Identifying meteorological drivers of extreme impacts: an application to simulated crop yields

Abstract. Compound weather events may lead to extreme impacts that can affect many aspects of society including agriculture. Identifying the underlying mechanisms cause impacts, such as crop failure, is crucial importance improve their understanding and forecasting. In this study, we investigate whether key meteorological drivers be identified using least absolute shrinkage selection operator (LASSO) in a model environment, method allows for automated variable able handle collinearity between variables. As an example impact, failure annual wheat yield simulated by Agricultural Production Systems sIMulator (APSIM) driven 1600 years daily data from global climate (EC-Earth) under present-day conditions Northern Hemisphere. We then apply LASSO logistic regression determine which during growing season failure. obtain good performance central Europe eastern half United States, while regions Asia western States are less accurately predicted. Model correlates strongly with mean variability yields; is, highest relatively large variability. Overall, nearly all grid points, inclusion temperature, precipitation vapour pressure deficit predict addition, predictors seasons required prediction. These results illustrate omnipresence compounding effects both different periods creating events. Especially indicators diurnal temperature range number frost days selected statistical relevant at most underlining overarching relevance. conclude useful tool automatically detect compound could applied other wildfires or floods. detected relationships purely correlative nature, more detailed analyses establish causal structure impacts.