Natural Hazards https://doi.org/10.1007/s11069-022-05601-7 ORIGINAL PAPER GIS‑based forest fire risk determination for Milas district, Turkey Mehmet Cetin1  · Özge Isik Pekkan2  · Mehtap Ozenen Kavlak2  · Ilker Atmaca3  · Suhrabuddin Nasery2  · Masoud Derakhshandeh4  · Saye Nihan Cabuk5 Received: 27 July 2022 / Accepted: 30 August 2022 © The Author(s), under exclusive licence to Springer Nature B.V. 2022 Abstract Forest fires are highly destructive phenomena in both ecological and economic terms. Therefore, it is significant to develop measures to detect and mitigate them. In this study, the forest fire risk map of the Milas district of Turkey was studied using geographical information systems and remote sensing methods. In the first part of the study, the for- est fire risk map of the area was developed via a weighted overlay technique with analy- sis of stand characteristics, topographic features, distance from intermittent streams and built-up environment. According to the resulting forest fire risk map, extremely low-, low-, medium-, high- and extremely high-risk classes covered 0%, 0.5%, 65%, 30% and 0.5% of the forested areas in Milas district of Turkey, respectively. In the second part, the location of a major forest fire, which took place in 2007 in the study area, was determined using the normalized difference vegetation index, the normalized burn ratio, and the burn area index. When compared with the forest fire risk map, it was revealed that 45% of the burned areas in 2007 fell into the high-risk class, while 51% of it was from the extremely high-risk zones. Moreover, the forest risk map was compared with eleven forest fire cases between 2013 and 2019. The results show that eight of these fires took place in high-risk territories. According to these results, it was concluded that the created risk map coincides with the fire incidents. Keywords Burn area index · Forest fire · GIS · Normalized burn ratio index · Risk assessment * Mehmet Cetin mehmet.cetin.landscape.architect@gmail.com; mehmet.cetin@omu.edu.tr 1 Department of City and Regional Planning, Faculty of Architecture, Ondokuz Mayis University, Samsun, Turkey 2 Department of Remote Sensing and Geographical Information Systems, Institute of Graduate Programs, Eskisehir Technical University, İki Eylül Campus, 26555, Tepebaşı, Eskişehir, Turkey 3 Department of City and Regional Planning, Faculty of Engineering and Architecture, Yozgat Bozok University, Erdoğan Akdağ Campus, 66900 Yozgat, Turkey 4 Department of Civil Engineering, Faculty of Architecture and Engineering, Istanbul Gelisim University, 34310, Avcılar, Istanbul, Turkey 5 Department of Geodesy and Geographical Information Technologies, Institute of Earth and Space Sciences, Eskisehir Technical University, İki Eylül Kampüsü, 26555, Tepebaşı, Eskişehir, Turkey Vol.:(0123145 63789) Natural Hazards 1 Introduction Large fire cases release huge amounts of gas and particulate matter into the atmosphere, which increase the direct and indirect impact of global climate change and result in serious problems for public health and even security (Patz et al. 2014; Wang et al. 2019). Wildfires around urban areas threaten residential areas as they may destroy the built-up environment and cause injury and death (Cosgun et al. 2019; Modugno et al. 2016). Based on the Global Fire Monitoring Center Report, wildfires killed 297 firefighters and civilians between 2008 and 2015 worldwide (Doerr and Santín 2016). Forest fires also destroy precious for- est ecosystems, since they damage the vegetation and increase soil erosion and flood risks (Nasi et al. 2002; Verma and Jayakumar 2012). In Greece, the huge forest fire in July 1997 burned almost 1100 hectares, roughly two-thirds of the forest area, which had been the largest source of oxygen for the city of Thessaloniki (Siachalou et al. 2009). The forest fire that took place in Chile in 2017 destroyed forest plantation and shrubland ecosystems. The resulting air pollution also influenced almost 74% of the Chilean popula- tion (de la Barrera et al. 2018). During a series of mega forest fires in eastern Australia, between September 2019 and January 2020, nearly 5.8 million hectares of forests burned (Boer et al. 2020; Nolan et al. 2020). The smoke of the fire in the Amazons on August 19, 2019, turned the city of São Paulo, 2000 km away from the centre of the fire, into darkness at 3 pm (Bencherif et al., 2020). Over the last two decades, 63,724 forest fires, of which 90% were human-induced, were recorded in Turkey (Çolak and Sunar 2020). Considering the devastating consequences of the forest fires, a good number of research studies have been conducted to help prevent and manage forest fire risks. As highlighted in these studies, it is necessary to determine high fire risk locations before developing preven- tive measures. In this regard, a forest fire risk map can be prepared via the analysis and evaluation of different natural and cultural factors (Akinola and Adegoke 2019; Chuvieco 2009; Çolak and Sunar 2020; Ghorbanzadeh et al. 2019a; Ljubomir et al. 2019; Naderpour et al. 2019; Nami et al. 2018). Forest fire risk maps are a valuable basis for locating fire- fighting resources in suitable and easily accessible spots (Romero-Calcerrada et al. 2008). There are multiple factors affecting the risk of fire outbreaks, and these must be con- sidered, based on the nature of the study area. They can be categorized into two groups: dynamic and non-dynamic factors. Dynamic factors are temporal and may change season- ally, daily or even hourly. Factors such as temperature, precipitation, radiation and sea- sonal travel rates are in this category (Amiro et al. 2004; Bar Massada et al. 2009; Busico et al. 2019; Holsinger et al. 2016; Nuthammachot and Stratoulias 2019). In contrast, non- dynamic factors, such as topography, forest accessibility and stand characteristics, do not change, or their change is only significant over a longer time span (Adab et al. 2013; Akinola and Adegoke 2019; Busico et al. 2019; Mota et al. 2019). Therefore, two types of forest fire risk maps (dynamic and non-dynamic) can be produced, based on the nature of the variables. The dynamic risk maps are useful for quickly responding to instant or changeable events, such as daily fire warnings. The non-dynamic maps are helpful to develop long-term preventive measures and strategies, such as the locating of fire watch- towers (Bao et al. 2015) and firefighting stations, constructing water storage facilities, plan- ning forest roads, determining recreational sites (Molina and y Silva 2019) and implement- ing public training programmes (Gençay and Mercimek 2019). As they are more stable, non-dynamic factors are evaluated within the scope of this study. 