Can satellite-based data substitute for surveyed data to predict the spatial probability of forest fire? A geostatistical approach to forest fire in the Republic of Korea

Figures & data

Figure 1. Schematic diagram of the applied methodology.

Figure 2. Study area with administrative boundaries and forest fire point data from two sources.

Table 1. Description of related factors used this study.

Factor	Time scale	Data type	Unit	Data source
Rainfall	2007–2016	point	mm	Korea Meteorological Administration (KMA)
Temperature	2007–2016	point	°C	KMA
Effective humidity	2007–2016	point	%	KMA
Maximum wind speed	2007–2016	point	m/s	KMA
Forest type	2009–2012 (5th map)	polygon	–	Forest Type Map from Korea Forest Service (KFS)
National protected area	2015	polygon	–	Environmental Geographic Information Service of Republic of Korea (ROK)
Population density	2015	census	people per sq. km	Population census of Korean Statistical Information Service (KOSIS)
Distance from road	2015	Raster (unit: m)	m	Intelligent Transport Systems Standard Node Link of ROK

Figure 3. Forest fire occurrence frequency on a 5-km² grid (a: MODIS active fire data and b: KFS fire survey data).

Figure 4. Forest fire hotspot analysis by fire intensity (a: MODIS active fire data and b: KFS fire survey data).

Figure 5. High-intensity forest fire hotspot analysis by fire intensity (a: MODIS active fire data and b: KFS fire survey data).

Figure 6. Spatial autocorrelation results by forest fire occurrence frequency (a: MODIS active fire data and b: KFS fire survey data). Note that the range and partial sill values are shown in m.

Figure 7. Spatial autocorrelation result by high-intensity forest fire (a: MODIS active fire data and b: KFS fire survey data). Note that the range and partial sill values are shown in m.

Figure 8. Spatial autocorrelation results by related factors for the MODIS active fire point data (a: rainfall, b: temperature, c: effective humidity, d: maximum wind speed, e: forest type, f: national protected area, g: population density, and h: distance from road). Note that the range and partial sill values are shown in m.

Figure 9. Spatial autocorrelation results by related factors for the KFS surveyed fire point (a: rainfall, b: temperature, c: effective humidity, d: maximum wind speed, e: forest type, f: national protected area, g: population density, and h: distance from road). Note that the range and partial sill values are shown in m.

Figure 10. Spatial modeling results of forest fire occurrence probability (a: MODIS active fire data and b: KFS fire survey data).

Figure 11. Evaluation of spatial modeling performance for forest fire occurrence probability using ROC and AUC (a: MODIS active fire data and b: KFS fire survey data).

Table 2. Analysis of factor contributions for spatial modeling of fire occurrence probability (%).

	MODIS active fire data		KFS fire survey data
Factors	Percent contribution	Permutation importance	Percent contribution	Permutation importance
Rainfall	14.9	28.8	10.0	15.0
Temperature	17.8	21.8	29.4	36.2
Effective humidity	9.2	2.6	6.8	13.3
Maximum wind speed	20.3	22.7	6.5	5.7
Forest type	6.4	3.9	6.1	4.1
National protected area	8.8	3.7	30.0	17.3
Population density	18.7	10	10.0	7.3
Distance from road	3.8	6.5	1.2	1.0