Modelling forest fire susceptibility in response to changing climatic scenarios in Indian western Himalaya

Sunil Kumar^1,2, and Amit Kumar^1,2*

*Corresponding author e-mail: amitkr@ihbt.res.in

¹Environmental Technology Division, CSIR-Institute of Himalayan Bioresource Technology, Palampur Himachal Pradesh − 176 061, India

² Academy of Scientific and Innovative Research (AcSIR), Ghaziabad, 201002, India.

Highlights

• The RCP2.6 was found as the most susceptible scenario for the forest fires among all climatic change scenarios (such as RCP2.6, RCP4.5, RCP6.0, and RCP8.5) during the year 2041 to 2060, while during the year 2061 to 2080, RCP6.0 was observed most susceptible in the Indian western Himalayan forests.

• In all the scenarios, the model predicted shifting of fire susceptible towards higher elevation and longitudinal region, but a small shift was also observed towards lower latitudinal ranges.

• Forest fires were found to be significantly dependent upon the precipitation and temperature of this region.

Abstract

Identifying the spatial and temporal attributes which are favouring forest fire susceptibility necessary for biological conservation. The adverse effects of climate change on the forest has increased wildfire. The rise in global temperature and alteration of rainfall patterns have produced appropriate conditions for forest fires. A non-parametric ‘Random Forest Algorithm’ for modelling the spatial distribution of forest fires was applied to predict the susceptibility of Indian western Himalayan forest due to fires. The forest fire susceptibility was simulated in the present (years 1970–2000) and future (years 2041–2060 and 2061–2080) environmental gradients. The real-time distribution of the fire susceptibility was evaluated and modelled using forest fire history data with an overall accuracy of more than 0.9. To derive the fire susceptible region in future, we have applied the model statistics of the present time to the future climatic scenario. The magnitude of increase of fires was predicted relatively more along longitudinal and elevational gradient as compared to the latitude. The high sensitive forest fires susceptible area was found as 35376.18 sqkm in the present conditions, while it occupied 61440.03 sqkm, 57181.76 sqkm, 57662.82 sqkm and 56612.11 sqkm respectively in 2041–2060 in the four projected climatic scenarios Representative Concentration Pathways (i.e., RCP2.6, RCP4.5, RCP6.0 and RCP8.5). During 2061–2080, a decline in RCP2.6 and RCP4.5 (56241.95 sqkm and 56668.29 sqkm) and an increase in RCP6.0 and RCP8.5 (61199.50 sqkm and 57510.15 sqkm) were predicted. The results clearly show the fire susceptible area will be higher in the RCP2.6 for the year 2041–2060 and RCP6.0 in 2061–2080. The current study thus provides scientific conclusions that the forest fire susceptibility is climate driven in the western Himalayas.

Keywords:

Forest fires

RCP

Random forest

Susceptibility model

Climate change

Indian western Himalaya

Remote sensing

1. Introduction:

Forest fires are one of the significant drivers of forest ecosystems. Forest fires cause biodiversity, economic loss, and biochemical changes. Several studies have been carried out for forest fire events concerning climatic factors like air temperature, precipitation, etc., and topographical parameters using models available for predicting current and future climatic conditions. The Himalayas, the youngest mountain ranges are the most susceptible to forest fires in the world. The western Himalayan forest is experienced frequent forest fires, hence high susceptible to fires incidences as compared with Eastern Himalayan forests which later gain high rain density. The occurrence and intensity of the forest fires have increased in recent decades in the Himalayas with the spreading out of Pinus roxburghii (Chir Pine) forests in a wide range. Mitigation measures to prevent fire incidences are usually seasonal, which starts in the dry period and by such ample safety measures, the fires can be prevented. The hot summer usually overlaps with fire season which extends from January to May in India (Bahuguna and Singh, 2002) and there are high risks due to high temperatures of spring and summer(Westerling et al., 2006). Forest fires are noticed to increase at the rate of 0.2Mha/yr (+ 2.5%/yr ) in Southeast Asia from since 1997(Giglio et al., 2013). In the last few years, the relationship of forest fires with meteorological variables (such as air temperature, precipitation, relative humidity, wind speed) has been studied by several researchers globally(Bedia et al., 2012; Krawchuk et al., 2009; San-Miguel-Ayanz et al., 2013). They have also proposed the future status of forest fires concerning changing climatic scenarios (Flannigan et al., 2005, 2000). It was observed that the higher the combustible material and air temperature, the higher will be the risk. Globally about 3–4% earth surface effects by forest fires every year. About 37,300 sqkm forests are suffered from forest fire annually in India(Chandra et al., 2015). Emission produces from fires impacts regional as well as global air quality and rainfall patterns. Also, it contributes to global warming by releasing of greenhouse gases (mainly carbon dioxide and soot), which is recognized as SLCP(Short-Lived Climate Pollutant) leading towards land degradation globally through several complex processes(Martin, 2019). Forest fire changes the structure and composition of vegetation communities(Kittur et al., 2014) enabling invasion of fire adaptive exotic species and removal of non-fire-resistant species. Long-term projected trends of weather scenarios have also been studied in the Himalaya (Choudhary et al., 2018; Thibeault et al., 2014) which reported that there has been a significant increase in the rates of the minimum (0.176°C), maximum (0.177°C ) and mean (0.104°C) temperature per decades respectively during 1901–2014 and decline in precipitation(Dimri and Dash, 2012; Ren et al., 2017). Also, about 1.6°C rises in temperature in the last century have been observed with warming in winters more rapidly in the north-western Himalayas (Bhutiyani et al., 2010; Dimri and Dash, 2012). Mean emissions due to wildfire was estimated as 1.5PgCyr^− 1 (Van Der Werf et al., 2009). An increase in global warming and a change in precipitation trends have made forest fires more vulnerable in the Himalayan region. The chapter on ecosystem impact of IPCC AR4 has also projected fire increases (Fischlin et al., 2007) through a model of vegetation dynamics which was driven by projections from various global climate models (GCMs) (Scholze et al., 2006). The climate of a region imparts plays a significant role in regulating forest fire (Harrison et al., 2010). The native communities are dependent on the local forest for their livelihood as they collect fodder, fuel wood, and several non-timber products. The change in trends in forest fires strongly affects these communities and thus spatial distribution modelling is required to improve local fire prevention action (Chen et al., 2015). Some studies was carried out for forest fires in future climatic conditions with regards to potential effects in fire regimes on Pacific Northwest forests its disturbance and stress, structure and composition and ecological processes (Halofsky, J.E. 2020). These models use a set of local observations to predict fire risk as a function of external explanatory variables (Chuvieco, 2003). The non-parametric models such as the random forest (RF) machine learning technique is superior to the above models for the fire susceptibility modelling. It is perfect for modelling ecological systems over traditional methodologies such as GLM (generalized linear models). A unique advantage of the modern machine learning technique has to discover compound associations and spatial patterns in an enhanced way than the conventional possibility data model which assumes familiarity in the data distribution (Evans et al., 2011). The sensitivity of random forest classifier is less to the over fitting and quality of training samples than other streamlined machine learning classifiers. This is due to by a random selection of training samples produced large numbers of decision trees (Belgiu and Drăgu, 2016). The methodology of machine learning in the ecological study has been used commonly (Olden et al., 2008). In the present study, a single machine learning model (RF) has more focus on generalization and assessment of four climatic scenarios. Since the occurrence of forest fires in present conditions may be misguiding in future because the climatic conditions after a few decades will be dissimilar from the present conditions. Hence, the statistical models for present fire susceptibility regions have been implemented to the future projection of fire susceptibility distribution in this study. There is a lack of this kind of study in Indian western Himalayan regions which is one of the most floral diversified forest ecosystems in the world due to diverse climatic conditions and topographical conditions and vulnerable to forest fires. The Major forest types of this region are alpine forests, semi-evergreen, deciduous, sub-tropical broad-leaved, sub-tropical pine forests, and sub-tropical montane temperate forests. The present study has been carried out to simulate the current forest fires in suitable regions in the western Himalayas and also to predict its future conditions during 2041–2060 and 2061–2080. The susceptible regions due to forest fires in the current time and future have been modelled and evaluated for the identification and understanding of such regions to support decision making by adopting the best adaptation strategies. We have modelled the fire susceptibility regions to predict current and future climatic conditions in four RCP scenarios i.e. RCP2.6, RCP4.5, RCP6.0, and RCP8.5 for the year 2041–2060 and 2061–2080 during the fires seasons (February to June). Four pathways have been suggested (IPCC Fifth Assessment Report(ARS5),2014) for describing projected future climate labelled as RCP2.6, RCP4.5, RCP6.0, and RCP8.5 having the possibility of a continuous rise in the greenhouse gas emission during the year 2050 and 2070 and future climatic conditions are assumed to be normal during the year 2041–2060 and 2061–2080 respectively. Representative Concentration Pathways (RCPs) are scenarios that describe alternative trajectories for carbon dioxide emissions and the resulting atmospheric concentration from 2000 to 2100. RCP 2.6 which denotes the biggest declines in GHGs, temperatures will likely increase by between 0.3°C and 1.7°C by 2100 and under RCP4.5, temperatures are expected to rise between 1.1°C and 2.6°C. In RCP 6.0 climatic scenario temperature rises 1.3°C to 2.2°C and in RCP8.5 assumes high and growing GHG emissions global temperatures would rise between 2.6°C and 4.8°C by the end of the century.

