Using farm accountancy data to calibrate a crop model for climate impact studies

Process-based crop models are widely used in decision support systems or to assess impacts of climate change on agriculture at different spatial scales. They include crop and/or cultivar-specific parameters that need to be calibrated. However, the availability of reference data is often limited. An alternative is to use yield records from widely available Farm Accountancy Data Network (FADN). The goal of this study was therefore to propose and test a crop model calibration procedure that makes use of FADN data. To account for the lack of management information in the FADN databases, the concept of ‘Meteorologically Possible Yield’ (MPY) was adopted. This concept is particularly relevant in the context of climate impact studies as MPYs are by definition only driven by climate variability. As an example, the procedure was applied to calibrate the generic crop model CropSyst for a region located in north-eastern Switzerland. Validation using data from a long-term field trial with detailed information on fertilizer applications showed that the proposed procedure provides robust simulation results and is therefore suitable for climate impact studies in regions where detailed experimental data are scarce. In a case study application, the transferability of the local calibration to a site with drier conditions was tested and simulation results for this new site were compared to results obtained using local recalibration. Results showed that predicted yields can differ substantially and the differences can be strongly amplified when impacts of climate change are considered. This highlights the need for adjusting model calibrations to local site conditions and for considering parameter uncertainties in climate impact studies.

► We propose a crop model calibration approach based on widely available yield data. ► Estimated ‘Meteorologically Possible Yields’ are used as calibration reference. ► Potentials are provided for calibrating crop models for wide range of locations. ► Local model recalibration is needed to account for differences in site conditions. ► Climate change impact assessment can highly depend on the crop model parametrization.