1 3 Natural Hazards The forest fire risk assessment studies in the literature primarily concentrate on inap- propriate weather conditions, which may favour fire ignition and propagation (Akay and Şahin 2018; Çolak and Sunar 2020; Fatma and Vedat 2018; Finney 2005; Mota et al. 2019; Nuthammachot and Stratoulias 2019), as well as a variety of other natural factors, such as topography, plant and soil moisture, stand types, canopy density, and factors relating to human behaviour, such as proximity to settlement areas and roads, demographic charac- teristics and so on, which have been adopted for the development of either fire risk index (FRI) (Çolak and Sunar 2020; Naderpour et al. 2019; Sivrikaya et al. 2014) or fire danger index (FDI) (Castro and Chuvieco 1998; Chuvieco and Salas 1996). In studies dealing with environmental factors, it may be necessary to work with variety of data layers. GIS and RS are important tools in many current issues that require the anal- ysis of a large number of spatial data, such as soil organic carbon research (Pekkan et al., 2021), ecological footprint research (Cetin et al., 2021a, b) and urban development process research (Cetin et al., 2021a, b). In addition to the variety of data layers, different meth- ods embracing geographical information systems (GIS) and remote sensing (RS) capabili- ties have also been utilized by the researchers for the realization of spatial analyses and production of forest fire risk/danger maps (Akay and Erdoğan 2017; Chuvieco and Salas 1996; Romero-Calcerrada et al. 2008). For example, Massada et al. (2009) studied fuel and topography, weather and structure data as main inputs to analyse the FRI. Chuvieco and Salas. (1996) considered fuel type, temperature, humidity, compactness, plant moisture, topography and human activity to estimate the FDI. Siachalou et al. (2009) integrated five factors, vegetation, slope, aspect, roads and settlements, to generate a fire risk map showing high-risk zones in areas with high slope, a south-facing aspect and low moisture. Akbulak et al. (2018) considered both human and natural factors, such as slope, elevation, aspect, vegetation type, canopy density, distance from roads, settlements, agricultural areas and population density based on the Canadian forest fire weather index (FWI) (Wotton et al. 2005). Vadrevu et al. (2010) studied tropic forest fires in India and developed a fire hotspot map using the GIS and analytical hierarchy process (AHP). The risk and danger indexes and maps in the majority of these studies are based on different risk level classes, which identify the severity of forest fire risk. In a number of studies (Akay and Erdoğan 2017; Akbulak et  al. 2018), a 4-level (medium, high, very high, extreme) fire risk classification was also adopted. A. G. McArthur, who developed the McArthur Forest Fire Danger Index (FFDI) (McArthur 1967), is one of the leading scientists working on forest fires. In their study, published by Luke and McArthur in 1978, they defined FFDI in five classes (Luke and McArthur 1978). Accordingly, in many studies including forest fire risk and danger, the classifications were evaluated in five groups. In this sense, Akay and Sahin (Akay and Şahin 2018) used a 5-level (very low, low, medium, high, very high) fire risk classification in their study. From this perspective, the main goal of this study is to determine the forest fire risks in the Milas district of Muğla province, Turkey, which is both a popular tourist destination and a delicate landscape, in terms of natural characteristics and values. To fulfil this aim, a weighted overlay analysis was performed with GIS capabilities and a non-dynamic forest fire risk map was proposed. 1 3 Natural Hazards 2 Material and methods 2.1 Study area The Milas district of the Muğla province, Turkey, was selected as the study area due to the high number of forest fires in the region, as reported by the General Directo- rate of Forestry (OGM 2018) (Fig. 1). Milas is the second largest district of the Muğla province and is located in southwest Anatolia, between 27° 30ꞌ–28° 30 ꞌ east longitude and 37° 00ꞌ–37° 30ꞌ north latitude. The total area of the district is 2067  km2. Milas lies within the Mediterranean climate zone where summers are hot and dry, and winters are mild and rainy. In summer, the average temperature is between 32 °C and 34 °C and sometimes even exceeds 40 °C. Milas is mostly open to northern and southern winds. According to statistics derived from the General Directorate of Forestry (OGM 2018), 66% of the Muğla province is covered by forests and the forests in Muğla province con- stitute 4% of the forests in Turkey. This rate refers to a total forest area of 829,309 hec- tares and ranks the province in fourth place in terms of forested area size. The dominant stand type in the Milas forests is Turkish pine (Pinus brutia). There are other species, such as the European Black Pine (Pinus nigra), Stone pine (Pinus pinea), and deciduous plants in smaller proportions. Fig. 1 Location of the study area, Milas district, Muğla/Turkey 1 3 Natural Hazards 2.2 Research variables 2.2.1 S tand characteristics The quality of the fuel, more specifically the characteristics of biomass, depends on the stand types. Therefore, stand type is a significant variable that affects ignition and also the behaviour of fire propagation. For example, needle leaved trees, such as conifers and dried leaved plants, have a greater potential to burn, whereas wet broad leaved species, such as beech, do not catch fire easily (Karabulut et al. 2013). A ground fire could easily turn into a top layer fire where the fuel type is highly inflammable (Akkaş et al. 2008). According to the literature, a significant proportion of forest fires occur in young stand regions (Akkaş et al. 2008). Generally, older trees are large in size and possess a porous structure where they can retain greater amounts of moisture. As the moisture increases, the burning effect decreases (Hood 2010). There- fore, more energy is needed to burn old trees, so they are less likely to catch fire. An extremely dense canopy corresponds to a higher moisture content of fuel beneath the canopy, which decreases the risk of fire. On