GIS-based Probability Assessment of Fire Risk in Grassland and Forested Landscapes of Golestan Province, Iran

Hamed Adab1+, Kasturi D. Kanniah1 ,and Karim Solaimani2 1 Remote Sensing Dept, Faculty of Geoinformationscience and Engineering, Universiti Teknologi Malaysia,

Skudai, Malaysia. 2 Remote Sensing & GIS Center of Sari University of Agric. & Natural Resources. Iran

Abstract. Forest and grassland fires are one of the major natural hazards in the West and North of Iran. The aim of the present study is to set up, assess and validate Hybrid Fire Index as a new fire risk index over Golestan province. This site frequently faces grassland and forest fire. The model deals to combine geo-spatial data by GIS technology to construct the fire risk index. The following variables were derived for the study area: vegetation moisture, slope, aspect, elevation, distance from roads, and vicinity of settlements. Those variables were weighted based on their impact on the fire risk. According to probability distribution analysis of hot spot data derived from MODIS and Hybrid Fire Index, low, moderate, and high risk were defined. Then the generated map of Hybrid Fire Index was classified into three classes. The agreement between the predicted high risk areas and the hot spots was assumed to be a major test about the Hybrid Fire Index reliability of the study approach. Based on the validation result, the Hybrid Fire Index presented a valuable fire risk map.

Keywords: Fire Risk, Hybrid Fire index, Subjective weights, spatial assessment

1. Introduction Wildfires are considered as a serious problem that distressing many terrestrial ecosystems in the Earth

system and causing economic damage for people (Butry et al., 2001) such as missing income relative to the land use, destruction and lost property, damages to agriculture, and loss of biodiversity.( Merlo and Briales, 2000. Pettenella et al.,2009) Also It is one of the most important parts of land degradation that is caused by deforestation and desertification. (Hernandez-Leal et al., 2006). Chuvieco and Congalton (1989) defined the fire risk as “the union of two components of fire hazard and fire ignition.” Based on forest terminology of FAO, “fire risk is the chance of a fire starting as determined by the presence and activity of any causative agent”.

Three groups of indices are proposed by the European Union’s (EU) Joint Research Center based on their temporal scale (JRC, 2002) as fallows;

• Structural or long-term indices. They do not change in short time such as topography, vegetation type, land-cover, land-use, slope, aspect, distance to roads and vicinity to urban areas, population density, climate, and soils, (Pelizzari et al., 2008, Chuvieco and Congalton, 1989; Aranha et al., 2001).

• Dynamic or short-term indices which change moderately continuously over time such as vegetation or weather condition. This index intend to detect the flammability of forest fuels during the fire season and hence it uses variables that changing in a short period of time (Chuvieco., et al 2005) .

+ Tel.: +60178403117. E-mail address: adabgeo@gmail.com.

170

2011 International Conference on Environmental and Computer Science IPCBEE vol.19(2011) © (2011) IACSIT Press, Singapore

• Integrated or Advanced indices that contain both structural and dynamic factors as mentioned above. But the most critical issue in this approach is how to combine effectively the relevant variables to get a coherent criterion.( Caetano., et al 2004)

Erten , et al (2002) used a structural index using topography, vegetation type, and distance from roads and settlements for presenting the forest fire risk zone map in Gallipoli Peninsula as a pilot area in Turkey. Alonso-Betanzos , et al (2002) used a dynamic index based on artificial neural network for the spatial prediction of forest fire risk in Galicia Spain. Five meteorological variables were selected for a daily scale, temperature, humidity and rainfall. Caetano, et al (2004) combined structural and dynamic indices to map a forest fire risk that can be updated in daily scale in central Portugal in the fire season.

In Middle East and North of Africa, Iran is regarded as the fourth country forest fire occurrence. In the north of Iran most fires occur from August until the end of December because there is a decrease in humidity and increase in winds speed. (Allard, 2001) The ECE/FAO reported the average number of fires per year is 130 and the average area burnt per year is 5 400 ha. In this region the largest fire occurred in 1993 which destroyed 33000 ha. (Alexandrian and Esnault, 1998). Golestan province (North East Iran) is one of the most important forest areas in Iran. Forest fires are an unfortunate and frequent hazard in Golestan, especially in Golestan National Park in the Alborz Mountains. In order to map a fire risk zone in Golestan province and minimize future possible fires, an attempt is made to set up a risk model.

