Forest Fire Susceptibility Mapping Using Analytical Hierarchy Process Approach for Chandoli National Park of Maharashtra, India


 The forest fire management starts with identifying fire potential areas. This study suggests a new approach based on Geographic Information Systems (GIS), Remote Sensing (RS), and Analytical Hierarchy Process (AHP) to map forest fire susceptibility of Chandoli National Park of Maharashtra. Influencing factors like past burnt area, Land Surface Temperature (LST), forest type, agriculture area, road network, and the vicinity of settlements taken into consideration. All variables assigned a weight based on their impact on the fire susceptibility. The final map categorized, ranging from very high to very low, into five fire susceptible regions. The result indicated that 12.99% (6506.89 ha) land area of the study region came in a very high susceptible region, while 18.70% (9361.10 ha) of high. The forest fire susceptibility map shows that 12.26%, 44.42%, 11.61% area comes under moderate, low, and very low- susceptible areas, respectively. Afterward, an accuracy assessment carried out with existing records of the forest fire. The mostly very high and high susceptible forest fire region comes where high road density, settlements and agriculture fields dominate. The result reveals that the anthropogenic factors and its activities in the forest region responsible for the frequent forest fire. The unification of remote sensing and the Analytical Hierarchy Process into GIS is beneficial to determining forested areas with high fire susceptibility and also to plan forest management after a forest fire.


INTRODUCTION
Forest fires are one of the significant natural and human-made hazards which evince effects on species of biodiversity include wildlife, environment, and also on humans (Buyyani et al. 2014). It increases the intensity of suppression in the forest, which is one of the richest According to FAO (Food and Agriculture Organization, 1986), 'Forest Fire Risk' is the chance of a fire starting as determined by the presence and activity of any causative agent.
The entire forest fire susceptibility depends on the proportion of available fuel and its sensitivity to burning and on the presence of anthropogenic and natural causes leading to fire inflammation.
The natural parameter includes topographic factors (Slope, elevation, and Aspect), biologic factors (vegetation type and density), and the climatic factors (temperature, relative humidity, precipitation, and wind velocity). Human related activities in the forest areas are the anthropogenic factors such as agricultural activities, recreational activities, the existence of forest roads, and settlements. In India, 2-3 % of the forest area affected annually by the fire, and on average, over 34000 ha forest areas are burnt by fire every year (Thakur et al. 2014). To understand the forest fire effects and ecosystem response is a challenge to researchers and management (Eskandari. 2018). it is not possible to control nature; still, it is possible to map the forest fire susceptibility to reduce the frequency of forest fire. Forest fire zones are locations where the forest fire is likely to start, and from where it easily be spared to other areas. An accurate assessment of forest fire problems and decide on solutions can only be satisfied when a  After the sanctuary area became National Park, it added ten sq. Km, and now the park has 318.86sq. Km core area and buffers with 181.72 sq km. The total area is 500.58 sq.km.

RESEARCH METHODOLOGY
For the present research, the following methodology has been used (Fig.2).
In the present investigation Analytical hierarchy process (AHP) method has been successfully applied for mapping forest fire susceptibility in the study area. The first hierarchy structure defined with variables (Past Burnt Area, Vegetation Types, LST, Road Network, Settlements and Agricultural Area). It uses the multi-level hierarchical structure of criteria and sub-criteria. The AHP method is an ordinal pairwise comparison of all useful criteria in a unique objective (Saaty. 2000). According to the importance given rate to a variable based on a ninepoint scale (Table.1 (2)}.A random index extracted from the mean RI table (Table.2) based on the number of criteria Where, max = Maximum eigenvalue of comparison matrix N= Number of variables (6 in the study).

GIS layers (Factors affecting forest fire Occurrence)
For the forest fire susceptibility mapping last five-year past burn area(PBA), Settlements, Road Network, Forest type, Agriculture area, and Land surface temperature (LST), these map layers were generated.