The main research objective of this study is to understand the fire susceptible regions in present and future climatic scenarios which is needful for conservation and management policies. Previously, these type of studies were carried out using conventional weighted-based methods. The model used in the present study ensembles various set of algorithms which are quite precise on a global or large scale. As the sensitiveness of climatic conditions on Himalayan forest ecosystem, it is important to monitor these resources for policy makers and to maintain or conserve them for the future.

2. Material and methods:

2.1. Study area:

The present study is the Indian western Himalayas which is spread across three Indian states namely Jammu and Kashmir, Ladakh, Himachal Pradesh, and Uttarakhand having geographical extents ranges 72.5°E- 80.9°E longitude to 28.8°N-37.0°N latitude. The total geographical area in Indian western Himalaya is 2,10,561sqkm. The elevation in this region varies from 186 to 8246 meters (Fig. 1). The Major forest types of this region are alpine forests, semi-evergreen, deciduous, sub-tropical broad-leaved, sub-tropical pine forests, and sub-tropical montane temperate forests. The foothills of Indian western Himalaya are having high prone to frequent forest fires particularly during March to May because of favourable prevailing climatic conditions, i.e., high temperature, prolonged dry spell, and suitable forest types prone to fires.

Fig. 1

Historic forest fire occurrence locations depicted in red points from the year 2000–2010 draped over digital elevation model in Indian Western Himalaya. The density plot on the top and right margin shows the relative proportion of elevation ranges along the longitude and latitude, respectively.

2.2. Data:

Standard bioclimatic data (BIO1 to BIO19) (Table 1) and topographical data (DEM, Slope, and aspect) have been used for the forest fire distribution prediction (Table 1). The climatic datasets are available at the ground resolution of 1 sqkm or 30 arc second for present (1970–2000) and future (2041–2060 and 2061–2080) climatic scenarios were downloaded from WorldClim (https://worldclim.org). Since the study area is large (2, 10,561 sq km) for which 1 Km spatial resolution modelling is optimum and has been used in many such cases (Verma. et al 2018). We have used bioclim datasets as base layer for present and future predictions. This is widely used datasets for forest modelling studies. The bioclim datasets which is representative of current weather scenario is for the year 1970 to 2000 only. It is not available up to 2010. It has also been assumed that there are negligible changes in the climatic condition during 2000 to 2010.

The digital elevation model (DEM) was obtained from SRTM 90 m. The slope and aspect maps for the study region were prepared using the terrain function of the raster library in ‘R studio’.

Land use/land cover map is also a major factor for a forest fire in this study area. The anthropogenic factor is responsible for the change in land use/land cover pattern. Also, humans influence the forest fires directly by igniting and controlling fires, and indirectly by modifying the forest structure and composition and bringing changes to the landscape (Bowman et al., 2011). The anthropogenic related activities are influencing forest fires (Guyette et al., 2002). initially burned areas are increases with population density (Bistinas et al., 2013). For modelling of fire distribution, we use Global 1-km downscaled Population Projection Grids for the Shared Socioeconomic Pathway 4(SSP4), which develop based on demographic and socioeconomic assumptions, according to their demand and utilization of resources concerning future scenario (Calvin et al., 2017), at the resolution of 1 km (Gao, 2019) was utilized from Socioeconomic Data and Applications Center(sedac) (https://sedac.ciesin.columbia.edu) and projected population data for 2041–2060 and 2061–2080 and present condition was used in the model.

Table 1
Spatial data used for the suitability modelling of the forest fire (https://worldclim.org).
Code	Variable
BIO1	Annual mean temperature
BIO2	Mean diurnal range [mean of monthly (max temp − min temp)]
BIO3	Isothermality (BIO2/BIO7) (× 100)
BIO4	Temperature seasonality (standard deviation × 100)
BIO5	Max temperature of the warmest month
BIO6	Min temperature of the coldest month
BIO7	Temperature annual range (BIO5–BIO6)
BIO8	Mean temperature of wettest quarter
BIO9	Mean temperature of driest quarter
BIO10	Mean temperature of warmest quarter
BIO11	Mean temperature of coldest quarter
BIO12	Annual precipitation
BIO13	Precipitation of wettest month
BIO14	Precipitation of driest month
BIO15	Precipitation seasonality (coefficient of variation)
BIO16	Precipitation of wettest quarter
BIO17	Precipitation of driest quarter
BIO18	Precipitation of warmest quarter
BIO19	Precipitation of coldest quarter
DEM	Digital elevation model
Slope	The slope in degree unit
Aspect	Aspect in degree unit
Population	Population density

The current scenario for fire distribution modelling was carried out using WorldClim Version2 dataset (Fick and Hijmans, 2017) (Table 1). The common downscaled, bias-corrected global climate model (GCM) dataset for HadGEM2-AO (Cho et al., 2013) was used for 2041–2060 and 2061–2080.

Fire point data from MODIS C6 was downloaded from NASA site (http://Firms.modaps.eosdis.nasa.gov) for the year 2000 to 2010 having geographic information for the fire season. These fire points were overlaid on land use/land cover classes (Roy et al., 2016) other than forest and grassland and were removed from the analysis to avoid the false prediction by the model. Fire point data for February to June is used for forest fire suitability modelling. The combined fire location from 2001 to 2010 of the study region was used for prediction of fire susceptibility for current and future climatic scenarios. There is very little change in the climatic conditions in ten years, thus we assume the climatic condition remain same as 1970–2000 climatic scenario and in the time of forest fire events. Multiple points spatially overlapping each other were removed. Also, multiple collinear points were discarded using Principal Component Analysis (PCA) in ‘R Studio’ to avoid over prediction of the fire distribution models. A total of 652 fire points were selected for forest fire susceptibility modelling.

