Exploring soil organic carbon dynamics through a multi-model simulation of multiple long-term experiments 

Agricultural system models are widely recognized as valuable tools for identifying best management practices and addressing the challenges posed by climate change. In this context, the use of model ensembles has been recently recommended for their enhanced performance and accuracy. However, assessing their effectiveness over a large geographical area, such as national scale is often currently lacking. This study focuses on simulating soil organic carbon (SOC) dynamics using an ensemble of models comprising DSSAT, CropSyst, EPIC, and APSIM models, utilizing data derived from five Long-Term Experiments (LTEs) spread across a north-to-south pedoclimatic range transect in Italy. This region is of particular importance as it represents a significant hotspot for climate change. The LTEs featured a robust array of 63 unique experimental protocols, incorporating variation effect in fertilization rates, cropping rotations, and tillage prescriptions. This resulted in a total of 2184 years of simulated data for each model. The dataset employed included SOC stocks and crop yield and biomass. Models underwent independent calibration, with crop and SOC parameters selected based on expert knowledge. Main crop cultivars, such as maize, soybean, sugarbeet, and wheat, were further categorized and calibrated by maturity classes. A similar approach was used for cover crops. The extensive dataset enabled a nuanced exploration of the models’ performance across varied agro-ecological contexts. The models proved capable of accurately reproducing the varied pedo-climatic conditions of the Italian peninsula, contributing to the advancement of our understanding of SOC dynamics.