Crop model inter comparison study for the same soil ecosystem under different climatic conditions

Comparisons of agroecosystem models have mostly been conducted to test their predictive ability under previous climatic conditions and for soils of the same site. However, the ability of individual agroecosystem models to predict the effects of changes in climatic conditions on soil ecosystems beyond the range of site-specific variability is unknown. The objective of this study was to test the predictive ability of agroecosystem models using weighable lysimeter data for the same soil, but under different climate conditions and to compare simulated plant growth and soil ecosystem response to climate change between these models. For this purpose, data from the TERENO-SOILCan lysimeter network for a soil ecosystem at the original site (Dedelow) and data from the lysimeters transferred to Bad Lauchstädt and Selhausen with Dedelow soil monoliths were evaluated. Soils were transferred to a drier and warmer site (Bad Lauchstädt) and to a warmer and wetter site (Selhausen) compared to the original site of the soils in Dedelow with the same crop rotation. After model calibration for the data from the original Dedelow site, plant growth and soil water balances of the transferred Dedelow soil monoliths were predicted using the site-specific boundary conditions and compared to the observations in Selhausen and Bad Lauchstädt. The overall simulation results of the models were divided into a plant-related part, ecosystem productivity (grain yield, biomass, LAI) and an environmental-related part, ecosystem fluxes (evapotranspiration, net drainage, soil moisture). The results showed that when the soil was moved to a drier region, the agronomic part of the crop models made good predictions, and when the soil was moved to wetter regions, the environmental flow part of the models seemed to make better predictions. The results suggest that accounting for climate change scenarios, more consideration of soil properties, and testing model performance for conditions outside the calibrated range and site-specific variability will help improve the models