Deliverable D2.2 – CycloNet: a tool for tropical cyclone storm-surge prediction

This report is Deliverable D2.2 of the TRANSCLIMA project. It presents CycloNet, a tool developed to estimate storm surges generated by tropical cyclones. CycloNet is based on AI models trained on hydrodynamic simulations. It takes tropical cyclone track and intensity parameters and produces peak surge levels along the coastline. The report describes the methodology, the training and validation data, the performance of the tool, and example applications. CycloNet corresponds to Module 2 of the MAWE framework described in the project’s Annex 1 (originally “CYCLONe-predictor”) and was presented at EGU 2026 and ICCE 2026. The simulations used to develop the tool were run on MareNostrum 5 (BSC) through RES computing projects.