Cloud-Native AI Ecosystem for Season-Aware Cyclone Forecasting: Track Prediction, SST-Integrated Rainfall Modeling, and Risk-Based Early Warning over the North Indian Ocean

Real-time operational forecasting of cyclones in the North Indian Ocean (NIO) requires an infrastructure that can handle heterogeneous, high-volume input data streams of meteorological parameters and multiple forecasting models in a parallel manner, which cannot be achieved by season-blind, single-model setups. Here, we report a cloud-native AI ecosystem that orchestrates elastic cloud resources, multimodal satellite data ingestion, and meteorological stream data for cyclone track prediction, rainfall forecasting, and risk assessment within a unified, scalable framework. We ingest and spatially align 1,500 thermal infrared satellite images (INSAT and METEOSAT) in near real-time to International Best Track Archive for Climate Stewardship (IBTrACS) coordinates by orchestrating elastic cloud services and, for cyclone track forecasting, apply a monsoon-sensitive strata, which categorizes input sequences two groups (SW / NE monsoon), reflecting different atmospheric steering dynamics. Within this structure, a season-aware multi-modal Convolutional Neural Network fuses visual and seasonal context streams to forecast cyclones tracks and recurvature (prediction accuracy 94.8%, precision 0.92, recall 0.91, F1-score 0.91 for 5-fold cross-validation with accuracy > 90% for both monsoons). To demonstrate regional generalization of our ecosystem, we input a local dataset covering 10 sites in the Chennai metropolitan area through the same cloud pipeline and couple with Sea Surface Temperature (SST) data to predict rainfall associated with cyclones. We also deploy a parallel cloud-native physics-guided LSTM model that uses domain knowledge such as minimum SST and low-pressure criteria to estimate the evolving cyclone risk profile, and the output is presented on a dashboard with geospatial risk information and time-to-impact indices for prompt decision-making.