Forest Fire Risk Assessment using Fuzzy Analytic Hierarchy Process

The goal of present investigation was to generate forest fire risk zones in Solan district of Himachal Pradesh. The methodology applied was based on Fuzzy Analytic Hierarchy Process (FAHP) technique which involved socio-economic and bio-physical factors for risk assessment. Risk factors were selected on the bases of occurrence of forest fire in the area during past few years. Results revealed highest weight for fuel type (0.3109) followed by aspect (0.2487), agricultural workers (0.1341), nutritional density (0.1244), population density (0.0622), slope (0.0524),elevation (0.0311), literacy rate (0.0207) and distance from road (0.0155), respectively. Out of total geographical area, 4.15% area was classified under very high risk, while 40.63% and 54.00% area was under high and moderate risk, respectively. Area under low risk (0.84%) and very low risk (0.37%) were extremely less. The results were in agreement with actual fire occurrences in the area. Current World Environment Journal Website: www.cwejournal.org ISSN: 0973-4929, Vol. 13, No. (3) 2018, Pg. 307-316


Introduction
Forest fires are extensive and critical facet of the world. The annual global area burned due to forest fire ranges from 300 and 450 Mha. 1 Over 80 percent of the global area burned occurs in grasslands and savannahs, primarily in South Asia, Africa, Australia and South America. Globally fires are frequent over most of the earth except in areas of scant vegetation and near the poles. 2 India witnesses most of severe forest fires during the summer season in the hills of Himachal Pradesh. 3 Quercus leucotric hophora, Acacia catechu, bamboos and other broad leaved tree species. Average daily mean temperature, relative humidity and annual rainfall were 18.4 °C, 1038.2 mm and 51.2 %, respectively.

Materials and Methods
In this investigation Saaty's (1998) 7 Fuzzy Analytic Hierarchy Process (FAHP) was used. FAHP is Multicriteria Decision Making methodology which involves decision-making framework to rank and prioritize the forest fire risk factors. Table 1 summarizes the related work done over the world.

Hierarchical Structure Development of Fire Risk Criteria
We used population density (PD), agricultural workers (AGRI-W), literacy rate (LR) nutritional density (ND), distance from road (DR), fuel type (FT), aspect (A), slope (S) and elevation (E) for evaluating the fire risk in the study area (Fig. 2).Fuzzy Analytic Hierarchy model was followed in order to construct the hierarchical structure, for reckoning fire risk (Fig. 2).
Relevant socio-economic data for sub-districts of Solan were collected from District Census Handbook. Road maps, Terrain maps and fuel type maps were generated using Shuttle RADAR Topographic Mission (90m), GLOBE COVER (300m) and GLCF, respectively.

Fig. 1
Forest fires have caused extensive damage in recent years leading to loss of wildlife habitat and biodiversity, change in micro-climate, adverse effect on livelihood of people, addition of greenhouse gases etc. Average estimated loss due to forest fire in Himachal Pradesh is INR 113 million per annum. 4 The forests of Himachal Pradesh are mainly comprised of Chir, Oak, Deodara, Khair, Saal, Bamboo and other broad-leaved tree species. Out of above species area occupied by Chir is highly prone to forest fires due to shedding of highly inflammable chir needles. 5 The forests of the Solan district are occupied by pure and mixed stands of chir pine and mostly conform to lower Shiwalik chir pine (9C 1 a) forest type and covers 7.68 per cent of total area of district. [5][6] There was need to generate forest fire risk zone for the study area in order to carry out prevention and management measures.
Common practice of Forest Fire Risk Zones has been delineated by assigning knowledge base weights to the risk factor classes according to their sensitivity to fire. Fuzzy Analytic Hierarchy Process (FAHP) has been used as multi-criteria decision analysis (MCDA) tool for weight estimation. [7][8][9]

Study Area
The study was carried out in Solan district of Himachal Pradesh, India. Solan occupied 10 percent area of the state i.e. 1,93,600 hectares. The area was primarily occupied by Pinus roxburghii,    (Table 2).

Fuzzy Analytic Hierarchy Process (FAHP)
FAHP was used for determining weights for the parameters. A judgmental pair wise comparison matrix 'A', was formed using the comparison scales (Table 3). Each entry a ij of the matrix 'A' was formed comparing the row element a i with the column element a j . 29 A = (a ij ) (i,j …n = 1,2…n; n= number of criteria) The entries a ij in matrix 'A' were done following rules given below: Standardized matrix 'W' was formed by using following equation: Final weights were derived by taking row average of matrix 'W'.

Consistency of comparisons
The value of ʎ max was required to calculate the consistency ratio (CR). 24 Consistency index (CI) = (ʎ max -n) / (n-1) Where, ʎ max = largest eigen value and n = number of criteria The final consistency ratio was calculated by dividing the consistency index with the random index CR = CI / RI Where, RI = Random index and CI = Consistency index Consistency ratio was designed such a way that shows a reasonable level of consistency in the pair wise comparisons if CR < 0.10 and CR ≥ 0.10 indicated inconsistent.
The resulting weights from Fuzzy Analytic Hierarchy Process were applied in the Cumulative Forest Fire Risk Index model. Table 5 (Fig. 4a).Accuracy of the Forest Fire Risk map was tested using NASA FIRMS forest fire dataset for the year 2018 (Fig. 4b). The Forest Fire Risk map for the three classes alone viz. moderate, high and very high predicted 99.4% of the total fire pixels (1012).The moderate class predictive capability was highest (60.77%), followed by high (33.99%) and very high (4.64%) fire risk class.