2. Data and Methodology

2.1. The area of study The case study is Golestan province. It has an area of 20,380.7 square km., is in the southeast of the

Caspian Sea (Figure 1). It has a variable climate based on certain factors such as geographical location, influenced by the latitude and geographical elevation; the Alborz Mountain Ranges, its distance from the sea, the desert areas south of Turkmenistan, local winds and forest density. The most part of Golestan province is moderate caspian climate (Mediterranean). But Gorgan plains have a semi-desert climate and warm .Because it is near to the desert of Turkmenistan, lower elevations, and away from the sea.

Fig. 1: Location of Golestan province (False Color Composite of Landsat ETM+)

2.2. Vegetation Moisture Dryness increases the flammability of the forest due to the fact that the moisture influences the spreading

of the fire. (Siachalou et al, 2009).The Normalized Difference Moisture Index (NDMI) is commonly used to evaluate the moisture of the vegetation. (Hemmleb et al, 2006).The following equation defines it;

(1)

Where; NIR is the near infrared spectral wavelength, and MIR is mid infrared spectral wavelength.

2.3. Elevation Elevation is a crucial physiographic variable that is associated to wind behavior and therefore it affects

fire capability, (Rothermel, 1983) so it has a big role in fire spreading. (Jaiswal et al., 2002). Elevation 171

influences vegetation structure, fuel moisture and air humidity. (Castro and Chuveico,1996, 1998). It is mentioned that humidity and temperature have higher influence on fire at upper altitude areas than lower ones. (Hernandez-Leal et al., 2006).

2.4. Slope Slope is one the parameters that influences fire behavior.(Rothermel, 1991). Fire moves most quickly up

slope and least quickly down slope. (Jaiswal et al., 2002.). Also, in steeper slopes, rate of fire spread might rise, since flames are angled closer to the surface of ground and wind effects can supply the process of heat convection for the fire produced .( DeBano et al, 1998).

2.5. Aspect South aspects have received more sun light and exposure in the North hemisphere, because of that, drier

soil is more capable to ignition (Noon, 2003). South aspect slopes are higher temperatures, robust winds, minor humidity and lower fuel moistures because Southern aspects receive more direct heat from the sun. In the earlier day East aspects get more ultraviolet and direct sunlight than west aspect, as a consequence east aspect drier faster. (Prasad et al, 2008).

2.6. Distance from roads and urban areas Forests regions are more fire prone where they are located near to roads and road density. Because roads

and paths allow local people, graziers, and tourists to go in to the forest and cause fire. Also forested regions near to settlements and houses are more capable to fire ignition because the accidental fire can be caused by housing density and inhabitants who settle within the forest. (Jaiswal et al., 2002)

2.7. Fire Risk Model Based on above information; elevation, slope, aspect, vegetation moisture, and distance from roads and

urban areas were obtained from different sources and were classified based on subjective weights. Each class has an own rate that shows the degree of fire sensitivity. (Table 1) The input raster of each variable was reclassified by changing the original values to rate values in a raster. The value of each map represents the rate of fire sensitivity. (Figure 2)

Table. 1: Weights assigned to variables and classes for forest fire risk modelling

Variables Classes Ratings of hazard

Fire sensitivity

NDMI (MODIS) >0.36,0.26_0.36,0.16_0.26,0_0.16,<0 1,2,3,4,5 Very Low ,Low, Medium, High, Very High Elevation (ASTER)

>2000,1000_2000, ,500_1000 , 200_500, <200 1,2,3,4,5 Very Low ,Low, Medium, High, Very High

Slope <5%, 10_5%, 25_10%, 35_25%, >35% 1,2,3,4,5 Very Low ,Low, Medium, High, Very High Aspect North, East, West, South 2,3,4,5 Low, Medium, High, Very High Distance from roads

>400 , 300_400, 200_300, 100_200, <100 m 1,2,3,4,5 Very Low ,Low, Medium, High, Very High

Distance from settlements

>2000 , 1500_2000, , 1000_1500, 500_1500, <500 m 1,2,3,4,5 Low, Medium, High, Very High

In order to map fire risk, a new fire risk index was introduced based on six parameters. This is called

Hybrid Fire index. The model is described by following equation; HFI=(100v+50s+25a+10×(r+c)+5e)/10 (2) Where; v, s, a, r, c, and e indicate vegetation moisture, slope, aspect, distance from road, and distance

from settlement, and elevation. A rate of 1, 2, 3, 4 or 5 is given to each level from low to high fire danger as an input. The range of output is from 20 (Low risk) to 100 (High risk).

