Sensitivity and uncertainty analysis for the annual phosphorus loss estimator model

Area of Science:

• Agricultural Science
• Environmental Modeling
• Soil Science

Background:

• Predictive models are crucial for estimating phosphorus (P) loss from agricultural fields.
• Model prediction uncertainties are often acknowledged but infrequently quantified in P loss modeling.
• The Annual P Loss Estimator (APLE) is a commonly used model for predicting annual P loss.

Purpose of the Study:

• To assess the impact of model input errors on APLE predictions of annual P loss.
• To conduct sensitivity analyses of APLE input variables to identify key drivers of prediction uncertainty.
• To compare the accuracy of the first-order approximation (FOA) method with Monte Carlo simulation (MCS) for estimating model uncertainties and evaluate APLE performance against measured data.

Main Methods:

• Sensitivity analysis was performed on all APLE input variables.
• The first-order approximation (FOA) method was compared with Monte Carlo simulation (MCS) for uncertainty estimation.
• APLE model performance was evaluated using measured P loss data, incorporating uncertainties in both predictions and measurements.

Main Results:

• Input errors in APLE can lead to prediction uncertainties ranging from ±2% to 64%.
• The FOA method provides reasonable uncertainty estimates for low to moderate input uncertainties, but may be less accurate with specific manure solid content (14-17%) due to model discontinuities.
• Monte Carlo simulation (MCS) offers a more accurate approach for uncertainty estimation in certain scenarios.

Conclusions:

• Quantifying prediction uncertainties is essential when using models like APLE for estimating agricultural phosphorus loss.
• The choice of uncertainty estimation method (FOA vs. MCS) can influence the accuracy of predicted uncertainties, particularly with specific input ranges.
• Future P loss modeling efforts should integrate robust uncertainty analyses to improve prediction reliability and inform management decisions.