UNEP/GRID-Sioux Falls


R.W. Klaver         A. Singh
E.A. Fosnight       United Nations Environment Programme
RSTX**       EROS Data Center
EROS Data Center       Sioux Falls, SD, US
Sioux Falls, SD, US            



Uncontrolled wildfires have an immense impact on both human population and the environment, as witnessed in the 1997 Indonesian wildfires. Little practical assistance was possible beyond fire suppression and humanitarian aid to affected populations during and after the fire. However, the technology exists to map fire potential, detect fire starts, monitor fire movement, and assess the impact of the wildfires at both regional and global scales.

A "Global Forest Fire Watch System" that provides early warning, monitoring, and assessment of wildfires can be implemented using current earth-observing technologies and local expert knowledge. The creation of such a system relies on the existence of a high-quality vegetation map, near-real-time low-cost satellite images, close working relationships with local sources of up-to-date weather information and expertise, and a local requirement for fire-related services and products.


Fires, whether of human or natural origin, have profound effects on land cover, land use, production, local economies, global trace gas emissions, and health. Uncontrolled wildfires can have an immense impact on the human population and the environment, as was witnessed in the wildfires in Indonesia during 1997. A fire analysis cycle can be defined that moves from mapping the potential for a fire start if there is ignition, to detecting the start of a fire, through monitoring the progression of a fire, to mapping the extent of the fire scars and the progression of vegetation regeneration. Such information would be useful to managers, policy makers and scientists interested in mitigating and evaluating the effects of wildfires.

Requirements vary widely at each stage of the fire analysis cycle (Figure 1). Fire potential quantifies the likelihood of a fire when there is ignition. Fire potential requires collecting baseline vegetation information, daily to weekly monitoring of vegetation condition or vigor, daily monitoring of weather conditions, and acquiring risk management information. Fire detection requires daily monitoring of fire starts. Fire monitoring requires daily monitoring of fire scars and of smoke and haze during the burn. Fire assessment requires analysis of fire burns and periodic monitoring of the vegetation transitions in the fire burns. Each of the four stages of the cycle is discussed in more detail below.

Figure 1 Fire Analysis Cycle

Figure 1. Fire Analysis Cycle


Fire potential depends on the amount of dead and live vegetation, the moisture in the live vegetation, and the moisture in the dead vegetation. The amount of dead and live vegetation is estimated from a high quality land cover map derived from (ideally) a high-resolution sensor, such as the Landsat Thematic Mapper or SPOT multispectral scanner, otherwise from a lower resolution sensor, such as National Oceanic and Atmospheric Administration’s Advanced Very High Resolution Radiometer (AVHRR) or Moderate Resolution Imaging Spectrometer (MODIS). Given this baseline land cover map, low-spatial and high-temporal resolution satellites, such as AVHRR, can be used to monitor near real-time changes in the vegetation vigor, which is correlated with the moisture of the live vegetation. The moisture in the dead vegetation is estimated from knowledge of local weather conditions. Thus, a baseline land cover map and a real-time estimate of the vegetation condition are needed.

The U.S. Geological Survey’s (USGS) EROS Data Center, in cooperation with the U.S. Forest Service Intermountain Fire Laboratory and the Pan American Institute of Geography and History (PAIGH), developed a method to assess and map broad areas to estimate the potential for fires. PAIGH sponsored an international project to develop a method to predict fire danger.

Local firefighters quickly control most fires while they are still fairly small. However, a significant number of fires exceed the ability of the first fire suppression forces to contain them and spread to cause loss of life and substantial damage to natural resources and property. To minimize this threat of loss from wildfires, fire managers must be able to plan protection strategies that are appropriate for local areas. A prerequisite for this planning is the ability to assess and map, for broad areas, the local potential for a major fire to occur. Using such geospatial information, managers can establish priorities for prevention activities to reduce the risk of wildfire spread and for allocating suppression forces to improve the probability of quickly controlling fires in areas of high concern.

Using these requirements, Burgan et al. (In Press) and Klaver et al. (1997) developed a Fire Potential Index (FPI). This index is based on the moisture of the live vegetation, the moisture of the dead vegetation, and the amount of the live and dead vegetation. The moisture of the live vegetation is derived from the relative greenness of the Normalized Difference Vegetation Index (NDVI) from the AVHRR sensor (Burgan and Hartford 1993). The moisture of the dead vegetation is calculated from temperature, relative humidity, and the state of the weather (Fosberg and Deeming 1971). The amounts of live and dead fuels are derived by reclassifying existing baseline land cover maps to the National Fire Danger Rating System’s (NFDRS) fuel models (Deeming and others 1978), which provide information on the loadings of live and dead fuels (Bradshaw et al. 1984).

