Nitrate (NO⁻3) export can vary widely among forested watersheds with similar nitrogen loading, geology, and vegetation, which suggests the importance of understanding differing internal retention mechanisms. Transport should be studied at the hillslope scale because the hillslope is the smallest unit with spatial and temporal resolution to reflect many relevant NO⁻3 retention and transport (flow-generation) processes, and headwater forested watersheds are largely comprised of sections of hillslopes. I conducted two experiments to elucidate subsurface flow dynamics and NO⁻3 transport and retention mechanisms on a constructed experimental hillslope model. In the first experiment, I tested whether decadal pedogenetic changes in soil properties in the experimental hillslope used by Hewlett and Hibbert (1963) would lead to changes in recession flow. I repeated (twice) their seminal experiment, whose results led to the development of the Variable Source Area paradigm, by also saturating, covering, and allowing the experimental hillslope to drain until it no longer yielded water. In the historical experiment there was fast drainage for 1.5 d, followed by slow drainage for ~140 d, which led the authors to conclude that recession flow in unsaturated soil could sustain baseflow throughout droughts. This long, slow drainage period was not reproduced in my experiments. Shapes of the drainage curves in my experiments were similar to the historical curve, but slow drainage was truncated, ending after 17 and 12 d, due likely to a leak in the boundary conditions, rather than to pedogenetic changes since the historical experiment. Leakage to bedrock, analogous to the leak in the hillslope model, is a commonly observed phenomenon and this study highlights how that can reduce drainage duration and the contribution of moisture from soils to support baseflow. In the second experiment, I tested whether movement of NO⁻3, which is considered a mobile ion, would be delayed relative to movement of water through a hillslope. I added concentrated pulses of ¹⁵NO⁻3 and a conservative tracer (²H₂O) on the same experimental hillslope, which was devegetated and irrigated at hydrologic steady state. Retention of the ¹⁵NO⁻3 tracer was high in the soil surface (0–10 cm) layer directly where the tracer was added. The portion of the ¹⁵NO⁻3 tracer that passed through this surface layer was further retained/removed in deeper soil. The reduction in the peaks in δ¹⁵N breakthrough was an order of magnitude larger than in δ₂H breakthrough at the outlet 5 m downslope of the tracer addition. The peaks in δ¹⁵N were also delayed relative to the peaks in δ₂H by 1, 6, 9 and 18.5 d for slope distances of 0, 2, 4, and 5 m, respectively, from tracer addition to the outlet. The excess mass of ¹⁵NO⁻3 recovered at the outlet was less than 3% of the original tracer mass injected. Nitrification and denitrification were estimated to be roughly 1:1 and were large fluxes relative to lateral transport into and out of the riparian zone. This tracer experiment shows that bedrock leakage, coupled with multiple retention/removal mechanisms can significantly delay export of added NO⁻3 with implications of additional NO⁻3 sink strength at the watershed scale. / Ph. D. / Nitrate (NO₃⁻) export can vary widely among forested watersheds with similar nitrogen loading, geology, and vegetation, which suggests the importance of understanding differing internal process mechanisms. I conducted two experiments to illustrate how water and NO₃⁻ moved on a constructed hillslope model. In the first experiment, I quantified differences in soil properties in the hillslope model used by Hewlett and Hibbert (1963). Then I repeated (twice) the seminal drainage experiment described in Hewlett and Hibbert (1963). The same hillslope (21.8°; 40%) was wetted up, covered, and allowed to drain until water stopped exiting at the outlet. In the historical experiment there was fast drainage for 1.5 d, followed by slow drainage for ~140 d, which led the authors to hypothesize that slow drainage in surface soil could continually contribute water to streams even during droughts. This long, slow drainage period was not reproduced in my experiments. Drainage was similar at the beginning of drainage between my experiments and the historical experiment, but in my experiment the slow drainage ended earlier (after 17 and 12 d) due likely to a leak in the constructed hillslope model, rather than to significant changes that occurred in the soil itself since the original experiment. This leak in the hillslope model is similar to leakage to bedrock, which is commonly observed in natural hillslopes. In the second experiment, I tested whether NO₃⁻ and water would move through a hillslope at the same rate. I added concentrated pulses of NO₃⁻ (as ¹⁵NO₃⁻ and water (as ²H₂O) on the same devegetated experimental hillslope. Retention of the ¹⁵NO₃⁻ tracer was high in the surface (0–10 cm) where the tracer was added, with little change in the immediately surrounding soil, despite high rates of water input immediately after tracer addition and throughout the experiment. The portion of the ¹⁵NO₃⁻ tracer that passed through the surface layer was further processed by microbes in deeper soil as it traveled downslope. This body of work shows that bedrock leakage, coupled with multiple retention mechanisms throughout the soil profile, can significantly delay export of added NO₃⁻ at the watershed scale.

Rainfall-runoff modeling in the United States has made extensive use of the Soil Conservation Service (SCS) curve number method for computing infiltration losses from rainfall. Even though the method is well established and may be applied to a wide range of environments, it often results in highly erroneous runoff estimates for shallow water table environments. Flat topography, wetlands, and fine sands are characteristics that make places like Florida very different from the environments where the SCS method was originally developed. The SCS method arose from experiments with soils that are dominated by infiltration excess (Hortonian mechanism), where runoff occurs after rainfall intensity exceeds the infiltration capacity of the soil. In contrast, Florida is likely dominated by saturation excess runoff (Dunne mechanism), where the soil storage capacity between a shallow water table and the ground surface is filled, and all remaining rainfall becomes runoff. The sandy soils of Florida have very high infiltration capacities, and thus infiltration excess is less likely than saturation excess. As a consequence of the saturation-excess mechanism, wetlands expand in the wet season as the soil moisture storage around the perimeter is filled. A modified form of the SCS method is proposed with the objective that it is more suitable than the current method in flatly sloped, humid environments. Initial conditions, such as the pre-storm soil moisture profile and depth to water table, are critical when predicting runoff in these areas. Air encapsulation is addressed because its presence causes the soil storage capacity to be filled significantly faster than in its absence. Equations are presented that provide an estimate of the average depth to water table and average soil storage capacity in a catchment. Two Florida catchments and one runoff test bed were selected for testing the new methodology. The runoff test bed demonstrated the saturation-excess mechanism while the catchments provided larger-scale testing of the method. Though more data is needed to fully assess the performance of the method, the approach offers a more plausible mechanism for runoff estimation in shallow water table environments with sandy soils.

