A geospatial dataset containing 300 m embedded environmental feature tiles for the Korean Peninsula, designed for wildfire spread prediction and reinforcement-learning research.
This dataset provides model-ready 16-channel tensors representing static and quasi-static environmental conditions needed for wildfire simulation and RL inference.
It is paired with the wildfire episode dataset (wildfire-korea-episodes-300m) and… See the full description on the dataset page:
https://huggingface.co/datasets/chaseungjoon/wildfire-korea-embedded-300m.
