Evaluation of multi-year droughts in global SMILEs

In recent years, consecutive drought years have affected large areas of the world, such as Europe in 2018-2020 and Northern America in 2020-2023.  Drought conditions on any given year pose considerable risk to agriculture, forestry, ecosystems or water and energy supply. Multi-year droughts bear the potential to aggravate such impacts due to water deficit build-up over a longer period without insufficient recovery during wet seasons. Adaptation strategies usually work for limited time and rely on recovery periods (e.g., storage lakes). Multi-year droughts thus strongly challenge current drought preparedness, adaptation and mitigation measures. 

With changing climate, droughts are projected to increase worldwide in duration and frequency. Due to legacy effects of depleted soils and self-intensification processes, the risk for full years of water deficits rises further. For well-informed adjustment of adaptation to multi-year droughts, a comprehensive assessment of their risks under current and future climate conditions is required. Therefore, it is crucial to assess the skill of climate models in simulating multi-year droughts globally. 

In this study, we identify models that perform best over different hotspot and regions where models share high, or more worryingly low skill in representing multi-year droughts. We also assess the sensitivity of multi-year drought definition to different drought metrics used, including the 6-month and 12-month standardized precipitation evapotranspiration index (SPEI). For the analysis, we focus on a minimum duration of 12 months with SPEI < -1.

To sufficiently sample climate variability and produce large enough samples of extreme, multi-year droughts, we use a range of CMIP6 single-model initial condition large ensembles (SMILEs). SMILEs provide multiple runs of shared forcing and model configurations, but different starting conditions and evolutions that sample the range of internal climate variability. Here, SMILEs are evaluated regarding their capability to depict multi-year droughts against reanalysis data for the recent past (1991-2020), and based on this evaluation we provide projections of the change in multi-year droughts based on best-performing models. This work presents first results on regional and global scales, acknowledging internal climate variability of the representation of multi-annual droughts.