GIS‐based spatiotemporal analysis of forest fires in Turkey from 2010 to 2020

Forests are essential in contributing to the continuity of the natural balance. Therefore, their protection and sustainability are vital. However, all over the world, forest fires occur, and forests are destroyed due to both human factors and unknown causes. It is necessary to carry out studies to prevent this destruction. At this point, GIS‐based location–time relationship‐based hot spot clustering analysis can provide significant advantages in detecting risky spots of forest fires. In this study, GIS‐based emerging hot spot clustering analysis was carried out to determine the risky areas where forest fires will occur and to carry out preventive studies in the relevant areas. Turkey was chosen as the pilot region, and analyses were carried out using the data obtained from the official statistics of the Ministry of Agriculture and Forestry General Directorate of Forestry according to the causes of the fires (negligence, intentional, accidental, unknown cause and natural) between the years 2010 and 2020. Spatial autocorrelation analysis was conducted for each fire type, and threshold distances were determined {with a number of distance bands = 20,000, distant increment = 10,000}. Emerging hot spot analyses were then conducted, and the results were presented as maps and statistical outputs. According to all fire types, 15 new hot spots, 14 persistent hot spots, 33 sporadic hot spots, 9 consecutive hot spots, 15 intensifying, and 2 diminishing hot spot regions were obtained throughout the country.

to detecting hot spot areas and conducting studies to reduce forest fire rates in designated areas. Many GISbased hot spot clustering methods have been uncovered throughout the literature, and then different uses are discussed. Considering these methods and GIS can be straightforward in detecting forest fire hotspots. In recent years, GIS-based hot spot clustering methods have emerged with new methods. While classical hot spot clustering methods only make a spatial evaluation when performing the analyses, hot spot clustering analyses considering the location-time relationship can examine both location and temporal relationships under a single framework and present the results depending on time. When the literature is examined, we encounter the space-time cube analysis and the emerging hot spot analysis based on the space-time cubes obtained with this analysis. Emerging hot spot analysis explains the phenomenon "…everything is related to everything else, but near things are more related than distant things" (Tobler, 1970), in which Tobler defines geography as "…everything is related to everything else, but near and recent things are more" related than distant things," revealing the importance of time and distance. Thus, adding time to classical statistical methods deals with obtaining hot spot regions more meaningfully.
Thus, it eliminates the burden of classical clustering methods based on the location-time relationship. Emerging hot spot analyses can provide convenience by establishing a location-time relationship to detect risky areas where forest fires will occur and taking the relevant precautions.
In this study, a GIS-based emerging hot spot clustering analysis was carried out to determine the risky areas where forest fires will occur and to carry out preventive studies in the relevant areas. Turkey was chosen as the pilot region for the study, and analyses were carried out using the data obtained from the official statistics of the Ministry of Agriculture and Forestry General Directorate of Forestry according to the causes of the fires (negligence, intentional, accidental, unknown cause, and natural) between the years 2010 and 2020. The study primarily carried out clustering trend analysis, and the threshold distances required for the analysis were determined. Emerging hot spot analyses were carried out using the required distance values, and the results were analyzed.

