Mapping and Inventory of Forest Fires in Andhra Pradesh, India: Current Status and Conservation Needs

Analyzing the spatial extent and distribution of forest fires is essential for sustainable forest management. The present study appraises the distribution of forest fires in one of the largest states in India, Andhra Pradesh, using satellite remote sensing. Advanced Wide Field Sensor (AWiFS) onboard on Indian Remote Sensing Satellite (IRS P6) was used for mapping and analyzing the spatial extent of burnt areas. Comparative analysis was carried out with respect to different forest types, protected areas and across elevation zones to demarcate and identify the fire-affected areas. The results show that about 19% (8594 km2) of forest area were burnt in the state during 2009. Burnt area statistics for Protected Areas reveal that 24% of forest cover was affected by fire. Nagarjunasagar Srisailam Tiger Reserve, the largest tiger reserve of the country, shows an area of 793 km2 (22%) under forest fire. Higher elevation areas which are predominantly dominated by savannah and woodlands experienced higher fire occurrence in comparison with lower elevation areas. Similarly, fires were prevalent near edges compared to core forest. Results of the study suggested that forests of Andhra Pradesh are prone to high fire occurrences and current fire regime poses a severe conservation threat to biodiversity both within and outside the Protected Areas.


Introduction
Fires affect the species composition, structure, pattern of vegetation, biomass, and consequently cause significant damage to the tropical forests [1,2].Globally, more than 350 million ha of forests were burned in 2000, an area of nearly equal to the size of India [3].In many tropical deciduous ecosystems, fire is part of the natural regeneration process, stimulating the germination of certain species (Teak) and clearing space for the invasion [4].Increased burn frequencies can therefore progressively weed out more fire-vulnerable components of biodiversity.Frequent fires may cause mortality of smaller trees, opening up the canopy and leading to the gradual conversion of once-closed forests to woodland savannahs with low tree density and a grassy understory [5].
In India, large areas of tropical dry forests have already been converted to grasslands or other land use as a result of fire and other practices [6,7].According to the Ministry of Environment and Forests, Government of India, 3.73 Mha of forests are affected by fires annually in India [8].The extent of these fires depends on various factors such as the frequency of human disturbances and the climate of the region [9].Forest dependent communities burn forests for shifting cultivation or to promote the growth of fresh fodder for grazing, and to facilitate the collection of non timber forest products [10].Beddome [11] described the effects of increased burn frequencies under slash-and-burn cultivation in the states of Andhra Pradesh and Orissa.These forests have a wretched and stunted appearance, are very dry, and more or less impenetrable from a tangled rank undergrowth, and there are no seedlings, nothing, in fact, but the coppice growth generally of not only the quicker growing but also poorer sorts of timber.Brandis [12] observation from Indian forests suggested that the dominance of certain thick-barked species, or species that are able to resprout vegetatively in fire affected forests.
There is no comprehensive data available on different dimensions of fire on a regular basis in India [8].There is a requirement of national level precise burnt area assessment using high-and moderate-resolution satellite data in addressing the various issues like reliable assessment in terms of area, number of patches and spatial spread, base line information for temporal monitoring and fire recurrence assessment, identification and prioritization of vulnerable areas, understanding the impacts on regeneration, succession, biodiversity and wildlife, assessment in terms of Green House Gas Emissions towards national communication, scope for developing web enabled national forest burnt area reporting system, and fire mitigation planning [1][2][3][4][5][6][7][8][9][10][11][12][13].
Technological advancement of satellite remote sensing both in spatial and spectral sensors is a cost-and time-effective solution in identifying areas for biodiversity conservation.Currently several satellite-based sensors like AWiFS, LISS-III, MODIS, ETM + , SPOT, AATSR, AVHRR, and MODIS provide synergistic datasets that have potential in forest fire detection, inventory, mapping, and damage assessment [1].Forest burnt areas have a typical spectral signature, especially if analysed by a multitemporal approach because of the different ground coverage between prefire (vegetation) and postfire (white ash, black ash, bare soil, and dead vegetation) conditions [14].
Forest fires, invasive species, and shifting cultivation are recognized as serious threat to biodiversity of many Protected Areas (PAs).Recent findings from the PAs of the tropics confirm that Lantana may be favored by fire and grazing [15].A critical issue is that without a clear understanding of the distribution and dynamics of forest fires, it is impossible to effectively manage them [16].With the current trend of increasing rate of forest fires, there is an urgent need to generate a spatial database for planning, decision making, and further objective-oriented requirements.With respect to the burnt area extent and forest-type detailed information is needed in order to assess postfire ecological effects on the biodiversity.
Among the 10 biogeographic zones of India, Deccan Peninsula was found to be prone to the highest fire frequency.Among the states of Deccan Peninsula, Andhra Pradesh, Chhattisgarh, and Madhya Pradesh are regularly affected with human-induced fires.Fire occurrence information as per AATSR satellite shows maximum fires during March and April in Deccan Peninsula [17].The estimated forest burnt area of Andhra Pradesh state was 6369 km 2 during 2000 [18], which is based on coarse resolution (1.1 km 2 ) SPOT-VGT satellite data.In view of scanty research work on forest fires in Andhra Pradesh, the present study was focused on assessment and inventory of forest fires covering number of fires, burnt area size, annual area burned, and distribution across forest types using IRS P6 AWiFS data for the 2009.In the present study, an attempt was also made to generate the information on forest burnt areas across the districts, Protected Areas, and elevation zones of Andhra Pradesh.There are 27 Protected Areas declared for conservation in Andhra Pradesh, comprising 22 wildlife sanctuaries and 5 national parks.The total area protected for biodiversity is about 12,555 km 2 or 4.56% of geographical area of Andhra Pradesh.Nagarjunasagar Srisailam Tiger Reserve is largest tiger reserve of the country, it occupies 28% of the total Protected Area of Andhra Pradesh [19]. of digital number to Top-of-Atmosphere (TOA) reflectance as suggested by Chavez [21] was followed.The raw digital data were enhanced using contrast stretching and ratiobased techniques to facilitate better discrimination of forest burnt areas and other land cover.The four spectral bands of AWiFS sensor can be used for a different observational task during and after a fire incidence.SWIR, NIR, and Red spectral bands were assigned to respective red, green, and blue gun to highlight source point of active fire (Figure 3).More specifically, fire plumes and burnt areas can be better distinguished in the SWIR spectral channel for detecting high temperature targets.Since 1.65 µm in the electromagnetic spectrum is very sensitive to flame and flaming energy and not very sensitive to smoldering and its energy, FCC with the SWIR, NIR, and R combination highlights the active fire pixels.SWIR channel is considered more suitable for identifying hotspots, since the maximum Planck black body radiance shifts to shorter wavelengths as temperature increases.The usefulness of the SWIR channel has long been recognized for monitoring forest fires [1,13].analysis [21].IRS P6 AWiFS Imagery was classified digitally by the technique of supervised classification using maximum likelihood classifier, with appropriate signatures/training sets for mapping forest burnt areas and land cover categories.The spectral separability and contingency analysis of training sites indicated that burnt pixels were distinct, and more than 96% of the pixels in the training sites were correctly classified.All the nonforest classes, namely, scrub, agriculture, settlement, barren, fallow, and water bodies were masked out using vegetation-type map.Finally, burnt pixels overlaid on different forest types to assess area under forest fire.The spatial information generated on forest burnt areas was utilized to understand the distribution of fires across the districts, Protected Areas.Fire occurrence analysis was carried out to understand the fire frequency from the forest edge to interior and disturbance sources.The Aster global DEM was used to find out distribution of forest fire across elevation zones [22].Testing samples have been selected from the reference field data for assessing the classification accuracy.