No description available.

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 / Water pollution in the United States costs billions of dollars every year. Surface water pollution is caused by excess nutrients and effects the value of fisheries, recreational activities, and commercial operations, and can even lead to health hazards such as dangerous algal blooms. A major source of water pollution is from agricultural activities such as fertilizing crops. This type of pollution is called non-point source, as there is no obvious point where excess nutrients from fertilizers or manure enters the water, such as a discharge pipe, instead the pollution flows over the land first and then into the waterways following the rainfall-runoff patterns. One way to prevent non-point source pollution from agricultural activities is to give farmers tools to optimize operations in a way that can help lower the chance that pollution will occur. Scientific models, like a weather forecast, can help, but there is a lack of tools made specifically for reducing water pollution that are available to farmers. This work focuses on creating a free to use, high resolution and rapid update forecast delivered over the internet, capable of informing agricultural management practices to reduce water pollution. Over the course of this study, published advances in watershed modeling were made filling gaps in existing forecast technology. The final product combines cutting edge watershed science, machine learning and statistical models, web mapping tools, and terabytes of data to meet design goals.

Water quality management for agricultural production is a complicated and interesting problem. Hydrological and economic factors must be considered when designing strategies to reduce nutrient runoff from agricultural activities. This dissertation is composed of three chapters that investigate cost-effective ways to mitigate water pollution from agricultural nonpoint pollution sources and explore farmers' incentives when participating in water quality trading programs. Chapter 1 investigates landscape targeting of best management practices (BMPs) based on topographic index (TI) to determine how targeting would affect costs of meeting nitrogen (N) loading goals for Mahantango watershed, Pennsylvania. We use the results from two climate models and the mean of the ensemble of seven climate models to estimate expected climate changes and the Soil and Water Assessment Tool-Variable Source Area (SWAT-VSA) model to predict crop yields and N export. Costs of targeting and uniform placement of BMPs across the entire study area (4.23 km2) are compared under historical and future climate scenarios. We find that with a goal of reducing N loadings by 25%, spatial targeting methods could reduce costs by an average of 30% compared with uniform BMP placement under three historical climate scenarios. Cost savings from targeting are 38% under three future climate scenarios. Chapter 2 scales up the study area to the Susquehanna watershed (71,000 km2). We examine the effects of targeting the required reductions in N runoff within counties, across counties, and both within and across counties for the Susquehanna watershed. We set the required N reduction to 35%. Using the uniform strategy to meet the required N reduction as the baseline, results show that costs of achieving a regional 35% N reduction goal can be reduced by 13%, 31% and 36% with cross-county targeting, within-county targeting and within and across county targeting, respectively. Results from Chapters 1 and 2 suggest that cost effectiveness of government subsidy programs for water quality improvement in agriculture can be increased by targeting them to areas with lower N abatement costs. In addition, targeting benefits are likely to be even larger under climate change. Chapter 3 investigates the landowner's nutrient credit trading behavior when facing the price uncertainty given the credits are allowed to be banked for future use. A two-step decision model is used in this study. For the first step, we determine the landowner's application level of a BMP on working land in the initial time period. The nutrient credits awarded to the landowner depend on the nutrient reduction level at the edge of field generated by the BMP application. For the second step, we use an intertemporal model to examine the landowner's credit trading behavior with stochastic price fluctuations over time and with transaction costs. The theoretical framework is applied with a numerical simulation incorporated with a hydro-economic model and dynamic programming. Nutrient Management (NM) is selected as the BMP on working land to generate N credits. We find that gains to the landowner from credit banking increase with higher price volatility and with higher price drift, but that gains are larger with price volatility. However, for a landowner holding a small amount of nutrient credits, the gains from credit banking are small due to transaction costs. / Doctor of Philosophy / Two considerations are critical for efforts to mitigate nutrient runoff from nonpoint sources: cost effectiveness of strategies to reduce nutrient runoff and landowners' incentives to participate in these programs. This dissertation is composed of three manuscripts, aiming to evaluate the cost effectiveness of government subsidy programs for water quality management in agriculture and investigate the landowner's incentives to participate in water quality trading programs for the Chesapeake Bay watershed. Chapter 1 investigates gains from targeting Best Management Practices (BMPs) under current and future climate conditions based on the soil characteristics relative to uniform BMP application for a small experimental watershed (4.23km2). Chapter 2 scales up the study area to a 71,000 km2 watershed and treats each county within the watershed as a representative farm to explore economic gains from targeting within county and across county based on counties' physical conditions and agricultural patterns. Both Chapters show that cost-effectiveness of government subsidy programs can be improved by spatial targeting BMPs to areas with lower abatement costs. Gains from targeting increase under climate change. In Chapter 3 we shows how a landowner's revenues from nutrient credit selling will be affected if the credits are allowed to be banked for future use when she faces price uncertainty. We find that gains to the landowner from credit banking increase more with higher price volatility than with higher price drift. Gains from banking are largely reduced by transaction costs associated with trading.