Development of a Decision Support Model for Establishing Response Strategies to Large-scale Wildfires Using GIS

Minseok Kim; Yoonha Lee

doi:10.7731/KIFSE.c3023a49

Abstract

Every year, wildfire disasters of various scales occur in Korea, causing damage to forests, facilities, and human lives owing to the difficulties in early suppression. Effective wildfire response requires collaboration between multiple agencies, necessitating a decision support system for strategy formulation. However, while there is active research on technology development related to situation prediction and suppression equipment, discussions on response technologies and strategies are insufficient. In this study, we propose a GIS-based decision-making methodology for establishing initial response strategies through prioritization of suppression zones during wildfires. A case study on past wildfires was conducted to evaluate the applicability of the proposed methodology. The methodology suggested in this study integrates wildfire suppression elements using the equal weight method to derive suppression priority scores by combining these elements with the distance to wildfire occurrence points. In this study, we aim to provide a decision-making foundation that considers spatial factors in initial wildfire responses, which can serve as a crucial basis for effective allocation of firefighting resources and strategy formulation in the field.

Keywords: Wildfire; Decision support; GIS

1. Introduction

In South Korea, wildfires of varying scales occur annually [1]. If not suppressed in the early stages, these fires can spread rapidly owing to strong winds [2], leading to extensive damage to forests and facilities, and even human casualties. While preventing wildfires is paramount, it is challenging owing to their diverse causes, climate change, aging populations in forest-adjacent areas, and the difficulty of monitoring vast regions. Additionally, given the expansive damage area and rugged terrain during wildfire spread, much of the firefighting efforts rely heavily on helicopters [3]. Consequently, the importance of devising measures to minimize wildfire damage has increased, necessitating the establishment of national response systems and the enhancement of the capabilities of the field commanders [1]. Thus, there has been a surge in research utilizing sensors, drones, smart thermal imaging CCTV, big data, and AI [4].

Despite active research on technology development related to situation prediction and suppression, discussions on response technologies and strategies remain insufficient. For instance, during large-scale wildfires, the Korea Forest Service, as the primary disaster management agency, and the National Fire Agency, as a supporting agency, mobilize many personnel and equipment. However, because of the simultaneous and extensive spread of wildfires, there is a need to improve the systematic operational capabilities to manage these resources [5]. According to a Yonhap News article [6], during the April 2023 Gangneung wildfire, a forest authority official remarked, "With embers carried by strong winds of 20-30 m/s flying several kilometers and spreading to pensions and homes in various locations, the entire city was engulfed in black smoke, and it was overwhelming to decide where to start extinguishing the fire." This highlights the need for a decision support system in formulating response strategies as wildfire response requires collaboration between various agencies and organizations to adapt to changing weather conditions. Therefore, this study proposes a GIS-based decision support system for establishing initial response strategies during wild-fire occurrences. While wildfire response encompasses a wide range of activities from a disaster management perspective, this study focuses on prioritizing suppression zones.

2. Theoretical Background

2.1 Wildfire response systems and strategies

In South Korea, the Framework Act on the Management of Disasters and Safety designates the Korea Forest Service as the primary disaster management agency for wildfires, and the Forest Protection Act requires the National Fire Agency to actively cooperate in wildfire suppression [7,8]. Additionally, laws pertaining to the National Fire Agency's wildfire response are partially stipulated in the Framework Act on Firefighting Services, and other established rules can be found in Regulations on the Duties and Roles of Wildfire Suppression Institutions [5]. During an actual wildfire event, the National Fire Agency is responsible for defending homes and facilities near forests, while the Korea Forest Service handles aerial firefighting in forested areas [2]. Other agencies such as the local governments, Ministry of National Defense, National Police Agency, Korea Meteorological Administration, Ministry of Environment (including the Korea National Park Service), and Cultural Heritage Administration provide support and cooperation.

The wildfire suppression process is divided into several stages as outlined in the wildfire suppression action plan (Annex 2), where Stage 1 is ignition, Stage 2 is spread, Stage 3 is main fire suppression, Stage 4 is mop-up, Stage 5 is completion of suppression, and Stage 6 is post-fire monitoring. According to Article 14 of the Regulations on the Duties and Roles of Wildfire Suppression Institutions, wildfire suppression plans must prioritize the following during a wildfire incident:

① Protection of human life
② Protection of military facilities, national infrastructure, and cultural heritage
③ Protection of homes and other properties
④ Protection of critical forest resources, such as protected forests, forests for seed collection, genetic resource protection forests, and experimental forests
⑤ Prevention of wildfire spread in other forest areas