| LITER ATURE BACKG ROUND
Emerging hot spot clustering analyses reveal more effective results based on location-time than classical clustering analyses. These methods, which have been tried to be used frequently in many areas in recent years, have also gained importance in determining the risk points of forest fires. At this point, although studies are being carried out, such a study has yet to be discussed, especially in Turkey, and the lack of many studies in the forest fire field has been influential in the study. Literature studies have been handled by examining the difference between this study from other studies and its beneficial contributions to the literature. Table 1 shows which methods each study deals with and whether they have standard methods with this study. In this table, the studies (Moran's I, Getis-Ord Gi*, Anselin Local Moran's I, space-time cube analysis, emerging hot spot analysis, or other methods) were examined, and the differences of this study were handled.
Therefore, based on the location-time relationship, this study aims to execute the change of forest fires depending on the years examined according to the causes of the fires. This study will primarily contribute to many studies in Turkey. It will set an example for the countries and literature where the causes of forest fires are similar to Turkey. will be between 0 and þ1, indicating a positive relationship between the variables and their cluster distribution.
The Moran's I statistic for spatial autocorrelation is given as: where, z i is the deviation of an attribute for feature i from its mean (xi -x), w i,j is the spatial weight between feature i and j, n is equal to the total number of features, and S o is the aggregate of all spatial weights: The z I -score for the statistic is computed as: where 2.1.1 | Incremental spatial autocorrelation The incremental spatial autocorrelation tool measures spatial autocorrelation for a series of distances and optionally creates a line graph of those distances and their corresponding z-scores. In other words, it runs the spatial autocorrelation (Global Moran's I) tool for a range of increasing distances and measures the intensity of the spatial clustering for each distance.
The density of the cluster is determined by the z-score returned. As the distance increases, so does the z-score, which indicates the aggregation concentration, and the z-score usually peaks at a given distance. At this point, the peak levels of more than one z-score can be determined. Peaks reflect distances where the spatial processes promoting clustering are most pronounced ( Figure 1).
The color of each point on the graph corresponds to the statistical significance of the z-score values (https:// pro.arcgis.com/en/pro-app/lates t/tool-refer ence/spati al-stati stics/ incre mental-spati al-autoc orrel ation.htm). In this study, incremental spatial autocorrelation analysis is considered to calculate the threshold distance to perform emerging hot spot analysis for forest fires.

| Space-time cube analysis
Space-time cube analysis provides visualization and analyses spatiotemporal data through time-series analysis, integrated spatial and temporal pattern analysis, and 2D and 3D visualization techniques. The space-time cube is comprises space-time bins with the x and y dimensions representing space and the t dimension representing time.
Every bin has a fixed position in space (x, y) and in time (z) (Esri, 2022a Due to the use of these point data, the analysis process is discussed. The operation of the method when using aggregating point data is as follows; an illustration of this analysis is shown in Figure 2.
a. In each cube bin, the points are counted, and summary field statistics are calculated. Thus, the data are expressed in cubes in terms of distance and time, and the t dimension is added to the x and y points of the cubes. These cubes are combined on the horizontal axis using the Getis-Ord Gi* statistic, and a spatial data set is created for a specific period.
b. The trend for bin values across time at each vertical axis is measured using the Mann-Kendall statistic, c. A grid cube is created by aggregating using a fishnet or hexagon grid.
d. After all the operations are done, the data are stored in multidimensional NetCDF (Network Common Data Form) format in the ArcGIS program.
e. At least 10 time periods (seconds, minutes, hours, days, weeks, months, seasons, years) must be used to create cubes. A minimum of 10 years is required for trend analysis.
In this study, space-time cube analysis is used in the emerging hot spot analysis to be carried out to determine the intensity of forest fires in the selected pilot region.

| Emerging hot spot analysis
Tobler defined geography as "…everything is related to everything else, but near things are more related than distant things" (Tobler, 1970). Classical statistical methods assume that a variable's data are independent. Therefore, distance is kept in the foreground while explaining the relationships between the variables. However, in recent years, time and distance have become critical. For this reason, it would be more meaningful to add time to the concept of proximity in Tobler's approach. In other words, the expression of Tobler's approach as "… everything is related to everything else, but near and recent things are more related than distant things" would be more appropriate.
While classical hot spot analyses were only analyzed separately in spatially different years in previous studies conducted based on Geographic Information Systems, emerging hot spot analyses were developed to handle location and time relationships under a single analysis. The space-time cube method is the analysis that can perform the specified analyses in one go, and it can detect hot and cold spots thanks to trend analysis (Esri, 2022b).
Emerging hot spot analysis uses the net.CDF files are generated from space-time cube analysis and detect hot and cold spot areas based on time ( Figure 3).
The hot spots in the emerging hot spot analysis are collected in 8 categories for hot and cold spots, 17 categories in total, according to their trend over time ( Table 2).

