analysis results, the Margin of Safety (MOS) of Total Maximum Daily Load (TMDL) were explicitly quantified as corresponding to 7.0 % and 21.3 % of the average of the simulated runoff and sediment load for ORD at significance level of 0.05. In conclusion, the HYSTAR model provided a new way to explicitly simulate runoff and sediment load of a watershed in a distributed manner. The approach developed here retains the simplicity of a unit hydrograph approach without employing numerical methods. Uncertainty analysis found that parameter uncertainty had greater impact on the model output than did expected Geographic Information System (GIS) data errors. In addition, the impact of the topographic data error on the model output was greater than was that of the land cover data error. Finally, this study provided a proof that a 5 to 10 % MOS that many TMDL studies consider underestimates modeling uncertainty. / Ph. D.

Few current modeling tools are designed to predict short-term, high-risk runoff from critical source areas (CSAs) in watersheds which are significant sources of non point source (NPS) pollution. This study couples the Soil and Water Assessment Tool-Variable Source Area (SWAT-VSA) model with the Climate Forecast System Reanalysis (CFSR) model and the Global Forecast System (GFS) model short-term weather forecast, to develop a CSA prediction tool designed to assist producers, landowners, and planners in identifying high-risk areas generating storm runoff and pollution. Short-term predictions for streamflow, runoff probability, and soil moisture levels were estimated in the South Fork of the Shenandoah river watershed in Virginia. In order to allow land managers access to the CSA predictions a free and open source software based web was developed. The forecast system consists of three primary components; (1) the model, which preprocesses the necessary hydrologic forcings, runs the watershed model, and outputs spatially distributed VSA forecasts; (2) a data management structure, which converts high resolution rasters into overlay web map tiles; and (3) the user interface component, a web page that allows the user, to interact with the processed output. The resulting framework satisfied most design requirements with free and open source software and scored better than similar tools in usability metrics. One of the potential problems is that the CSA model, utilizing physically based modeling techniques requires significant computational time to execute and process. Thus, as an alternative, a deep learning (DL) model was developed and trained on the process based model output. The DL model resulted in a 9% increase in predictive power compared to the physically based model and a ten-fold decrease in run time. Additionally, DL interpretation methods applicable beyond this study are described including hidden layer visualization and equation extractions describing a quantifiable amount of variance in hidden layer values. Finally, a large-scale analysis of soil phosphorus (P) levels was conducted in the Chesapeake Bay watershed, a current location of several short-term forecast tools. Based on Bayesian inference methodologies, 31 years of soil P history at the county scale were estimated, with the associated uncertainty for each estimate. These data will assist in the planning and implantation of short term forecast tools with P management goals. The short term modeling and communication tools developed in this work contribute to filling a gap in scientific tools aimed at improving water quality through informing land manager's decisions. / PHD