Understanding, interpreting, and communicating future extreme precipitation risk using flow precursors

Many of the most societally impactful weather events in Europe occur on short timescales and there is a growing demand for improved projections of how such extremes will change in the future. That is, how will global climate change over decades impact extreme weather over days? The multiscale nature of this question challenges the capabilities of current earth system models, and this is especially the case for hydrometeorological extremes. Accurately simulating the hazards posed by extreme precipitation requires faithfully resolving interactions between the large-scale circulation, synoptic dynamics, the local boundary-layer, and hydrological and land surface conditions.

 

This is not only a quantitative modelling challenge, but a challenge of interpretation and narrative: the dynamics of extreme precipitation are diverse across space and time, and the statistics of the highest impact events are necessarily poorly constrained. These challenges are complicated further by the evergrowing size and hetereogeneity of multi-model datasets How can we explain model biases and trends in extreme precipitation? When models project similar changes in hydrometeorological risk do they do so for the same reasons? What implications do these factors have for regional downscaling and impact modelling? Can we relate future extremes quantitatively and robustly to historical high-impact events, as often requested by societal stakeholders?

 

We tackle these questions through a novel flow-precursor framework, applied to observational data, large ensemble climate simulations and subseasonal weather forecasts. We decompose extreme event risk into contributions from different scales and flow conditions, using regionally specific synoptic flow precursors which are directly associated with individual high-impact extremes or classes of extreme. These precursors are algorithmically identified and can be easily computed in large datasets, allowing us to obtain a physical interpretation of changing extreme risk across Europe without obscuring regional or seasonal diversity in precipitation dynamics.

 

We show how climate model biases and forced changes in extreme precipitation can be explained, categorised, and visualised in a succinct way that highlights important differences in their suitability for use in downscaling, impact modelling and storyline development. We demonstrate how dynamical decomposition can extract usable climate information even from heavily biased models, and how insights from models at different scales–such as from large climate ensembles and high-resolution weather forecasts–can be quantitatively synthesised to provide new insights on future hazards and plausible worst-case scenarios. Finally, we show how the method can be used to reframe complex, probabilistic climate projections and weather forecasts in terms of individual high impact historical events, aiding scenario visualisation, and allowing stakeholders to leverage their experience and domain knowledge when preparing for future high-impact extremes.