Wilson, Grace2018-09-212018-09-212018-07https://hdl.handle.net/11299/200177University of Minnesota Ph.D. dissertation.July 2018. Major: Land and Atmospheric Science. Advisor: David Mulla. 1 computer file (PDF); vii, 150 pages.In the first part of this research, the field-scale hydrologic model, DRAINMOD-NII, was used to make predictions on nitrogen losses for fields utilizing different nitrogen fertilizer management practices. DRAINMOD-NII was used to predict nitrate loads for different fertilizer application rates and timing to corn at Waseca, Lamberton, and Willmar in Southern Minnesota. The fertilizer practices simulated included a single application in the spring before planting, a split application with half applied pre-plant and half at approximately the V6 stage for corn, and split and variable rate N practice (VRN) which utilized the split timing and a lower rate based on in-season monitoring of plant N requirements. Results of simulations at all three locations were used in a regression analysis to develop equations to predict nitrate-N loses for the region more generally as a function of fertilizer timing and application rate. Model results showed that fertilizer application applied as a VRN practice has the potential to reduce nitrate loads in southern Minnesota. In the second part of this project, a new framework was developed which would take into account uncertainty in the evaluation of hydrologic and water quality models. Model performance is evaluated using statistical goodness-of-fit (GOF) measures that compare the observed data measured in the field to model predictions. Though both the observed and predicted data contain uncertainty due to measurement error and model input parameter error, current methods of calculating GOF do not take these uncertainties into account. Here, a framework was developed that allows the observed and predicted data to be described in terms of the uncertainty due to measurement and model parameter error. This framework was used to adjusted mean square error (MSE) and normalized mean square error (NMSE) to account for errors in the observed and predicted data. Additionally, the accuracy of this approach was evaluated.enThesis or Dissertation