Modeling fire spread is critical in fire risk management. Creating data-driven models to forecast spread remains challenging due to the lack of comprehensive data sources that relate fires with relevant covariates. We present the first comprehensive and open-source dataset that relates historical fire data with relevant covariates such as weather, vegetation, and topography. Our dataset, named WildfireDB, contains over 17 million data points that capture how fires spread in the continental USA in the last decade. The paper accompanying this dataset is part of the 2021 Neural Information Processing Systems (NeurIPS) Dataset and Benchmark Track. The paper describes the algorithmic approach used to create and integrate the data, describe the dataset, and present benchmark results regarding data-driven models that can be learned to forecast the spread of wildfires. Please see https://colab.research.google.com/drive/1cm2Z4E0HzXMAcuUrE26wHXL2FS_pIj3t?usp=sharing for an introduction about how to load the database using python (pandas).

Author SS was funded by Agriculture and Food Research Initiative Competitive Grants no. 2019-67022-29696 and 2020-69012-31914 from the USDA National Institute of Food and Agriculture, AM was funded by the Center of Automotive Research at Stanford (CARS) and National Science Foundation Award Number IIS181495, MW was funded by National Science Foundation Award Number IIS181495, and TD was funded by the Department of Management Science and Engineering at Stanford University.