February 22, 2024
Research Highlight

Footprint of Large-Scale Climate Variabilities on Regional Weather Systems

For the first time, a weather system clustering scheme based on a deep learning model reflects the influence of ENSO and MJO on regional weather and hydrological conditions


Clouds approach mountains near Seattle. Using a deep learning-based clustering model, scientists identified 12 unique winter weather systems that influence the Puget Sound area.

The Science                                 

Weather system clustering provides a high-level summary of regional weather conditions and connects them to both large-scale climate variabilities and regional hydrological processes. Using regional meteorological data from an atmosphere reanalysis product, scientists identified 12 unique winter weather systems in the Puget Sound area, featuring differing precipitation and temperature responses to large-scale climate variabilities. By connecting the large-scale drivers to regional hydrologic conditions, the weather system clustering highlights two types of flood-inducing weather in the regionone causes excess precipitation, and the other causes enhanced precipitation and intense snowmelt under warm temperature.

The Impact

Precipitation and temperature are important drivers of the land hydrologic process in the snow-dominated Puget Sound region. Although they are known to be modulated by some well-known modes of large-scale climate variabilitysuch as the El Nino-Southern Oscillation (ENSO) and the Madden-Julian Oscillation (MJO)existing weather system classification schemes fail to reflect such modulation on weather timescales. This study introduces a novel weather system clustering scheme that, for the first time, successfully links regional precipitation and temperature to ENSO/MJO and allows a process-level understanding of the effects of these modulations on regional hydrology. This study also demonstrates that new insights in regional hydroclimate can be achieved with deep learning.


Researchers analyzed daily atmospheric conditions surrounding the Puget Sound area from a reanalysis product during the cold season (October–March) of 19812020 using a clustering model with a deep learning encoder. Tuned for optimal concurrent precipitation and temperature predictive skills, the clustering model identified 12 unique regional weather systems, each featuring differing atmospheric temperature, relative humidity, and wind patterns. Some of these systems resemble the well-recognized meteorological systems in this region, such as atmospheric rivers and the Aleutian low-pressure system.

These weather systems can reflect the regional modulations of ENSO and MJO: weather systems associated with high precipitation show similar responses to ENSO/MJO phases as the patterns of precipitation itself. This suggests that these weather systems can be used to understand the physical mechanism of ENSO/MJO modulation effects on precipitation in this region. Meanwhile, they also reflect the unique precipitation and surface temperature combination in the Puget Sound region, highlighting two types of flood-inducing weatherone causes intense precipitation, and the other causes high precipitation and warm temperatures, which often results in intense snowmelt contribution to flood. The weather systems identified in this study can be used to evaluate the global/regional climate model performance and provide a foundation for improved high-level summary and analysis of regional weather conditions.


Ning Sun, Pacific Northwest National Laboratory, ning.sun@pnnl.gov


This research is supported by the Department of Energy, Office of Science, Biological and Environmental Research program as part of the Earth System Model Development, Regional and Global Modeling and Analysis, and MultiSector Dynamics program areas.

Published: February 22, 2024

Chen, X., Leung, L. R. & N. Sun. “Weather Systems Connecting Modes of Climate Variability to Regional Hydroclimate Extremes”. Geophysical Research Letters,50, e2023GL105530 (2023). [DOI: 10.1029/2023GL105530]