Centre for Remote Imaging, Sensing and Prcessing (CRISP)	CRISP's Research
Contact: crisp@nus.edu.sg	Copyright © CRISP, 2001

Forest Fire Detection and Monitoring

This section is based on an article originally published as a chapter in a book:
S. C. Liew, L. K. Kwoh, O. K. Lim, and H. Lim (2001), Remote sensing of fire and haze, in "Forest fires and regional haze in Southeast Asia", ed. P. Eaton and M. Radojevic (New York: Nova Science Publishers), Chapter 5, pp. 67-89.

Forest/Land Fires in Southeast Asia

Biomass burning has been the traditional method of clearing land for shifting cultivation in Southeast Asia. However, the small scale clearing of land is increasingly being replaced by modern large-scale conversion of forests into plantations/agricultural land, usually by fires (Goldammer 1997). In periods of extreme drought, especially those associated with the El Nino Southern Oscillation (ENSO) phenomenon, the fires get out of control resulting in severe transboundary air pollution in the form of smoke haze.

In Southeast Asia, forest fire events occurred during the dry seasons in 1982-83, 1987, 1991, 1994 and 1997-98. In the 1982-83 forest fire event, it has been reported that 5 million hectares of primary and secondary rain forests in Borneo was affected by the fires. The recent one in 1997 was especially severe, causing many countries in Southeast Asia to be affected by thick smoke haze.

Remote Sensing Satellites For Fire Detection and Monitoring

Satellite remote sensing offers a useful tool for fire monitoring, management and damage assessment. In satellite remote sensing, information about the earth's surface and atmosphere is acquired using sensors mounted on-board satellites orbiting the earth.

In comparison to the conventional methods of information gathering, satellite remote sensing provides the following advantages,

large area coverage,
frequent and repetitive coverage of an area of interest,
quantitative measurement of ground features using radiometrically calibrated sensors,
synoptic views of events in relation to the environment.
semiautomated computerised processing and analysis, and
relatively low cost per unit area of coverage.

Several remote sensing satellites are currently available, providing imagery suitable for forest fire research and fire monitoring operations. Each of these satellite-sensor platform is characterised by the wavelength bands employed in image acquisition, spatial resolution of the sensor, the coverage area and how frequently a given location on the earth surface can be imaged by the imaging system.

The following satellite-sensor systems are commonly used in fire detection and monitoring:

NOAA-AVHRR

Spectral Bands	1: red 2: NIR 3: MWIR 4, 5: thermal
Swath	about 2000 km
Spatial Resolution	1.1 km
Temporal Frequency	Daily
Applications	Detection of thermal emissions from active fires (hot spots)

DMSP-OLS

Spectral Bands	1: visible 2: thermal
Swath	about 2000 km
Spatial Resolution	2.7 km
Temporal Frequency	Daily
Applications	Detection of light emission from night-time fires

Terra-MODIS

Spectral Bands	36 bands covering visible, NIR, SWIR, MWIR and thermal IR
Swath	about 2000 km
Spatial Resolution	Two 250m bands and five 500m bands in visible, NIR and SWIR; 1km in MWIR and thermal IR
Temporal Frequency	Daily
Applications	Detection of thermal emissions from active fires; burn scars mapping; vegetation/land cover mapping

Landsat-TM

Spectral Bands	1: blue 2: green 3: red 4: NIR 5, 7: SWIR 6: thermal Panchromatic Band
Swath	about 100 km
Spatial Resolution	30 m (bands 1, 2, 3, 4, 5, 7) 15 m (panchromatic band) 60 m (band)
Temporal Frequency	Once in 16 days
Applications	Detection and mapping of burn scars; vegetation classification and mapping

SPOT-HRV/HVIR

Spectral Bands	1: green 2: red 3: NIR 4: SWIR (SPOT 4 only) Panchromatic Band
Swath	about 100 km
Spatial Resolution	20 m (bands 1, 2, 3, 4) 10 m (panchromatic band)
Temporal Frequency	Almost daily with 3 satellites
Applications	Detection of active fires; accurate location of fires; detection and mapping of burn scars; vegetation classification and mapping

Detection of fires and fire-affected areas

Heat Emission

During a fire event, the heat emitted by active fires can be detected by infrared sensors on-board a satellite. This procedure is commonly known as 'hot spot detection'. Fire hot spots can be detected and monitored during a fire event to provide information on the general locations, spatial distributions and temporal evolution of fire.

The AVHRR sensors on board the NOAA Polar Orbiting Environment Satellites are commonly used to detect heat emission from fires for the construction of fire hotspot distribution maps.

The amount of thermal radiation emitted at a particular wavelength from a warm object depends on its temperature. If the earth's surface is regarded as a blackbody emitter, its apparent temperature and the spectral radiance are related by the Planck's blackbody formula, plotted on the left. For a surface at a brightness temperature around 300 K, the spectral radiance peaks at a wavelength around 10 mm. The peak wavelength decreases as the brightness temperature increases. For typical fire temperatures from about 500 K (smouldering fire) to over 1000 K (flaming fire), the radiance versus wavelength curves peak at around 3.8 mm, which coincides with Band 3 of AVHRR. For this reason, Band 3 of AVHRR is often used for detection of 'fire hotspots'.

