Causal analyses reveal changing land-atmosphere patterns and soil moisture control under warming

Dian Indrawati, Somnath Mondal, Poulomi Ganguli, Auroop Ganguly

Climate change is projected to modify the global water cycle and land-atmosphere interactions. However, warming-induced changes in multivariate dependence remain insufficiently explored. Here we examine the interdependence among temperature, precipitable water, precipitation, evaporation, soil moisture, and runoff in all five earth system models within the latest IPCC CMIP6 ensembles that generate relevant simulated data. Our causal network analyses, augmented by artificial intelligence, suggest a near universal statistically significant shift in interdependence structures, along with a growing role of soil moisture, under mid- and end of century scenarios, including pronounced regional and seasonal heterogeneity. Our results have implications for food security, including a 3% decline in global domestic production and 8% in the northern subtropics. Our findings motivate reliable observations and credible early warning systems.