the other hand, where the canopy den- sity is severely low, the level of produced heat is hardly enough to burn any adjacent fuel (Ray et al. 2005). However, a sparse canopy allows the wind to flow more easily, resulting in an increased fire spread rate. 2.2.2 Topographic factors Terrain elevation affects temperature and moisture. The wind effect also varies at dif- ferent elevations (Holsinger et  al. 2016; Sullivan et  al. 2014). Vegetation type, fuel moisture, and air humidity vary at different elevations (Castro and Chuvieco 1998). It is reported that humidity and temperature effects become more significant at higher alti- tudes (Hernandez-Leal et al. 2006). In contrast, it is also reported that forest fires may be less severe at higher elevations due to increased precipitation (Adab et al. 2013). Terrain slope is a significant factor in the fire spread rate and pattern (Butler et  al. 2007; Xavier Viegas 2004). The fire propagation rate increases with an increase in slope. The effect of slope varies in accordance with the direction of fire spread direc- tion, downhill or uphill. In an experimental study, Wagner (Van Wagner 1988) conclude that a downhill fire spread rate against the slope follows a parabolic pattern where it decreases when the slope increases from zero to 22° and then increases to the initial flat level rate when the slope increases to 45°. This phenomenon is explained by the effect of fire head radiation and falling burned debris. However, for uphill fire propagation, there is consensus in the literature that a steeper slope corresponds to a faster fire spread rate (Dupuy 1995; Viegas 2004). Another significant topographical factor is aspect, where south-facing land receives more radiation, and therefore is drier and more prone to fire ignition compared to north- facing land (Noonan 2003). 1 3 Natural Hazards 2.2.3 H uman intervention The distance between settlements and forests is a determining risk factor, because human activities, such as camping, hunting and recreation in or near forests, mainly take place when the forests are close to the settlement areas (Fatma and Vedat 2018). The impact of human activity starting a fire can be quantified in terms of proximity to/ distance from roads and settlements (Rogan et  al. 2006). Forest areas in the vicinity of roads are more prone to fire risk, because roads allow local people, shepherds and tourists to reach forests more easily (Jaiswal et al. 2002). Another significant factor causing forest fire risk is the existence of human settlements nearby or close to forested areas (Ghorbanzadeh et al. 2019b). There are higher risks of fire in forested regions that are close to the settlements and houses (Jaiswal et al. 2002). Care- lessly thrown unextinguished cigarette butts and campfires are amongst the sources of fire risk. Intermittent streams also play an important role in the start and spread of a fire, espe- cially in dry seasons, as they may contain dry fuel (leaves, branches, roots, barks, seeds and suchlike) or other flammable materials (Townsend and Douglas 2000), such as human produced garbage. Intermittent streams are also places through which the wind blows faster compared to dense forest areas, and therefore, a fire can spread extremely quickly causing more damage. 2.3 Data sources The main materials used in this study include a combination of vector and raster data covering the Milas district. A 1/25000 scale stand map, covering the 2003–2013 period and detailing road networks and settlement areas, was obtained in vector format from the Regional Directorate of Muğla Forestry and was used to determine and reclassify the types of stand, stand age and canopy density. Raster data, with Landsat 5 images of the study area, were obtained from USGS Earth Explorer. A digital elevation model (DEM) of the study area was provided from ASTER available on the EARTH DATA platform of NASA and used to produce elevation, slope and aspect layers (Table  1). Two Landsat 5 images (dated 12.06.2007 and 15.08.2007) (Table 1) were utilized to calculate NDVI, NBR and BAI indexes for the detection of the burned forests in Güvercinlik, Milas, in 2007. The forest fire risk map of the Milas district Table 1 Satellite imagery data sources Acquisition date Product id Satellite, sensor Spatial resolution (m) 06.12.2007 LT05_L1TP_180034_20070612_20180126_ Landsat 5, TM 30 01_T1 15.08.2007 LT05_L1TP_180034_20070815_20180123_ Landsat 5, TM 30 01_T1 01.03.2000, and 30.11.2013 ASTGTMV003_N38E028 ASTER, V003 30 ASTGTMV003_N38E027 ASTGTMV003_N37E028 ASTGTMV003_N37E027 1 3 Natural Hazards was compared both with the resulting map (Güvercinlik fire map) and the locations of eleven large fire cases, which occurred between 2013 and 2019. The data related to the fire locations were provided by the General Directorate of Forestry (Milas, Turkey), and the exact locations of these fires were detected using Google Earth images. 2.4 Weighted overlay analysis The main method adopted for the development of the forest fire risk map of the study area is weighted overlay analysis, conducted in a GIS environment. Weighted overlay analy- sis has been used by researchers for different spatial analysis applications. This method is mainly preferred when the cumulative interaction of different variables with their rela- tive importance is of interest (Jabbar et al. 2019; Mandal and Mondal 2019; Yathish et al. 2019). A work flowchart for forest fire risk map development is provided in Fig. 2. As can be seen in the figure, different factor layers were weighted and overlaid. To fulfil this aim, each factor layer was reclassified based on information in the literature (Çolak and Sunar 2020; Nolan et al. 2020; Yathish et al. 2019) and the speculations of experts in the field of forestry and forest fire management. Since multiple factors do not have equal significance on fire risk, weight factors, which were determined using the AHP, were allo- cated to each category (Table 2). The same process was also applied for sub-categories. A ten-score scale was used for quantifying risk factors where ‘ten’ was the highest risk class and ‘one’ referred to the lowest risk. Equation one shows the calculation procedure (Eq. 1): Fig. 2 Flowchart for fire risk map development and assessment 1 3 Natural Hazards 1 3 Table 2 Risk scores and weight percentage for risk variables Main categories Weight(i) Subcategories Weight (j) Subcategory classes Risk score (Pπj) Stand characteristics 42.85% Stand type 