| Methodology
The study consists of eight stages ( Figure 4).
1. Selecting the study area.
2. Obtaining the fire data from the official statistics of the Ministry of Agriculture and Forestry General Directorate of Forestry (according to the causes of the fires of the years 2010-2020; the number of fires, negligence, intentional, accidental, unknown cause, and natural).
3. Organizing the provided data associated with the coordinate, and conversion to the WGS_1984_Web_Mercator_ Auxiliary_Sphere projection system.
4. Collecting the edited data in the geodatabase in ArcGIS 10.8.
5. Performing the spatial autocorrelation analyses and determining the threshold distance to be used in the analyses to be carried out according to each fire type.
6. Generating space-time cubes based on years for each fire type (net.CDF).
7. Performing an emerging hot spot analysis using space-time cubes.
8. Presenting all outputs as visual maps.

F I G U R E 3
Emerging hot spot analysis using NetCDF files generated from space-time cube illustration (Esri, 2022b;Vale, 2018).

TA B L E 2
Emerging hot spot categories and definitions (Esri, 2022b).

Figure Hot spot category Definition
No pattern detected Does not fall into any of the hot or cold spot patterns defined below New hot spot A location that is a statistically significant hot spot for the final time step and has never been statistically significant hot spot before Consecutive hot spot A location with a single uninterrupted run of at least two statistically significant hot spot bins in the final time-step intervals. The location has never been a statistically significant hot spot prior to the final hot spot run and less than 90% of all bins are statistically significant hot spots Intensifying hot spot A location that has been a statistically significant hot spot for 90% of the time-step intervals, including the final time step. In addition, the intensity of clustering of high counts in each time step is increasing overall and that increase is statistically significant Persistent hot spot A location that has been a statistically significant hot spot for 90% of the time-step intervals with no discernible trend in the intensity of clustering over time (Continues)

| Study area
This study was carried out to cover 81 provinces of Turkey ( Figure 5). Turkey is located in the northern hemisphere and the middle belt, between longitudes 26°-45° east and latitudes 36°-42° north. Turkey is a Eurasian country with territories on two continents. It is 1600 km long and 800 km wide. It covers an area of 783,562 km 2

Figure Hot spot category Definition
Diminishing hot spot A location that has been a statistically significant hot spot for 90% of the time-step intervals, including the final time step. In addition, the intensity of clustering in each time step is decreasing overall and that decrease is statistically significant Sporadic hot spot A statistically significant hot spot for the final time-step interval with a history of also being an on-again and off-again hot spot. Less than 90% of the time-step intervals have been statistically significant hot spots and none of the time-step intervals have been statistically significant cold spots Oscillating hot spot A statistically significant hot spot for the final time-step interval that has a history of also being a statistically significant cold spot during a prior time step. Less than 90% of the time-step intervals have been statistically significant hot spots Historical hot spot The most recent time period is not hot, but at least 90% of the time-step intervals have been statistically significant hot spots New cold spot A location that is a statistically significant cold spot for the final time step and has never been a statistically significant cold spot before

Consecutive cold spot
A location with a single uninterrupted run of at least two statistically significant cold spot bins in the final time-step intervals. The location has never been a statistically significant cold spot prior to the final cold spot run and less than 90% of all bins are statistically significant cold spots Intensifying cold spot A location that has been a statistically significant cold spot for 90% of the time-step intervals, including the final time step. In addition, the intensity of clustering of low counts in each time step is increasing overall and that increase is statistically significant Persistent cold spot A location that has been a statistically significant cold spot for 90% of the time-step intervals with no discernible trend in the intensity of clustering of counts over time Diminishing cold spot A location that has been a statistically significant cold spot for 90% of the time-step intervals, including the final time step. In addition, the intensity of clustering of low counts in each time step is decreasing overall and that decrease is statistically significant Sporadic cold spot A statistically significant cold spot for the final time-step interval with a history of also being an on-again and off-again cold spot. Less than 90% of the time-step intervals have been statistically significant cold spots and none of the time-step intervals have been statistically significant hot spots Oscillating cold spot A statistically significant cold spot for the final time-step interval that has a history of also being a statistically significant hot spot during a prior time step. Less than 90% of the time-step intervals have been statistically significant cold spots Historical cold spot The most recent time period is not cold, but at least 90% of the time-step intervals have been statistically significant cold spots TA B L E 2 (Continued) | 1299