172

A. The classification of the slope B. The classification of the aspect D. The classification of the NDMI

Fig. 2: The classification of the map variables

2.8. Zonation of fire risk The definition of risk level (high, moderate and low) is a crucial subject to indicate the zonation of the

areas which are in high, moderate and low risk zones with regard to forest fire in the each region. For this purpose, the outcome of HFI was normalized between 0 with low risk and 255 with high risk. The risk value of each hot spot was extracted from 0 to 255. Based on probability distribution of risk value of hot spots, the best distribution was selected for index. It shows which value of risk is more probable to occur. (Figure 3). For definition of risk, we arranged three classes of risk, Low, Moderate, and High respectively. Based on probability distribution, low risk was defined between 1- 10, moderate 10-50, and high 50-100 and the fire risk map were classified based on that. (Figure 4). The final fire risk map of index was classified and mapped based on level of Low, Moderate, and High risk and its probability. (Figure 5).

3. Results For assessment of fire risk zone maps, the MODIS hot spot dataset was obtained since 2001 to 2003

from the Fire Information for Resource Management System (FIRMS) that provides by NASA . (Davies, et al. 2009.). For evaluation of index, the number of hot spot in each risk class was counted. It would be justified, that the most hot spots should be located in high risk zone and the lowest one in low zone. In other words, high risk zone have majority hot spots than low risk zone. Using histogram by zone showed that the HFI presented reasonable risk zonation because the most hot spots are located in high risk zone. Based on

Fig. 3: Probability distribution of fire risk based on HFI

Fig. 4: Definition of Level of Low, Moderate, and High risk

Fig. 5: The fire risk zone maps

C. The classification of distance from roads

E. The classification of the vicinity of settlements

F. The classification of the elevation

173

areas that are affected by the fire, it can be concluded that HFI model provided valuable information about forest fire risk. The fact that the number of hot spot in moderate risk areas is not very low does not mean that our model failed. (Figure 6) It should be noted that especially within the low risk areas, there were only the 9 % of hot spots in this zone. Therefore, the model succeeded in predicting which areas are more risky by fire.

Fig. 6: the percentage of hot spots that corresponds to low, medium and high risk zones.

4. Conclusions The present study was an attempt to integrate remote sensing data and GIS concept to determine risky

places and to plan forestry management. Fire risk models are a great approach for precautionary measures for the environmental protection of the forests. The objective of this paper was to assess Hybrid Fire index as a new fire risk index for forests in northeastern Iran. The model variables were a set of vegetation moisture (NDMI), slope, aspect, elevation and distance from roads and settlements. Suitable weight variables were selected for each variable and try to adjust outcome of the model to the characteristics of the study area. The high risk areas were compared with the hot spots from MODIS. The result showed that HFI can be considered as more reliable. We used subjective weights for forest risk zonation. Since subjective weights are based on personal judgments, using data mining models can explore the pattern of affected factors on fire and improve the results. The model can be further improved by using climatic data such as wind speed and direction, atmosphere’s dryness, rainfall etc. All in all, we cannot use same weights and same variables in different regions because forest fire in each part of earth has an own characteristics. And indices should be modified over areas with different environmental conditions. However more simulation results need to compare the other methods.

5. Acknowledgements The authors are grateful to the USGS and NASA data archive for providing the ASTER, and MODIS

hot spot data.

6. References [1] D.T. Butry, D.E. Mercer, J.R. Prestemon, J.M. Pye, T.P.Holmes. What is the price of catastrophic wildfire? Journal

of Forestry. 2001.99: 9–17.

[2] M. Merlo, E. Briales . Public goods and externalities linked to Mediterranean forests: economic nature and policy, Land Use Policy .2000.17: 197-208.

[3] D. Pettenella, M. Marchetti, D. Marino, A. Marucci, M .Ottaviano, B.Lasserre. Final reports of studies on forest fires done under Forest Focus Regulation. Economic Damages Study: "Proposal for a harmonized methodology to assess socio-economic damages from forest fires in Europe." Joint Research Centre (JRC) and the Directorate General for Environment (DG ENV) of the European Commission (EC). 2009.

[4] P.A. Hernandez-Leal, M. Arbelo, A. Gonzalez-Calvo. Fire risk assessment using satellite data. Advances in Space Research 2006.37:741–746.

[5] E. Chuvieco , R.G. Congalton. Application of Remote Sensing and Geographical Information System to Forest Fire Hazard Mapping. Remote Sensing for Environment, 1989.29: 147-159.