2.3. Variable selection

Before building a distribution model checking for collinearity in the predictor is necessary (Dormann et al., 2013). On removing highly correlated variables the ability has enhances to understand each variable’s effect on fire probabilities. Pearson's linear model was used to check the collinearity in bioclimatic, topographical, and projected population variables. The ecologically meaningful variables were selected if they correlated less than 0.7 and the Variance Inflation Factor (VIF) was used for the finding of hidden relationship (Guisan et al., 2017). Thus variables having a relationship of less than 0.7 in the person's model and VIF less than 10 (Fig. 2) were used for predict forest fire susceptibility model.

Fig. 2

Pearson's correlation coefficients of a linear relationship in present explanatory variables

Thus the variables were selected for the model are having low correlation and ecologically more significant. Finally, 03 topographic, 05 bioclimatic, and grided population variables were used for forest fire susceptibility modelling (Fig: 3).

Fig. 3

Scatterplot depicting the relationship between the selected explanatory variables used in the model for the current time.

2.4. Prediction of forest fire distribution using the random forest algorithm

For a random forest algorithm, pseudo-absence or background data is necessary for the (Liu et al., 2016), which allows forecasting the possibility of presence. We have used historic fire points data as Presence-background for fire susceptibility modelling and pseudo-absence was pick arbitrarily in the proportion 2:1 for the simulation process to retrieve high accuracy of the Random Forest prediction (Liu et al., 2016). The modelling was done in R Programming languages (R Core Team, 2018) with biomod2 package. However, Biomod 2 is extensively used by the ecologists to predict species distribution, but it can also be used to model other binomial data including gene, markers, and ecosystems in function of its explanatory variables (Thuiller W. et al 2016).

The uncertainty between different predictor variables were minimised following various steps like multiple co-linearity, ensemble modelling, etc. The first removes the redundant variables from the analysis and later assigns weightage to the variables based on their relevance in the analysis. The difference in the resolution of the spatial data were also resembles to similar scales.

The area under the Receiver Operating Characteristics (ROC) curve was applied to evaluate models in distributional modelling (McPherson et al., 2004). True skill statistics (TSS) were applied to the evaluation of model performance during ensemble modelling (bootstrap aggregation) which is an easy and intuitive evaluate of the presentation of distribution models (ALLOUCHE et al., 2006). Like kappa, TSS calculates both commission and omission errors, and success as a result of random suggestion, which ranges from − 1 to + 1, where towards + 1 shows ideal agreement and values towards zero or less shows poor presentation. However, in comparison with kappa, TSS is unaffected by occurrence and also the size of the validation set. A 2*2 confusion matrix was applied with true positive(a), false positive(b), false negative(c), and true negative(d), numbers. The specificity, sensitivity and TSS were calculated using Eq. (1–3).

𝑆𝑒𝑛𝑠𝑖𝑡𝑖𝑣𝑖𝑡𝑦 =

$\:\frac{\varvec{a}}{\varvec{a}+\varvec{c}}$

(1)

𝑆𝑝𝑒𝑐𝑖𝑓𝑖𝑐𝑖𝑡𝑦 =

$\:\:\:\frac{\varvec{d}}{\varvec{b}+\varvec{d}}$

(2)

𝑇𝑆𝑆 = 𝑆𝑒𝑛𝑠𝑖𝑡𝑖𝑣𝑖𝑡𝑦 + 𝑆𝑝𝑒𝑐𝑖𝑓𝑖𝑐𝑖𝑡𝑦 −1 (3)

Training datasets are used for the calculation of the predicted model sensitivity, specificity, and TSS. As the output results are a probabilistic and required to be converted in the terms of high and low fire suitability region. The continuous probability of occurrence was divided into four classes (High, Medium, Low, and Not susceptible) with an equal interval threshold.

3. Results and discussion

3.1. Evaluation and prediction of the model

The sensitivity, specificity, and TSS of the ensemble model in RF were evaluated using the training set predicted model for fire season is 0.969, 0.956, 0.925, and 0.991 respectively which is satisfactory in model prediction for further analysis. The important variables for classifying fire susceptibility in the Indian western Himalayan region were evaluated as elevation followed by precipitation of warmest quarter, population density, slope, precipitation of the driest month, mean temperature of driest quarter, aspect, precipitation seasonality (coefficient of variation), and mean temperature of the wettest quarter. This indicates that the rainfall and temperature in the summer seasons play a major role in the forest fire in the western Himalayas. Among topographical factors, elevation has high importance than slope and aspect. Population density also governs the forest fires because native peoples put forests under fires for good fodder production in monsoon seasons.

3.2 Land Use and land cover

Land use/land cover pattern shows that the deciduous broadleaf forest (80.13%) is highly sensitive to fires followed by evergreen needle leaf forest (76.62%), fallow land (50.5%), plantations (41.75%), mixed forest (29.12%), wasteland (27.12%), shrubland (11.11%), Evergreen broadleaf forests (9.13%) and others 1% (Fig. 4). Field observation shows that mixed forests and other forests at lower elevations possess scattered Pinus roxburghii tree species causing major forest fires.

Fig. 4

Landuse/Landcover statistics in the Indian western Himalaya (Roy et. al.,2015).

Table 2
Forest type area (%) susceptible for forest fires under current climatic conditions in the Indian western Himalaya.
Forest class	High%	Moderate%	Low%
Deciduous Broadleaf Forest	80.14	7.08	0.15
Mixed Forest	29.12	38.57	6.44
Shrub Land	11.11	15.66	5.37
Barren Land	0.04	0.17	0.19
Fallow Land	50.59	45.32	3.57
Wasteland	27.12	4.84	0.57
Plantations	41.76	40.94	10.25
Grassland	0.12	2.92	1.55
Evergreen Broadleaf Forest	9.14	9.66	6.40
Evergreen Needle leaf Forest	76.62	46.62	14.71

3.3 Prediction of forest fire susceptible areas.

Fire susceptible areas during the fire season were showed an increasing trend in all the future climatic scenario (Fig. 5, and 6). The total high fire-sensitive area calculated in present climatic conditions is 35376.18 sqkm and potential susceptible area in the future climatic condition increased to 61440.03 sqkm and 57181.76 respectively in RCP2.6 and RCP4.5 for





	A Figure 5 : Forest Fire Susceptibility map of the western Himalayan region for (a) Current climatic conditions, RCP 2.6 for the year 2041–2060 (b) and 2061–2080, (c) RCP 4.5 for the year 2041–2060 (d) 2061–2080, (e) RCP 6.0 for the year 2041–2060 (f) 2061–2080, (g) RCP 8.5 for the year 2041–2060(h), and 2061–2080(i).

the year 2041–2060. The area was estimated lesser, i.e., 56241.95 sqkm and 56668.29 sqkm respectively during 2061–2080. In the case of RCP6.0 and RCP8.5, a continuous raise in fire susceptible areas was predicted during 2041–2060(57662.82 sqkm and 56612.11 sqkm) and 2061–2080 (61199.50 sqkm and 57510.15 sqkm). Largely, our results show that for the year 2041–2060 and 2061–2080 in RCP2.5 and RCP6.0 climatic scenario is more susceptible to forest fires respectively as compare to that RCP8.5, which is projected for more emission and global warming. This could be due to the projected better precipitation in RCP8.5 that may refuse to fire sensitivity.

Fig: 6: Area statistics for forest fire prediction for present climatic conditions (yellow column) and future RCP's (2041–2060 and 2061–2080)

Fig. 7

Histograms representing the current and projected high forest fire-sensitive areas concerning (a) latitude, (b) longitude, and (c) elevation gradients.

