Estimating Wildfire Risk from Space

EarthzineDEVELOP Spring 2017 Article Session, DEVELOP Virtual Poster Session

This article is a part of the NASA DEVELOP’s Spring 2017 Article Session. For more articles like these, click here

Scientists collaborated with fire managers in the Great Plains to produce an interactive wildfire risk map derived from NOAA and NASA data.

Authors List:

Kathy Dooley

Alec Courtright

Kimberly Berry

Ben House III

Mike Kruk (science adviser)

A prairie wildfire in South Dakota. Image Credit: Darren Clabo

Grasslands in the Missouri River Basin are the most prominent land cover type and are essential for livestock grazing and agriculture. However, they are susceptible to wildland fires, with approximately 6.6 million acres (1) burned annually in this region. These fires spread rapidly throughout the Great Plains, making them difficult for local wildland fire management groups to contain. Wildland firefighters are most successful in containing a large fire if they are near the fire when it starts. But rural areas don’t have fire stations every few miles like your average U.S. city does. Instead, fire managers continuously evaluate fire potential and may position emergency response equipment and fire teams to regions that are most at risk to wildfire. In addition to fighting fires, fire managers also ignite fires to conduct prescribed burns. Wildfires play an integral role in ecosystem health and biodiversity, so carefully placed controlled burns help revitalize the landscape (2). However, when conducting a controlled burn, it is essential to ensure that the landscape isn’t too dry or windy, as these burns have the potential to spread rapidly and unpredictably.

As part of their decision-making process, fire managers in the Great Plains rely on weather station and drought condition data to assess daily fire risk. However, weather stations in the region are sparse and manually combining weather information takes considerable time and effort. Additionally, they do not currently have a reliable method for estimating vegetation moisture. åÊDuring periods of extreme wildfire risk, the Bureau of Indian Affairs (BIA) charters a plane to assess vegetation conditions. However, these flights are costly and provide only a cursory glimpse into fuel conditions. Fire managers currently use fire danger warnings, such as South Dakota Grassland Fire Danger Maps and the Red Flag Warning System to assess statewide fire danger. However, fire managers would like to have finer resolution maps that include precipitation and drought conditions. The BIA Great Plains Region and South Dakota State Fire Meteorologist Darren Clabo asked NASA DEVELOP if they could estimate vegetation conditions from satellite data to provide fire managers with reliable, åÊsub-county level spatial information on the driest, most flammable areas. In order to address these issues, NOAA National Centers for Environmental Information (NCEI) and NASA DEVELOP collaborated with the U.S. Army Corps of Engineers, the BIA, and Clabo to produce a user-friendly tool to assess fire risk in South Dakota and across the Missouri River Basin.

David Martin, Assistant Regional Fire Manager at the BIA Great Plains Region, provided the DEVELOP team with a data on fire indicators he uses to assess wildfire risk based on conditions that resulted in large fires over the past decade.

The DEVELOP team used Google Earth Engine as a platform for their interactive wildfire risk map in order to ensure quick processing speeds and the ability to easily share the output. The project partners in South Dakota will be able to perform the necessary analyses without downloading and processing each individual variable.

Table 1. Original fire potential indicators provided by regional experts for the spring fire season.

The resulting map features NOAA and NASA data. NOAA’s Real Time Mesoscale Analysis (RTMA) provided wind speed, relative humidity, and temperature data; the Palmer Drought Severity Index (PDSI) yielded drought conditions; the Normalized Difference Infrared Index (NDII) and Normalized Difference Snow Index (NDSI) derived from the Moderate Resolution Imaging Spectrometer (MODIS) sensor onboard NASA’s Aqua and Terra satellites approximated fuel moisture and snow cover, respectively, and Global Precipitation Measurement (GPM) mission estimated precipitation. These data products were chosen based on their spatial and temporal resolution (Table 2) and their accessibility through Google Earth Engine.

Table 2. Datasets used to derive the wildfire potential indicators defined by regional fire experts in the Great Plains.