60.80% Pinus brutia 10 Pinus nigra 7 Pinus pinea 9 Deciduous plants 4 Stand age 27.20% I 5 II 9 III 8 IV 7 V 6 VI 5 Unspecified (damaged stand) 4 Stand canopy density 12.00% 1–10% 5 10–40% 8 40–70% 10 > 70% 7 Topographical characteristics 14.30% Slope 6.66% > 35 degree 6 35–15 degree 5 15–10 degree 4 10–5 degree 3 0–5 degree 2 Natural Hazards 1 3 Table 2 (continued) Main categories Weight(i) Subcategories Weight (j) Subcategory classes Risk score (Pπj) Aspect 46.66% South 10 Southwest 8 Southeast 8 Northwest 3 Northeast 3 North 2 East 4 West 4 Flat 5 Elevation 46.66% 0–200 m 10 200–400 m 9 400–600 m 8 600–800 m 7 800–1000 m 6 1000–1200 m 5 1200–1400 m 4 Human intervention 42.85% Distance from roads 28.95% 0–50 m 10 50–100 m 9 100–200 m 8 200–300 m 7 Natural Hazards 1 3 Table 2 (continued) Main categories Weight(i) Subcategories Weight (j) Subcategory classes Risk score (Pπj) 300–400 m 6 400–12,500 m 5 Distance from intermittent streams 6.42% 0–50 m 8 50–100 m 7 100–200 m 5 200–300 m 3 300–400 m 2 400–7000 m 1 Distance from settlements 64.63% 0–50 m 10 50–100 m 9 100–200 m 8 200–300 m 7 300–400 m 6 400–500 m 5 500–1000 m 4 1000–1500 m 3 1500–30,000 m 3 Natural Hazards ∑n ∑c nsc ijP=1 j=1 j Riskscore = (1) nc where Pπj is the factor score, j is sub-categories weight factor, i is the main categories weight factor, nsc is the number of sub-categories under that category, and nc is the number of main categories, which is three in this study. The AHP was implemented for the determination of the weights shown in Table 2. The AHP is a decision-making approach, which was first introduced by Thomas L. Saaty (Saaty 1990). It is a process based on the quantification of binary choices according to an allo- cated degree of significance, which provides a scored list of choices. The method basically originated from Eigenvalue problems. In this process, the results of pairwise comparison of variables are inserted into a matrix. In the AHP, a certain degree of inconsistency is still acceptable (Shim 1989). In this study, score ranges of 0–2, 2–4, 4–6, 6–8 and 8–10 are classified as extremely low risk (VLR), low risk (LR), medium risk (MR), high risk (HR) and extremely high risk (VHR), respectively. ArcGIS Desktop software was used to analyse the data and to obtain maps. 2.5 Burned area detection and fire risk map assessment In addition to determining the forest fire risk map of the Milas district, a major fire in the study area in 2007, and eleven other forest fires between 2013 and 2019, were detected. The location of these past eleven fires was overlaid with the forest fire risk map to evalu- ate the risk potential and categories of the previously burned areas. Since point values are inadequate in defining the exact locations of the forest fires, as well as the total boundaries of the burned land, a number of indexes, based on remotely sensed data, including NDVI, NBR and BAI methods, were utilized to determine the forest areas burned and damaged vegetation during the Güvercinlik fire in July 2007. This forest fire deeply affected the local people who were followed by the media for months. The NDVI method basically shows the vegetation health based on differences between the near infrared (NIR) band, which is strongly reflected by vegetation, and the red band, which is absorbed by vegetation (Eq.  2). NBR calculation is similar to NDVI, but uses short wave infrared (SWIR) instead of the red band. It is reported that the NBR method is a more sensitive index compared to NDVI for fire severity quantification (Escuin et al. 2008). The BAI method, which also uses NIR and red bands, relies on the designation of a ‘conversion point’ depending on the radiative characteristics of charcoal-char (Martín et al. 2005) (Eq. 4). The BAI method is considered a more sensitive index when compared to NDVI and NBR (Chuvieco et al. 2002). These three indexes have been used by researchers for burned area and fire severity detection studies. For the NDVI and NBR methods, band ratios were used, as in the following equations (Eqs. 2–4), respectively: NDVI = (NIR−R)∕(NIR + R) (2) NBR = (NIR−SWIR)∕(NIR + SWIR) (3) For the BAI, digital numbers (i.e. pixel value) of red and near infrared (NIR) bands need to be calibrated for the top of atmosphere (TOA) reflectance to calculate BAI according to 1 3 Natural Hazards Fig. 3 Reclassified stand characteristics maps Fig. 4 Forest fire risk map based on stand characteristics (a), risk class distribution (b) (Chander and Markham 2003). After this, BAI was calculated based on the work of (Chu- vieco et al. 2002) using Eq. 4 as follows: ( ) ( )2 ( )2 BAI = 1∕ cr − r + cnir − nir (4) 3 Results and discussion 3.1 F ire risk map based on stand characteristics Stand characteristics, including stand type, age and canopy density, were reclassified and are mapped as shown in Fig.  3. The weight factors (Table  2) were set to 60.8%, 27.2% and 12.0% for stand type, age and canopy density, respectively. The weighted overlay of these sub-categories resulted in a fire risk map based on stand characteristics (Fig. 4). The estimated risk scores ranged between 3.79 and 9.52. Since the area is mainly covered with 1 3 Natural Hazards Fig. 5 Reclassified topographic characteristics maps Fig. 6 Forest fire risk map based on topographic characteristics (a) and risk class distribution (b) stand types of easily ignitable properties (Fig. 3), almost 83% of the district was found to comprise areas of high- and extremely high-risk classes (Fig. 4). The distribution of the high, extremely high, low and medium risk classes was detected as 61%, 22%, 15% and 2%, respectively. 3.2 Fire risk map based on topographic characteristics The topographic characteristics within the study area are reclassified in line with the weighted overlay objectives (Fig. 5). The resulting map shows that the study area has con- siderably high risks for forest fires in terms of elevation, especially at elevations lower than 200 m, because the highest weight is allocated for less than the 200 m class, and the majority of the region possesses an elevation below 200 (Fig. 5c). The aspect map, on the other hand, reveals that the topography is mainly southeast, south and southwest facing. The reclassified slope characteristics map shows that sharp slopes above 35° are extremely 1 3 Natural Hazards Fig. 7 Reclassified human intervention maps rare, while slopes between 15° and 35° are very dense in the area. However, the latter is observed mostly at higher elevations. The fire risk map based on topographic characteristics is illustrated in Fig. 6. Accord- ing to Fig. 6, the Milas district is a highly risky region in terms of forest fires when the topographic characteristics are considered. The area distribution of risk classes in the study area is shown in the pie chart in Fig. 6. Figure 6b shows that almost 67% of the district falls into the high- and extremely high-risk classes with a fire risk score ranging of between 6 and 10. 