The forest fire susceptibility tends to increase and sifted to the higher elevations, longitude, and latitudes in all the future climatic scenarios (Fig. 7). The mean latitude for the fire susceptibility in the present time is 31.13488°N. In RCP2.6 it will be 31.10391°N during 2041–2060 and 31.0676°N in 2061–2080. Similarly, it is 31.16132°N and 31.16132°N in RCP4.54.5 in the years 2041–2060 and 2061–2080 respectively. The mean latitudes would be 31.06184°N and 31.11324°N in RCP6.0 for the year 2041–2060 and 2060–2080 respectively, and 31.08203°N and 31.11122°N for RCP8.5 for the year 2041–2060 and 2061–2080 respectively. The mean longitude for the fire susceptibility in the present time was observed as 77.13154°E. In RCP2.6 it is 77.35608°E during 2041–2060 and 77.37916°E in 2061–2080. The mean longitude was projected as 77.23744°E and 77.33165°E in RCP4.5 in the year 2041–2060 and 2061–2080 respectively. It is 77.39549°E and 77.34587°E in RCP6.0 for the year 2041–2060 and 2061–2080 respectively. Similarly in 2041–2060 and 2061–2080, it will be 77.35999°E and 77.31041°E for RCP8.5 respectively. The mean elevation for the fire susceptibility in the present time was found as 691.41 m. In RCP2.6 it will be 1035.51 m in 2041–2060 and 982.87 m in 2061–2080. It would be, 962.28 m and 982.60 m in RCP4.5 in the year 2041–2060 and 2061–2080 respectively. Similarly, the mean elevations would be 1014.6 m and 1026.33 m in RCP6.0 for the year 2041–2060 and 2061–2080 respectively. In RCP8.5, these are predicted as 997.17 m and 995.99 m for the year 2041–2060 and 2061–2080 respectively.

Results clearly show a shifting of fire-sensitive regions towards higher latitudes in RCP6.0 and RCP8.5 and declining in RCP 2.5 and RCP4.5 climatic scenarios from the projected the year 2041–2060 to 2061–2080 but, opposite trends were observed for the longitudes. In the case of elevation, it showed increasing trends for all projected climatic scenarios except RCP 2.5.

Fig. 8

(a) mean annual minimum temperature (b) Mean annual maximum temperature, and(c) Mean annual precipitation of the study area for present climatic conditions and future climatic scenarios(RCP2.6, RCP4.5, RCP6.0, and RCP8.5) for the years 2041–2060 and 2061–2080.

The mean annual minimum temperature (Fig. 8(a)) of the study area is -0.72°C which tends to increase in all climatic scenarios as 2.80°C, 3.36°C, 2.61°C, and 3.35°C for RCP2.6, RCP4.5, RCP6.0, and RCP8.5 respectively in the year 2041–2060. But in the year 2061–2080 minimum annual temperature of RCP2.6 and RCP4.5 tends to decline to 2.29°C and 3.26°C, respectively as compared to the year 2041–2060. The mean annual maximum temperature of the present climatic condition was observed as 21.19°C and found to be continuously increasing in all future climatic conditions. In the year 2041–2060 the mean annual maximum temperature (Fig. 8(b)) of RCP2.6, RCP4.5, RCP6.0, and RCP8.5 will be 22.00°C, 22.64°C, 21.83°C, and 23.01°C respectively. In the year 2061–2080, it will remain 30.25°C, 31.30°C, 30.62°C, and 31.91°C respectively. Similarly, the present mean annual precipitation (Fig. 8(c)) of the study region is 1262.80 mm which will increase in RCP2.6, RCP6.0, and RCP8.5 to 1385.19 mm, 1270.94, and 1364.46 respectively for the year 2041–2060 expect RCP4.5 which showed a declining trend to 1248.62 mm and in the year 2061–2080. The predicted annual precipitation tends to increase inclined in RCP4.5(1261.67mm), RCP6.0(1418.05mm), and RCP8.5(1485.06mm) and whereas it appeared to decline in RCP2.5 (1386.52mm).

Fig. 9

Regression showing an increase and decrease fire susceptible areas in four future climatic scenarios for the year 2041–2060 to 2061–2080.

As per the model prediction, the forest fire susceptibility area tends to decline in 2061–2080 as compared to 2041–2060 (Fig: 9) in RCP2.6 with a high rate in higher elevation regions. In RCP4.5 decrease of the susceptible area in lower elevation ranges up to 1700 meters and increases in higher regions have been noticed. The predicted areas for fire susceptibility in RCP6.0 tend to increase at all elevations at higher rates as compared to RCP8.5 in the year 2080 as compared to 2041–2060. These prediction results are influenced due to the rise in temperature and precipitation also.

Table 3
State-wise classified in High, Medium, and Low forest fire susceptible areas in Indian western Himalayas.
High Susceptible Area(sq KM)
	Present	RCP2.6 (2041–2060 )	RCP2.6 (2061–2080 )	RCP4.5 (2041–2060 )	RCP 4.5 (2061–2080 )	RCP6.0 (2041–2060 )	RCP6.0 (2061–2080 )	RCP8.5 (2041–2060 )	RCP8.5 (2061–2080 )
HP	9709.00	20131.23	19205.31	18792.22	18104.50	19874.81	21145.16	18389.42	17593.71
UK	15234.46	27409.47	25053.39	24429.31	25403.62	25828.02	26587.81	25371.69	25783.62
J&K	10432.72	13899.33	11983.25	13960.23	13160.17	11959.99	13466.54	12851.00	14132.82
Medium Susceptible Area(sq KM)
	Present	RCP2.6 (2041–2060 )	RCP2.6 (2061–2080 )	RCP4.5 (2041–2060 )	RCP 4.5 (2061–2080 )	RCP6.0 (2041–2060 )	RCP6.0 (2061–2080 )	RCP8.5 (2041–2060 )	RCP8.5 (2061–2080 )
HP	12307.54	5346.88	5236.70	5304.39	5994.26	4600.83	4295.50	6202.38	6648.13
UK	12949.16	3599.87	5468.58	5832.96	4669.25	4500.74	4466.89	4530.98	4722.53
J&K	5776.79	5157.49	5287.83	3986.57	5831.52	6279.43	4708.85	6036.03	5157.49
Low Susceptible Area(sq KM)
	Present	RCP2.6 (2041–2060 )	RCP2.6 (2061–2080 )	RCP4.5 (2041–2060 )	RCP 4.5 (2061–2080 )	RCP6.0 (2041–2060 )	RCP6.0 (2061–2080 )	RCP8.5 (2041–2060 )	RCP8.5 (2061–2080 )
HP	3730.93	2549.94	2670.20	2449.84	2273.41	2246.05	2208.60	2050.90	3149.80
UK	3019.46	1729.72	1661.31	1568.42	1563.38	1559.78	1400.63	1863.67	1968.08
J&K	3354.31	7233.59	5546.35	5639.25	3098.67	8469.31	5091.24	6252.79	7357.45
Ladakh	2.16	237.64	69.13	61.21	332.69	68.41	259.24	35.29	1012.49

3.4 State wise Forest fire prediction.

State-wise forest fire susceptible areas were calculated in three vulnerability classes as high, medium, and low susceptible areas. At present, the Uttarakhand (UK) possess the highest highly fire susceptible area category followed by Jammu and Kashmir (J&K) and Himachal Pradesh (HP) (Table 4). Also, for the projected the year 2041–2060 and 2061–2080 all future climatic situation RCP2.6, RCP4.5, RCP6.0 & RCP8.5 scenario higher susceptible fire area was estimated in after Uttarakhand (UK) followed by in Himachal Pradesh (HP) and Jammu & Kashmir (J&K). In the case of medium susceptible area category, higher forest fire susceptible areas were noticed in the UK at present, RCP2.6 (2061–2080), RCP4.5(2041–2060) scenario as compared to HP and J&K. Low susceptible area category were observed higher in HP at present condition. This tends to decrease in RCP2.6, RCP4.5, RCP6.0 & RCP8.5, whereas it tends to increase in J&K in RCP2.6, RCP4.5, RCP6.0 & RCP8.5 condition in the year 2041–2060 and 2061–2080.