Results and Discussion
IRS P6 AWiFS False Colour Composite image of Andhra Pradesh was shown in Figure 2. The forest types found in Andhra Pradesh are semievergreen, moist deciduous, dry deciduous, dry evergreen, thorn, teak mixed, bamboo mixed, riverine forest, mangrove, and savannah.Most abundant forest type was dry deciduous forest which comprises 32524 km 2 of the total forest area, followed by moist deciduous forest of 4865.9 km 2 (Table 1).The forest fires in Andhra Pradesh are mostly ground fires which are usually confined to the vegetation of the lower canopy destroying shrubs and herbs and also scorch the tree bases.Surface fires are also frequent and associated with ground fires, which burn the shrubs and trees of the middle canopy.The spatial analysis of fire occurrence and extent of forest fires in 2009 was presented below.more burnt area has been accounted in moist deciduous forests as compared to dry deciduous forests.The study also found dry deciduous forests are more vulnerable to fires than thorn forests.The reason could be thorn forests are receiving very low rainfall and hence insignificant amount of combustible matter is available.About 11.7% of area of semi evergreen forests are also undergone fire indicates conservation threat to high biologically richness areas of Andhra Pradesh.Dry evergreen forests which are distributed in Krishna and Guntur are less affected due to fire.Riverine (riparian) forests are also affected by fire in summer due to drying of streams.

District
Wise Forest Burnt Area.Administratively, Andhra Pradesh has 23 districts which were grouped into three zones: (1) Coastal Andhra, (2) Rayalaseema, (3) Telangana.On the whole, forest burnt area in Telangana region was very high (4637.7 km 2 ), followed by Coastal Andhra (2502.1 km 2 ) and Rayalaseema (1453.7 km 2 ).As compared to Rayalaseema region, Telangana receives moderately high rainfall, but represents more forest fires reveals severe anthropogenic pressure (Figure 5).Among the 23 districts, Adilabad, Khammam, Warangal, Kurnool, Visakhapatnam, Vizianagaram, and Prakasham were evidenced with largescale fires (   3).Prescribed fire is an important tool for managing habitats and the wildlife populations.Tender shoots from resprouting saplings and shrubs are highly nutritious and attract many herbivores.But uncontrolled fires in tiger reserve and other protected areas can have extreme effects on native biodiversity and may provide an opportunity to invasive alien species to colonise.

