Home Articles Application of Remote Sensing and GIS for forest fire susceptibility mapping using...

Application of Remote Sensing and GIS for forest fire susceptibility mapping using likelihood ratio model

Biswajeet Pradhan
Remote Sensing and GIS Division
Cilix Corporation Sdn Bhd.
Lot L4-I-6, Enterprise 4
Technology Park Malaysia
Bukit Jalil Highway, Bukit Jalil
57000, Kuala Lumpur, Malaysia
Tel. +603- 8996 6250
Fax. +603- 8996 4087
[email protected]

Mohamad Arshad Bin Awang
Remote Sensing and GIS Division
Cilix Corporation Sdn Bhd.
Lot L4-I-6, Enterprise 4
Technology Park Malaysia
Bukit Jalil Highway, Bukit Jalil
57000, Kuala Lumpur, Malaysia
Tel. +603- 8996 6250
Fax. +603- 8996 4087

This paper presents capability of remote sensing and Geographical Information Systems (GIS) to evaluate forest fire susceptibility. Forest Fire locations were identified in the study area from historical hotspots data from year 2000 to 2005 using AVHRR NOAA 12 and NOAA 16 satellite images. Various other supported data such as soil map, topographic data, and agro climate was collected and created using GIS. These data were constructed into a spatial database using GIS. The factors that influence to fire occurrence, such as fuel type and Normalized Differential Vegetation Index (NDVI) were extracted from classified Landsat-7 ETM imagery. Slope and aspect of topography, were derived from topographic database. Soil type was extracted from soil database and dry month code from agroclimate data. Forest fire susceptibility was analyzed using the forest fire occurrence factors by likelihood ratio method. The results of the analysis were verified using forest fire location data. The validation results show satisfactory agreement the susceptibility map and the existing data on forest fire location. The GIS was used to analyze the vast amount efficiently, and statistical programs were used to maintain the specificity and accuracy. The result can be used for early warning, fire suppression resources planning and allocation.

1.0 Introduction
Fire has been identified as one of the major threats causing the loss of forests in several states in Malaysia. According to Forestry Department of Peninsular Malaysia (JPSM) and Forest Research Institute Malaysia (FRIM) statistics show that during the last 7 years (between 1992 and 1998), more than 1600 ha of peat swamp forests of Peninsular Malaysia have been destroyed by fire (Wan Ahmad, 2002). In the event of a prolonged spell without rain, and a lowering of the water table in the peat swamp forest, the organic layers becomes completely dry and is very prone to fire. Fires in these peat swamp forest create much more smoke per hectare than other types of forest fires and are difficult to extinguish. Therefore, the understanding of the areas at risk to fire needs to be closer concentration in peat swamp forests. A precise evaluation of forest fire problems and decision on solutions can only be satisfactory when a fire hazard zone mapping is available (Jaiswal et al, 2002).

Geospatial technology, including Remote Sensing and Geographic Information Systems (GIS), provides the information and the tools necessary to develop a forest fire susceptibility map in order to identify, classify and map fire hazard area. Before, during and after disaster, the accurate sharing of information is important. Making the information available via the world-wide web, people can share information to assess the situation and make decisions. In this study we want to develop and produce a forest fire hazard model and map for Sungai Karang and Raja Muda Forest Reserve, Selangor (Peat Swamp Forest) using frequency ratio, which is a statistical model.

2.0 Study Area
The study area is located approximately between Upper Left (3° 23′ 53.6″E and 101° 3′ 36.3″N) and Lower Right (3° 45′ 18.05″E and 101° 30′ 55.33″N). The area located within the Kuala Selangor District, northern part Selangor. The landuse at the study area is mainly peat swamp forest, plantation forest, inland forest, scrub, grassland and ex-mining area. The landform of the area ranges from very flat terrain, especially for the peat swamp forest, ex mining, grassland and scrub area, to quite hilly area for the natural forest ranging between 0- 420 meter above sea level. Based on Malaysian Meteorological Services Department, the temperature of northern part of Selangor is between 29° C to 32° C and mean relative humidity of 65% to 70%. The highest temperature is between April to June while the relative humidity is lowest in June, July and September. The rainfall about 58.6mm to 240mm per month was recorded in the study area (Tanjung Karang weather station provided by Malaysian Meteorological Services Department).