3.3 Fire risk map based on human intervention In previous studies, authors have highlighted that intentional or unintentional human inter- vention was the main reason for forest fires (Caldararo 2002; Ganteaume et al. 2013; Mol- licone et al. 2006; Sanford et al. 1985). Examples of unintentional actions include throw- ing glassware or high radiation absorbing black objects into forests. Human intervention is directly related to accessibility to forests and population density. Depending on the distance from roads and intermittent streams, the entire study area was grouped into six different classes (Fig. 7) and weight factors were set to 6.42%, 28.95% and 64.63%, respectively, for the distance from intermittent streams, distance from roads and distance from settlements. The forest fire risk map, based on human intervention, is shown in Fig. 8. As can be seen in the figure, areas with medium fire risk are homogeneously distributed all over the Milas district. Areas with an extremely high-risk potential (8–10 scores) comprise only 3% of the total forest areas and 8% of the total area fall into the high-risk class. Sixty-six per- centage of the forested lands are at low-risk and no extremely low-risk areas are detected in terms of human intervention. 3.4 Final forest fire risk map The final forest fire risk map for the Milas district was produced via a weighted overlay of three forest fire risk maps, based on stand characteristics, topographic characteristics and human intervention, respectively, as shown in Figs. 4, 6 and  8. According to the final forest fire risk map, it can be seen that 35,050.72 hectares of for- ested land, corresponding to 30% of the total 114,920 hectares of forested area in the Milas 1 3 Natural Hazards Fig. 8 Forest fire risk map based on human interference (a) and risk class distribution (b) Fig. 9 Final forest fire risk map (a) and risk class distribution (b) district, falls into the high- and extremely high-risk classes (Fig. 9). The other risk classes are found to be 63.8% (medium risk), 4.8% (low risk), and 1.3% (extremely low risk). 3.5 Assessment of risk maps in comparison with past fire data As can be seen in Table 3, topographic characteristics alone are not sufficient for fire risk assessment, because seven of the eleven forest fires took place in medium-risk class zones. There are just three cases (number 1, number 5, number 9) in the extremely high-risk class. 1 3 Natural Hazards Table 3 Risk scores for the past eleven large fires during the period 2013–2019 No. Coordinates Risk score Risk score Risk score Risk score (Max 10) (Max 10) (Max 10) (Max 10) Topography Stand Human Final map 1 573,076.625 4,132,390.385 m: Akgedik Barajı 8.5 7.7 3.32 5.94 2 547,211.179 4,123,544.705 m: Kıyıkışlacık 6.34 9.52 4.8 7.32 3 549,289.618 4,115,358.292 m: Boğaziçi 5.8 7.44 6.9 7.12 4 565,108.858 4,108,365.237 m: Kısırlar 5.8 9.25 4.67 6.82 5 567,601.295 4,106,975.955 m: Demirciler 9.2 7.4 5.03 6.78 6 565,213.143 4,106,169.327 m: Demirciler 5.9 8.4 3.77 6.07 7 567,481.477 4,104,171.308 m: Fesleğen 5.8 8.44 3.58 5.98 8 569,162.259 4,103,807.003 m: Fesleğen 5.4 9.52 3.58 6.39 9 569,667.017 4,102,961.884 m: Fesleğen 8.2 7.16 4.16 6.06 10 568,216.866 4,102,566.795 m: Demirciler 4.9 7.44 2.94 5.16 11 576,096.325 4,102,965.774 m: Akçakaya 5.5 7.17 7.55 7.09 a The scores represent risk classes of extremely low risk (0–2), low risk (2–4), medium risk (4–6), high risk (6–8) and extremely high risk (8–10) As to the human intervention factor, Table 3 reveals that the majority of forest fires took place in medium-risk areas. However, when the risk values in the table are ana- lysed, it can be concluded that human-induced risk factors are relatively low. Considering the eleven forest fires risk scores for stand characteristics, it is possible to explain that the forest fires burned the stand types in the high- and extremely high- risk classes. One reason for this is the high-risk score of Pinus brutia (10), which is the most common stand type throughout the study area. Similarly, regarding the final risk map, Table 3 illustrates that eight of the eleven fire cases took place in high-risk zones. This number of correct matches actually sup- ports the consistency of the produced forest fire risk map. However, there is no doubt that the point data are sufficient to make a full comparison and to justify the success of the risk map because, for large fires which cover a wide area, the starting point is not clear. Therefore, the consistency of the map is also evaluated via comparison of the forested land burned in Güvercinlik, Milas district, in 2007. The burned area in 2007, in the Güvercinlik fire, was determined using NDVI, NBR and BAI indexes, as shown in Fig. 10. The results show that all three indexes were able to detect the burned area. The difference in index values, both before and after the fire, provides a high distinction between the burned and unburned areas. The burned area is determined as 271, 277 and 332 hectares using the BAI, NDVI and NBR indexes, respectively. The forest fire risk map of the Milas district is overlaid with both the territories of the burned area in the Güvercinlik fire (Fig. 11) and the locations of the eleven forest fires in the study area, between 2013 and 2019 (Table 3). To fulfil this aim, the forest fire risk map (Figure  9) is clipped according to the burned areas in Güvercinlik (Fig. 11). The results reveal that 51% of the burned areas were from the extremely high-risk class and 45% of the area fell into the high-risk zones. 