IndiaWeatherBench: A Dataset and Benchmark for Data-Driven Regional Weather Forecasting over India

ICLR 2026 Conference Submission14427 Authors

Keywords: weather forecasting, regional weather forecasting, india, limited area modeling, climate change

TL;DR: We propose a novel dataset and benchmark for weather forecasting over the Indian region

Abstract: Regional weather forecasting is a critical problem for localized climate adaptation, disaster mitigation, and sustainable development. While machine learning has shown impressive progress in global weather forecasting, regional forecasting remains comparatively underexplored. Existing efforts often use different datasets and experimental setups, limiting fair comparison and reproducibility. We introduce IndiaWeatherBench, a comprehensive benchmark for data-driven regional weather forecasting focused on the Indian subcontinent. IndiaWeatherBench provides a curated dataset built from high-resolution regional reanalysis products, along with a suite of deterministic and probabilistic metrics to facilitate consistent training and evaluation. We establish strong baselines by adapting state-of-the-art global models, including FourCastNet, Pangu-Weather, GraphCast, and Stormer, to the regional domain. To enable this adaptation, we propose two simple yet effective boundary conditioning strategies: boundary forcing and coarse-resolution conditioning. We conducted a thorough empirical evaluation of these baselines under different settings and metrics, complemented by a case study on predicting extreme heatwaves in India. While focused on India, we designed IndiaWeatherBench to be easily extensible to other geographic regions. We will open-source all raw and preprocessed datasets, model implementations, and evaluation pipelines to promote accessibility and future development in regional weather forecasting research.

Primary Area: datasets and benchmarks

Submission Number: 14427