There is a high potential of danger of fire in the dry season especially in the peat swamp forest and plantation forest. Most of fires are caused by human activities, either due to carelessness or burning activities in crop plantations. On 1995 and 1999, fire was occurred in the peat swamp forest area within the study area. Figure 1 shows Raja Muda Musa and Sg Karang Reserve Forest, Selangor.

Figure 1: Sg Karang and Raja Muda Musa Forest Reserve, Selangor

3.0 Data using GIS and Remote Sensing
Accurate detection of the location of hotspots is very important for probabilistic forest fire susceptibility analysis. Recent advances in remote sensing, GIS and computer technologies provided an opportunity to assess and monitor the land cover changes in a near real time basis. NOAA AVHRR satellite data with a spatial resolution of 1.1 km at nadir was found to be extremely useful for national-scale assessment and monitoring of major land cover types (Giri & Shrestha, 1996). Historical forest fire data were collected from satellite remote sensing NOAA AVHRR 12 and NOAA 14 sensors for last 5 years. To assemble a database to assess the surface area and number of hotspots in the study area, a total of 112 hotspots were mapped in a mapped area of 616 km2. The imagery from Landsat-7 ETM of path 157 and row 058 acquired on 21 September 2001 was used in this study. The spatial resolution for Landsat-7 ETM was 30 meter x 30 meter. Fuel map were extracted from satellite imagery.

GIS data and ancillary data consist of biophysical and socio-economic variable is based on 1: 25,000 scale. Contour, administrative boundaries, water resources, settlement, transportation infrastructure are based on the topographic map from Survey Department (JUPEM). Forest fire reports have been collected from Forest Department Peninsular Malaysia (JPSM). Hotspots prone areas, fire occurrence map, peat swamp map and soil maps have been acquired and digitized. Socio-economic data such as population data and socio-economic data were obtained from Statistical Department. Meteorological data such as temperature and relative humidity and Fire Danger Rating System (FDRS) map were obtained from Malaysian Meteorological Services Department. Image processing was carried out using ERDAS Imagine 8.7 and PCI Geomatica 9.0.

To apply the probabilistic method, a spatial database that considers forest fire-related factors was designed and constructed. These data are available in Malaysia either as paper or as digital maps. The spatial database constructed is shown in Table 1.

There were six factors that were considered in calculating the probability, and the factors were extracted from the constructed spatial database. The factors were transformed into a vector-type spatial database using the GIS, and forest fire-related factors were extracted using the database. A digital elevation model (DEM) was created first from the topographic database. Contour and survey base points that had elevation values from the 1:25,000-scale topographic maps were extracted, and a DEM was constructed with a resolution of 20 m. Using this DEM, the slope angle and slope aspect were calculated. The soil map is obtained from a 1:100,000-scale soil map. Landsat-7 ETM, 30 meter x 30 meter resolution was used for extracting fuel map in the Sg Karang and Raja Muda Forest Reserve, Selangor. Multi-parametric analyses or overlays was carried out using GIS which severity zones and the prioritize was based on frequency ratio approach. Land use and land cover data was classified using a Landsat-7 ETM image employing an unsupervised classification method and topographic map. The fuel type has been classified into ten classes, such as peat swamp forest, mangrove, inland forest, rubber plantation, grassland, oil palm plantation, paddy, shrub, cleared land and unclassified (water bodies and urban) were extracted for fuel type mapping. Finally, the NDVI map was obtained from Landsat-7 ETM satellite images. The NDVI value was calculated using the formula NDVI = (IR – R) / (IR + R), where IR value is the infrared portion of the electromagnetic spectrum, and R-value is the red portion of the electromagnetic spectrum. The NDVI value denotes areas of vegetation in an image.

The factors were converted to a raster grid with 30 m × 30 m cells for application of the frequency ratio model. The area grid was 2,418 rows by 1,490 columns (i.e., total number is 3033610) and 112 cells had forest fire occurrences.

4.0 Methodology

4.1 Frequency ratio model and its application
Frequency ratio approaches are based on the observed relationships between distribution of hotspot and each hotspot-related factor, to reveal the correlation between hotspot locations and the factors in the study area. Using the frequency ratio model, the spatial relationships between hotspot-occurrence location and each factors contributing hotspot occurrence were derived. The frequency is calculated from analysis of the relation between hotspot and the attributing factors. Therefore, the frequency ratios of each factor’s type or range were calculated from their relationship with hotspot events as shown in Table 2. In the relation analysis, the ratio is that of the area where hotspots occurred to the total area, so that a value of 1 is an average value. If the value is greater than 1, it means a higher correlation, and value lower than 1 means lower correlation.

