Crop growth and soil water fluxes at erosion-affected arable sites: A model inter-comparison based on weighing-lysimeter observations

Agro-ecosystem models have been developed to study effects of agricultural management on crop production, mostly from an agronomic point of view. Based on a biophysical process representation, their most prominent advantage is the coupled modelling of crop development and yield formation, as well as water and nutrients fluxes in the plant-soil system. Crop models have previously been calibrated based on experimental data with a focus on plant observations. Less attention has been given to soil water and solute dynamics despite the importance of plant nutrient availability and chemical leaching, particularly for arable soils often affected by erosion. The question was whether the description of soil processes and properties play an important role in the crop simulations.

The aim of this study was to compare the ability of agro-ecosystem models to predict crop development and water fluxes under changing environmental conditions. Observations on crop growth and soil water dynamics were obtained from four weighable lysimeter of the TERENO-SOILCan lysimeter network in the northeast of Germany (Dedelow). The intact soil monoliths are representative for the spatial soil variability of erosion-affected hummocky agricultural landscape. Twelve agro-ecosystem models (AgroC; DailyDayCent; Daisy; HERMES; MONICA; Theseus, Theseus-HydroGeoSphere; Theseus-Hydrus-1D; Expert-N coupled to CERES, GECROS, SPASS, and SUCROS) were tested. Crop development stages were used to calibrate the agro-ecosystem models. The model performance was tested against observed grain yield, aboveground biomass, leaf area index, actual evapotranspiration, drainage, and soil water content.

Model descriptions were highly diverse for both crop development and water fluxes. Crop growth and soil water fluxes were better predicted by the Multi Model Mean simulations than by any individual model. Results demonstrate that i) the hydraulic properties of erosion-affected soil profiles controlled the observed interactions between crop yield, plant development, and water fluxes, ii) data on phenological stages contained insufficient information content to calibrated agro-ecosystem models for soils affected by erosion, and iii) neither an individual model nor the Multi Model Mean could describe the observation on crop development and water dynamics, when using phenological stages only for model calibration. The results suggest that soil does matter in agro-ecosystem models and that weighable lysimeter can provide such soil related observation.