MEMISOGLU BAYKAL
with its lakes. While 755,688 km 2 of this constitutes the territory of Asia, the remaining 23,764 km 2 constitutes Europe's territory. The reason for choosing Turkey in the study is to seek answers to the questions such as the occurrence of forest fires in the country in recent years, how these fires have changed over the years according to the causes of these fires, and how the planning will be developed by determining the areas where forest fires will occur.

| Data
Forest fire data obtained from the official statistics of the Ministry of Agriculture and Forestry General Directorate of Forestry were used in the study. The data were obtained from the relevant directorate page and arranged in Microsoft Excel format (.xls). Since there is a need for at least 10 time periods every year in the emerging hot spot analysis, 10 years of non-spatial data for 2010-2020 were obtained from the relevant page for the analysis. These non-spatial data are the number of fires, negligence, intentional, accidental, unknown cause, and natural according to fire causes. The recorded provincial-based annual forest fire statistics are discussed in Figure 6. Subsequently, location information was added to the non-spatial data provided, and the data were made related to the location. All data were imported into ArcGIS 10.8 and converted into a single projection system F I G U R E 4 Methodology.
(WGS_1984_Web_Mercator_Auxiliary_Sphere). All the edited data were finally combined in the geodatabase and made ready for analysis ( Figure 7).

| Spatial autocorrelation analysis
In the first stage, whether the data showed a clustering tendency or not was tested. Incremental spatial autocorrelation analysis was performed on the organized data for four interval years to determine the distance bandwidth used in clustering analyses, and analyses were carried out using the obtained threshold distance values.
While executing the incremental spatial autocorrelation, the minimum distance for each feature in the data set with at least one neighbor is set as the default value for the beginning distance. As a result, the beginning distance for incremental spatial autocorrelation analysis increases since more significant regions (polygons) have longer distance to reach at least one neighbor from their centroids. In this study, incremental spatial autocorrelation analysis was performed using classified forest fires {with a number of distance bands = 20,000, distant increment = 10,000}. Statistical results were handled for the study area in Table 3. Since the analyses conducted in the F I G U R E 5 Study area.

| Space-time cube analysis
At this study stage, the space-time cubes required to be used in the emerging hot spot analysis were created. 100 × 100 km grids were determined so all analyses could be carried out simultaneously. This situation is because the minimum threshold distance value in the spatial autocorrelation analysis outputs is 130,000, and all neighborhood distances should be smaller than this value in the analyses. Given 10 years was needed F I G U R E 7 Associated forest fire data.

TA B L E 3
Statistical results of forest fire data types in Turkey. for the analysis, the space-time cubes were created to include 2010-2020. Forest fires falling into each cube were counted separately according to each fire type (number of fires, negligence, intentional, accidental, cause unknown, natural), and fire totals were obtained. The outputs were stored in *net.CDF formats and arranged for use in emerging hot spot analysis. The hot spot changes of the space-time cubes obtained according to each forest fire type according to the years were created and visualized in 3D in ArcScene in Figure 8.

Forest fires types
ArcScene is a 3D visualization application that allows viewing GIS data in three dimensions. It is an efficient program that allows multiple layers of data to be superimposed in a 3D environment by providing height information from the 3D surface (https://deskt op.arcgis.com/en/arcma p/lates t/exten sions/ 3d-analy st/3d-analy st-and-arcsc ene.htm). Figure 8 is the map where, due to the space-time cube analysis performed, the hotspot regions of all forest fire types over the years are visualized in 3D. The resulting maps represent statistically significant points at 90, 95, and 99% confidence levels (red dots) and statistically insignificant points (white dots). Thus, it can be seen on a single map in which province a forest fire is at possible risk. However, emerging hot spot analysis is carried out to detect the change in hot spot regions over time, and a single map represents the resulting product. Therefore, after this stage, an emerging hot spot analysis was carried out to determine where forest fires pose a more extreme risk over the years.