The predicted fire susceptible area has a high and significant correlation with the rainfall and minimum temperature of that area. Hence, it can be concluded that the forest fire susceptibility of the region is climatic driven.

The active period for a forest fire in the region is February to June (Fire season) due to high temperatures in summer and prolonged drought conditions. From July onwards, forest fire incidents generally decrease due to the arrival of monsoon leading to wet and humid conditions unfavourable for forest fires.

During the field survey, we observed that Chir pine (Pinus roxburghii) forest (most vulnerable due to forest fire) are mixed with Shorea robusta and Acacia Catechu at the lower elevation and mixed with Quercus leucotrichophora at the upper elevation. Most of the fire incidents are noticed between elevation range 400 to 1800 amsl. The litterfall of the chir pine forests which gets started in February and March contains a lot of oil content that is highly inflammable. Therefore, with the rise of ambient temperature, forest fire incidents increase. The study has revealed that at the lower elevation forest continues to face major changes of monoculture plantation of fire-sensitive tree species like Pinus roxburghii and Acacia Catechu in the replacement of Quercus leucotrichophora and other non-fire sensitive broadleaf native species (Shah and Sharma, 2015). This will lead to an increase in fire-sensitive forest areas at lower elevations in the future. Also, our study is in agreement with the observation made by (Chitale and Behera, 2019) that the expansion of the distribution range of fire-sensitive above species and shrinking of non-fire sensitive species (like Quercus spp.) in future climatic scenario (Saran et al., 2010) may also lead to the forest fire sensitivity in higher elevation. The change in climatic condition patterns such as a change in rainfall pattern, and delay in the onset of monsoon, change in phenological pattern, etc., may lead to a shift of the fire season. The change in temperature pattern with low winter season span will also result in the extension of the fire season in the future climate towards upper and lower both the elevation directions. It has been observed that the changing climatic patterns have influenced the weather conditions in the Himalayan region and thus forest fires in these regions have relation with the climate change and weather conditions. The number of the forest fires increases due to climate change increases by 50% by 2011 (UNDP 2022). Results clearly shows the increase in the temperature in all future climatic scenarios which leads the increase in the forest fire events and also extension of the fire prone area. In the future climatic conditions precipitation and minimum temperature play important role for the forest fire events in all RCPs. With the increase in the temperature, change in the pattern of rainfall and also, shifting of the fire sensitive plant species cause widely spread and high intense of forest fires (Borunda A., 2020). A study claim that the forest fire area is doubled in 2050 as compare to the present in USA (Abatzoglou, J.T et al 2021). Increase in the CO₂ level in future climatic scenarios can increase in productivity of the forests (Hickler T. et al 2015) which leads more fuel load results in higher intense fire. A study proves the disappearing of the high altitudinal species which is less adapted to the forest fires (Werner R. et al 2021). It means the fire sensitive forests like chir pine may invade to the higher altitudes due to suitable climatic conditions like higher temperature and that area became more fire sensitive in future.

Table 4
Summary statistics of high fire susceptible areas with temperature and rainfall for present and four future climatic scenarios for the year 2041–2060
	Present Climatic conditions				RCP 2.6(2041–2060 )				RCP 2.6(2061–2080 )				RCP 4.5(2041–2060 )				RCP 4.5(2061–2080 )
Elevation (m)	Area (Sqkm)	Min Temp (°C)	Max Temp (°C)	RF (mm)	Area (Sqkm)	Min Temp (°C)	Max Temp (°C)	RF (mm)	Area (Sqkm)	Min Temp (°C)	Max Temp (°C)		Area (Sqkm)	Min Temp (°C)	Max Temp (°C)	RF (mm)	Area (Sqkm)	Min Temp (°C)	Max Temp (°C)	RF (mm)
Below300	4874.0	6.5	37.2	1318.2	4525.5	8.9	38.8	1530.9	4564.3	8.8	39.5	1600.4	4881.9	9.8	39.9	1442.0	4411.0	9.8	40.9	1443.6
301–500	8309.1	5.4	37.9	1336.4	7873.5	8.4	39.5	1628.6	7712.1	8.2	40.0	1641.9	8451.0	9.1	39.9	1443.1	8273.9	9.2	41.0	1455.5
501–700	7679.7	5.1	36.6	1486.5	8396.3	8.3	38.2	1836.7	8278.2	8.2	38.7	1840.2	8330.0	9.1	38.6	1612.6	8189.6	9.2	39.7	1641.5
701–900	5849.8	4.8	35.0	1503.6	7284.4	8.0	36.5	1840.8	7083.4	7.8	37.0	1842.3	7180.7	8.7	37.0	1624.5	6994.9	8.9	38.1	1664.0
901–1100	3028.3	5.2	33.2	1627.4	6151.6	7.4	34.4	1747.9	5850.5	7.1	35.0	1756.3	6160.9	8.0	35.2	1563.3	5880.1	8.2	36.2	1603.9
1101–1300	2966.3	4.1	32.2	1437.9	5817.4	6.6	32.5	1667.0	5584.8	6.2	33.1	1678.1	5998.9	7.2	33.5	1504.7	5818.1	7.4	34.3	1540.4
1301–1500	2076.9	3.2	31.0	1409.0	5827.5	5.8	30.9	1584.1	5831.1	5.4	31.4	1592.8	5935.5	6.4	31.9	1432.7	5672.7	6.6	32.7	1464.9
1501–1700	561.7	0.9	30.3	1251.6	5438.6	4.3	29.9	1350.7	5324.8	3.8	30.4	1357.4	5101.6	4.9	30.7	1238.6	5017.3	5.1	31.5	1247.8
1701–1900	30.2	0.3	29.0	1319.3	4543.5	3.9	28.6	1415.4	3798.1	3.3	29.1	1416.2	3576.3	4.3	29.3	1286.3	3831.9	4.5	30.1	1298.5
1901–2100	0.0	-0.9	27.8	1341.3	2908.7	3.0	27.6	1426.8	1437.4	2.3	28.0	1418.9	1461.2	3.4	28.1	1285.1	1712.5	3.5	29.0	1297.1
2101–2300	0.0	-3.0	26.9	1265.8	1199.1	1.3	26.8	1316.7	376.6	0.6	27.2	1301.6	51.9	1.6	27.3	1180.9	430.7	1.8	28.1	1189.2
2301–2500	0.0	-4.7	25.8	1197.7	1199.1	-0.3	25.9	1219.4	376.6	-1.1	26.3	1201.6	51.9	0.0	26.2	1093.4	430.7	0.1	27.1	1096.1
2501–2700	0.0	-6.5	24.8	1118.3	249.9	-2.1	25.0	1105.1	23.8	-2.9	25.3	1087.2	0.0	-1.7	25.2	995.4	5.0	-1.6	26.1	992.4
2701–2900	0.0	-8.5	23.9	1002.3	25.2	-4.2	24.1	967.7	0.0	-5.0	24.4	951.6	0.0	-3.7	24.4	878.5	0.0	-3.7	25.3	871.2
2901–3100	0.0	-10.6	23.0	868.3	0.0	-6.2	23.4	831.4	0.0	-7.0	23.7	816.9	0.0	-5.7	23.9	759.3	0.0	-5.7	24.6	750.7
3101–3300	0.0	-12.9	22.2	721.4	0.0	-8.2	22.9	694.0	0.0	-9.1	23.1	681.0	0.0	-7.7	23.4	637.8	0.0	-7.7	24.1	629.9
3301–3500	0.0	-14.8	21.2	621.0	0.0	-9.9	22.0	606.8	0.0	-10.7	22.2	594.6	0.0	-9.3	22.6	558.8	0.0	-9.4	23.3	552.7

