Long-term assessment of arable and grassland systems in Veneto: EPIC modeling and regional soil representativeness calibration

Applying biogeochemical models at the regional scale is difficult because soils, climate, land use, and agricultural management can vary considerably across space. In addition, model parameters that perform well at one site may not necessarily be suitable for other environmental conditions. This study addresses these challenges by evaluating the Environmental Policy Integrated Climate (EPIC) model across several long-term experimental sites in the Veneto region of northeastern Italy. Five experimental sites were included, covering both arable cropping systems and permanent grasslands and representing a range of environmental conditions, from lowland agricultural areas to mountainous environments. Information from the experimental sites was complemented with soil data from the LUCAS (Land Use/Cover Area frame Survey) database, particularly to improve the representation of permanent grassland soils, which are less well represented in long-term experimental datasets. Before calibration, a Morris sensitivity analysis was performed to identify the parameters with the strongest influence on model outputs and to reduce the number of parameters included in the calibration. The selected parameters were then calibrated using a multi-objective Non-dominated Sorting Genetic Algorithm (NSGA), combining observations from different experimental sites and considering both crop yield and soil organic carbon. By calibrating the model across several sites at the same time, the aim was to identify parameter sets that could perform reasonably well under different soil, climate, and management conditions, rather than producing highly site-specific calibrations. The model was evaluated under a range of agricultural practices, including conventional and conservation management and cover cropping, as well as across arable and permanent grassland systems. This allowed EPIC performance to be compared across management systems and soil conditions. Overall, the results show the importance of using long-term observations from contrasting agricultural systems when calibrating process-based models for regional applications. The multi-site approach adopted in this study provides a practical framework for improving EPIC parameterization and for supporting its application in heterogeneous agricultural regions.