Skip to main content
ARS Home » Midwest Area » Bowling Green, Kentucky » Food Animal Environmental Systems Research » Research » Publications at this Location » Publication #287984

Title: Sensitivity and uncertainty analysis for a field-scale P loss model

item Bolster, Carl
item Vadas, Peter

Submitted to: American Water Resources Association Conference Proceedings
Publication Type: Abstract Only
Publication Acceptance Date: 12/2/2012
Publication Date: 3/26/2013
Citation: Bolster, C.H., Vadas, P.A. 2013. Sensitivity and uncertainty analysis for a field-scale P loss model. American Water Resources Association Conference Proceedings. Abstract.

Interpretive Summary:

Technical Abstract: Models are often used to predict phosphorus (P) loss from agricultural fields. While it is commonly recognized that there are inherent uncertainties with model predictions, limited studies have addressed model prediction uncertainty. In this study we assess the effect of model input error on predictions of annual P loss by the Annual P Loss Estimator (APLE) model, an empirically-based spreadsheet model developed to describe field-scale P loss from surface runoff. A sensitivity analysis for all APLE input variables was conducted to determine which variables the model is most sensitive to. Two methods – first-order approximation (FOA) and Monte Carlo (MC) simulation – were compared to determine whether the FOA method is appropriate for estimating uncertainties with APLE. APLE predictions were then evaluated against measured P loss data accounting for errors in both the model input variables and the observed data. Results showed that for low to moderate uncertainties in APLE input variables, the FOA method yields reasonable estimates of model prediction uncertainties, though for cases where manure solid content is ~ 15%, the FOA method may not be as accurate as the MC method due to an artifact in APLE. The estimated uncertainties in APLE predictions based on assumed errors in the input variables ranged from ± 2 to 64% of the predicted value. Incorporating uncertainties in both measured data and model predictions improved model performance. Results from this study highlight the importance of including reasonable estimates of model uncertainty when using models to predict P loss.