Â鶹ÊÓƵ

Skip to main content

Characterizing Extreme Monsoonal Rainfall Trends with Self Organizing Maps

This study investigates spatial and temporal patterns of Indian summer monsoon rainfall (June–September) and their relationship to low pressure system (LPS) tracks, with a focus on identifying and characterizing extreme rainfall events. Daily gridded rainfall data from the India Meteorological Department and LPS track data were combined to explore how distinct configurations of rainfall are organized across India when LPSs make landfall and traverse the subcontinent. By aggregating rainfall at each grid point over the duration of individual LPS events and applying self-organizing maps (SOMs), we identified classes of rainfall distributions, including those representing the most intense precipitation. Composites of additional variables such as atmospheric pressure and temperature were employed to examine potential correlates and drivers of these heavy rainfall regimes.

The SOM approach successfully grouped rainfall patterns into coherent clusters, illustrating clear spatial distinctions and suggesting possible mechanisms—such as moisture pathways and large-scale circulation features— which contributed to the heaviest monsoonal rainfall. In further analysis, these rainfall classes were assessed for their alignment with synoptic and large-scale teleconnection patterns, providing insights into how the monsoonal system modulates extremes via LPS tracks. Results indicate that certain meteorological variables, including geopotential height, exhibit consistent anomalies during high-intensity rainfall events, pointing to broader-scale circulation controls.

This work demonstrates the efficacy of SOMs in classifying and analyzing sub-seasonal rainfall extremes over India, revealing both local and large-scale climatic processes that contribute to extreme precipitation. The methods and findings presented here could be extended to other gridded datasets and time scales, enhancing our ability to interpret and predict extreme processes.