Data from: Are AI weather models learning atmospheric physics? A sensitivity analysis of cyclone Xynthia

This collection contains the data for the manuscript titled "Are AI Weather Models Learning Atmospheric Physics? A Sensitivity Analysis of Cyclone Xynthia", which is necessary to reproduce the results presented. The data is provided in NetCDF format and organized into three directories: gradients, perturbations, and predictions. The gradients directory contains the sensitivity fields of kinetic energy over the Bay of Biscay for Cyclone Xynthia, relative to the input variables at the initial time. The perturbations directory includes sensitivity-based initial condition perturbations at a 36-hour forecast lead time for Cyclone Xynthia. Finally, the predictions directory holds the control and perturbed forecasts for Cyclone Xynthia.