The U.S. Forest Service calculated FPI daily for the 1997 fire season (Figure 2). Scientists in Chile, Mexico, and Spain are currently calculating the FPI for the Mediterranean ecosystems of their countries; future work will expand to comparable areas in Argentina. The FPI is anticipated to be robust for most ecological systems throughout the world.

FPI daily for the 1997 fire season

Figure 2. This graphic from URL http://www.fs.fed.us/land/wfas/exp_fp_4.gif (February 16, 1998) shows the relative greenness and 10-hr fuel moisture used to calculate the Fire Protection Index (FPI) for October 2, 1997. The lower-right graphic shows the traditional National Fire Danger Rating System (NFDRS). The FPI is of higher spatial resolution than the NFDRS.

The fire potential and actual fires need to be modeled in concert with socioeconomic information, such as population and land use, to determine costs to the human population in the actual burned areas and in neighboring areas affected directly by the smoke and haze and indirectly by economic losses, such as drops in forest or range productivity and tourism.


Satellite-borne sensors are available that can detect fires in the visible, thermal, and mid-infrared bands. Active fires can be detected by their thermal or mid-infrared signature during the day or night or by the light from the fires at night. The sensors must also have frequent overflights and the data from the overflights must be available in near real-time. Two sensors that meet the criteria are the AVHRR sensor, which has a thermal sensor and daily overflights, and the Defense Meteorological Satellite Program’s Optical Linescan System (OLS) sensor, which has daily overflights and operationally collects visible images during its nighttime pass.

Researchers at the EROS Data Center are currently investigating the utility of the OLS nighttime visible band for monitoring the location of fires in Madagascar. Nightly data were available in near real-time during the 1997 "burn season" (August-December). Historical fire data will be analyzed for the 1992-1997 "burn seasons" in an attempt to relate brightness values to fire intensity, vegetation cover, and area burned.

Data from the OLS sensor need to be adjusted to account for the locations of alternative sources of light, such as city lights or gas fires in oil fields (Elvidge and others 1997a, Elvidge and others 1997b). Once these light sources have been identified, the remaining signal can reasonably be associated with vegetation fires (Cahoon and others 1992). These fires comprise a combination of controlled agricultural-related fires and wildfires. In the Indonesian example, the city lights on Java are clearly visible, and a comparison of the population distribution with the location of many of the detected lights in Borneo and Sumatra suggests the locations of the wildfires (Figure 3).

Figure 3 DMSP OLS nighttime visible image from NOAA
Figure 3 Estimated 1995 population density

Figure 3. September 30, 1997, DMSP OLS nighttime visible image from NOAA WWW site (http://www.ngdc.noaa.gov/dmsp/dmsp.html 16 Feb. 1998) and estimated 1995 population density, (http://na.unep.net/globalpop/ 16 Feb. 1998).

Band 3 of AVHRR, in combination with bands 4 and 5, has been shown to be effective for detecting wildfires (Robinson 1991; Langaas 1992; Chuvieco and Martin 1994; Kennedy, Belward, and Grégoire 1994; Malingreau and Justice 1997; Pozo, Olmo, and Alados-Arboledas 1997). As with OLS images, significant agricultural fires were detected. These controlled agricultural fires can often be excluded through the use of nighttime thermal images. An inspection of a composite of the AVHRR bands produced by NOAA shows both the extent of the smoke and haze from the fire and many of the actual fire locations (Figure 4).

Figure 4 26 September 1997 AVHRR

Figure 4. 26 September 1997–AVHRR GAC. AVHRR images can be acquired at http://www.saa.noaa.gov/


Fire monitoring differs from fire detection in emphasis rather than in fundamental methods. Fire monitoring measures and describes the growth of known fires; three characteristics of interest are the growth of the fire, extent of the smoke plume, and mapping of the fire scar.

Monitoring the movement and dispersion of fires is a variant of fire detection, where the focus is the analysis of changing fire patterns. As in fire detection, thermal and nighttime visible images are effective for mapping changing fire patterns. Monitoring the extent of the smoke plume requires analysis of visible and near-infrared wavelengths (Figure 5). Tracking the smoke plume allows the impact of fires on neighboring human populations to be estimated. Radar can be used to monitor the extent of fire scars through moderate smoke and haze.

