Exploring Climate Variability and Worst-Case Drought Storylines using Large Ensembles and Ensemble Boosting

Understanding and robustly sampling climate variability is vital for producing reliable near- and long-term projections of water availability and drought. Climate variability on interannual to multi-decadal timescales can substantially influence precipitation, temperature, or humidity, shaping the intensity, frequency, and persistence of extreme hydrological events. Particularly for multi-year variability, such influences can lead to consecutive years of extreme hydrological stress, challenging the resilience of natural and human systems. Furthermore, sampling the worst-case, most extreme yet plausible conditions of extreme drought, potentially compounding with other system stressors, is crucial for producing comprehensive risk assessments. Regionally, climate variability can amplify or dampen the anthropogenic signal of global warming. Therefore disentangling its contribution from such anthropogenic changes is crucial to understand observed changes and how they may continue into the future, as well as to determine worst-case or unprecedented conditions plausible today.

Here, we showcase how climate variability sampling techniques such as Single Model Large Ensembles (SMILEs) and Ensemble Boosting can be used to assess how soon unprecedented extreme heat and drought stress could occur over Europe, whether it could happen successively year after year, and how intense worst-case heat and drought stress could become already today. SMILEs consists of several simulations from one climate model under the same forcing to capture the effect of freely evolving internal variability and generate a range of possible climate outcomes, from daily to centennial scales. Ensemble Boosting uses extreme conditions in a SMILE as a starting point, which are then re-run under a small butterfly-effect like perturbation to produce hundreds of physically consistent storylines that explore worst-case extremes, by amplifying the chaotic nature of climate variability around the original parent event itself. Together, these approaches are extremely powerful tools to produce risk storylines that remain physically consistent across time, space, and across variables, and that can be used to assess hydrological impacts to better prepare for the challenges posed by accelerating climate change and its influence on water resources.