CycloneNet: A Deep Learning Framework for Atmospheric Singularity Mapping via Spatio-Temporal Attention

CycloneNet is a high-performance software engineering framework released under the CC BY-NC 4.0 license. Developed by a Specialist in Mission-Critical Systems, it represents a novel, hybrid approach for the forensic analysis of tropical cyclones. The core innovation is a physics-guided deep learning system designed not to predict, but to diagnostically pinpoint—with sub-pixel spatial accuracy—the precise geographic locations of thermodynamic energy sources that fueled historical storm intensification. By bridging the gap between atmospheric physics and scalable software architecture, the framework achieves a ROC-AUC of 0.97 and a Recall of 0.92. This record (v1.0.1) contains the following scientific artifacts: Technical Paper (PDF): A Deep Learning Framework for Atmospheric Singularity Mapping. Full methodology and architectural design. Scientific Dataset (CSV): cyclonenet_scientific.csv. Complete audit trail of 18 major hurricanes (1989-2024), including "Target Lock" coordinates. Validation Report (TXT): validation_report.txt. Summary of key performance metrics and storm-level MAE (Mean Absolute Error). Execution Log (TXT): pipeline.log. Full-trace audit logs (extensive) detailing the end-to-end pipeline execution, from Copernicus CDS API data retrieval to final tensor calculations. The framework serves as a forensic tool for atmospheric auditing, providing a transparent and fully verifiable "black box" reconstruction of past extreme weather events. Contact: estefano.senhor@gmail.com GitHub: https://github.com/estefano-ferreira/cyclone-net