Landslides are major forest disasters in the Republic of Korea that result in significant loss of life and property damage. Since the 1980s, the National Institute of Forest Science has conducted various studies on nonstructural landslide mitigation, including early warning systems and susceptibility mapping, to reduce damage induced by landslides. A landslide early warning system based on the TANK model was developed and used for decision-making on the preparedness level and for ordering evacuations for local people. A nationwide 5-level landslide susceptibility map was developed based on logistic regression in the 2010s, considering geological, geographical, and forest characteristics. In addition, we provide primary data for determining evacuation priorities by developing debris-flow hazard zonation and societal vulnerability assessments in that area. For public awareness, all of this information is provided through the WebGIS platform “Landslide Information System.”

Understanding the properties of wildfire-disrupted catchments is crucial for managing river floods and landslide risks. Using a hydrophone, we investigated the changes in sediment supply conditions in small mountainous catchments (30.8 ha) in southern Korea for 6 years (March 2014–December 2019). Bedload transport rates mostly increased in burned catchments 1–2 years post-wildfire (early post-fire sediment regime) but decreased 3 years post-wildfire owing to the coverage of burned slopes (late post-fire sediment regime). Landslides triggered 5 years post-wildfire increased the bedload transport rates by approximately one order of magnitude (post-landslide sediment regime). Although sediment activity decreased in wildfire-disturbed catchments after recovery of the ground-cover layer 1–3 years post-fire, thy increased during events such as landslides. Furthermore, even in undisturbed mountainous catchments (139.7 ha), bedload transport rates increased after intense rainstorm events (≥100.0 mm), with this change lasting for approximately 2 years. Our observations showed that the forest restoration after a wildfire should be planned and implemented from a min- to long-term perspective. In addition, during rainstorms, the relationship between flow and bedload transport showed large variability even in undisturbed catchments. Furthermore, surrogate monitoring using a hydrophone was useful in understanding the changes in bedload transport characteristics according to various supply conditions of the catchment.

When landslides and debris flows occur, predicting sediment discharge is important for soil erosion control. In this study, aerial photography and GIS spatial analysis were carried out on a landslide and debris flow area. The sediment discharge was most affected by the flow length, flow width, and the amount of slope failure within the source area. In particular, the sediment discharge was highly correlated with the flow length. We have developed an equation that can estimate sediment discharge. The developed equations are grouped and four estimation equations are proposed. The developed sediment discharge equation can predict the amount of soil runoff downstream when a landslide and debris flow occur in a forest watershed. In addition, it can be used to make design decisions concerning the scale and location of erosion control dams, contributing to disaster prevention in downstream areas.

The estimation of the sediment yield in Korea relies primarily on universal soil loss equation-based empirical models. However, validating these models remains difficult, and their application is limited to mountain streams where bedload transport is dominant. This study applied machine-learning techniques to predict bedload yields in mountain streams as an alternative to empirical models. Analysis of global datasets revealed that tree-based models achieved high accuracy (R ≥ 0.94). When the best-performing models were applied to three domestic mountain streams, significant predictive capability was confirmed, particularly in the Gimcheon site (R > 0.87, NSE > 0.73). These findings suggest that machine-learning-based models hold sufficient potential as quantitative decision-making tools for the design and management of check dams.