Machine Learning‐Based Detection of Weather Fronts and Associated Extreme Precipitation in Historical and Future Climates
Extreme precipitation has significant consequences and impacts and is expected to increase in intensity with climate change. Extreme precipitation also originates from many different sources, and it is important to understand these sources, their associated mechanisms, and how they might change in a warming climate. In the midlatitudes, extreme precipitation is often associated with weather fronts, pointing to a need for analyzing regional and seasonal changes in fronts of different types as an important component to understanding how total and extreme precipitation, and their associated impacts, might change in the future. In this work, we use a machine learning-based detection algorithm (DL-FRONT) to identify weather fronts over the contiguous United States (CONUS) in high resolution Community Earth System Model (CESM) simulations. We compare detected fronts in different modeled climates to study the impact of climate change on weather fronts. We further associate the detected fronts with total and extreme precipitation, studying responses across seasons and front types, to understand how climate change impacts the intersection of synoptic-scale features with extreme events.
We show success in applying DL-FRONT to CESM output over CONUS, despite the algorithm being trained on observational and reanalysis data. We find that detected front frequencies in CESM have seasonally varying spatial patterns and responses to climate change and are found to be associated with modeled changes in large scale circulation such as the jet stream. We also find that total and extreme frontal precipitation mostly decreases with climate change, with some seasonal and regional differences. While CONUS mean and extreme precipitation generally increase during all seasons in these climate change simulations, the likelihood of frontal extreme precipitation decreases, demonstrating that extreme precipitation has seasonally varying sources and mechanisms that will continue to evolve with climate change.
Our results demonstrate how machine learning-enabled automated detection of synoptic weather features, specifically fronts, can enable greater understanding of seasonal and regional precipitation sources and mechanisms. Leveraging the power of automated front detection, we discuss the mechanisms related to the seasonality of different front types and projected changes due to climate change. We find seasonal differences in total frontal precipitation changes across CONUS driven by the relative importance of changes in frontal precipitation frequency and intensity. By investigating modeled changes in total and extreme frontal precipitation, we advance the understanding of extreme precipitation events, their intersection with frontal systems of different types, and how those associations are changing with climate change.