Simulating Potential Yield If Industry Is Disabled: Applying a Generalized Linear Modelling Approach to Major Food Crops

Modern civilization is highly dependent on industrial agriculture. Industrial agriculture in turn has become an increasingly complex and globally interconnected system whose historically unprecedented productivity relies strongly on external energy inputs in the shape of machinery, mineral fertilizers, and pesticides. It leaves the system vulnerable to disruptions of industrial production and international trade. Several scenarios have the potential to damage electrical infrastructure on a global scale, including electromagnetic bursts caused by solar storms or the detonation of nuclear warheads in the higher atmosphere, as well as a globally coordinated cyberattack. The current COVID-19 pandemic has highlighted the importance of crisis preparation and the establishment of more resilient systems. To improve preparation for high-stake risk scenarios their impact especially on critical supply systems must be better understood. To further the understanding of consequences for the global food production this work aims to estimate the effect the global inhibition of industrial production could have on the crop yields of maize, rice, soybean, and wheat. A generalized linear model with a gamma distribution was calibrated on current crop-specific gridded global yield datasets at five arcmin resolution. Gridded datasets on the temperature regime, the moisture regime, soil characteristics, nitrogen, phosphorus and pesticide application rates, the fraction of irrigated area and a proxy to determine whether farm activities are mechanized were chosen as explanatory variables. The model was then used to predict crop yields in two phases following a global catastrophe which inhibits the usage of any electric services. Phase 1 reflects conditions in the year immediately after the catastrophe, assuming the existence of fertilizer, pesticides, and fuel stocks. In phase 2 all stocks are used up and fertilizer, pesticides and fuel are not available anymore. While the fit varies dependent on the crop, the model agrees well with the data based on McFadden's ρ² (maize: 0.45, rice: 0.41, soybean: 0.34, wheat: 0.38). The predictions showed a reduction in yield of 10-30% in phase 1 and between 34 and 43% in phase 2. Overall Europe, North and South America and large parts of India, China and Indonesia are projected to face major yield reductions of up to 95% while most African countries are scarcely affected. The findings clearly indicate hotspot regions which align with the level of industrialization of agriculture. Further, it is shown that the yield reduction is likely III to be substantial, especially in industrialized countries. The analysis also provides insights on major factors influencing crop yield under losing industry circumstances. Due to data unavailability some crucial factors could not be included in the model, but their qualitative discussion leads to the conclusion that the presented results can be considered optimistic, and that further research is needed to quantify the impact of the omitted aspects.