[6] FAO (Food and Agricultural Organization). Wild land fire management terminology. Report number 70.FAO Forestry Paper, 1986.M-99.

174

[7] Joint Research Center (JRC). Pilot Projects on Forest Fires.2002 [On-line]: http://natural-hazards.aris.sai.jrc.it/fires/,

[8] A. Pelizzari, R. A. Goncalves, M. Caetano. Information Extraction for Forest Fires Management. Book Chapter of Computational Intelligence for Remote Sensing. Intelligence, 2008, 133: 295-312.

[9] J.Aranha, G.Alves , D.Lopes. Burnt areas identification and analysis by means of remotely sensed images classification. A case study in Northern Portugal, Proceedings of the RSPS 2001 – Geomatics, Earth Observation and the Information Society, 2001. 1:629-641.

[10] E. Chuvieco, , A. Camia, , G. Bianchini, , T. Margaleff, , N. Koutsias, J. MartInez. Using Remote Sensing and GIS for Global Assessment of Fire Danger. 40th annual IEEE International Conference on Communications. ICC 2005.

[11] M. R. Caetano, S. Freire , H. Carrão. Fire Risk Mapping by Integration of Dynamic and Structural Variables, Remote Sensing in Transition (R. Goossens, editor), Millpress, Rotterdam, 2004. 1:319-326.

[12] D. Alexandrian, F. Esnault. Public policies affecting forest fires in the Mediterranean Basin. In: Proceedings of FAO Meeting on Public Policies Affecting Forest Fires, 1998, 1: 39-58.

[13] G. Allard . The fire situation in Islamic Republic of Iran. In: FRA Global Forest Fire Assessment 1990- 2000. Forest Resources Assessment Programme, Working Paper 55, 198-202.

[14] E. Erten, V. Kurgun, Ne. Musaoglu. Forest fire risk zone mapping from satellite imagery and GIS case study. International Journal of Applied Earth Observation and Geoinformation , 2002.4: 1–10.

[15] A. Alonso-Betanzos , O. Fontenla-Romero , B. Guijarro-Berdinas ,E. Hernandez-Pereira, J. Canda, E. Jimenez, J. L. Legido, Su. Muniz, C.Paz-Andrade, M.I.Paz-Andrade. A Neural Network Approach for Forestal Fire Risk Estimation. 15th European Conference on Artificial Intelligence 2002.643-647.

[16] S. Siachalou, G. Doxani , M. Tsakiri-Strati. Integrating Remote Sensing Processing and GIS to Fire Risk Zone Mapping: A Case Study for the Seih-Sou Forest of Thessaloniki. Proceeding of ICC 2009.

[17] M. Hemmleb, F. Weritz, A. Schiemenz, A. Grote, C. Maierhofer. Multi- spectral data acquisition and processing techniques for damage detection on building Surfaces. Image Engineering and Vision Metrology, ISPRS, 2006.

[18] R.C. Rothermel . How to predict the spread and intensity of forest and Range fires. Gen.Tech.Rep.INT-143. USDA Forest Service. Intermountain Forest and Range Experiment Station. 1983.

[19] R. C. Rothermel. Predicting behavior and size of crown fires in the Northern Rocky Mountains. General Technical Report INT-438. US Department of Agriculture, Forest Service, Intermountain Forest and Range Experiment Station, Ogden, UT, 1991. P. 46.

[20] R.K. Jaiswal, S. Mukherjee , K.D. Raju , R. Saxena. Forest Fire Risk Zone Mapping through Satellite Imagery & Geographical Information System. International Journal of Applied Earth Observation and Geoinformation, Elsevier Publications, 2002.4:1 -10.

[21] L.F. DeBano , D.G. Neary , P.F. folliott. Fire’s Effects on Ecosystems. John Wiley and Sons. 1998.

[22] E. K. Noon .A Coupled Model Approach for Assessing Fire Hazard at Point Reyes National Seashore: Flam Map and GIS ,Proceedings of 2nd International Wildland Fire Ecology and Fire Management Congress, Springs Resort.2003.

[23] V. K. Prasad , K.V.S. Badarinath, A. Eaturu. Biophysical and anthropogenic controls of forest fires in the Deccan Plateau, India. Journal of Environmental Management. 2008.86: 1–13

[24] D.K. Davies, H.F. Vosloo, S.S. Vannan, P.E. Frost. Near real-time fire alert system in South Africa: from desktop to mobile service. In Proceedings of the 7th ACM conference on designing interactive systems. 2008. 315-322.

175