Source code and test data of the multi-model wind stress Drag Coefficient-Inversion Neural Network (Cd-INN)

This is the source code of the multi-model wind stress Drag Coefficient-Inversion Neural Network (Cd-INN), along with the related test data. The wind stress drag coefficient (Cd) is crucial for predicting storm surges, but current Cd parameterization schemes are uncertain due to limited observational data, leading to biases. To address this, we have developed the multi-model wind stress Drag Coefficient-Inversion Neural Network (Cd-INN) to quickly and continuously integrate observational data, avoiding the high computational costs of traditional data assimilation methods. We utilized 1,127 numerical simulations to train the Cd-INN, mapping storm surges to varying Cd values in the South China Sea. Once trained, generating an "optimal" Cd value for a new storm surge event requires only a time series of obseverd surge (sequence length ≥ 6) and standard standard meteorological inputs (U10/V10 and MSL).