Simple scaling of climate inputs allows robust extrapolation of modelled wheat yield risk at a continental scale

Abstract

Climate change increases variability and uncertainty of crop performance. Process-based crop growth models represent the complex spatio-temporal interactions between plants, atmosphere, and soils and enable realistic climate risk assessments of future crop yield. But they require continuous, detailed daily weather data. Probability distributions of crop model results provide risk profiles of yield and serve to assess the impacts of long-term climate variability and change on crop yields. This paper tests to what extent a simple method for adjusting daily weather data using seasonal and monthly factors can produce robust estimates of risk profiles at a continental scale.We examined the predictability of risk profiles of modelled wheat grain yield across the Australian grain belt. Snowtown, in the middle of the South Australian grains belt (33.8°S, 138.2°E) was selected as the reference site, and 49 wheat-growing sites spanning from 23.5 to 42.8°S of latitude and 115–151.8°E of longitude were used for testing the adjustments of precipitation, maximum and minimum temperatures and global solar radiation. Adjustment factors were calculated as the difference in long-term average of a given climate variable between a test site and the reference site. For each test site, we compared risk profiles modelled with observed weather data with step-wise adjusted weather data.Simple adjustments of both rainfall and temperatures produced good matching of risk profiles (root mean square error, RMSE < 0.5 t/ha) in 80% of the sites. Adding the adjustment of the temperatures – with monthly factors- and solar radiation improved the match of risk profiles in the most climate-contrasting sites. In regions with limited availability of high-quality climate data, simple scaling of climate inputs used in this study can provide basic climate data for modelling and generating robust risk profiles of crop yield. Keywords: Spatial analogues, Climate risk, Crop modelling, APSIM, High-quality climate data, Limited climate data, Poor-data environments, Risk profile