2.2 Previous research on wildfire response technologies and strategies

Kim et al. (2019) [9] developed a wildfire response decision tree to address issues, such as command problems, confusion due to a diversified command systems, and reduced disaster communication functionality, that were observed during the 2005 Yangyang wildfire response. Their goal was to improve the command system of the existing wildfire response manual. Kwak et al. (2020) [1] created a decision support checklist for wildfire response, encompassing four categories: disaster response, evacuee relief, emergency facility restoration, and resource support. The disaster response checklist included items such as the current situation at the wildfire site, scale, spread speed, direction, evacuee evacuation and damage status, and resource status. In 2021, Kwak et al. [10] developed a GIS-based algorithm for optimal resident evacuation routes, providing multiple routes to increase evacuation efficiency during wildfires. In the same year, Kwak et al. [11] proposed a defense area setting algorithm to protect facilities within wildfire spread ranges by establishing primary defense areas around these facilities and using road areas within these zones as defensive lines. Son et al. (2023) [12] developed an integrated urban wildfire emergency response system by combining real-time big data, AI, and GIS technology to minimize the time required for situation assessment and decision-making from wildfire monitoring to situation dissemination. This system was tested and its applicability was confirmed in Gangneung, Gangwon State special self-governing province.

While these prior studies have employed various approaches and cutting-edge technologies to provide practical solutions for wildfire response and have evaluated their applicability through empirical studies, they are restricted to specific situations or regions and lack quantitative evaluation methodologies for suppression priorities. To overcome these limitations, this study proposes a generalized model using national spatial data. The proposed methodology, which secures objectivity through GIS-based quantitative evaluation and considers various elements, is designed to be flexible and applicable from initial response to long-term response.

3. Methodology Development and Application to the Study Area

To minimize wildfire damage through various strategies and tactics, such as constructing firebreaks, designating evacuation shelters, and directing aerial firefighting, predicting the wildfire direction in advance is crucial [13]. A comparison of past wildfires, such as the Yangyang wildfire, to successful responses like the Goseong wildfire highlights the importance of rapid decision-making centered on a central control tower, systematic responses based on structured manuals, and accurate situational assessment as critical factors for swift and appropriate decision-making [1]. Therefore, in this study, we aim to develop a decision support system for establishing response strategies during wildfires by leveraging national spatial data and GIS to create initial response strategies for wildfire suppression.

3.1 Selection of study area and establishment of spatial data

In this study, we studied the April 2023 Gangneung wildfire to establish phased suppression zones for initial wildfire response. Using drone aerial footage acquired from April 21-23, 2023, the fire line and spatial information of the target site were constructed, as shown in Figure 1.

To construct the spatial information for the population, cultural heritage, buildings, and forest resources within the study area, national spatial information data provided in *.shp or *.txt format was utilized, as shown in Table 1. The population data was sourced from the population census data of Statistics Korea. For cultural heritage, spatial information provided by the Cultural Heritage Administration was used. Building information was sourced from the National Geographic Information Institute, and forest resource information was obtained from the ecological and natural maps provided by the Korea Forest Service. The spatial information constructed based on these elements—population (a), cultural heritage (b), buildings (c), and forest resources (d)—is illustrated in Figure 2.

3.2 Development of a GIS-based model for formulating initial wildfire response strategies

n this study, a model for establishing initial wildfire response strategies was developed using GIS. The process of the model proposed in this study is shown in Figure 3, and the analysis for each step is as follows.

Step A involves spatial mapping of the wildfire outbreak point by generating point-based feature data from the coordinates of the wildfire location. Using this data, a grid (50 m × 50 m) is created with a fishnet. Step B involves the spatial mapping of the wildfire suppression priority factors. The attribute information of these priority factors is entered into the grid created in Step A, based on the wildfire outbreak point. Step C is the weight sum stage, where the grid with the attribute information of each factor is reclassified and summed. In this study, the equal weight method was used for the weight sum analysis. The equal weight method assigns the same importance to all factors, producing results with minimal input and knowledge. This method has been widely applied in many decision-making problems [14]. Step D is the spatial partitioning stage. To consider fire truck accessibility, the analysis area is partitioned based on roads with a width of at least 2.5 m and the fire line. Step E involves assigning weights and normalizing to calculate the priority scores. The distance from the wildfire outbreak point and the sum of weights for priority factors are used to calculate the priority scores using min-max normalization (Eq. (1)). Min-max normalization has the advantage of converting various data ranges and units into a common scale for comparative analysis [15]. In terms of distance from the wildfire outbreak point, smaller values indicate higher priority, whereas larger sums of priority factor weights indicate higher priority.