Table 5
Summary statistics of high fire susceptible areas with temperature and rainfall for present and four future climatic scenarios for the year 2061–2080 along the elevation gradients.
	RCP6.0(2041–2060 )				RCP6.0(2061–2080 )				RCP8.5(2041–2060 )				RCP8.5(2061–2080 )
Elevation (m)	Area (Sqkm)	Min Temp (°C)	Max Temp (°C)	RF (mm)	Area (Sqkm)	Min Temp (°C)	Max Temp (°C)	RF (mm)	Area (Sqkm)	Min Temp (°C)	Max Temp (°C)	RF (mm)	Area (Sqkm)	Min Temp (°C)	Max Temp (°C)	RF (mm)
Below300	4117.8	9.6	39.3	1366.7	4866.1	10.0	40.2	1552.1	4536.3	9.9	39.8	1551.3	4866.1	11.1	41.1	1751.3
301–500	7950.5	8.9	39.6	1396.3	8337.2	9.5	40.3	1627.4	7869.1	9.5	40.5	1557.5	7962.0	10.8	41.4	1774.4
501–700	8271.7	8.8	38.3	1587.6	8385.5	9.5	39.0	1850.3	8085.2	9.4	39.2	1755.4	7588.3	10.9	40.1	1981.6
701–900	7053.9	8.4	36.7	1625.1	7064.7	9.1	37.3	1867.5	6803.3	9.1	37.5	1777.8	6973.3	10.6	38.5	1987.8
901–1100	5802.3	7.7	34.7	1586.2	5861.3	8.4	35.4	1786.2	5804.5	8.3	35.4	1720.6	6088.9	9.8	36.6	1904.7
1101–1300	5458.8	6.8	32.8	1539.5	5646.0	7.6	33.6	1711.1	5731.0	7.5	33.5	1660.2	6032.7	9.0	34.7	1818.6
1301–1500	5711.6	5.9	31.2	1475.0	5903.8	6.8	32.0	1628.0	5792.2	6.6	31.8	1583.3	5919.0	8.2	33.1	1722.3
1501–1700	5385.3	4.3	30.0	1278.8	5615.8	5.3	30.8	1399.9	5292.4	5.0	30.9	1359.5	5257.1	6.8	32.0	1453.5
1701–1900	4439.0	3.6	28.7	1333.7	4706.2	4.7	29.5	1462.3	4305.1	4.4	29.6	1414.4	3992.5	6.2	30.7	1514.1
1901–2100	2235.4	2.6	27.6	1338.8	2612.7	3.7	28.3	1470.1	1855.8	3.4	28.5	1412.2	1871.0	5.3	29.6	1508.8
2101–2300	617.2	0.7	26.7	1236.0	1014.7	1.9	27.4	1355.7	268.6	1.5	27.8	1295.3	467.4	3.6	28.9	1376.4
2301–2500	617.2	-1.0	25.8	1146.9	1014.7	0.3	26.5	1256.0	268.6	-0.2	27.0	1196.9	467.4	2.0	28.0	1264.8
2501–2700	2.2	-2.9	24.8	1044.9	170.7	-1.5	25.5	1140.3	0.0	-2.0	26.0	1086.3	24.5	0.2	27.0	1140.6
2701–2900	0.0	-5.0	23.9	920.3	0.0	-3.6	24.7	1001.1	0.0	-4.1	25.2	953.9	0.0	-1.9	26.2	995.9
2901–3100	0.0	-7.2	23.3	793.9	0.0	-5.7	24.0	861.4	0.0	-6.2	24.5	820.8	0.0	-4.0	25.6	854.1
3101–3300	0.0	-9.4	22.7	665.4	0.0	-7.8	23.4	719.5	0.0	-8.4	23.9	686.1	0.0	-6.1	25.0	712.2
3301–3500	0.0	-11.1	21.8	582.8	0.0	-9.5	22.6	628.7	0.0	-10.0	23.0	599.7	0.0	-7.8	24.2	621.8

4. Conclusions

The study predicted a forest fire susceptible region in the western Himalayas and determined the potential of such prediction using the Random Forest model. The historic fire susceptibility region for the present time have been utilized for the prediction of forest fires in climate change projections based on IPCC5 for the years 2041–2060 and 2060–2080. The potential of change in forest fire susceptibility in elevation, latitude, and longitude in the western Himalayas was explored. The forest fire-sensitive area was observed less in 2080 as compared to 2041–2060 in all RCP2.6 and RCP4.5 scenarios. However, in RCP6.0 and RCP8.5, it showed increasing trends. The reason may be the increase in the rainfall trends in lower altitudes. The results provide a better understanding of forest fires pattern and may be useful for forest fire planning and the preparedness for the control measures. The study indicated not only the shifting of fire susceptible regions but it also predicted the shift of fire-sensitive forest to the higher elevation due to global warming. Fire susceptibility depended upon temperature and precipitation during the fire season.

Other factors such as wind speed, wind direction, stand structure, and fuel load are also responsible for forest fires (Kumar et al., 2015) but have not been targeted in this study. The forest fire susceptibility directly or indirectly depends on the distance of the forest from the road, habitation, and water channel network. These variables are available only for present conditions and thus could not be used in the present work for future predictions and can be a possible research issue. The present study on the temporal and spatial distribution of forest fire susceptibility due to changing climatic conditions can provide key knowledge to make forest fire management action plans and also guidelines for the adaptation policies for forest fire at the regional level. Also, the use of high spatial resolution remote sensing data will improve the fire susceptibility model for the detection of the potential of fire dynamics in the Himalayan region. In the Himalayan region, varied climatic microclimatic conditions require such high-resolution variables for precise the prediction of the fire model. Since forest fires depend upon the forest types of the region, the study indicated a change in the forest type composition due to changes in population patterns. Our results have predicted not only key indicators for the distribution of fire susceptible region but also suggested shifting of fire susceptible forest types like Chir pine to the higher elevation in future climatic scenarios.

Acknowledgement

The authors are thankful to Dr Sanjay Kumar, Director, CSIR-IHBT, Palampur for his support and providing the facility. Author Sunil Kumar acknowledges the Council of Scientific and Industrial Research (CSIR), New Delhi for providing SRF fellowship. We also thank HoD and the staff members of the Environmental Technology division of CSIR-IHBT, Palampur for their help during this research. This is CSIR-IHBT Publication No. 4761

Author Contribution

Sunil Kumar developed the theoretical formalism, performed the analytic calculations, and performed the numericalsimulations. Both Sunil Kumar and Amit Kumar, authors, contributed to the final version of the manuscript. Amit Kumar supervised theproject.

