Monitoring of Forest Fires And Oil Pollution from Space
A typical SPOT scene of forest fires is shown in Fig.3.This scene was acquired on 15 February,19998, near Samarinda in East Kalimantan. In the visible red band (XS2) image, smoke plumes can be seen as bright strips against a generally dark forest background. Precise locations of the spots can be determined from the source of the plumes in the full-resolution image.The smoke plumes are less visible in the near infrared band (XS3) image but the burnt areas can be readily distinguished as dark patches in this band. Burnt scars can be delineated in the false-color composite SPOT images and regional maps of burnt areas can be generated using mosaics of several SPOT images over the regions of interest [5].These burnt scar maps are valuable assessing damage to the forests land also help in the study of forest regeneration processes. Fig.4 shows an example of a burnt scar map derived from mosaics of 100-m resolution SPOT quicklook images(from July to December 1997) over the southern coast of Kalimantan. For comparison, a cumulative hot spot distribution image derived from daily NOAA-14 AVHRR images(from May to October 1997) is shown in Fig.5.The hot spot distribution derived from NOAA imagery generally agree with the burnt scar maps fro the eastern part of the image. However, the Western part of the image shows an abundance of detected hot spot in Fig.5.but few burnt scars are delineated in the SPOT mosaic (Fig4).This area was covered by clods in the SPOT images acquired during the 1997 fire period. The burnt scar map could be update by using possible cloud-free images acquired after the fire period. However , the forest regeneration process may hinder the use of post-fire image in delineating burnt scars due to the possibility of vegetation regrowth in the burnt areas.
Fig.3: Band XS2 (left)and band XS3 (right) images of a SPOT scene (K302 J350) near Samarindra, East Kalimantan on 15 February,1998.ãCNES 1998.
Fig 4 (left): Burnt scar map over the southern coast of Kalimantan derived from SPOT quick look image mosaics (June to December 1997). This map covers a width of about 400 km.
Fig .5(right): cumulative hot spot distribution derived from daily NOAA-14AVHRR images( May to October 1997)
One major limitation of optical/infrared remote sensing imagery lies in the inability of optical / infrared radiation in penetrating clouds and thick haze. It is possible to use the cloud-penetrating Synthetic aperture radar (SAR) to complement optical imagery in mapping of burnt areas. SAR backscatter intensity and interferometric coherence have been used in forest mapping and monitoring [6,7,8]. Tropical forests are known to have a constant backscattering coefficient (
so) between -7and -6 dB in C-band SAR. The interferometric coherence of the vegetated areas is typically low compared with the clear cuts or sparsely vegetated areas. If multitemporal SAR data of an area of interest are acquired, clearings of forests/vegetation can be detected by an observed change in
so and/or an increase in coherence of the area. Unlike optical/infeaed sensors, SAR is unable to detect hot spots or smoke plumes directly associated with fire. It is thus not able to tell whether the clearings are due to fire or other means. However, if fires have been known to occur in an area of interest, the extent of fire affected areas can be mapping using SAR backscattered intensity and interferometric coherence signatures.
Fig. 6(a) Interferometric coherence map derived from ERS-1/2 tandem data in April 1996.(b) interferometric coherence map derived from ERS-1/2 tandem data in October 1997.(c) burnt scar map derived from the coherence and amplitude data of the two data sets in 1996 and 1997. Black; new clearing, possible affected by fires in 1997, Dark gray: Old clearings with vegetation regrowth. Light gray: Vegetated areas, white: old clearing.(d) Location map of the study area.
Interferometric SAR imagery acquired during the ERS-1/2 tandem missions in April/May1996 and October1997 over the southeast corner of Kalimantan has been used to perform change detection and to delineate areas affected by fires during the 1997 fires. The study area is covered by four ERS scenes of about 100 km cover the study area by 100km each (see Fig.6d for the location). The two coherence maps in 1996 and 1997 are shown in Fig. 6a and 6b respectively while the burnt scar map derived from coherence change is shown in fig. 6c. The possible fire burnt scars are characterized by a low interferometric coherence in 1996 and an increase in coherence between the 1996 and 1997 imagery. About 15%(550kha) of the total area surveyed has been found to be affected by fires. Most of these burnt areas occur in the ERS scene adjacent to the town of Banjarmasin .