The fire indicators are aggregated in a single tool, åÊFire Risk Estimation (FIRE). The tool facilitates four major tasks:

  1. Access the datasets in Table 2.
  2. Determine whether they meet or exceed the defined thresholds.
  3. Weigh the contributing indicators.
  4. Sum the weighted contribution from each indicator to determine an overall fire potential.

Relative humidity was derived from RTMA data using the August-Rush-Roche Approximation (Equation 1). Fuel moisture conditions were determined using the Normalized Difference Infrared Index (NDII) and derived from the MODIS 8-Day Level 3 Surface Reflectance data. Based on an existing publication by Yebra, et al. (2013), NDII was calculated using both near infrared (Band 2) and shortwave infrared (Band 5) imagery due to strong absorption features in the near infrared and shortwave infrared wavelengths (3).

Snow cover was estimated using the Normalized Difference Snow Index (NDSI) Modis product available in Google Earth Engine. NDSI is the only fire potential indicator that inversely contributes to the risk matrix, eliminating the estimated potential for fire if snow is present. All remaining data were processed in Google Earth Engine to produce a single, easily accessible map of wildfire risk throughout the Missouri River Basin. The map also includes optional layers that allow users to toggle between the various factors contributing to fire potential.

Wildland fire teams respond to a prairie fire in the Great Plains. Image Credit: Darren Clabo

Some indicator thresholds, such as temperature, change greatly from one season to the next. To determine the impact of temperature on fire potential outside of the spring fire season, temperature anomalies were computed using NOAA nClimGrid 30-year monthly average temperature and calculating daily averages through linear approximation. The current temperature was then subtracted from the corresponding daily average to get the seasonally-relative daily temperature anomaly. All other seasonally-dependent weights (Table 3) were determined through conversations with Clabo and the BIA.

Table 3. Seasonal weights for fire seasons within the Great Plains Region. Fire seasons are based on South Dakota grasslands and were determined by the project team and partners.

FIRE is updated on a daily basis, and fire managers will use it to determine where to position their emergency response equipment. This tool provides a unique analysis to regional fire managers that incorporates more fire weather variables than products currently used in the region and outputs a higher spatial resolution. In fall 2017, this map and its corresponding code will become freely open to the public. Currently, the NASA DEVELOP team is working with end-users to produce daily fire potential maps throughout the spring and summer fire seasons. Clabo will compare these daily results with other fire potential analyses to validate and test the matrix. This tool can also be expanded to other regions in the United States, as the relative importance of variables can be modified to fit any regional climate and landscape. Understanding when and where fires are most likely to occur will allow fire response teams to be better prepared when the next fire ignites.

Example of daily output map of fire (March 13, 2017) potential in the Missouri River Basin. Image Credit: NASA DEVELOP


[1]E. Crouch, “Wildfires – Annual 2012 | State of the Climate | National Centers for Environmental Information (NCEI)”,, 2017. [Online]. Available: [Accessed: 19- Feb- 2017].

[2]V. Salomonson, “Estimating fractional snow cover from MODIS using the normalized difference snow index”,Remote Sensing of Environment, vol. 89, pp. 351-360, 2004.

[3]M. Yebra, P. Dennison, E. Chuvieco, D. Ria̱o, P. Zylstra, E. Hunt, F. Danson, Y. Qi and S. Jurdao, “A global review of remote sensing of live fuel moisture content for fire danger assessment: Moving towards operational products”, Remote Sensing of Environment, vol. 136, pp. 455-468, 2013.


Alec Courtright graduated from the University of South Carolina-Columbia in 2015 and is the center lead at NOAA NCEI.

Kimberly Berry graduated from the South Dakota School of Mines in 2015 and is working with DEVELOP at NOAA NCEI as an independent research consultant on the Missouri River Climate II project.

Ben House III is a student at University of North Carolina at Asheville and is working with DEVELOP at NOAA NCEI as an independent research consultant on the Missouri River Climate II project.

Kathy Dooley graduated from Carleton College in 2015 and is working with DEVELOP at NOAA NCEI as an independent research consultant on the Missouri River Climate II project.