References:

Abatzoglou JT, Battisti DS, Williams AP et al (2021) Projected increases in western US forest fire despite growing fuel constraints. Commun Earth Environ 2:227. https://doi.org/10.1038/s43247-021-00299-0

ALLOUCHE O TSOAR, A., Katistic (TSS). J Appl Ecol 43, 1223–1232. https://doi.org/10.1111/j.1365-2664.2006.01214.x

ALEJANDRA BORUNDA (2020) The science connecting wildfires to climate change. National geographic. https://www.nationalgeographic.com/science/article/climate-change-increases-risk-fires-western-us

Verma AK, Namitha Nhandadiyil Kaliyathan NS, Bisht R, Nautiyal SD, Sharma (2018) Predictive distribution modeling of forest fire in pine zone of Uttarakhand Himalayas of India. Advances in Forest Fire Research 2018 - D. X. Viegas (Ed.) Chap. 5 - Decision Support Systems and Tools https://doi.org/10.14195/978-989-26-16-506_120

Bahuguna VK, Singh S (2002) Fire situation in India. Int Fire News 26:23–27

Bedia J, Herrera S, Gutiérrez J, Zavala G, Urbieta I (2012) Sensitivity of fire weather index to different reanalysis products in the Iberian Peninsula

Belgiu M, Drăgu L (2016) Random forest in remote sensing: A review of applications and future directions. ISPRS J Photogramm Remote Sens. https://doi.org/10.1016/j.isprsjprs.2016.01.011

Bhutiyani MR, Kale VS, Pawar NJ (2010) Climate change and the precipitation variations in the northwestern Himalaya: 1866–2006. Int J Climatol 30:535–548. https://doi.org/10.1002/joc.1920

Bistinas I, Oom D, Sá ACL, Harrison SP, Prentice IC, Pereira JMC (2013) Relationships between Human Population Density and Burned Area at Continental and Global Scales. PLoS ONE 8:e81188. https://doi.org/10.1371/journal.pone.0081188

Bowman DMJS, Balch J, Artaxo P, Bond WJ, Cochrane MA, D’Antonio CM, Defries R, Johnston FH, Keeley JE, Krawchuk MA, Kull CA, Mack M, Moritz MA, Pyne S, Roos CI, Scott AC, Sodhi NS, Swetnam TW (2011) The human dimension of fire regimes on Earth. J Biogeogr. https://doi.org/10.1111/j.1365-2699.2011.02595.x

Calvin K, Bond-Lamberty B, Clarke L, Edmonds J, Eom J, Hartin C, Kim S, Kyle P, Link R, Moss R, McJeon H, Patel P, Smith S, Waldhoff S, Wise M (2017) The SSP4: A world of deepening inequality. Glob Environ Chang 42:284–296. https://doi.org/10.1016/j.gloenvcha.2016.06.010

Chandra K, Bhardwaj AK, Chandra KK, Kumar Bhardwaj A, of Soil STUDY OF STRUCTURE AND FUNCTION OF DIFFERENT AGRO FORESTRY PRACTICES EXISTING IN CHHATTISGARH PLAIN ZONE View (2015) Incidence of Forest Fire in India and Its Effect on Terrestrial Ecosystem Dynamics, Nutrient and Microbial Status project Restoration of coal mine dumps using AMF View project Incidence of Forest Fire in India and Its Effect on Terrestrial Ecosystem Dynamics, Nutrient and Microbial Status of Soil. Int. J. Agric. For. 5, 69–78. https://doi.org/10.5923/j.ijaf.20150502.01

Chen F, Du Y, Niu S, Zhao J (2015) Modeling Forest Lightning Fire Occurrence in the Daxinganling Mountains of Northeastern China with MAXENT. Forests 6:1422–1438. https://doi.org/10.3390/f6051422

Chitale V, Behera MD (2019) How will forest fires impact the distribution of endemic plants in the Himalayan biodiversity hotspot? Biodivers Conserv 28:2259–2273. https://doi.org/10.1007/s10531-019-01733-8

Cho S, Marzin C, Baek H-J, Lee J, Lee H-S, Hyun Y-K, Cho C, Kwon W-T, Gan S-Y, Kim M-J, Choi D-H, Lee J, Lee, Jaeho, Boo K-O, Kang H-S, Byun Y-H (2013) Climate change in the 21st century simulated by HadGEM2-AO under representative concentration pathways. Artic Asia-Pacific J Atmos Sci. https://doi.org/10.1007/s13143-013-0053-7

Choudhary A, Dimri AP, Maharana P (2018) Assessment of CORDEX-SA experiments in representing precipitation climatology of summer monsoon over India. Theor Appl Climatol 134:283–307. https://doi.org/10.1007/s00704-017-2274-7

Chuvieco E (2003) Wildland fire danger: estimation and mapping: the role of remote sensing data

Dimri AP, Dash SK (2012) Wintertime climatic trends in the western Himalayas. Clim Change 111:775–800. https://doi.org/10.1007/s10584-011-0201-y

Dormann CF, Elith J, Bacher S, Buchmann C, Carl G, Carré G, Marquéz JRG, Gruber B, Lafourcade B, Leitão PJ, Münkemüller T, Mcclean C, Osborne PE, Reineking B, Schröder B, Skidmore AK, Zurell D, Lautenbach S (2013) Collinearity: A review of methods to deal with it and a simulation study evaluating their performance. Ecography (Cop) 36:27–46. https://doi.org/10.1111/j.1600-0587.2012.07348.x

Evans JS, Murphy, Melanie A, Holden, Zachary A, Cushman, Samuel A (2011) Modeling species distribution and change using random forest [Chapter 8]. Springer 139–159. https://doi.org/10.1007/978-1-4419-7390-0_8

Fick SE, Hijmans RJ (2017) WorldClim 2: new 1-km spatial resolution climate surfaces for global land areas. Int J Climatol Int J Clim 37:4302–4315. https://doi.org/10.1002/joc.5086

Fire Information for Resource Management System (FIRMS) https://firms.modaps.eosdis.nasa.gov/

Fischlin A, Midgley G, Price J, Leemans R, Gopal B, Turley C, Rounsevell M, Dube P, Tarazona J, Velichko A (2007) Ecosystems their properties goods and services. climate change 2007: Impacts adaptation and vulnerability. contribution of working group ii to the fourth assessment. Contrib. Work. Gr. II to Fourth Assess. Rep Intergov Panel Clim Chang. 213–252

Flannigan MD, Logan KA, Amiro BD, Skinner WR, Stocks BJ (2005) FUTURE AREA BURNED IN CANADA. Springer 72:1–16. https://doi.org/10.1007/s10584-005-5935-y

Flannigan MD, Stocks BJ, Wotton BM (2000) Climate change and forest fires. The Science of the Total Environment

Gao J (2019) Global Population Projection Grids Based on Shared Socioeconomic Pathways (SSPs), Downscaled 1-km Grids, 2010–2100 | NASA Open Data Portal [WWW Document]. Palisades, NY NASA Socioecon. Data Appl. Cent. https://doi.org/https://doi.org/10.7927/H44747X4

Giglio L, Randerson JT, Van Der Werf GR (2013) Analysis of daily, monthly, and annual burned area using the fourth-generation global fire emissions database (GFED4). J Geophys Res Biogeosciences 118:317–328. https://doi.org/10.1002/jgrg.20042

Guisan A, Thuiller W, Zimmermann NE (2017) Habitat Suitability and Distribution Models, Habitat Suitability and Distribution Models. Cambridge University Press. https://doi.org/10.1017/9781139028271

Guyette RP, Muzika RM, Dey DC (2002) ECOSYSTEMS[ Dynamics of an Anthropogenic Fire Regime. https://doi.org/10.1007/si0021-002-0115-7. Springer

Harrison SP, Marlon JR, Bartlein PJ (2010) Fire in the Earth System, in: Changing Climates, Earth Systems and Society. Springer Netherlands, pp. 21–48. https://doi.org/10.1007/978-90-481-8716-4_3

Halofsky JE, Peterson DL, Harvey BJ (2020) Changing wildfire, changing forests: the effects of climate change on fire regimes and vegetation in the Pacific Northwest. USA fire ecol 16:4. https://doi.org/10.1186/s42408-019-0062-8