(1)

Score=1-x-xminxmax-xmin*0.5+y-yminymax-ymin*0.5

x=Distance Case
x_min=Minimum Distance Case
x_max=Maximum Distance Case
y=Sum of Weights for Priority Factors
y_min=Minimum Sum of Weights for Priority Factors
y_max=Maximum Sum of Weights for Priority Factors

4. Results and Discussion

To assess the applicability of the proposed methodology, the spatial information of the four wildfire suppression priority factors previously established was used to analyze the suppression priority scores for the Gangneung wildfire area in 2023. Table 2 shows the classification criteria and weights for each factor based on their area. For the census tract population (Figure 4(a)), the population was calculated by defining the grid area as a proportion of the total census area, and reclassified into equal intervals. For the cultural heritage element (Figure 4(b)), the significance of the heritage was identified as higher if the same cultural heritage was registered multiple times under different categories, such as National Registered Heritage, National Designated Heritage, and Local Heritage. For the building element (Figure 4(c)), reclassification was based on the building area within each grid. For the forest resource element (Figure 4(d)), the area was based on plantations of grades 1-3, excluding artificial plantations of grades 4-5, according to the vegetation conservation grade assessment and classification criteria (related to Article 13). Figure 4(e) shows the results of the weight sum analysis for the wildfire suppression priority factors.

The results of weight assignment and normalization for calculating priority scores are shown in Figure 5 and Table 3.

Based on the forest fire line, priority scores were calculated for 2626 grids. The range for the distance factor, segmented in 100 m intervals from the wildfire origin, was from 1 to 26, while the range for the sum of priority weights was from 1 to 9. The normalized priority scores had an average of 0.42, a maximum of 0.855, and a minimum of 0.08. Table 3 shows the average values of the summed weights and priority scores based on distance. The number of grids per distance was the highest for distance case 13 (1200-1300 m) with 162 grids and lowest for distance case 1 (within 100 m) with 34 grids. The average of the summed weights was 2.912, with the highest in the region 100-200 m from the wildfire origin (distance case 2) at 3.896, and the lowest in the region 1300-1400 m (distance case 14) at 1.837. The average priority score was 0.416, with the highest in the region within 100 m (distance case 1) at 0.675 and the lowest in the region 2300-2400 m (distance case 24) at 0.174.

The analysis of the locations within the top 10% priority score (0.65 or higher) from the wildfire outbreak point is shown in Figure 6. In the eastern (E) direction of the wildfire outbreak point, 78 locations (27.3%) were identified, holding the highest proportion and showing strength across all distance intervals. Within Distance Case 1-2 (within 200 m), there were 55 grids with a priority score of 0.65 or higher, and among these, 13 grids were located to the east (E), comprising the largest proportion. For Distance case 3-5 (200-500 m), the analysis showed 24 grids to the west (W) and 19 grids to the east (E). For Distance Case 6-10 (500-1000 m), the analysis revealed 33 grids to the east (E) and 26 grids to the northwest (NW). The directions of the top 10% priority grids based on the wildfire outbreak point at various distances are shown in Table 4. In other words, for the 2023 Gangwon wildfires, regions in the east (E), west (W), and northwest (NW) directions were identified as higher priority areas regardless of the distance compared to that in the north (N) direction when the spread direction was not considered.

The application of the proposed methodology to the 2023 Gangneung wildfire-affected area revealed that regions close to the fire point were excessively prioritized. This was owing to using the equal weight method, assuming equal weights for all priority factors (population, cultural assets, buildings, forest resources), and overly considering the distance factor from the wildfire outbreak point in the normalization stage. Therefore, while the proposed methodology can provide relatively meaningful reference data for initial wildfire response, it is deemed unsuitable for developing mid- to long-term response strategies as the wildfire progresses. To address this, future research will focus on selecting appropriate weights for each factor and developing methods to adjust weights based on the spread direction using wildfire spread simulations, thereby improving the proposed methodology.

5. Conclusion

In this study, we proposed a GIS-based decision-making system for establishing initial response strategies (prioritizing fire suppression areas) during a wildfire. The proposed methodology uses the national spatial data of population, cultural assets, buildings, and forest resources, employing the equal weight method to sum weights. Additionally, it utilizes min-max normalization to reflect fire suppression priority score and distance factor of each grid, proposing a method to select priority areas for wildfire suppression. A case study was conducted on the Gangneung wildfire that occurred in 2023 to examine the applicability of the proposed methodology. Priority scores were calculated for 2626 grids in the Gangneung wildfire-affected area, allowing identification of the locations with the highest priorities based on distance from the wildfire outbreak. Analysis of the locations within the top 10% priority score revealed that, without considering the spread direction, regions to the east (E), west (W), and northwest (NW) were prioritized. This suggests that the methodology could contribute to the establishment of initial response strategies. However, it was observed that priority sharply decreases with increasing distance from the wildfire outbreak point, and the lack of weighting for priority factors suggests limited applicability for mid- to long-term response strategies. Therefore, further research is needed to make the methodology a more useful tool for response strategy development. Despite these limitations, the proposed methodology, based on national spatial data, can effectively identify priority fire suppression areas based on the distance from the wildfire outbreak point. This approach is expected to be valuable for efficient allocation of firefighting resources and the development of phase-specific wildfire suppression plans.

