Learning Regional Monsoon Patterns with a Multimodal Attention U-Net
This work addresses the challenge of accurate monsoon forecasting for India's agriculture and climate planning, though it is incremental as it builds on existing deep learning approaches with new data and architectural tweaks.
The authors tackled monsoon rainfall prediction in India by developing a multimodal deep learning framework using a 1 km resolution dataset, achieving state-of-the-art results with consistent outperformance over existing methods, especially in extreme rainfall categories.
Accurate monsoon rainfall prediction is vital for India's agriculture, water management, and climate risk planning, yet remains challenging due to sparse ground observations and complex regional variability. We present a multimodal deep learning framework for high-resolution precipitation classification that leverages satellite and Earth observation data. Unlike previous rainfall prediction models based on coarse 5-50 km grids, we curate a new 1 km resolution dataset for five Indian states, integrating seven key geospatial modalities: land surface temperature, vegetation (NDVI), soil moisture, relative humidity, wind speed, elevation, and land use, covering the June-September 2024 monsoon season. Our approach uses an attention-guided U-Net architecture to capture spatial patterns and temporal dependencies across modalities, combined with focal and dice loss functions to handle rainfall class imbalance defined by the India Meteorological Department (IMD). Experiments demonstrate that our multimodal framework consistently outperforms unimodal baselines and existing deep learning methods, especially in extreme rainfall categories. This work contributes a scalable framework, benchmark dataset, and state-of-the-art results for regional monsoon forecasting, climate resilience, and geospatial AI applications in India.