The nexus between detection uncertainty and quantifying meteorological extremes of the Euro-Atlantic atmospheric rivers

Abstract While orographic and regional climate effects play a crucial role in how Atmospheric Rivers (ARs) affect the weather, it is essential to reduce uncertainty in detecting ARs to understand their impact on weather patterns better. This study delves into the relationship between AR detection uncertainty and the characterisation of meteorological extremes over the Euro-Atlantic region using long-term historical data from ERA5 during 1940-2022. Leveraging the Bayesian AR detection (BARD), a form of statistical machine learning model in the Toolkit for Extreme Climate Analysis (TECA), we examine AR detections and variability and their consequential impact on precipitation, wind speeds and temperature patterns over Europe, the UK and Scandinavia. There is widespread uncertainty and disagreement among AR detections from 1024 sets of AR priories (detectors) from the model at a given instance, which led to a large spread in the aggregated detected AR probabilities (ARP) spatially and temporally. The study reveals significant differences in ARs’ attributes, such as frequency, intensity, and their impact on weather phenomena at selected probability thresholds across space and time. The magnitude of meteorological phenomena associated with ARs over land varies based on the chosen deciles of ARPs. Although AR frequencies and intensities are increasing in many parts of Europe and the UK, the effects of AR detection uncertainties are more prominent over inland Europe and Scandinavia than over coastal Europe and the UK. Using the entire probability spectrum is an alternative to reduce uncertainty but may result in computational constraints and imprecise discernment of extreme events. Understanding meteorological extremes induced by AR uncertainties can help us develop effective strategies to mitigate hazards and adapt to changing climate conditions.