Spatiotemporal Analysis of Soil Drought Evolution in France: Attribution to Atmospheric Drivers

Drought is a dry period characterized by an abnormal water deficit relative to local climatology, propagating through the land surface hydrological cycle. Soil drought, in particular, refers to a deficit of accessible water for vegetation, affecting ecosystems and societies through activities such as agriculture and infrastructure stability. Soil droughts are expected to evolve under climate change since their two meteorological drivers, precipitation and reference evapotranspiration (which represent the atmospheric water demand), are evolving, too. Under climate change, reference evapotranspiration is projected to increase, and precipitation patterns are expected to shift. However, the evolution of droughts in France remains uncertain, and understanding these changes brings information for adaptation strategies. Since drought events unfold over both space and time and their impacts depend on these spatiotemporal characteristics, this study analyzes them as contiguous spatiotemporal phenomena.

This study proposes a methodological framework to (1) identify spatiotemporally contiguous soil drought events, (2) analyze changes in their characteristics under climate change, and (3) attribute these changes to meteorological drivers. The detection method, adapted from existing algorithms for identifying large-scale spatiotemporal extreme events, is here tailored to study soil droughts at a regional scale using high-resolution data. This method connects contiguous points where standardized water deficits exceed a predefined threshold in space and time. Additionally, the framework integrates an attribution approach adapted from Zscheischler et al. (2013) that links detected changes in drought characteristics to meteorological drivers, here precipitation and evapotranspiration, offering a detailed perspective on the mechanisms underlying these changes.

The framework is applied to France using high-resolution monthly data (8 km × 8 km) from the SAFRAN atmospheric reanalysis (1958-2020) and 12 climate simulations under greenhouse gases emission scenario RCP 8.5 (1950–2100) from the EXPLORE2 project, which drive a Land Surface Model offline. Precipitation, reference evapotranspiration, and soil wetness are standardized relative to the 1960–2020 baseline using the Standardized Precipitation Index method. Uncertainty is assessed by evaluating the spread across the ensemble of 17 climate simulations and comparing simulated historical events against reanalysis data. Results show that simulations reproduce past drought characteristics with sufficient accuracy to analyze future trends. Projections indicate an increase in drought intensity by the end of the 21st century, primarily driven by rising reference evapotranspiration.

Zscheischler, Jakob, Miguel D. Mahecha, Stefan Harmeling, and Markus Reichstein. 2013. “Detection and Attribution of Large Spatiotemporal Extreme Events in Earth Observation Data.” _Ecological Informatics_ 15 (May):66–73. [https://doi.org/10.1016/j.ecoinf.2013.03.004](https://doi.org/10.1016/j.ecoinf.2013.03.004).