| Emerging hot spot analysis
At this study stage, emerging hot spot analyses were carried out according to each type of forest fire, using the *net.CDF formatted analysis outputs obtained in the previous stage. The statistical distribution of the emerging hot spot results according to each forest fire type is shown in Figure 9. According to all fire types, 15 new hot spots, 14 persistent hot spots, 33 sporadic hot spots, 9 consecutive hot spots, 15 intensifying, and 2 diminishing hot spots for the selected pilot region when Figure 9 is examined.
The results obtained were examined one by one depending on the causes of the fires.
For the number of fires: Four persistent hotspots and seven sporadic hotspots were obtained in the selected region ( Figure 10). When the resulting map is examined, • Persistent hot spots were encountered in the southeast of Balıkesir, Bursa, Kütahya, Uşak, western parts of Denizli, and most of the provinces of İzmir, Manisa, and Aydın.
• The sporadic hot spot was seen in the Anatolian side of Istanbul, Western Kocaeli, most of Yalova, and northern part of Bursa and a small part of Çanakkale, southwestern part of Balıkesir, north of İzmir and north-west of Manisa. In addition, the south of Kayseri, the western part of Kahramanmaraş, the east of Adana, the whole of Osmaniye and the western part of Gaziantep, and a small part of Kilis have also been identified as sporadic hot spots.
• In the remaining parts of Turkey, statistically significant points could not be determined according to the number of fires depending on the years.
When surface-area evaluations of hot spot regions were made for a number of fires, persistent hot spot areas constituted 1.8% of the total area, and sporadic hot spot areas constituted 2.9% of the total area.

F I G U R E 9
Distribution of total emerging hot spots in Turkey (2010-2020).
For negligence from fire types: one consecutive hot spot region, four intensifying hot spot regions, one new hot spot region, and eight persistent hot spot regions were obtained in the selected region ( Figure 11). When the resulting map is examined, • Persistent hot spot regions were encountered in the southeast of Balıkesir, a small part of Bursa, west of Kütahya, almost all parts of Uşak, Manisa, İzmir, Aydın and a small part of Muğla and Denizli provinces.
• Intensifying hot spot areas were seen east of Çanakkale, more than half of Balıkesir, and a small part of Manisa and İzmir provinces.
• A consecutive hot spot region has been detected in a southern part of İzmir.
• A new hot spot region was detected in most parts of Bursa and a small part of Balıkesir.
• In the remaining parts of Turkey, a statistically significant result could not be obtained over the years due to negligence. Therefore, it can be seen that forest fires due to negligence are intensified in western Turkey.
When surface-area evaluations of hot spot regions were made for negligence from fire types, consecutive hot spot zones make up 0.23% of the entire area, followed by intensifying hot spot zones at 2.35%; new hot spot zones at 0.76%; and persistent hot spot regions at 5.2%.

MEMISOGLU BAYKAL
• When the consecutive hot spot analysis results are evaluated on a regional basis, İzmir constitutes 92.5% of consecutive areas, while Aydın has 7.5%.
• When Intensifying hot spot analysis results are evaluated on a regional basis, Balıkesir ranks first with 50.8%, followed by İzmir with 26.79%; Çanakkale at 11.44%; Manisa with 10.89% and Bursa with 0.06%.
• When the new hot spot analysis results are evaluated on a regional basis, Bursa ranks first with 83.47%, and Balıkesir ranks second with 16.5%.
For intentional fire types: two diminishing hot spot regions were detected (Figure 12). When the resulting map is examined, • Diminishing hot spot areas have been detected in the south of İzmir, a small part of Manisa, Uşak, and Denizli, and a large part of Aydın.
• In the remaining parts of Turkey, a statistically significant result could not be obtained over the years for the intentional fire type.
When surface-area evaluations of hot spot regions were made for intentional fires, the diminishing hot spot regions make up 1.53% of the entire area.

