Multi-Location Response and Calibration Stability of Potato Models to Changes in Atmospheric Carbon Dioxide Concentration

The accuracy of models to predict the impact of changing climate factors on crop growth is influenced by data availability and quality, model structure, and model calibration. In this context, we previously evaluated the ability of an ensemble of ten potato crop models to simulate the effect of ambient (aC) and elevated (eC) atmospheric carbon dioxide concentration on yield at 8 experimental locations across Europe. Each modeling group developed a single cross-location calibration parameter set using aC data from just two locations. Simulations were then conducted using this ‘limited’ calibration across all locations for aC and eC responses. Results indicated that the mean of all model responses were within the range of observed variation when averaged across all locations, but this accuracy varied substantially with experimental site. In the next phase of this study, modelers developed site-specific calibration parameters for each individual location, as well as one single cross-location calibration set, using the full set of aC data. This was viewed as a ‘full’ calibration approach since data from all 8 experiments were made available to the modelers. Model simulations were then conducted for eC response at each location using these new within- and cross-location calibration parameter sets. This presentation will focus on a) the differences in the accuracy of model response to eC across all sites using within- versus cross-location calibration, b) the variation in predicted yields among individual locations, and c) comparison of cross-, and within-, location responses to eC and aC between the ‘full’ and ‘limited’ calibration approaches. Multi-model accuracy to aC and eC responses, and the geospatial stability of the different model calibration parameter sets, will be quantified. Insights regarding the influence that data availability and calibration methodology have on these results will also be assessed.