Modeling soil organic carbon evolution in long-term arable experiments with AMG model

Abstract

Reliable models predicting soil organic carbon (SOC) evolution are required to better manage cropping systems with the objectives of mitigating climate change and improving soil quality. In this study, data from 60 selected long-term field trials conducted in arable systems in France were used to evaluate a revised version of AMG model integrating a new mineralization submodel. The drivers of SOC evolution identified using Random Forest analysis were consistent with those considered in AMG. The model with its default parameterization simulated accurately the changes in SOC stocks over time, the relative model error (RRMSE = 5.3%) being comparable to the measurement error (CV = 4.3%). Model performance was little affected by the choice of plant C input estimation method, but was improved by a site specific optimization of SOC pool partitioning. AMG shows a good potential for predicting SOC evolution in scenarios varying in climate, soil properties and crop management.