Datasets for Article: Generative Algorithms for Wildfire Progression Reconstruction from Multi-Modal Satellite Active Fire Measurements and Terrain Height

Training and testing datasets used to obtain results presented in the article Generative Algorithms for Wildfire Progression Reconstruction from Multi-Modal Satellite Active Fire Measurements and Terrain Height (https://arxiv.org/abs/2506.10404). training_data and testing_data are derived from WRF-SFIRE wildfire progression simulations, as described in the associated manuscript. Training data is separated into 7 files, training_data_1.tar.gz - training_data_7.tar.gz, which contain 2500 training samples each and can be concatenated along the 0 dimension to obtain the final training array of size 17500x512x512x3. File real_fire_test_samples.tar.gz contains non-synthetic active fire measurement data (VIIRS+GOES) and terrain height data for the Bobcat, Tennant, Oak, Barnes, and WIlliams Flats wildfires, which are examined in the associated manuscript.