Network Divergence Reveals Predictable Pathways of Extreme Rainfall in Central India

Abstract Extreme rainfall events over Central India have increased, reflecting intensified monsoon precipitation driven by factors like urbanization and land use change. With projections indicating further intensification and flood vulnerability, better prediction is critical. We present a physically grounded, operationally actionable framework using network divergence on directed networks from nonlinear synchronization of extreme events. Applied to high‐resolution rainfall data, the method predicts over 60% of events exceeding the 95th percentile in Central India, based on two conditions: preceding extreme rainfall over the east coast along with a low‐pressure anomaly and moisture convergence over Central India. West‐northwestward propagating low‐pressure systems from the Bay of Bengal and vertically integrated moisture convergence are identified as primary drivers. Unlike numerical weather prediction models, our approach is observationally driven, mechanistically interpretable, and computationally efficient, offering a promising route for early warning systems and demonstrating the potential of network‐based methods for predicting climate extremes beyond Central India.