1 3 Natural Hazards Fig. 10 Burned area detection using the BAI, NDVI and NBR indexes for the Güvercinlik fire, Milas, Tur- key, 2007 4 Conclusions Forest fires can cause irreversible effects on forest resources, have major economic conse- quences and threaten human life. In this context, it is extremely important to take the nec- essary precautions to identify fire hazards and risks in forest areas and to mitigate damage. Based on the study of Giannakopoulos et al. on global warming effects during 2031–2060, Turkey may be one of the most affected countries for this period in terms of the increase in the number of summer days (Giannakopoulos et al. 2009). In this context, it is important to identify the factors that cause fires and to obtain the necessary data for risk management. In this study, the GIS-based weighted overlay method is performed to produce a forest fire risk map in the study area regarding stand characteristics (stand type, stand age, stand density), topographic features (slope, aspect, elevation) and human interven- tion factors (distance from roads, distance from settlements, distance from intermittent streams). As the necessary layers in a weighted overlay process, as well as their assess- ment in terms of the weights and influence factors, may vary depending on the geo- graphic context, local, regional and national policies and strategies, related legislation 1 3 Natural Hazards Fig. 11 The estimated forest fire risk classes for the Güvercinlik mega fire area, Milas, Turkey, 2007(a), Risk class distribution (b) and so on, the obtained forest risk map is specific to the study area and the estimated risk scores cannot be used as a scheme directly for other similar studies without consid- ering and evaluating the unique characteristics and variables of the region. The study research shows that forest fire risks can be mapped based on non-dynamic factors. In addition, possible future studies may include additional dynamic fire risk factors, such as climate parameters, for the development of fire risk maps. The forest fire risk produced for the Milas district is supposed to be a valuable basis and input for further planning processes within the province. One recent precaution to mitigate and prevent forest fires is the plantation of fire resistant tree species. To fulfil this aim, the determination of risky areas, in terms of forest fires, is of significance in order that a proper plantation plan is implemented. Responsible organizations and firefighters need to consider potential forest fire haz- ards to identify local forest fire threats and to assess risks for communities. Moreo- ver, forest fire risk maps are important for training the public and local people, raising awareness, and increasing community participation in forest fire mitigation and preven- tative action. Forest fire risk maps can also facilitate pre-attack planning, detection of areas to be treated in terms of forest fire risk reduction, application of fuel treatment applications, locating fire towers and water tank construction. Acknowledgements We sincerely thank Akın Akman, Murat Boztürk and Adem Kurtipek for sharing their valuable forest fire fighting experience. We also thank the Milas District General Directorate of Forestry in Turkey (Orman Genel Müdürlüğü) for providing data. Authors’ Contributions Mehmet designed the study and performed the experiments; Ozge, Mehtap, ilker performed the experiments, analysed the data, and wrote the manuscript; and Suhrabuddin, Masoud, Saye performed the experiments, analysed the data, and wrote the manuscript. Funding There is no funding resource to be declared. 1 3 Natural Hazards Data Availability The authors confirm that the data supporting the findings of this study are available within the article. Code availability No extra code has been created. Declarations Conflicts of interest There are no conflicts of interest to be declared. References Adab H, Kanniah KD, Solaimani K (2013) Modeling forest fire risk in the northeast of Iran using remote sensing and GIS techniques. Nat Hazards 65:1723–1743. https://d oi.o rg/ 10.1 007/ s11069- 012-0 450-8 Akay A, Erdoğan A (2017) GIS-based multi-criteria decision analysis for forest fire risk map- ping. ISPRS Ann Photogramm Remote Sens Spatial Inf Sci 4:25. https://d oi. org/1 0.5 194/ isprs-a nnals-I V-4- W4-2 5- 2017 Akay AE, Şahin H (2018) Forest fire risk mapping by using GIS techniques and AHP method: a case study in bodrum (turkey). Eur J For Eng 5:25–35. https://d oi.o rg/ 10. 33904/ ejfe. 579075 Akbulak C, Tatlı H, Aygün G, Sağlam B (2018) Forest fire risk analysis via integration of GIS, RS and AHP: the Case of Çanakkale, Turkey. J Hum Sci 15:2127–2143 Akinola OV, Adegoke J (2019) Assessment of forest fire vulnerability zones in Missouri, United States of America. Int J Sust Dev World 26:251–257. https://d oi. org/ 10. 1080/1 3504 509. 2018. 15518 15 Akkaş M, Bucak C, Boza Z, Erkonat H, Bekereci A, Erkan A, Cebeci C, (2008) Büyük Orman Yangınlarının Meteorolojik Veriler Işığında İncelenmesi. TC Çevre ve Orman Bakanlığı, Ege Ormancılık Araştırma Müdürlüğü, Teknik Bülten 36. Amiro BD, Logan KA, Wotton BM, Flannigan MD, Todd JB, Stocks B, Martell D (2004) Fire weather index system components for large fires in the Canadian boreal forest. Int J Wildland Fire 13:391–400. https:// doi.o rg/ 10.1 071/W F030 66 Bao S, Xiao N, Lai Z, Zhang H, Kim C (2015) Optimizing watchtower locations for forest fire monitoring using location models. Fire Saf J 71:100–109. https:// doi. org/ 10.1 016/j.fi resa f. 2014.1 1.0 16 Bar Massada A, Radeloff VC, Stewart SI, Hawbaker TJ (2009) Wildfire risk in the wildland–urban inter- face: a simulation study in northwestern Wisconsin. For Ecol Manage 258:1990–1999. https://d oi. org/ 10. 1016/j. foreco. 2009.0 7. 051 Bencherif H, Bègue N, Kirsch Pinheiro D, du Preez DJ, Cadet JM, da Silva Lopes FJ, Clerbaux C (2020) Investigating the long-range transport of aerosol plumes following the amazon fires (august 2019): a multi-instrumental approach from ground-based and satellite observations. Remote Sens 12(22):3846 Boer MM, de Dios VR, Bradstock RA (2020) Unprecedented burn area of Australian mega forest fires. Nat Clim Chang 10:171–172. https://d oi. org/1 0.1 038/ s41558-0 20-0 716-1 Busico G, Giuditta E, Kazakis N, Colombani N (2019) A hybrid GIS and AHP approach for modelling actual and future forest fire risk under climate change accounting water resources attenuation role. Sus- tainability 11:7166. https://d oi. org/1 0. 3390/ su112 47166 Butler B, Anderson W, Catchpole E, 2007. Influence of slope on fire spread rate, In: Butler, Bret W.; Cook, Wayne, comps. The fire environment--innovations, management, and policy; conference proceedings. 