Visit ParkNet NPS Geographic Information Systems National Park Service, U.S. Department of the Interior
[Go Back]

Space-Based Burn Severity Mapping in the National Parks

Fire Management Program Center
Brian Sorbel
In any given year, numerous national parks experience significant fire activity. From the sawgrass prairies in Everglades National Park to the expansive black spruce forest tracts of Denali National Park and Preserve in Alaska, fires burn in and affect a variety of vegetation and community types. Even within a single park, fires typically burn across a highly heterogeneous landscape typified by both variable topography and fuels. Similarly, the effects of fire across the landscape are also highly heterogeneous and varied. Within any given fire, some areas are radically changed due to intense scorching and other areas are completely untouched. An infinite variety of potential fire effects occur within these two extremes. An example of this ‘fire mosaic’ can be seen in Figure 1. This degree of environmental change caused by fire, or burn severity, is of interest to NPS fire and resource managers. Burn severity maps capture the heterogeneous nature of fire effects and offer a more complete description and quantification of fire’s effect on the landscape. Burn severity mapping is the effort to map the fire mosaic and the resultant variety of fire effects. Since 2000, the National Park Service Fire Management Program, in conjunction with the USGS EROS Data Center (EDC) and the USGS Northern Rocky Mountain Science Center (NRMSC) has mapped burn severity on all large NPS fires using Landsat satellite imagery and the Normalized Burn Ratio.

The National Park Service Fire Management program is interested in mapping burn severity on large wildland fires for a variety of reasons. Burn severity maps identify unburned areas within the fire perimeter as well as provide a measure of the likely effect of the fire on the vegetation and fuels condition of burned areas. Burn severity maps can be used to estimate post-fire erosion potential, predict the susceptibility of a burned area to invasion by non-native species and assess the post-fire vegetation and fuels condition. Several methods present themselves as potential means to assess burn severity across a large wildland fire. These include hand mapping, photo-interpretation of aerial reconnaissance and the use of satellite imagery. Since 2000, the NPS, in conjunction with the USGS NRMSC and EDC, has investigated burn severity mapping with Landsat satellite imagery. Landsat imagery is cost-effective and timely. Its area of coverage is adequate for capturing many large wildland fires within a single scene (Figure 2). The 30-meter resolution of Landsat imagery and derived burn severity products compares favorably with that of NPS vegetation and fuels mapping products. Landsat-derived burn severity products are also immediately GIS-ready and can be used for GIS analysis as well as cartographic output. Finally, Landsat imagery is standardized worldwide, offering comparable metrics which can be used to map, assess and compare burn-severity on a system wide level across all NPS units as well as on an interagency basis across federal, state and private lands.

The EROS Data Center generates burn severity products for the NPS by applying the Normalized Burn Ratio (NBR) to pre and postfire Landsat imagery. The NBR uses pre and postfire Landsat imagery to develop a continuous index of burn severity. The NBR is derived in a manner similar to the Normalized Difference Vegetation Index (NDVI). Whereas NDVI is calculated by generating an index of Landsat bands 3 and 4, the NBR is calculated using an index of Landsat bands 4 and 7, the two bandwidths that show greatest response to burning. The formula to derive the NBR uses at-satellite reflectance of the two bandwidths, as follows:

NBR = (TM Band 4 – TM Band 7) / (TM Band 4 + TM Band 7)

The Normalized Burn Ratio is calculated for both pre and postfire Landsat scenes. Derivation of the NBR yields a floating point dataset with values ranging from –1.0 to +1.0. A final differenced NBR (∆NBR) dataset is derived as follows:

∆NBR = NBRprefire – NBRpostfire

The ∆NBR is the final burn severity product. Derivation of the ∆NBR yields a floating point dataset with values ranging from –2.0 to +2.0. Delta-NBR datasets are scaled by 1000 to yield a final burn severity dataset with possible values ranging between –2000 and +2000. Generally, a threshold exists between about -100 and +150 ∆NBR units that marks an approximate breakpoint between burned and unburned areas. Areas with ∆NBR values below this threshold are unburned; areas with ∆NBR values above this threshold are burned. Furthermore, within the burned area, increasing ∆NBR values correspond to increased burn severity on the ground. A grayscale ∆NBR burn severity product for the 2000 Bircher and Pony fires in Mesa Verde National Park is shown in Figure 2.

