Vibrant Planet's Pyrologix division provides the information hosted on the Vibrant Planet Data Commons site (www.vpdatacommons.org) (the “Site”) as a PUBLIC SERVICE for NONCOMMERCIAL USES only. The information is provided as-is, as-available, and without warranty of any kind, whether express, implied, or statutory, concerning the accuracy, completeness, reliability, or suitability of the information. Further, we do not warrant that this site will (i) be uninterrupted or error-free; or (ii) result in any desired outcome. We are not responsible for content of other services, data or public information that may be used by or linked to this site. By accessing this website and/or the data contained within, you assume the entire responsibility and liability related to the use of the information it contains, and you release Vibrant Planet Data Commons, Pyrologix and all data providers from liability. You may use the Site only for lawful purposes Users also should note that property boundaries included on any map do not represent an on-the-ground survey suitable for legal, engineering, or surveying purposes. They represent only the approximate relative locations. We reserve the right to withdraw or amend this Site, and any service or material we provide on the Site, in our sole discretion without notice. We will not be liable if for any reason all or any part of the Site is unavailable at any time or for any period. From time to time, we may restrict access to some parts of the Site, or the entire Site to users or visitors to the site. The Site and its entire contents, features, and functionality (including but not limited to all information, software, text, displays, images, video, and audio, and the design, selection, and arrangement thereof) are owned by Vibrant Planet Data Commons, its licensors, or other providers of such material and are protected by United States and international copyright, trademark, patent, trade secret, and other intellectual property or proprietary rights laws. This site is operated by Vibrant Planet Data Commons.
Wildfire ignition probability data provides spatially explicit estimates of the likelihood that a wildfire will start in a given location. The resulting datasets, specific to the Western and Southeastern U.S. regions, offer geospatial estimates of wildfire ignition probabilities, distinguishing between human-caused and natural (lightning) ignitions, as well as providing combined probabilities for both. The authors employ Random Forest machine learning, customized for probabilistic predictions, to model ignition likelihood based on spatial trends in observed fire occurrences, topographic features, climatic factors, vegetation characteristics, and human development patterns. The resulting datasets are scaled to recent observed ignition rates (e.g. 2006-2020 fire occurrence database) and have a spatial resolution of 120 meters. These datasets are a valuable resource for wildfire risk assessments (QWRA), risk mitigation planning, and decision support in land management, policy development, and other fire-related contexts.
Wildfire ignition probability data helps us better understand the complex dynamics of the likelihood that a wildfire will start in a specific location by cause, and provides a tool for managers to make informed decisions about how to manage these ecosystems to meet their goals. For example:
Ignition probability data is a fundamental input for QWRAs, which inform land management decisions, resource allocation, and risk mitigation strategies.
Identifying high-probability ignition areas enables targeted prevention efforts (e.g., public outreach in human-caused hotspots) and the strategic allocation of mitigation resources.
These datasets can guide policy development, zoning regulations, and building code requirements to enhance community resilience to wildfire.
This study offers several improvements over traditional wildfire ignition data:
The study covers two major fire-prone regions of the United States: the Western U.S. and the Southeastern U.S..
This study employed a multi-step methodology to develop wildfire ignition probability datasets. First, ignition data from the Fire Occurrence Database (FOD) was filtered, applying region-specific size thresholds to focus the analysis on fires with significant growth potential. In the Western U.S., this threshold was 100 hectares, while in the Southeastern U.S., it was 8 hectares.
A variety of spatial datasets representing topographic, climatic, vegetative, and human development factors were obtained and resampled to a consistent 120-meter resolution for modeling ignition likelihood. A spatial kernel density filter was applied to the FOD data to derive a spatial trend feature, aiming to capture subtle influences on ignition patterns not explicitly represented by other input variables. The Synthetic Minority Oversampling Technique (SMOTE) was used to address the inherent imbalance between positive (ignition) and negative samples in the dataset, improving model robustness and generalizability.
Separate random forest machine learning models were trained for human-caused and natural ignitions within each study region. Region-specific differences in variable importance were accounted for (Table 2). For example, topography played a more significant role in the Western U.S. models. The random forest algorithm was customized to produce probabilistic outputs, enabling more nuanced risk assessments.
Finally, model predictions were scaled to match the observed ignition rates from the FOD (2006-2020). This scaling step transformed the predictions into real-world units of mean annual ignitions per square kilometer, ensuring relevance for practical applications. Non-burnable areas were excluded during the scaling process to enhance accuracy.
Whether the particular situation calls for wildfire to be promoted, prevented, or mitigated, understanding and quantifying wildfire risk provides useful information for its management.
For a comprehensive exploration of the research, including detailed methodology, data descriptions, and model development, please refer to the original article by the authors:
Wildfire ignition probability datasets by cause for the Western and Southeastern United States
The wildfire ignition probability analysis produced three primary spatial datasets for both the western and southeastern US. These datasets are available in the following formats:
All datasets have a 120-meter pixel resolution with probability values ranging from 0 (very low likelihood) to a theoretical maximum of 1 (very high likelihood). However, it's important to note that in practice, the likelihood of a wildfire starting in any given pixel is quite low. Even in high-risk areas, the highest observed probability is around 0.0004, meaning a probability of 1, while theoretically possible, is extremely unlikely. All datasets are calibrated to match observed annual wildfire ignition rates from 2006-2020.
As described above, the datasets encompass three types of ignition probabilities:
Best for analysis in GIS software focused on the continental U.S. where preserving area measurements is important.
Ideal for cloud-based analysis, fast web visualization, and large datasets.
Optimized for web mapping applications.
Best for analysis in GIS software focused on the continental U.S. where preserving area measurements is important.
Ideal for cloud-based analysis, fast web visualization, and large datasets.
Optimized for web mapping applications.
What is a STAC Catalog? Think of the STAC catalog like a digital library index for our wildfire ignition datasets. It provides organized information (metadata) about each dataset, including location, time periods, and other important details. This makes it easier for you to find and use the data relevant to your research or projects.
Exploring the Data:
STAC Browser: Below we provide an easy-to-use STAC Browser for basic data exploration. This lets you quickly visualize the datasets and filter them based on your needs.
Moran, C. J., Scott, J. H., & Vogler, K. (2023). Wildfire ignition probability datasets by cause for the western and southeastern United States.
This wildfire ignition probability data was made possible in part by funding from the Strategic Analytics Branch, Fire and Aviation Management, National Headquarters, United States Forest Service.