Physics-Informed Fourier Neural Operators for Indian Summer Monsoon Rainfall Prediction: Mitigating Amplitude Smoothing Through Spectral Constraints
Proceedings of the AAAI Symposium Series · 06.11.2026 · 10 (1) · 538-540
DOI: 10.1609/aaaiss.v10i1.43451 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Symposium Series, online erschienen am 06.11.2026. Verfasst von Kiran Salunke.
Abstract (Originalsprache)
Predicting Indian Summer Monsoon Rainfall (ISMR) subseasonal variability is challenging due to non-linear atmospheric dynamics. Standard deep learning models optimized solely on Mean Squared Error (MSE) suffer from amplitude smoothing, over-damping spatial extremes toward the temporal mean.We propose a Physics-Informed Fourier Neural Operator (PINO) that integrates historical Sea Surface Temperature (SST) sequences and global climate teleconnections via FiLM conditioning. By enforcing an annealed spectral Laplacian (∇2) loss constraint in the Fourier domain, PINO preserves high-frequency convective gradients. Evaluated across a 560-day test period (2016–2020 JJAS), PINO outperforms IITM CFSv2 climate model, achieving a 15.2% reduction in RMSE and improving Anomaly Correlation Coefficient (ACC) from −0.0306 to 0.2182.
Zitieren
APA
Salunke, K. (2026). Physics-Informed Fourier Neural Operators for Indian Summer Monsoon Rainfall Prediction: Mitigating Amplitude Smoothing Through Spectral Constraints. Proceedings of the AAAI Symposium Series, 10(1), 538-540. https://doi.org/10.1609/aaaiss.v10i1.43451
Vancouver
Salunke K. Physics-Informed Fourier Neural Operators for Indian Summer Monsoon Rainfall Prediction: Mitigating Amplitude Smoothing Through Spectral Constraints. Proceedings of the AAAI Symposium Series. 2026;10(1):538-540. doi:10.1609/aaaiss.v10i1.43451
BibTeX
@article{Salunke2026,
author = {Salunke, Kiran},
title = {Physics-Informed Fourier Neural Operators for Indian Summer Monsoon Rainfall Prediction: Mitigating Amplitude Smoothing Through Spectral Constraints},
journal = {Proceedings of the AAAI Symposium Series},
year = {2026},
volume = {10},
number = {1},
pages = {538--540},
doi = {10.1609/aaaiss.v10i1.43451}
}
EUVIMED