To calculate the Forest Fire Susceptibility Index (FFSI), each factor’s frequency ratio values were summed to the training area as in equation (1). The hotspot susceptibility value represents the relative susceptible to forest fire occurrence. So the greater the value, the higher the susceptible to forest fire occurrence and the lower the value, the lower the susceptible to forest fire occurrence.

FFSI = Fr1 + Fr2 + …… + Frn          (1)
(FFSI: Forest Fire Susceptibility Index; Fr: Rating of each factors’ type or range) The forest fire susceptibility map was made using the FFSI values and for interpretation is shown in Figure 2.

This study consists of development of Forest Fire Susceptible Map. Figure 2 shows flowchart of methodology.

Figure 2: Flowchart of methodology

5.0 Conclusion and Discussion
In the present study, frequency analysis method was applied for the forest fire susceptibility mapping for Sungai Karang and Raja Muda Forest reserve. The validations results show that the frequency ratio model has predication accuracy of 3.52%. Here, the authors can conclude that the results of frequency ratio model had shown the best prediction accuracy in forest fire susceptibility mapping.

The frequency ratio model is simple, the process of input, calculation and output can be readily understood. The large amount of data can be processed in the GIS environment quickly and easily. Moreover, it is hard to process the large amount of data in the statistical package. Recently, forest fire susceptibility mapping has shown a great deal of importance for haze detection and fire prevention in forest area. The results shown in this paper can help the concerned authorities for forest fire management and mitigation. However, one must be careful while using the models for specific mitigation. This is because of the scale of the analysis where other forest fire related factors need to be considered. Therefore, the models used in the study are valid for awareness so that necessary prevention measures can be taken during the time of forest fire. In this paper, Forest fire susceptibility map was developed to determine the level of severity of forest fire hazard zones in terms of mapping susceptibility to fire by assessing the relative importance between fire factors and the location of fire ignition


  • Boyle.T (1999), ‘Malaysia: Conservation and sustainable use of tropical peat swamp forests and associated wetland ecosystems’, United Nations Development Programme (UNDP), Project Brief.
  • Chuvieco, E., Congalton. R.G., (1989). ‘Application of remote sensing and geographic information systems to forest fire hazard mapping’, Remote Sensing of the Environment, 29, pp.147-159.
  • Giri, C. and Shrestha, S. 1996. Land Cover Mapping and Monitoring from NOAA AVHRR data in Bangladesh. International Journal of Remote Sensing, Vol. 17, No. 14, pp. 2749-2759.
  • Goward, S.N.; A. Kerber, and V. Kalb. 1987. Comparison of North and South American Biomass from AVHRR Observations. Geocarto International, Vol. 2, No. 1, pp. 27-39.
  • Jaiswal, R.K., Saumitra, M., Kumaran, D.R. and Rajesh, S. (2002), ‘Forest fire risk zone mapping from satellite imagery and GIS’, International Journal of Applied Earth Observation and Geoinformation, Vol. 4, pp. 1-10.
  • John R. Jensen. (1996), ‘Introductory Digital Image Processing’, Prentice Hall, Upper Saddle River, New Jersey
  • Ketpraneet, S. (1991), ‘Forest fire and effects of forest fire on forest system in Thailand’, Kasetsart University, Bangkok.
  • Li, Z.; Nadon, S.; Cihlar, J. (2000), ‘Satellite detection of Canadian boreal forest fires: development and application of the algorithm’, International Journal of Remote Sensing 21(16): 3057-3069.
  • Narain, A. (2003), ‘Disaster management system: an integrated approach using satellite communication, education and remote sensing’, available at: https://home.iitk.ac.in/, ramesh/narain.do
  • Qadri. S.T 1 (2001), ‘Fire, Smoke and Haze, The ASEAN Response Strategy’ Asian Development Bank, Manila.
  • Salas, J. and Chuvieco, E. (1994), ‘Geographic information systems for wildland fire risk mapping’, Wildfire, Vol. 3 No. 2, pp. 7-13.
  • Stibig H-J., Beuchle, R, and Janvier, P. (2002), ‘Forest cover map of insular Southeast Asia at 1:5 500000’, TREES Publications Series D: N°3, EUR 20129 EN, European Commission, Luxembourg, available at:
  • Wan Ahmad, W.S. (2002), ‘Forest fire situation in Malaysia’, IFFN, No. 26, pp. 66-74. available at: www.fire.unifreiburg.de/