- 1Dept. of Energy Resources and Geosystems Engineering, Sejong University, Seoul, Republic of Korea (jhlee6086@gmail.com)
- 2Dept. of Energy Resources and Geosystems Engineering, Sejong University, Seoul, Republic of Korea (hjpark@sejong.ac.kr)
- 3Dept. of Earth and Environmental Sciences, Korea University, Seoul, Republic of Korea (youngjlee@korea.ac.kr)
Landslides, a major natural disaster in South Korea, are primarily triggered by heavy rainfall associated with global climate anomalies. In particular, the years 2020 and 2022 witnessed unprecedented torrential rains during the summer, resulting in the most severe landslide damages recorded in recent history, with significant human and economic losses.
Landslide susceptibility assessment involves the spatial analysis of direct triggering factors, such as rainfall, and conditioning factors both internal and external to slopes, to predict the likelihood and impact of landslide occurrences. Based on the mechanisms considered, assessment methodologies are typically classified into physically-based models and data-driven models. Physically-based models, which have been extensively studied globally, are well-suited for landslide susceptibility analysis in South Korea as they allow for the integration of engineering principles to address rainfall and internal slope conditions. However, their limitations in addressing the multifaceted interactions among diverse influencing factors necessitate the incorporation of data-driven approaches.
This study seeks to integrate physically-based models with data-driven models to capture both the engineering mechanisms of rainfall-induced landslides and the complex interrelationships among diverse influencing factors. Since these models operate as independent frameworks, a fusion approach is adopted to combine their outputs effectively. Fusion methodologies vary depending on the stage at which data or information is integrated. In this research, decision-level fusion is employed, which aggregates the independent decisions or outputs of multiple models to produce the final result. Specifically, serial decision fusion and parallel decision fusion, two representative decision-level fusion techniques, are implemented. The study evaluates the performance and applicability of the fusion models by comparing the outcomes of different fusion strategies.
Acknowledgements
This work was supported by the National Research Foundation of Korea(NRF) grant funded by the Korea government(MSIT) (No. RS-2024-00358026).
How to cite: Lee, J.-H., Park, H.-J., and Lee, Y.-J.: A Study on Landslide Susceptibility Fusion Models Using Decision-Level Fusion, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7971, https://doi.org/10.5194/egusphere-egu25-7971, 2025.