Limitations of AVHRR for Fire Detection

The resolution is about 1 km, hence, the number of fires within a detected hot spot cannot be determined.
The geolocation error is a few km, fire location is unprecise.
Due to the spatial resolution limitation, it generally overestimates fire area, and underestimates the number of fires.
The sensor saturates at about 47 deg C, so it is unable to determine fire temperature and size of fires.
Many false alarms need to be eliminated. False detection may arise due to sun glitter and warm land surface.

Light Emission

The US Air Force (USAF) Defence Meteorological Satellite Program (DMSP) operates a series of satellites which carry very sensitive light sensors known as the Operational Linescan System (OLS) that can detect light emission from the earth surface at night. The digital data from DMSP-OLS were archived at NOAA's National Geophysical Data Center (NGDC) since 1992. The USAF initially imposed a 72 hours embargo on the DMSP-OLS before they could be released by NOAA. Since 21 Dec 1999, the 72 hours hold has been relaxed to 3 hours.

The DMSP-OLS has a ground swath of about 3000 km. It has two broad spectral bands, one covering the visible-near infrared region (0.5 - 0.9 mm) and the other is in the thermal infrared region around 10 mm. The OLS data are acquired in two spatial resolution modes: 'fine' and 'smoothed'. The full resolution fine data have a nominal spatial resolution of 0.56 km. On board averaging of five by five blocks of fine data produces "smoothed" data with a nominal spatial resolution of 2.7 km. Most of the data received by NOAA-NGDC is in the smoothed spatial resolution mode. The DMSP-OLS data are also available from NOAA SAA. The DMSP-OLS has a unique capability to observe faint sources of visible- near infrared emissions present at night on the Earth's surface, including cities, towns, villages, gas flares, heavily lit fishing boats and fires. By analysing a time series of DMSP-OLS images, it is possible to define a reference set of "stable" lights, which are present in the same location on a consistent basis. Fires are identified as lights detected on the land surface outside the reference set of stable lights.

Due to the coarse resolution of the smoothed OLS data, they have similar limitations as the AVHRR data for fire detection. For example, it is not possible to determine the number of fires, locations of individual fires and fire size within an identified fire hotspot. Furthermore, smouldering fire or fire burning below the canopy will not be detected by OLS.

Smoke Plumes

Smoke plumes in optical remote sensing images are direct telltale signs of active fires. The can be visually identified in an image as some fuzzy bright strips, usually conical in shape converging to a point on the ground. If more than one plumes are visible, the plumes all orientated in a general direction along the direction of the wind. In high resolution images such as those acquired by the SPOT satellites, the precise locations of the active fires can be determined from the points of origin of the individual smoke plumes.

A SPOT quicklook image of part of the 'Mega Rice Project' area in Central Kalimantan, acquired on 8 September 1997, showing burned areas and smoke plumes arising from sites of active fires. The image covers an area of 60 km x 60 km. © CNES 1997, acquired and processed by CRISP.

Large smoke plumes can also be observed from relatively low-resolution sensors such as the ones on-board a meteorological satellite. The distribution of these smoke plumes, together with the prevailing wind conditions, provide early warnings for the possible spreading of smoke haze to the neighbouring regions.

Burned Area

Fire affected areas can be discriminated in a remote sensing image due to the strong contrast with the unburned areas. Burn scar maps derived from remote sensing images provide information on the spatial extent and distribution of the fire affected areas, and the total area burned. The burned area, together with land cover information, provide the basis for estimating emission of carbon dioxide and other gases implicated for global climate change, and for assessment of fire damage to the economy and environment.

Land cover maps can be derived from remote sensing images using reflectance signatures of the various land cover types. The land cover maps are useful in ascertaining the types of land affected by fires during a fire event for effective fire fighting and fire management, and for damage assessment.

Near Real-Time Forest/Land Fire Monitoring Operation

An on-going regional land/forest fire monitoring operation is conducted at CRISP, in collaboration with the Ministry of Environment, Singapore. This operation makes use of both the wide coverage AVHRR and high resolution SPOT images in detecting fires in the region. Based on the daily "hot-spot" data derived from NOAA-AVHRR, SPOT satellites are programmed to acquire "zoom-in" views of the fire areas. Daily reports on fires and smoke plumes, with selected annotated images, are forwarded to the Ministry of the Environment.

With SPOT images, it is possible to accurately locate each of the fires and smoke plumes, to tell if the fires are in forests, peat swamps, or plantations, and if they are associated with activities such as clearing of agricultural land or timber logging sites.

(i) SPOT Images for Near Real Time Forest Fire Monitoring ; (ii) Commonly used Satellite Sensors