Item

Abstract

Spatial correlation between climate variables may modulate concurrent regional crop failures and reduce global crop production. However, the influence of spatial correlation in crop production fields on globally aggregated production remains poorly understood. Systematically addressing this gap using observed crop production is challenging, as such observational datasets typically suffer from limited sample sizes and/or coarse spatial information. Here, using gridded global simulations from the Global Gridded Crop Model Intercomparison Phase 3 (GGCMI3), we quantify how spatial correlation between regional crop productions influences global production across different spatial scales for maize, wheat, soybean, and rice. By employing the mean of crop production from multiple crop models forced with reanalysis climate data, we find minimal influence of the correlations between the productions of major breadbasket regions on global breadbasket-aggregated production. This aligns with the fact that global major breadbasket regions are generally non-large and distant from each other, whereas spatial correlations in the crop production field influence global crop production through correlations between small and nearby areas. The correlation between crop production of areas characterized by small spatial scales (100–1000 km) enhances extremely low (5th percentile) global production by about 0.9-1.1 standard deviation of the global production on average. This correlation effect at small spatial scales is less important for weaker extremes of low global crop production. Finally, crop model simulations forced with bias-corrected climate simulations often are not able to reproduce the correlation effects seen in crop model simulations forced with reanalysis climate data, suggesting that bias-corrected climate model input may degrade correlation effects in GGCMI3 crop simulations. These model-based results highlight that spatial correlations are a critical driver of global production risk, stressing the need for improved cross-regional processes representation in crop models to enhance future food security risk assessments.