26–30 March 2007; Destin, FL. Proceedings RMRS-P-46CD. Fort Collins, CO: US Department of Agriculture, Forest Service, Rocky Mountain Research Station. CD-ROM. p. 75–82. Caldararo N (2002) Human ecological intervention and the role of forest fires in human ecology. Sci Total Environ 292:141–165. https://d oi.o rg/ 10.1 016/S 0048-9 697(01) 01067-1 Castro R, Chuvieco E (1998) Modeling forest fire danger from geographic information systems. Geocarto Int 13:15–23. https:// doi.o rg/ 10. 1080/ 10106 04980 935462 4 Cetin M, Aksoy T, Cabuk SN, Kurkcuoglu MAS, Cabuk A (2021a) Employing remote sensing technique to monitor the influence of newly established universities in creating an urban development process on the respective cities. Land Use Policy 109:105705 Cetin M, Agacsapan B, Cabuk SN, Senyel Kurkcuoglu MA, Isik Pekkan O, Baran Argun E, Cabuk A (2021b) Assessment of the ecological footprint of eskisehir technical university-iki eylul campus. J Indian Soc Remote Sens 49(10):2311–2327 1 3 Natural Hazards Chander G, Markham B (2003) Revised Landsat-5 TM radiometric calibration procedures and postcalibra- tion dynamic ranges. IEEE Trans Geosci Remote Sens 41:2674–2677. https:// doi. org/ 10. 1109/ TGRS. 2003. 818464 Chuvieco E (2009) Earth observation of wildland fires in Mediterranean ecosystems. Springer Chuvieco E, Salas J (1996) Mapping the spatial distribution of forest fire danger using GIS. Int J Geogr Inf Sci 10:333–345. https:// doi.o rg/1 0.1 080/0 2693 799608 9020 82 Chuvieco E, Martin MP, Palacios A (2002) Assessment of different spectral indices in the red-near-infrared spectral domain for burned land discrimination. Int J Remote Sens 23:5103–5110. https:// doi. org/ 10. 1080/ 01431 16021 01531 29 Cosgun U, Kavgaci A, Güngöroglu C, González-Cabán A, (2019) Socioeconomic factors affecting forest fires: a case study of Antalya, Turkey. Gen. Tech. Rep. PSW-GTR-261 (English). Albany, CA: US Department of Agriculture, Forest Service, Pacific Southwest Research Station: 96–97. 261. Çolak E, Sunar F (2020) Evaluation of forest fire risk in the Mediterranean Turkish forests: A case study of Menderes region, Izmir. Int J Disaster Risk Reduct 45:101479. https:// doi. org/1 0. 1016/j. ijdrr. 2020. 101479 de la Barrera F, Barraza F, Favier P, Ruiz V, Quense J (2018) Megafires in Chile 2017: monitoring multi- scale environmental impacts of burned ecosystems. Sci Total Environ 637–638:1526–1536. https://d oi. org/ 10.1 016/j.s cito tenv. 2018.0 5.1 19 Doerr SH, Santín C (2016) Global trends in wildfire and its impacts: perceptions versus realities in a chang- ing world. Phil Trans Rl Society b Biol Sci 371:20150345. https://d oi. org/ 10.1 098/r stb. 2015. 0345 Dupuy J-L (1995) Slope and fuel load effects on fire behavior: laboratory experiments in pine needles fuel beds. Int J Wildland Fire 5:153–164. https://d oi. org/ 10.1 071/ WF995 0153 Escuin S, Navarro R, Fernández P (2008) Fire severity assessment by using NBR (normalized burn ratio) and NDVI (normalized difference vegetation index) derived from LANDSAT TM/ETM images. Int J Remote Sens 29:1053–1073. https:// doi.o rg/ 10. 1080/ 014311 6070 12810 72 Fatma E, Vedat A (2018) Forest fire susceptibility analysis of kahramanmaras province. Bingöl Üniv Sos Bilim Enst Derg 8:335–356. https://d oi.o rg/ 10. 29029/ busbed. 437858 Finney MA (2005) The challenge of quantitative risk analysis for wildland fire. For Ecol Manage 211:97– 108. https://d oi.o rg/ 10. 1016/j. foreco. 2005.0 2. 010 Ganteaume A, Camia A, Jappiot M, San-Miguel-Ayanz J, Long-Fournel M, Lampin C (2013) A review of the main driving factors of forest fire ignition over Europe. Environ Manage 51:651–662. https:// doi. org/1 0.1 007/ s00267- 012- 9961-z Gençay G, Mercimek A (2019) Public consciousness and influence of law on forest crimes: insights from Kastamonu. Turk For Policy Econ 106:101978. https://d oi. org/ 10. 1016/j. forpol.2 019.1 01978 Ghorbanzadeh O, Blaschke T, Gholamnia K, Aryal J (2019a) Forest fire susceptibility and risk mapping using social/infrastructural vulnerability and environmental variables. Fire 2:50. https://d oi. org/1 0. 3390/ fire2 030050 Ghorbanzadeh O, Blaschke T, Gholamnia K, Aryal J, (2019b) Forest Fire Susceptibility and Risk Mapping Using Social/Infrastructural Vulnerability and Environmental Variables. MDPI. Giannakopoulos C, Le Sager P, Bindi M, Moriondo M, Kostopoulou E, Goodess CM (2009) Climatic changes and associated impacts in the Mediterranean resulting from a 2 °C global warming. Glob Planet Chang 68:209–224. https:// doi. org/ 10. 1016/j. glopl acha.2 009.0 6.0 01 Hernandez-Leal PA, Arbelo M, Gonzalez-Calvo A (2006) Fire risk assessment using satellite data. Adv Space Res 37(4):741–746 Holsinger L, Parks SA, Miller C (2016) Weather, fuels, and topography impede wildland fire spread in west- ern US landscapes. For Ecol Manage 380:59–69. https://d oi.o rg/1 0. 1016/j.f oreco.2 016. 08.0 35 Hood S (2010) Mitigating old tree mortality in long-unburned, fire-dependent forests: a synthesis (Fort Col- lins, CO: US Department of Agriculture, Forest Service, Rocky Mountain Research Station) 71. Gen Tech Rep Jabbar FK, Grote K, Tucker RE (2019) A novel approach for assessing watershed susceptibility using weighted overlay and analytical hierarchy process (AHP) methodology: a case study in eagle creek watershed, USA. Environ Sci Pollut Res 26:31981–31997. https://d oi. org/1 0. 1007/ s11356- 019-0 6355-9 Jaiswal RK, Mukherjee S, Raju KD, Saxena R (2002) Forest fire risk zone mapping from satellite imagery and GIS. Int J Appl Earth Obs Geoinf 4:1–10. https://d oi. org/1 0. 1016/S 0303- 2434(02) 00006-5 Karabulut M, Karakoç A, Gürbüz M, Kızılelma Y (2013) Coğrafi bilgi sistemleri kullanarak başkonuş dağinda (kahramanmaraş) orman yangini risk alanlarinin belirlenmesi. Uluslarar Sos Araştırmalar Derg 6:171–179 Ljubomir G, Pamučar D, Drobnjak S, Pourghasemi HR, 2019. 15 - Modeling the spatial variability of for- est fire susceptibility using geographical information systems and the analytical hierarchy process, In: 1 3 Natural Hazards Pourghasemi, H.R., Gokceoglu, C. (Eds.), Spatial Modeling in GIS and R for Earth and Environmental Sciences. Elsevier, pp. 337–369. Luke R, McArthur A (1978) Bushfires in Australia Canberra. Australian Government Publishing Ser- vice, Australia Mandal S, Mondal S (2019). Weighted overlay analysis (woa) model, certainty factor (cf) model and analytical hierarchy process (ahp) model in landslide susceptibility studies, Statistical approaches for landslide susceptibility assessment and prediction. Springer, pp 135–162 Martín M, Gómez I, Chuvieco E (2005) Performance of a burned-area index (BAIM) for mapping Medi- terranean burned scars from MODIS data, In: Proceedings of the 5th International workshop on remote sensing and gis applications to forest fire management: fire effects assessment. Paris, Uni- versidad de Zaragoza, GOFC GOLD, EARSeL, pp 193–198 McArthur AG, (1967) Fire behaviour in eucalypt forests Modugno S, Balzter H, Cole B, Borrelli P (2016) Mapping regional patterns of large forest fires in wild- land-urban interface areas in Europe. J Environ Manage 172:112–126. https://d oi.o rg/1 0. 