|dc.description.abstract||Agricultural productivity in the Lower Arkansas River Valley (LARV) in southeastern Colorado has been high over the last 100 years due to extensive irrigation practices. In the face of this high productivity, however, the LARV currently face many issues as a result of the long period of irrigation, including waterlogging and soil salinization, leading to a decline in crops yields and high concentrations of nutrients and trace elements. In particular, irrigation practices have led to high concentrations of selenium (Se) and nitrate (NO3) in groundwater, surface water, and soils, similar to other semi-arid irrigated watersheds worldwide. Environmental concerns due to these high concentrations include human health, health of fish and waterfowl, and eutrophication of surface water bodies. The objective of this thesis is to identify water management strategies that can lead to a decrease in the concentrations of Se and NO3 in groundwater and surface water in the LARV by evaluating the three-water management BMPs which is reduced irrigation (RI), lease fallowing of irrigated land (LF), and canal sealing (CS). This is accomplished by constructing and testing a computational model that simulates the fate and transport of Se and NO3 in a coupled irrigated stream-aquifer system, and then applying the model to evaluate selected best management practices (BMPs) to decrease the concentration of Se and NO3 to comply with Colorado water quality regulations. The modeling system consists of MODFLOW, which simulates groundwater and stream flow, and RT3D-OTIS, which simulates the reactive transport of the principal Se and nitrogen (N) species in groundwater and a connected stream network. RT3D-OTIS uses simulated flows from MODFLOW to exchange Se and N species' mass between streams and the aquifer on a daily time step. The coupled flow and reactive transport model is applied to an approximately 552 km² study region in the LARV between Lamar, Colorado and the Colorado-Kansas border. The model is tested against Se and NO3 concentrations measured in a network of groundwater monitoring wells and stream sampling site, and against return flows and mass loads to the river estimated from the mass balance. Model calibration was performed manually and by using PEST software tool, and the effects BMPs on Se and NO3 concentrations in groundwater, streams, and groundwater mass loadings to the Arkansas River within the stream-aquifer system are quantified. Three BMPs are considered RI, LF, and CS, which are simulated for a 40-year period and then compared to a baseline ("do nothing") scenario. The results indicate that implementation of the CS scenario might lead to lower groundwater concentrations of Se and NO3 by 40% and 38%, respectively, a reduction in groundwater mass loading to the Arkansas River by 100% and 60% for Se and NO3, and a reduction in stream concentrations of Se and NO3 by 30% and 40%, respectively. In contrast, the RI and LF scenario, while lowering the water table and in consequence the rate of groundwater return flow to the Arkansas River, leads to elevated groundwater concentrations of both Se and NO3 in the riparian areas, resulting in an overall increase in groundwater mass loading to the river. This may be due to changes in the rate of groundwater flow due to lower hydraulic gradients leading to longer residence times of NO3 in the aquifer, increasing the potential for the release of Se from the bedrock shale through oxidation processes. Also, lowering the water table due to reduced recharge from irrigation reduces the size of the saturated zone, perhaps contributing to a higher concentration of Se and NO3. Moreover, changes in water and mass flux between the saturated and unsaturated zone occur under RI and LF scenarios. As a consequence of these altered processes, the RI and LF scenarios do not decrease the in-stream concentrations of Se and NO3 in the Arkansas River, with values for Se and NO3 increasing by 15% and 8%, respectively under the RI scenario, and by 10% and 10.5% for the LF scenario. Further, the results are compared with results obtained from a modeling study in the Upstream Study Region of the Lower Arkansas River Valley, to determine the similarity and differences of BMP implementation in the two regions. Further assessment of localized BMPs should be performed to determine key regions where they should be implemented for the largest impact on Se and NO3. Combined water management BMPs and land management BMPs, like reduced fertilizer application and enhanced riparian buffers, should also be evaluated.