Figure 5 August 4, 1997 AVHRR browse

Figure 5. August 4, 1997 - AVHRR browse

Figure 5 September 22, 1997 AVHRR Browse

Figure 5. September 22, 1997 - AVHRR Browse

The AVHRR preview or browse images are received at the Australian Bureau of Meteorology (Darwin) and made available through the Department of Land Administration, Remote Sensing Services, Western Australian State Government (http://www.rss.dola.wa.gov.au 16 Feb. 1998).

The general extent of the fires can be measured from satellite images, but detailed information vital for fire fighting can only be acquired from airborne sensors closely integrated with resources on-the-ground.


Once fires are extinguished, a combination of low resolution images (AVHRR) and higher resolution images (SPOT, Landsat, and radar) can be used to assess the extent and impact of the fire. Radar has proved effective in monitoring and assessing the extent and severity of fire scars in the boreal forests (Kasischke, Bourgeau-Chavez and French 1994), for quantifying biomass regeneration in tropical forests (Luckman and others 1997), and for modeling ecosystem recovery in mediterranean climates (Viedma and others 1997). Low-resolution visible and infrared sensors, such as AVHRR, have proved useful for automated fire mapping (Fernández, Illera and Casanova 1997) and for evaluating the impact of fire on long-term land cover change (Ehrlich, Lambin and Malingreau 1997). Multiresolution studies incorporating both AVHRR and Landsat images reveal the scale-related influences of analyzing fire regeneration (Steyaert, Hall, and Loveland 1997).

Information related to new fire scars and vegetation succession within the scars can be used to update the baseline vegetation map used for fire prediction. Continued monitoring of the fire scars provides extensive information on land cover transitions involving changes in productivity and biodiversity, which, in turn, influence fire potential. Knowledge of the extent and intensity of the fire scars provides important information for the rehabilitation of the burn areas.


Significant research and applications of the elements of the fire analysis cycle are under way or in use. At the national level, many countries are implementing fire potential models to mitigate the impact of fire. Fire detection and monitoring are critical. Admittedly, the most important application is the minimizing direct harm to human populations, which is not included at this time in the context of global- or continental-scale fire models. Fire detection and monitoring as discussed here emphasize fire distribution, the relationships between fire, human populations, and ecosystems, and climate studies. Fire assessment actively feeds back into the fire prediction models, resulting in a dynamic baseline land cover database incorporating fire history and vegetation succession.

Data availability and long-term viability are critical to the success of an effective global fire analysis system. Effective access to AVHRR, Landsat, and OLS images and their long-term archive is finally becoming a practical reality, which does not imply that suitable images have been collected. Scheduling of the sensor and cloud cover Costs of images and image analysis still remain a substantial part of the overall system cost, particularly for higher resolution images, such as Landsat or SPOT. Extracting information from these data is or soon will be manageable. The far more difficult problem is access to weather data affecting fire potential. To produce a product of sufficient quality to serve the needs of global fire analysis and to assist in fire management requires significant, detailed, local, real-time knowledge of weather conditions. To be effective, a global fire analysis system requires access to low-cost data, an organization with long-term viability, long-term commitments with local sources of weather data, and local requirements for the products.



The support provided for this study by UNEP, NASA, and the USGS is gratefully acknowledged. The authors appreciate the valuable suggestions made by Steve Howard, Dave Meyer, and Jacqueline Klaver (USGS EROS Data Center), who reviewed the manuscript. All remaining omissions and errors rest solely with the authors.


L.S. Bradshaw, J.E. Deeming, R.E. Burgan, and J.D. Cohen, compilers. The 1978 national fire-danger rating system: technical documentation. Gen. Tech. Rep. INT-169. Ogden, UT: U.S. Department of Agriculture, Forest Service, 1984.

R.E. Burgan, R.W. Klaver, and J.M. Klaver. Fuel models and fire potential from satellite and surface observations. International Journal of Wildland Fire, In Press.

R.E. Burgan and R.A. Hartford. Monitoring vegetation greenness with satellite data. Gen. Tech. Rep. INT-297. Ogden, UT: U.S. Department of Agriculture, Forest Service, 1993.

D.R. Cahoon, Jr., B.J. Stocks, J.S. Levine, W.S. Cofer, III, and K.P. O’Neill, Seasonal distribution of African Savanna Fires: Nature, Vol. 359, pp. 812-815, 1992.

E. Chuvieco, and M.P. Martin, "Global Fire Mapping and Fire Danger Estimation Using AVHRR Images," Photogrammetric Engineering and Remote Sensing, Vol. 60, No. 5, pp. 563-570, 1994.