Hickler T, Rammig A, Werner C (2015) Modelling CO2 Impacts on Forest Productivity. Curr Forestry Rep 1:69–80. https://doi.org/10.1007/s40725-015-0014-8

Socioeconomic Data and Applications Center | SEDAC A Data Center in NASA's Earth Observing System Data and Information System (EOSDIS). https://sedac.ciesin.columbia.edu/

Global climate and weather data, https://worldclim.org

Kittur B, Kumar Jhariya M, Lal C, Parmar Y (2014) Is the forest fire can affect the regeneration and species diversity. Copyright@ EM International

Krawchuk MA, Moritz MA, Parisien M-A, Van Dorn J, Hayhoe K (2009) Global Pyrogeography: the Current and Future Distribution of Wildfire. PLoS ONE 4:e5102. https://doi.org/10.1371/journal.pone.0005102

Kumar S, Meenakshi, Bairagi G, Das, Vandana, Kumar A (2015) Identifying triggers for forest fire and assessing fire susceptibility of forests in Indian Western Himalaya using geospatial techniques. Nat Hazards 78:203–217. https://doi.org/10.1007/s11069-015-1710-1

Liu C, Newell G, White M (2016) On the selection of thresholds for predicting species occurrence with presence-only data. Ecol Evol 6:337–348. https://doi.org/10.1002/ece3.1878

Martin DA (2019) Linking fire and the United Nations Sustainable Development Goals. Sci Total Environ 662:547–558. https://doi.org/10.1016/j.scitotenv.2018.12.393

McPherson JM, Jetz W, Rogers DJ (2004) The effects of species’ range sizes on the accuracy of distribution models: Ecological phenomenon or statistical artefact? J Appl Ecol 41:811–823. https://doi.org/10.1111/j.0021-8901.2004.00943.x

Olden JD, Lawler JJ, Poff NL (2008) Machine Learning Methods Without Tears: A Primer for Ecologists. Source Q Rev Biol 83:171–193. https://doi.org/10.1086/587826

Ren YY, Ren GY, Sun XB, Shrestha AB, You QL, Zhan YJ, Rajbhandari R, Zhang PF, Wen KM (2017) Observed changes in surface air temperature and precipitation in the Hindu Kush Himalayan region over the last 100-plus years. Adv Clim Chang Res 8:148–156. https://doi.org/10.1016/j.accre.2017.08.001

Roy PS, Meiyappan P, Joshi PK, Kale MP, Srivastav VK, Srivasatava SK, Behera MD, Roy A, Sharma Y, Ramachandran RM, Bhavani P, Jain AK, Krishnamurthy. YVN (2016) Decadal Land Use and Land Cover Classifications across India, 1985, 1995, 2005, https://doi.org/10.3334/ORNLDAAC/1336 [WWW Document]. ORNL DAAC, Oak Ridge, Tennessee, USA. URL https://daac.ornl.gov/cgi-bin/dsviewer.pl?ds_id=1336 (accessed 12.15.20).

San-Miguel-Ayanz J, Moreno JM, Camia A (2013) Analysis of large fires in European Mediterranean landscapes: Lessons learned and perspectives. Ecol Manage 294:11–22. https://doi.org/10.1016/j.foreco.2012.10.050

Saran S, Joshi R, Sharma S, Padalia H, Dadhwal VK (2010) Geospatial modeling of Brown oak (Quercus semecarpifolia) habitats in the Kumaun Himalaya under climate change scenario. J Indian Soc Remote Sens 38:535–547. https://doi.org/10.1007/s12524-010-0038-2

Scholze M, Knorr W, Arnell NW, Prentice IC (2006) A climate-change risk analysis for world ecosystems. Proc. Natl. Acad. Sci. U. S. A. 103, 13116–13120. https://doi.org/10.1073/pnas.0601816103

Shah S, Sharma DP (2015) Land use change detection in Solan forest division, Himachal Pradesh, India. For. Ecosyst. 2. https://doi.org/10.1186/s40663-015-0050-7

R Core Team (2018) R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/

Thibeault J, Frey KE, Panday PK (2014) Changing temperature and precipitation extremes in the Hindu Kush-Himalayan region: An analysis of CMIP3 and CMIP5 simulations and projections. Int J Climatol Int J Clim 35:3058–3077. https://doi.org/10.1002/joc.4192

Thuiller W, D G, Breiner RE (2016) F, biomod2: Ensemble Platform for Species Distribution Modeling. [WWW Document]. R Packag. version 3.3-7. https//CRAN.R-project.org/package = biomod2. URL https://www.researchgate.net/publication/309762991_biomod2_Ensemble_Platform_for_Species_Distribution_Modeling (accessed 12.15.20)

UNDP. 23th Feb (2022) Number of wildfires to rise by 50% by 2011 and governments are not prepared. https://www.unep.org/news-and-stories/press-release/number-wildfires-rise-50-2100-and-governments-are-not-prepared

Van Der Werf GR, Morton DC, Defries RS, Olivier JGJ, Kasibhatla PS, Jackson RB, Collatz GJ, Randerson JT (2009) CO2 emissions from forest loss. Nat Geosci. https://doi.org/10.1038/ngeo671

Westerling AL, Hidalgo HG, Cayan DR, Swetnam TW (2006) Warming and earlier spring increase Western U.S. forest wildfire activity. Science (80-). 313940–943. https://doi.org/10.1126/science.1128834

Werner Rammer KH, Braziunas WD, Hansen Z, Ratajczak AL, Westerling MG, Turner (2021) Rupert Seidl. Widespread regeneration failure in forests of Greater Yellowstone under scenarios of future climate and fire. Glob Change Biol. 10.1111/gcb.15726

; Wilfried Thuiller; Bruno Lafourcade; Robin Engler, Miguel B, Araújo (2009) BIOMOD – a platform for ensemble forecasting of species distributions., 32(3), 369–373. 10.1111/j.1600-0587.2008.05742.x

Yes

Abstract

Identifying the spatial and temporal attributes which are favouring forest fire susceptibility necessary for biological conservation. The adverse effects of climate change on the forest has increased wildfire. The rise in global temperature and alteration of rainfall patterns have produced appropriate conditions for forest fires. A non-parametric ‘Random Forest Algorithm’ for modelling the spatial distribution of forest fires was applied to predict the susceptibility of Indian western Himalayan forest due to fires. The forest fire susceptibility was simulated in the present (years 1970–2000) and future (years 2041-2060 and 2061-2080) environmental gradients. The real-time distribution of the fire susceptibility was evaluated and modelled using forest fire history data with an overall accuracy of more than 0.9. To derive the fire susceptible region in future, we have applied the model statistics of the present time to the future climatic scenario. The magnitude of increase of fires was predicted relatively more along longitudinal and elevational gradient as compared to the latitude. The high sensitive forest fires susceptible area was found as 35376.18 sqkm in the present conditions, while it occupied 61440.03 sqkm, 57181.76 sqkm, 57662.82 sqkm and 56612.11 sqkm respectively in 2041-2060 in the four projected climatic scenarios Representative Concentration Pathways (i.e., RCP2.6, RCP4.5, RCP6.0 and RCP8.5). During 2061-2080, a decline in RCP2.6 and RCP4.5 (56241.95 sqkm and 56668.29 sqkm) and an increase in RCP6.0 and RCP8.5 (61199.50 sqkm and 57510.15 sqkm) were predicted. The results clearly show the fire susceptible area will be higher in the RCP2.6 for the year 2041-2060 and RCP6.0 in 2061-2080. The current study thus provides scientific conclusions that the forest fire susceptibility is climate driven in the western Himalayas.