The Journal of Engineering Geology ISSN:1226-5268(Print) 2287-7169(Online) 지질공학

This study aims to quantitatively analyze changes in landslide hazard and population exposure under different rainfall scenarios across Gyeongsangbuk-do, Republic of Korea, and to develop a WebGIS platform for integrated visualization and comparison of the results. The physically based TRIGRS (Transient Rainfall Infiltration and Grid-Based Regional Slope-Stability) model was applied to estimate the factor of safety (FS) under six rainfall scenarios ranging from 12 to 72 h in duration and from 50 to 300 mm in cumulative rainfall. In addition, a total of 467 landslide casualty-prone sites were analyzed using square analysis areas of 0.5, 1, 2, and 4 km² centered on each site. Areas with FS < 1.3 were spatially overlaid with 100 m census population grids to estimate potential population exposure. Differences associated with rainfall scenarios and analysis-area sizes were evaluated using the Friedman test, Wilcoxon signed-rank test, and Spearman rank correlation analysis. The proportion of the study area with FS < 1.3 increased from 2.29% under the 12 h / 50 mm scenario to 9.19% under the 36 h / 150 mm scenario and 15.34% under the 48 h / 200 mm scenario. The expansion of hazardous areas was particularly pronounced between the 24 h / 100 mm and 36 h / 150 mm scenarios and between the 36 h / 150 mm and 48 h / 200 mm scenarios. Beyond the 48 h / 200 mm scenario, only minor numerical variations in FS occurred in some grid cells, whereas no additional changes were observed in the spatial classification based on the FS < 1.3 threshold. At the 467 concern sites, the proportion of sites containing FS < 1.3 areas increased with analysis-area size. Under the 48 h / 200 mm scenario, the hazard detection rate increased from 87.6% for the 0.5 km² analysis area to 98.9% for the 4 km² analysis area. Although the effect of analysis-area size on the FS < 1.3 area ratio was statistically significant, its effect size was very small (Kendall’s W = 0.012), whereas the effect on population exposure was considerably larger (W = 0.492). For the 4 km² analysis area, the proportion of sites classified as having both landslide hazard and exposed population increased from 22.5% under the 12 h / 50 mm scenario to 68.7% under the 48 h / 200 mm scenario. A statistically significant positive relationship was observed between the FS < 1.3 area ratio and estimated exposed population (ρ = 0.323, p < 0.001); however, the relatively modest correlation indicates that physical landslide hazard alone does not fully explain population exposure. Finally, a WebGIS platform was developed to integrate and interactively compare TRIGRS-based landslide hazard maps, landslide human-damage concern sites, analysis-area information, and population exposure results across rainfall scenarios. The results demonstrate the importance of jointly considering physical landslide hazard, population exposure, and spatial analysis scale when evaluating landslide-prone areas. The proposed framework can provide useful spatial information for identifying areas requiring prioritized attention in landslide disaster management. However, the estimated exposed population represents potential spatial exposure rather than actual casualties. Further research should incorporate observed landslide inventories, higher-resolution population and building data, and vulnerability information to improve the applicability of the framework to comprehensive landslide risk assessment.