Drivers of compound drought-heatwave events: assessments of univariate extremes and causal soil moisture-temperature feedback

Compound drought-heatwave events (CDHEs), defined by the co-occurrence of soil moisture droughts and heatwaves, are among the most damaging climate extremes due to their impacts on ecosystems, agriculture, and humans. Previous studies have reported increasing CDHE occurrence in many regions. However, the extent to which CDHE trends are driven by long-term changes in the soil moisture–temperature (SM–T) feedback remains unclear, compared to their roles in single heat or drought events alone. In particular, traditional correlation-based approaches to quantify SM–T feedback are limited in their ability to resolve its causal roles.

We investigate how land–atmosphere feedback drives CDHEs using the normalized non-stationary Liang–Kleeman information flow. This framework allows us to quantify the strength of the coupling in both directions of the soil moisture–temperature feedback while considering common confounders to assess how these couplings have evolved over recent decades at the global scale. Using ERA5 and ERA5-Land, we find that the widespread increases in CDHE frequency cannot be fully explained by changes in heatwave or drought frequency alone. We identify significant trends in the coupling strength for directions of the SM-T feedback, with generally stronger trends in regions with higher water availability.

We further combine this causal analysis with anthropogenic attribution to disentangle the respective roles of anthropogenic forcing and natural climate variability, using an ensemble of CMIP6 models under historical and natural-only forcings. We find significant effects of anthropogenic emissions on CDHE frequency across most land areas. On the contrary, we find heterogeneous spatial patterns of the anthropogenic impact on the frequency of univariate extremes, emphasizing the need to investigate anthropogenic impacts on the dependence between climate variables. We therefore investigate and detect anthropogenic influences on the trends of the SM-T feedback across most land areas for both coupling directions, highlighting the role of evolving land–atmosphere feedbacks in shaping compound event likelihood. Our study provides a physically interpretable and reanalysis-based pathway toward improved understanding of compound extremes under ongoing climate change using causal, multivariate methods for identifying the compounding physical drivers of high-impact climate events.