Uncertainties in Climate Impacts on Grain Maize in Switzerland – Does the Choice of Crop Modelling Approach Matter?

Abstract

Crop models are commonly applied to estimate impacts of projected climate change and to anticipate suitable adaptation measures. Thereby, uncertainties from global climate models, regional climate models, downscaling techniques and impacts models cascade down to impact estimates. It is essential to quantify and understand uncertainties in impact assessments in order to provide informed guidance for decision making in adaptation planning. A question that has hardly been investigated in this context is how sensitive climate impact estimates are to the choice of the impact model approach.In a case study for Switzerland we compare results of three different crop modelling approaches to assess the relevance of impact model choice in relation to other uncertainty sources. The three approaches include an expert-based, a statistical and a process- based model. With each approach impact model parameter uncertainty and climate model uncertainty (originating from climate model chain and downscaling approach) are accounted for. ANOVA-based uncertainty partitioning is performed to quantify the relative importance of different uncertainty sources.Results suggest that estimated yield impacts can differ substantially depending on the choice of the crop modelling approach. Impact uncertainty originating from impact model choice is in fact larger than uncertainty originating from climate model chain. Uncertainty contributions from downscaling and crop model parameterization are small in comparison. While estimates of yield changes are highly uncertain, the directions of estimated changes in climatic limitations are largely consistent. This leads us to the conclusion that by focusing on estimated changes in climate limitations, more meaningful information can be provided to support decision making in adaptation planning – especially in cases where yield changes are highly uncertain.