Elevation Zone
Wise Forest Burnt Area.Forest fire occurrence was analysed across elevation (altitudinal) gradient (Table 4).Forest fires were very high in elevation range of above 1200 m, which mostly represents savannah/woodland ecosystems.The low elevation (<200 m) forests of Chittoor, Nellore, Krishna, Guntur, and Srikakulam were affected by less fire occurrence as compared to elevation zone of 200-600 m.The areas with southern aspect and high slope (>30%) were found to be more sensitive for fire spread.

Forest Burnt Area
Patches.The numbers of burnt area patches are found to be 9592.Forest burnt area patches of different size categories were analysed (Table 5).Patch size analysis of forest burnt areas revealed the information that the maximum number (6601) of burnt area patches are under 25 ha.Interestingly the burnt area class of >400 ha shows 5545.3 km 2 (64.5%) of total burnt area followed by <25 ha class with representation of 741.1 km 2 (8.6%) of area.3.6.Correlation of Fire Frequency to Forest Edge and Settlements.GIS analysis of fire frequency from the forest edge to interior and disturbance sources (settlements) reveals fires were prevalent in edge forests as compared to core forests.Fire frequency was very high at a distance of 500 m from forest edge (Figure 8) which fell down to nearly zero percentage of burnt area by 4000 m.This could be attributed to man made activities outside their settlement areas, and in general anthropogenic origin forest fire in the state.Using random statistics techniques, overall classification accuracy and kappa statistics were computed.The result shows an overall accuracy of classified map is 92.1% and kappa statistics is 0.91.

Conclusion
Assessment of forest fire using satellite remote sensing indicates that about one fifth of the state's forest was under burnt area in the year of 2009.Deccan plateau and Eastern Ghats were the most affected phytogeographic regions in the state.Smaller protected areas were less vulnerable in comparison with larger ones attributed to the effective management existing with the small areas.Higher elevation savannahs and woodlands experienced prevalent forest fire in comparison with lower elevation areas.The study developed a forest burnt area map using remote sensing data for the conservation planning of forests in general and protected areas in particular.Spatial information generated in the present work is useful for effective forest management and ecological damage assessment.

Figure 4 :Figure 5 :
Figure 4: Forest burnt areas overlaid on forest-type map of Andhra Pradesh.

Figure 8 :
Figure 8: Pattern of fire frequency in relation to distance from forest edges to interior.
Data.Indian Remote Sensing Satellite P6 (ResourseSat-1) AWiFS sensor has spatial resolution of 56 m and radiometric resolution of 10-bit.AWiFS operates It has a swath width of 740 km and a revisit period of five days.Cloud free IRS-P6 AWiFS data pertaining to January, February, March, and April 2009 were selected for the study.AWiFS data were geometrically rectified with an average of less than one pixel accuracy using ERDAS Imagine 9.3.Orthorectified Landsat ETM+ datasets were downloaded from Global Land Cover Facility for reference [20].To reduce the error due to various atmospheric conditions at different dates of image acquisition conversion Code PA (wildlife Sanctuary/National Park) Figure6: Forest burnt areas overlaid on Protected Area map of Andhra Pradesh.

Table 1 :
Areal extent of forest types and proportion of burnt area in each forest type of Andhra Pradesh, India.

Table 1
undergone a large-scale forest fire which was accounted as 34.6%, followed by bamboo mixed forest (32.2%) and teak mixed forest (29.2%).But, the highest spatial extent of forest fires was noticed in dry deciduous forests and moist deciduous forests.Presently the moist deciduous forests in northern parts of Andhra Pradesh are under threat due to slash and burn agriculture.So, proportionately slightly

Table 2
3.3.Protected Area WiseForest Burnt Area.Taking into account the wildlife sanctuaries and national parks, the total area affected by forest fires during 2009 are 2680.5 km 2 (23.6%).Of the 11 Protected Areas in Eastern Ghats, Papikonda wildlife sanctuary and Gundla Brahmeswaram wildlife sanctuary were affected 49.4% and 41.7% of

Table 2 :
Region-and district-wise forest burnt area statistics of Andhra Pradesh: 2009.

Table 3 :
Areal extent of forest burnt areas in Protected Areas of Andhra Pradesh.

Table 4 :
Areal extent of forest burnt area across elevation zones.

Table 5 :
Size class distribution of forest burnt areas in Andhra Pradesh.