Sub-seasonal Modulation and Predictability of Indian monsoon hourly Rainfall Extremes

Authors: Bijit Kumar Banerjee, Devabrat Sharma, Mahen Konwar, Simanta Das, Utpal Sarma, B. N. Goswami Date: 2026-03-25 Paper ID: arxiv:2603.24374

Summary

This paper investigates the physical predictability of highly destructive hourly rainfall extremes over the Indian subcontinent, which current numerical weather prediction models struggle to forecast. The authors establish a physical basis for medium-range predictability by identifying a trivariate clustering involving surface temperature, precipitable water vapor, and the phases of the Monsoon Intraseasonal Oscillation (MISO). They show that extremes organize into coherent storm systems embedded within larger mesoscale and synoptic features during active MISO phases. Based on this understanding, they developed an algorithm to track these systems, suggesting that physics-informed deep learning models could forecast these organized clusters with lead times exceeding one week.

Key Contributions

• Identified a trivariate clustering relationship between hourly rainfall extremes, surface temperature, and Monsoon Intraseasonal Oscillation (MISO) phases, providing a physical basis for medium-range predictability.
• Established that rainfall extremes organize into storm systems embedded within mesoscale convective clusters and synoptic low-pressure systems during active MISO phases.
• Developed an algorithm capable of identifying, tracking, and monitoring these organized storm systems linked to extreme rainfall.
• Provided a foundation for physics-informed training of data-driven models to forecast organized clusters of hourly rainfall extremes over a week in advance.

Limitations

Rapid error growth still limits the precise prediction of isolated hourly rainfall extremes, although the forecast for organized clusters shows promise.

Key Concepts

• Monsoon Intraseasonal Oscillation: A significant, recurring fluctuation in the strength and activity of the South Asian monsoon that occurs on sub-seasonal timescales (typically 30-60 days).

Limitations

Rapid error growth still limits the precise prediction of isolated hourly rainfall extremes, although the forecast for organized clusters shows promise.

Links

• ArXiv Abstract
• PDF