The ∆NBR burn severity mapping methodology was initially developed and tested by USGS NRMSC and NPS researchers on fires occurring in Glacier National Park. In comparisons between the NBR approach and other methods, the NBR approach scored favorably. The NBR approach was also initially applied with generally favorable results in Yellowstone National Park, Yosemite National Park, Yukon-Charley Rivers National Preserve and Bandelier National Monument. In 2001, the NPS Fire Management Program, sufficiently intrigued by results from the initial test cases, entered into a partnership with the USGS EROS Data Center to systematically produce ∆NBR burn severity datasets for all large fires (typically greater than 300 acres) occurring on NPS lands. This partnership has produced burn severity datasets for more than 100 fires occurring in nearly 40 NPS units. In this partnership, NPS Fire Management notifies the EDC of burns that it would like mapped. The EDC locates, acquires and processes appropriate pre and postfire Landsat imagery to generate the ∆NBR burn severity products. Using this data, the EDC also generates an initial satellite-derived final fire perimeter. Burn severity products including the ∆NBR dataset, satellite-derived final fire perimeter, pre and postfire Landsat imagery and associated metadata are sent to the park on CD-media. In addition, all datasets are posted on the USGS – NPS National Burn Severity Mapping Project website ( for download by NPS personnel, researchers and the public. Upon retrieval of burn severity data, NPS Fire Management assesses the accuracy of the burn severity products and the derived perimeter and notifies the EDC of any changes. Figure 3 displays a sample burn severity product from Yukon-Charley Rivers National Preserve.

The NPS Fire Management Program uses Composite Burn Index plots to validate the accuracy and quality of ∆NBR burn severity products. Composite Burn Index (CBI) plot methods were developed jointly by USGS NRMSC and NPS researchers to provide some capability to ground-truth remotely sensed burn severity products. Ocular estimates related to the degree of environmental change caused by fire in various forest and non-forest strata are made and marked on a field data sheet. As an example, crews make ocular estimates of the change caused by fire to fuels greater than eight inches in diameter with possible options being: unchanged; 5% loss, blackened; 15% loss with deep char; and >40% loss with deep char. A score between 0.0 and 3.0 is recorded for each component with 0 meaning that the component is unchanged and 3 meaning that the component has either been completely consumed by fire or has been radically changed by fire. In all, up to 22 severity scores are recorded in multiple strata for a variety of measures including change in litter, amount of new serals, % of tall shrubs consumed, etc. These scores are then averaged to yield CBI ratings for the understory, overstory, and the total plot. The overall CBI plot score is cross-referenced with the satellite measure of severity at that location to determine the degree of correlation.

Comparisons between satellite measures of burn severity (∆NBR values) and ground measures (CBI plot scores) are generally favorable. A comparison of CBI and corresponding ∆NBR values over 88 plots in Glacier National Park yielded an R2 value of 0.84. A similar comparison in Yukon-Charley Rivers National Preserve produced an R2 value of 0.75 (Figure 4). Comparable results have been obtained in many of the other 14 parks where CBI plots have been used to determine the accuracy of ∆NBR burn severity products. NPS and USGS researchers, with funding from the Joint Fire Sciences Program, will use this data in a comprehensive assessment of the relationship between satellite and ground measures of severity on a national scale across several community types.

The USGS – NPS Burn Severity Mapping Project continues to develop. Figure 5 shows the parks and other areas where burn severity products have been created. Work to this date has focused on method selection, refinement and testing as well as on data production and delivery mechanisms. Future efforts will focus on both the interpretation of burn severity data as well as its incorporation into fire management and monitoring applications. It is expected that burn severity datasets will become an NPS-wide corporate data layer available to NPS fire managers, resource managers and researchers. Just as a park can display all known fire ignition locations and areas burned using point and polygon fire history themes, comprehensive park-wide burn severity datasets will complete the landscape-level description of fire effects.
April 08, 2004