Find the data story here: https://www.vpdatacommons.org/stories-3/wildfire-ignition-probability Short Description: Mapping wildfire ignition probabilities provides useful information for fire and land management. We developed models of ignition probability for wildfires with growth potential (larger than 100 ha in size for the western US and 8 ha in size for the southeastern US) by ignition cause with a machine learning algorithm customized to produce unbiased probabilistic predictions. Input variables included spatial trends of observed fire ignitions and other topographic, climatic, vegetative, and human development features. For each region, we produced two spatial datasets of ignition probability at a 120 m pixel resolution with units of ignition probability per square kilometer per year: human and natural causes that are scaled to the observed annual rate of wildfire ignition from 2006-2020, which equaled 303 (human) and 353 (natural) for the western US, and 4337 (human) and 511 (natural) for the southeastern US at the respective firesize thresholds. These two layers can be summed to produce an all-cause annual ignition probability. These datasets are available in the Vibrant Planet Data Commons (www.vpdatacommons.org), whose mission is to provide the best available data for increasing the pace and scale of forest restoration.