Past Burn Area:
One of the most famous indices used for mapping burn area is the normalized burn ratio (NBR), which combines the reflectance in the NIR and SWIR bands (See Eq.3). NBR is the most sensitive index to emphasize burned areas and to determine the severity of the burns.
The temporal difference between pre-fire and post-fire NBR to produce an index is the Differenced Normalized Burn Ratio (dNBR) (See Eq.4).

dNBR = NBR pre -NBR post
In the present study, pre-fire and post-fire NBR of the last five years (from 2014 to 2018) calculated from Landsat 8 (OLI/TIRS) images(as per data availability) in the peak period of a forest fire, and then pre/post NBR ratios differenced to the mapping burn area. A total of 8.16 sq.km area burned in five years (Fig.3).

Settlements:
The forest fire and settlements have a positive relationship. In the settlement, lots of human activities can cause burning, which can spread a forest fire & cause a lot of havoc. Despite the natural fires, anthropogenic is a significant source of forest fire in India. About 95 percent of forest fires are human-made (Biswadip et al. 2010). The visual digitization process extracted settlements in the study area on Google image in ArcGIS (Fig: 4).

Road Network :
Similar to the village, the road also has enough human activities and vehicle movements, although less than communities. Any physical activity by man or vehicle on the road could cause an unwanted fire. Thus proximity to the road plays a vital role in the chance of fire (Saklani 2008). The road network also digitized by google image in ArcGIS and generated the map (Fig.5).

Forest Type
The grass and dry vegetation are more susceptible to fire in comparison to moist and evergreen ones. The moisture content of vegetation delays ignition ( Saklani 2008).Open forest with grass this vegetation class generated by a supervised classification technique, which is a vulnerable area to wildfire (Fig.6). For the generation of vegetation map, a combination of field knowledge, expert opinion, supervised classification, and other information obtained from forest departments was utilized.

Agriculture Area
The human interference frequency leads the forest fire; due to agriculture expansion, loss of habitat of wildlife has seen, and as a result, human-animal conflicts have also increased. The seasonality of cropping pattern, monitoring burning drive to LST and indirect forest fires. The agricultural field also visually digitized on Google image (Fig.7).

Land Surface Temperature (LST)
The land surface temperature plays a very crucial role in a forest fire; in fact, LST is a driver of wildfire. By using the thermal band of Landsat imageries (Band 10 & 11), Land surface temperature was calculated (Fig.8). In the first step, the Digital numbers were converted to spectral radiance using radiance rescaling factors provided in the metadata file, by the following formula (See Eq.5).

L(λ)= gain*DN +offset
The second step is to transform the spectral radiance to at-satellite brightness temperature using the thermal constant, by the following formula (See Eq.6). The last step is the calculation of land surface temperature (See Eq.7).
Land Surface Emissivity 0.004+ProVeg+0.986 Where the proportion of vegetation (Pv) can be calculated as (See Eq.9).

AHP for Forest Fire Susceptibility Mapping
The mathematical AHP approach used to prepare the map of forest fire susceptibility of Chandoli

ACCURACY ASSESSMENT OF THE FIRE RISK MAP
The results of overlaying the map of the actual fire (Burn area in 2019) on the forest fire susceptibility map in Chandoli National Park showed that the burned regions in the study area were highly accordant with the high and very high-hazard classes in the fire susceptibility map (Fig.10). Total 80% of the burned area located in the high and very high-susceptible classes in the fire susceptibility map. This may be related to the reciprocal relations of the essential active factors in the confines of the actual fire (Eskandari. 2018).

CONCLUSION
This research focuses on the forest fire susceptibility mapping using GIS and remote sensing in Chandoli National Park. However, this approach shows acceptable accuracy for forest fire susceptibility mapping. The results of this present study revealed that about 12.99% of the study area (6506.89 ha) a very high-susceptible region, and 18.70% (9361.10 ha) under a high susceptible region of fire occurrence. It demonstrates that the Chandoli National Park forest exposed to future fires. Prediction of the future fires in the Chandoli National Park forest will be possible using the fire susceptibility map obtained from this study. Therefore, the prevention measures of the future fires occurrence (establishment of scout towers, allocation of the required facilities for firefighting, etc.) should be performed in the high-susceptible area. On behalf of the study, it suggests to the forest department for fire management in the future to enhance the existing conventional fire control system (Kanga. 2017). The integration of the satellite data into GIS is beneficial to determine risky places and to plan forestry management after the fire.