Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: 3/28/2015
Publication Date: 9/22/2015
Publication URL: http://handle.nal.usda.gov/10113/62944
Citation: Bolster, C.H., Vadas, P.A., Boykin, D.L. 2015. Model parameter uncertainty analysis for an annual field-scale phosphorus loss model. Meeting Abstract #94.
Technical Abstract: Phosphorous (P) loss models are important tools for developing and evaluating conservation practices aimed at reducing P losses from agricultural fields. All P loss models, however, have an inherent amount of uncertainty associated with them. In this study, we conducted an uncertainty analysis with the Annual P Loss Estimator (APLE) model, an empirically-based spreadsheet model developed to describe annual, field-scale P loss when surface runoff is the dominant P loss pathway. We first estimated and evaluated model parameter uncertainties associated with five internal regression equations used by APLE to calculate total soil P from measurements of soil clay content, organic matter, and labile P; the P enrichment ratio determined from erosion rates; concentration of P in runoff calculated from labile soil P; and partitioning of P between runoff and infiltration for applied manures and fertilizers based on runoff ratio. Our analysis included calculating parameter uncertainties and 95% confidence and prediction intervals for five internal regression equations in APLE. We then predicted P loss while including uncertainties in both model parameters and model inputs and compared the relative magnitude of these sources of uncertainty to the overall uncertainty associated with predictions of P loss. Good correlations between predicted and observed data were found, though a significant amount of observed variability was not captured by the equations. This resulted in uncertainties in predicted P loss ranging from 5 to 50 %. Results from this study highlight the importance of including reasonable estimates of model parameter uncertainties when using models to predict P loss. Our results also demonstrate how the estimation of model parameter uncertainty can identify model limitations.