J.E. Deeming, R.E. Burgan, and J.D. Cohen, The national fire-danger rating system - 1978. Gen. Tech. Rep. INT-39. Ogden, UT: U.S. Department of Agriculture, Forest Service, 1978.

D. Ehrlich, E.F. Lambin, and J. Malingreau, "Biomass Burning and Broad-Scale Land Cover Changes in Western Africa," Remote Sensing of the Environment, Vol. 61, pp. 201-209, 1997.

C.D. Elvidge, K.E. Baugh, E.A. Kihn, H.W. Kroehl, and E.R. Davis, "Mapping City Lights With Nighttime Data from the DMSP Operational Linescan System," Photogrammetric Engineering and Remote Sensing, Vol. 63, No. 6, pp. 727-734, 1997a.

C.D. Elvidge, K. E. Baugh, E.A. Kihn, H.W. Kroehl, E.R. Davis, and C. Davis, "Relation between Satellite Observed Visible Near Infrared Emissions, Population, Economic Activity and Electric Power Consumption," International Journal of Remote Sensing, Vol. 18, No. 6, pp. 1373-1379, 1997b.

A. Fernández, P. Illera, and J.L. Casanova, "Automatic Mapping of Surfaces Affected by Forest Fires in Spain Using AVHRR NDVI Composite Image Data," Remote Sensing of the Environment, Vol. 60, pp. 153-162, 1997.

M.A. Fosberg, and J.E. Deeming, Derivation of the 1- and 10-hour timelag fuel moisture calculations for fire-danger rating. Research Note RM-207. Fort Collins, CO: U.S. Department of Agriculture, Forest Service, 1971.

E.S. Kasischke, L.L. Bourgeau-Chavez, and N.H.F. French, "Observations of Variations in ERS-1 SAR Image Intensity Associated with Forest Fires in Alaska," IEEE Transactions on Geoscience and Remote Sensing, Vol. 32, No. 1, pp. 206-210, 1994.

P.J. Kennedy, A.S. Belward, and J. Grégoire, "An Improved Approach to Fire Monitoring in West africa Using AVHRR Data," International Journal of Remote Sensing, Vol. 15, No. 11, pp. 2235-2255, 1994.

J.M. Klaver, R.W. Klaver, and R.E. Burgan, "Using GIS to Assess Forest Fire Hazard in the Mediterranean Region of the United States," 1997 ESRI Users Conference, http://www.esri.com/base/common/userconf/proc97/HOME.HTM (February 14, 1998), San Diego, CA, July 8-11, 1997.

S. Langaas, "Temporal and Spatial Distribution of Savanna Fires in Senegal and The Gambia, West Africa, 1989-90, Derived from Multitemporal AVHRR Night Images," International Journal of Wildland Fire, Vol. 2, No. 1, pp. 21-36, 1992.

A. Luckman, J. Baker, T.M. Kuplich, C.C.F. Yanasse, and A.C. Frery, "A Study of the Relationship between Radar Backscatter and Regenerating Tropical Forest Biomass for Spaceborne SAR Instruments," Remote Sensing of the Environment, Vol. 60, pp. 1-13, 1997.

J.P. Malingreau, and C. O. Justice, Definition and Implementation of a Global Fire Product Derived from AVHRR Data, 3rd IGBP-DIS Fire Working Group Meeting Report, Toulouse, France, 13-15 November 1996: Toulouse, France, IGBP-DIS Office, 1997.

D. Pozo, F.J. Olmo, and L. Alados-Arboledas, "Fire Detection and Growth Monitoring Using a Multitemporal Technique on AVHRR Midinfrared and Thermal Channels," Remote Sensing of the Environment, Vol. 60, pp. 111-120, 1997.

J.M. Robinson, "Fire from Space - Global Fire Evaluation Using Infrared Remote-Sensing," International Journal of Remote Sensing, Vol. 12, No. 1, pp. 3-24, 1991.

L.T. Steyaert, F.G. Hall, and T.R. Loveland, "Land Cover Mapping, Fire Regeneration, and Scaling Studies in the Canadian Boreal Forest with 1-km AVHRR and Landsat TM Data," Journal of Geophysical Research, Vol. 102, No. d24, pp. 29,581-29,598, 1997.

O. Viedma, J. Meliá, D. Segarra, and J. García-Haro, "Modeling Rates of Ecosystem Recovery after Fires by Using Landsat TM Data," Remote Sensing of the Environment, Vol. 61, pp. 383-398, 1997.

* Presented at the First International Conference on Geospatial Information in Agriculture and Forestry
** Work performed under U.S. Geological Survey contract 1434-CR-97-CN-40274.