Notes

Author Contributions

Conceptualization, M.-S.K. and Y.-H.L.; methodology, M.-S.K. and Y.-H.L.; software, M.-S.K. and Y.-H.L.; validation, Y.-H.L; formal analysis, Y.-H.L; investigation, M.-S.K.; resources, M.-S.K.; data curation, M.-S.K.; writing—original draft preparation, M.-S.K; writing—review and editing, M.-S.K. and Y.-H.L.; visualization, M.-S.K.; supervision, Y.-H.L; project administration, Y.-H.L.; funding acquisition, Y.-H.L. All authors have read and agreed to the published version of the manuscript.

Conflict of Interest

The authors declare no conflict of interest.

Acknowledgments

This study was supported by research fund from Chosun University (K209522001).

Figure 1.

Study area.

Figure 2.

Established spatial information.

Figure 3.

Decision support model for formulating response strategies in case of large wildfires.

Figure 4.

Layers and weight sum results of factors for wildfire suppression priority.

Figure 5.

Deriving priority scores through normalization.

Figure 6.

Top 10% priority regions.

Table 1.

Sources and Data Format of National Spatial Information Utilized

Geospatial Data	Source	Data Format
Census Tract Population	Statistics Korea	*.txt
Census Tract Population	Statistics Korea	*.shp
National Registered Heritage, National Designated	Korea Heritage Service	*.shp
Heritage, National Heritage Protection Area (NHPA)
Local Heritage, Local Heritage Protection Area (LHPA)
Building	Ministry of Land, Infrastructure and Transport	*.shp
Ecological and Natural Map	Korea Forest Service	*.shp

Table 2.

Weights of Factors for Wildfire Suppression Priority

Weight	Census Tract Population (unit: person)	Cultural Heritage (unit: m²)	Building (unit: m²)	Ecological and Natural Map (unit: m²)
0	0	0	0	0
1	35	1,000	500	500
2	43	2,000	1,000	1,000
3	58	3,000	1,500	1,500
4	87	4,000	2,000	2,000
5	175	5,000	2,500	2,500

Table 3.

Average Weights and Average Priority Scores by Grid based on Distance

Distance Case	Number of Grids	Average Weight	Average Priority Score	Rank
Distance Case	Number of Grids	Average Weight	Average Priority Score	1-10	11-20	21-30	31-40	41-50	Over than 51
1	34	3.794	0.675	15	0	4	7	0	8
2	67	3.896	0.661	17	9	10	0	11	20
3	88	3.5	0.616	17	14	11	0	13	33
4	120	3.2	0.577	22	15	0	16	0	67
5	122	3.27	0.562	0	19	16	15	0	72
6	123	3.366	0.548	0	17	25	0	16	65
7	156	3.058	0.509	1	19	0	13	0	123
8	138	3.384	0.509	3	0	18	18	0	99
9	140	3.071	0.469	0	2	21	0	13	104
10	147	2.673	0.425	0	1	0	14	0	132
11	152	2.914	0.424	0	0	1	17	0	134
12	156	2.968	0.409	3	2	10	0	5	136
13	162	2.704	0.379	0	4	0	6	9	143
14	116	2.19	0.334	0	0	1	3	0	112
15	123	1.837	0.301	0	0	0	0	2	121
16	116	2.517	0.345	0	0	0	2	3	111
17	95	2.937	0.354	0	0	0	0	0	95
18	77	3.403	0.366	0	0	0	0	5	72
19	89	3.045	0.324	0	0	0	0	1	88
20	75	2.693	0.278	0	0	0	0	0	75
21	74	2.689	0.256	0	0	0	0	0	74
22	79	2.671	0.227	0	0	0	0	0	79
23	84	2.833	0.221	0	0	0	0	0	84
24	66	2.182	0.174	0	0	0	0	0	66
25	22	2.409	0.171	0	0	0	0	0	22
26	5	2	0.125	0	0	0	0	0	5

Table 4.