1016/j. jenvm an.2 016. 02. 013 Molina JR, y Silva FR (2019) Valuation of the economic impact of wildland fires on landscape and recreation resources: a proposal to incorporate them on damages valuation. Gen Tech Rep PSW- GTR-261 (English). Albany, CA: US Department of Agriculture, Forest Service, Pacific Southwest Research Station: 261: 228–238 Mollicone D, Eva HD, Achard F (2006) Human role in Russian wild fires. Nature 440:436–437. https:// doi.o rg/ 10. 1038/4 4043 6a Mota PHS, Rocha SJSSd, Castro NLMd, Marcatti GE, França LCdJ, Schettini BLS, Villanova PH, San- tos HTd, dos Santos AR (2019) Forest fire hazard zoning in mato grosso state. Brazil Land Use Policy 88:104206. https:// doi. org/1 0. 1016/j.l andu sepol. 2019. 104206 Naderpour M, Rizeei HM, Khakzad N, Pradhan B (2019) Forest fire induced Natech risk assessment: a survey of geospatial technologies. Reliab Eng Syst Saf 191:106558. https://d oi. org/1 0. 1016/j.r ess. 2019. 106558 Nami M, Jaafari A, Fallah M, Nabiuni S (2018) Spatial prediction of wildfire probability in the hyrcan- ian ecoregion using evidential belief function model and GIS. Int J Environ Sci Technol 15:373– 384. https:// doi. org/1 0.1 007/ s13762- 017-1 371-6 Nasi R, Dennis R, Meijaard E, Applegate G, Moore P, (2002) Forest fire and biological diversity. UNA- SYLVA-FAO: 36–40. Nolan RH, Boer MM, Collins L, Resco de Dios V, Clarke H, Jenkins M, Kenny B, Bradstock RA (2020) Causes and consequences of eastern Australia’s 2019–20 season of mega-fires. Glob Change Biol. https://d oi.o rg/ 10. 1111/g cb.1 4987 Noonan EK (2003) A coupled model approach for assessing fire hazard at point Reyes national seashore: FlamMap and GIS. In: Orlando FL (ed) Second international wildland fire ecology and fire man- agement congress and fifth symposium on fire and forest meteorology. American Meteorological Society, pp 127–128 Nuthammachot N, Stratoulias D (2019) A GIS- and AHP-based approach to map fire risk: a case study of kuan kreng peat swamp forest. Thail Geocarto Int. https://d oi.o rg/ 10. 1080/ 101060 49.2 019. 16119 46 OGM (2018) General directorate of forestry. https://w ww. ogm. gov.t r/ ekutu phane/I stati stik ler/ Ormanc% C4% B1l% C4% B1k% 20% C4% B0sta tisti kleri/ Ormanc% C4% B1l% C4% B1k% 20% C4% B0sta tisti kleri%2 02018.r ar. Patz JA, Frumkin H, Holloway T, Vimont DJ, Haines A (2014) Climate change: challenges and opportu- nities for global health. JAMA 312:1565–1580. https:// doi. org/ 10.1 001/ jama.2 014.1 3186 Pekkan OI, Kurkcuoglu MAS, Cabuk SN, Aksoy T, Yilmazel B, Kucukpehlivan T, Cetin M (2021) Assess- ing the effects of wind farms on soil organic carbon. Environ Sci Pollut Res 28(14):18216–18233 Ray D, Nepstad D, Moutinho P (2005) Micrometeorological and canopy controls of fire susceptibility in a forested amazon landscape. Ecol Appl 15:1664–1678. https://d oi.o rg/ 10. 1890/ 05- 0404 Rogan J, Miller J, Wulder M, Franklin S (2006) Integrating GIS and remotely sensed data for mapping forest disturbance and change. Understanding forest disturbance and spatial pattern: Remote sens- ing and GIS approaches: 133–172 Romero-Calcerrada R, Novillo C, Millington J, Gomez-Jimenez I (2008) GIS analysis of spatial pat- terns of human-caused wildfire ignition risk in the SW of Madrid (Central Spain). Landscape Ecol 23:341–354. https://d oi.o rg/ 10.1 007/s 10980- 008- 9190-2 Saaty TL (1990) How to make a decision: the analytic hierarchy process. Eur J Oper Res 48:9–26 Sanford RL, Saldarriaga J, Clark KE, Uhl C, Herrera R (1985) Amazon rain-forest fires. Science 227:53. https://d oi. org/1 0. 1126/ scien ce.2 27. 4682. 53 1 3 Natural Hazards Shim JP (1989) Bibliographical research on the analytic hierarchy process (AHP). Socioecon Plann Sci 23:161–167. https://d oi. org/ 10.1 016/ 0038- 0121(89) 90013-X Siachalou S, Doxani G, Tsakiri-Strati M (2009). Integrating remote sensing processing and GIS to fire risk zone mapping: a case study for the Seih-Sou forest of Thessaloniki, In: Proceeding of ICC Sivrikaya F, Sağlam B, Akay AE, Bozali N (2014) Evaluation of forest fire risk with GIS. Pol J Environ Stud 23 Sullivan AL, Sharples JJ, Matthews S, Plucinski MP (2014) A downslope fire spread correction factor based on landscape-scale fire behaviour. Environ Model Softw 62:153–163. https:// doi. org/ 10. 1016/j.e nvso ft. 2014. 08. 024 Townsend SA, Douglas MM (2000) The effect of three fire regimes on stream water quality, water yield and export coefficients in a tropical savanna (northern Australia). J Hydrol 229:118–137. https://d oi.o rg/ 10. 1016/ S0022- 1694(00) 00165-7 Vadrevu KP, Eaturu A, Badarinath K (2010) Fire risk evaluation using multicriteria analysis: a case study. Environ Monit Assess 166:223–239. https://d oi. org/ 10. 1007/s 10661-0 09-0 997-3 Van Wagner C (1988) Effect of slope on fires spreading downhill. Can J for Res 18:818–820. https://d oi. org/ 10. 1139/ X07-1 59 Verma S, Jayakumar S (2012) Impact of forest fire on physical, chemical and biological properties of soil: a review. Proc Int Acad Ecol Environ Sci 2:168 Viegas DX (2004) Slope and wind effects on fire propagation. Int J Wildland Fire 13:143–156. https:// doi. org/ 10. 1071/W F0304 6 Wang N-Y, Wang LK, Wang M-HS, Wong JO (2019) Scientific studies of global warming, climate change, glacier melting and salmon protection. STEAM 1(9):1–110 Wotton B, Stocks B, Martell D (2005) An index for tracking sheltered forest floor moisture within the Cana- dian Forest fire weather index system. Int J Wildland Fire 14:169–182. https:// doi. org/1 0. 1071/ WF040 38 Xavier Viegas D (2004) A mathematical model for forest fires blowup doi: https:// doi.o rg/ 10.1 080/0 0102 20059 08836 24 Yathish H, Athira K, Preethi K, Pruthviraj U, Shetty A (2019) A comparative analysis of forest fire risk zone mapping methods with expert knowledge. J Indian Soc Remote Sens 47:2047–2060. https://d oi. org/1 0. 1007/ s12524-0 19- 01047-w Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Springer Nature or its licensor holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law. 1 3