F I G U R E 11
Emerging hot spot regions depending on negligence in Turkey.
• When the diminishing hot spot analysis results are evaluated on a regional basis, Aydın ranks first with 49.37%, followed by İzmir at 30.74%; Denizli at 13.06%, Manisa at 5.75%; and Uşak at 1.07%.
For the accidental fire type: 1 consecutive hot spot zone; 4 new hot spot zones, and 12 sporadic hot spot zones were detected (Figure 13). When the resulting map is examined; • Consecutive hot spot regions were provided in some parts of Malatya, Elazig, Diyarbakir, Sanliurfa, and Adıyaman provinces.
• New hot spot regions have been detected in a part of Erzincan, a part of Malatya, Bingöl, and Diyarbakır, and almost all parts of Tunceli and Elâzığ provinces.
• Sporadic hot spot regions were determined in Osmaniye, Gaziantep, and Kahramanmaraş provinces, south of Balikesir, a small part of Izmir, half of Manisa, a small part of Aydın, Uşak, Denizli, Burdur and almost all of Muğla, Antalya, Isparta, Konya, Sivas, Kayseri, Adana.
• In the remaining part of Turkey, a statistically significant result could not be obtained over the years for the accidental fire type.
When surface-area evaluations of hot spot regions were made for the accidental fire type, consecutive hot spot zones account for 0.77% of the entire area, followed by new hot spot zones with 3.08% and sporadic hot spot zones with 7.64%.
• When the results of the consecutive hot spot analysis are evaluated on a regional basis, Şanlıurfa ranks first with a rate of 42.84%, followed by Diyarbakır at 39.08%; Adıyaman at 9.58%; Malatya at 7.6%; and Elazig at 0.88%.

F I G U R E 1 2 Emerging hot spot regions for intentional fires in Turkey.
• When the results of the consecutive hot spot analysis are evaluated on a regional basis, Şanlıurfa ranks first with a rate of 42.84%, followed by Diyarbakır at 39.08%; Adıyaman at 9.58%; Malatya at 7.6%; and Elazig at 0.88%.
For the cause unknown fire type: 7 consecutive hot spot zones, 10 new hot spot zones, 14 sporadic hot spot zones were detected (Figure 14). When the resulting map is examined, • New hot spots have been seen in Bartın, Karabük, Kastamonu, and a small part of Sinop, Erzincan, Tunceli, Malatya, Elazığ, Bingöl, Diyarbakır, Adıyaman, most of Şanlıurfa, and a large part of Mardin.
• Consecutive hot spots have been identified in Kayseri, Adana, Osmaniye, a small part of Kahramanmaraş, the western part of Gaziantep, and the north of Kilis. In addition, while the fire has not changed over the years in a small part of Manisa, Uşak, İzmir Aydın, Denizli, Burdur, Muğla, Isparta, Antalya, and Konya, consecutive hot spot regions have been observed due to increases in recent years.
• Sporadic hot spot areas have been encountered in Kırklareli and Tekirdağ, especially in the European side of Istanbul and a small part of Yalova. In addition, a small part of Aydın and Denizli and almost all parts of Muğla, the southern part of Antalya, and small parts of Konya, Karaman, Mersin, Kayseri, Sivas, Kahramanmaraş, Malatya, Adıyaman, Elazığ, Diyarbakır, and Şanlıurfa.

F I G U R E 1 3 Emerging hot spot regions for accidental fires in Turkey.
• In the remaining parts of Turkey, a statistically significant result could not be obtained over the years according to the type of fire whose cause is unknown.
When surface-area evaluations of hot spot regions were made for the cause unknown fire type, consecutive hot spot zones constituted 3.85% of the entire area, followed by new hot spot zones at 6.54% and sporadic hot spot zones at 7.02%.

