Sensitivity analysis, calibration and validation of EPIC for modelling soil phosphorus dynamics in Swiss agro-ecosystems

Proper management of soil phosphorus (P) is essential to ensure sustainability in agriculture. To avoid excessive P accumulation in soil, a methodology for predicting long-term variations in soil P content is required. Using the Swiss Soil Monitoring Network (NABO) database, we tested the efficiency of the model EPIC to predict soil P temporal changes for typical Swiss arable and grassland sites. After performing a sensitivity analysis to identify the most influential model parameters for which calibration was needed, we calibrated the influential parameters on the available topsoil P data of 4 selected NABO sites and then we tested the calibrated model by comparison of predictions with data from a second set of 14 sites. We found that the model performance for grassland sites improved significantly when site-specific estimates of bioturbation depth were used. These site-specific estimates showed a close relationship to independently assessed subsoil hydromorphy.