Reza Nosratpour: Analysis of Extreme Precipitation Pathways using Climate Networks and Tail Dependence Coefficient
Heavy rainfall events have increased in both frequency and intensity in recent decades, producing major social and economic impacts worldwide. Extreme rainfall can be represented as an interconnected spatiotemporal process in which atmospheric systems propagate dependence structures across regions and time. This study proposes a novel extremal climate network framework for identifying propagation pathways of heavy rainfall events across multiple temporal lags. Unlike conventional climate network approaches that primarily focus on pairwise connectivity and diagnostic analysis, the proposed framework captures propagation structures and enables their integration into machine learning forecasting applications.
The framework estimates multivariate tail dependence coefficients to construct a weighted climate network, where nodes represent spatial locations and edge weights quantify extremal rainfall dependence. K-means++ clustering is then applied to identify homogeneous regions sharing similar sources and propagation characteristics of extreme precipitation. Multi-step propagation pathways are subsequently detected by progressively increasing the dimensionality of the dependence structure, enabling the identification of spatio-temporal relationships and directional rainfall propagation patterns.
The framework was implemented using IMERG precipitation estimates over Australia, with independent rain gauge observations used for validation. Results revealed strong extremal connectivity and enhanced predictability across northern and eastern Australia, associated with monsoonal systems, tropical cyclones, and El Niño–Southern Oscillation variability. The proposed framework achieved strong predictive skill, demonstrating its potential for extreme rainfall forecasting and climate adaptation planning.
This work relates directly to the ACCESS community through its integration of machine learning, climate diagnostics, and Earth system analysis for improving the understanding and prediction of extreme weather processes over Australia.