F I G U R E 1 4 Emerging hot spot regions for cause unknown fires in Turkey.
For the natural fire type: 11 intensifying hot spot zones, 2 persistent hot spot zones have been detected ( Figure 15). When the result map is examined, • Persistent hot spot regions have been detected in most parts of İzmir and Aydın and small parts of Manisa, Uşak, and Denizli.
• In the remaining parts of Turkey, a statistically significant result could not be obtained over the years according to the natural type of fire types.
When surface-area evaluations of hot spot regions were made for natural fires, intensifying hot spot zones make up 5.31% of the entire area, and persistent hot spot zones make up 1.53%.
• When the persistent hot spot analysis are evaluated on a provincial basis, Aydın ranks first with 49.37%, followed by İzmir at 30.74%; Denizli at 13.06%; Manisa at 5.75%; and Uşak at 1.07%.
The analysis results of all fire types in Turkey are summarized and shown in Table 4. When the table is examined, it is observed that the forest fire risk is mainly seen in İzmir, followed by the provinces of Manisa, Denizli, F I G U R E 1 5 Emerging hot spot regions for the natural fire type in Turkey.

TA B L E 4
Analysis results of all fire types.

Provinces
Uşak, Aydın, and Balıkesir. Therefore, it is thought that necessary planning should be done by taking more precautions in these provinces compared with other provinces.

| RE SULTS
Forests are significant natural resources contributing to ecological balance and economic sustainability. Therefore, their protection is an important requirement. However, forest fires are a fundamental problem that cause long-term effects on forests, damages ecosystems, and causes significant economic losses. Efforts should be made to prevent these fires. At this point, GIS-based location-time relationship-based hot spot clustering analysis, which has come to the fore in many areas in recent years, provides significant advantages in determining the risky locations of forest fires. Space-time cube analysis is essential in terms of being an analysis in which time and location information are evaluated together, being important in the transition from classical hotspot clustering methods to developing hotspot clustering methods, and performing all analyses simultaneously by hosting more than one time zone at the same time.
With the developing hot spot clustering analysis based on space-time cubes, risky points of forest fires can be determined based on the time-location relationship, and necessary measures can be taken.
In this study, GIS-based emerging hot spot clustering analysis was carried out to determine the risky areas where forest fires will occur and to carry out preventive studies in the relevant areas. Turkey was chosen as the pilot region for the study, and analyses were carried out using data obtained from the official statistics of the Ministry of Agriculture and Forestry General Directorate of Forestry according to the causes of the fires (negligence, intentional, accidental, unknown cause and natural) between the years 2010 and 2020.
The study used clustering trend analysis after the threshold distances required for the analysis were determined. Emerging hot spot analyses were carried out using the required distance values, and the results were analyzed. As a result of emerging hot spot analyses, the hot spot numerical results obtained are as follows: 4 persistent, 7 sporadic for the number of fires; 1 consecutive, 4 intensifying, 1 new, and 8 persistent for the negligent fire type; 2 diminishing for the intentional fire type; 1 consecutive, 4 new, 12 sporadic for the accidental fire type; 7 consecutive, 10 new, and 14 sporadic for the cause unknown fire type; 11 intensifying, 2 persistent for the natural fire type. According to all fire types, 15 new hot spots, 14 persistent hot spots, 33 sporadic hot spots, 9 consecutive hot spots, 15 intensifying, and 2 diminishing hot spot regions were obtained throughout the country. While all of these identified hot spot areas are important, it is important to focus on controlling the new hot spot areas occurring in some fire types and determining why such an increase has occurred within the specified years. Apart from this, it is essential to examine the hot spot regions that are continuous, sometimes hot spots and increasingly intensifying in terms of prevention, reduction and planning of forest fires.
This study will fill an essential gap in this field since no study considers the location-time relationship in forest fires throughout Turkey. In addition, very few studies have evaluated the change of forest fires over the years throughout the world. Therefore, the results obtained are important as they will shed light on similar studies and provide a basis for studies in similar countries. Thanks to this study, unlike the classical hot spot clustering analyses, an easier and faster analysis can be carried out by conducting the analyses under a single space-time framework; such that the change in forest fires over the years can be questioned through a single analysis. Thus, the change over the years will be examined, preventing the occurrence of forest fires and dealing with forest fires more effectively at the once they occur.

ACK N OWLED G M ENTS
Thanks, the General Directorate of Forestry of the Ministry of Agriculture and Forestry for sharing the forest fire on its official website.

CO N FLI C T O F I NTE R E S T S TATE M E NT
The author declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this article.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.