Top 10% of Azimuth Ranking Status

Distance Case	Azimuth
Distance Case	Total	W	SW	S	SE	E	NE	N	NW
1-2	55	8	6	5	6	13	4	8	5
3-5	103	24	10	8	15	19	13	3	11
6-10	107	12	4	0	10	33	9	13	26
11-26	21	0	0	0	0	13	2	6	0

References

1. J. H. Kwak, K. J. Bhang and M. I. Kim, “Developing a Decision Making Support Information Checklist Based on Analyses of Two Large-Scale Forest Fire Cases”, Crisisonomy, Vol. 16, No. 11, pp. 21-30 (2020), http://doi.org/10.14251/crisisonomy.2020.16.11.21.

2. K. O. Jeong and D. J. Kim, “A Study on the Improvement of Safety Management by Analyzing the Current Status and Response System of Forest Fire Accidents”, Journal of the Society of Disaster Information, Vol. 18, No. 3, pp. 457-469 (2022), https://doi.org/10.15683/kosdi.2022.9.30.457.

3. M. A. Storey, O. F. Price, M. Almeida, C. Riberio, R. A. Bradstock and J. J. Sharples, “Experiments on the Influence of Spot Fire and Topography Interaction on Fire Rate of Spread”, PLOS ONE, Vol. 16, No. 1, pp. e0245132(2021), https://doi.org/10.1371/journal.pone.0245132.

4. A. Bouguettaya, H. Zarzour, A. M. Taberkit and A. Kechida, “A Review on Early Wildfire Detection from Unmanned Aerial Vehicles Using Deep Learning-Based Computer Vision Algorithms”, Signal Processing, Vol. 190, pp. 108309(2022), https://doi.org/10.1016/j.sigpro.2021.108309.

5. K. W. Kang, J. H. Oh, H. B. Jeon and J. S. Lim, "Proposing a New Paradigm for Large-Scale Wildfire Response in Firefighting", The 35th Annual 119 Firefighting Policy Conference, pp. 1-59 (2023).

6. Yonhap News, "[One Month After Gangneung Mountain Fire] ①: 'Rising Again'... Full Efforts for Recovery of Homes and Forests Damaged", (2024). https://www.yna.co.kr/view/AKR20230509096700062.

7. Forest Protection Act, "No. 20309, Artice 38 Establishment and Operation of Consolidated Command Headquarters at Forest Fire Site" (2024).

8. Forest Protection Act, "No. 20309, Artice 39 Cooperation" (2024).

9. N. G. Kim, J. H. Kwak and M. I. Kim, "Development of Foreset Fire Response Decision Tree through Large Forest Fire Case Analysis Based on Response Sctors", Korean Society of Civil Engineers 2019 Convention, pp. 1188-1189 (2019).

10. J. H. Kwak, N. G. Kim and M. I. Kim, “Development of Evacuation Route Planning Algorithm for Responding to Urban Forest Fires”, Journal of the Korean Society of Hazard Mitigation, Vol. 21, No. 6, pp. 63-70 (2021), https://doi.org/10.9798/KOSHAM.2021.21.6.63.

11. J. H. Kwak, N. G. Kim and M. I. Kim and K. J. Bang, "Development and Application of Facility Defense Zone Setting Algorithm for Large Forest Fire Response", Korean Society of Civil Engineers 2021 Convention, pp. 369-370 (2021).

12. M. W. Son, B. H. Lee and B. S. Kim "Development of an Urban Forest Fire Information for Real Time Emergency Management And Prevention Support System(FIREMAP) Using Big Data", Korean Society of Civil Engineers 2023 Convention, pp. 209-210 (2023).

13. N. G. Kim, J. H. Kwak and M. I. Kim, “Numerical Simulations of Large Forest Fires in the East Coastal Region of Korea Considering the Yangganjipung Local Wind”, Journal of the Korean Society of Hazard Mitigation, Vol. 21, No. 4, pp. 39-48 (2021), https://doi.org/10.9798/KOSHAM.2021.21.4.39.

14. M. S. Kim, W. H. Hong, Y. H. Lee and S. C. Baek, “Selection of Take-Off and Landing Sites for Firefighter Drones in Urban Areas Using a GIS-Based Multi-Criteria Model”, Drones, Vol. 6, No. 12, pp. 412(2022), https://doi.org/10.3390/drones6120412.

15. S. Saichon, “Performance Comparison of New Adjusted Min-Max with Decimal Scaling and Statistical Column Normalization Methods for Artificial Neural Network Classification”, International Journal of Mathematics and Mathematical Sciences, (2022), https://doi.org/10.1155/2022/3584406.