Deep learning techniques have been widely applied to classify tree species and segment tree structures. However, most recent studies have focused on the canopy and trunk segmentation, neglecting the branch segmentation. In this study, we proposed a new approach involving the use of the PointNet++ model for segmenting the canopy, trunk, and branches of trees. We introduced a preprocessing method for training LiDAR point cloud data specific to trees and identified an optimal learning environment for the PointNet++ model. We created two learning environments with varying numbers of representative points (between 2048 and 8192) for the PointNet++ model. To validate the performance of our approach, we empirically evaluated the model using LiDAR point cloud data obtained from 435 tree samples scanned by terrestrial LiDAR. These tree samples comprised Korean red pine, Korean pine, and Japanese larch species. When segmenting the canopy, trunk, and branches using the PointNet++ model, we found that resampling 25,000–30,000 points was suitable. The best performance was achieved when the number of representative points was set to 4096.
Rainfall-induced landslide early warning systems require reliable estimation of soil moisture conditions. This study proposes a Soil Water Index (SWI) framework based on a three-stage TANK model. Through GLUE (Generalized Likelihood Uncertainty Estimation)-based behavioral parameter sampling and K-means clustering, SWI response characteristics were classified into two representative scenarios: slow drainage (Scenario 1) and fast drainage (Scenario 2). Two-stage thresholds—Watch (α = 0.40 × SWIpeak) and Warning (β = 0.70 × SWIpeak)—were established from SWI rise profile analysis at 500 m and 5 km resolutions, providing 20–27 and 4–5 h of lead time, respectively. Verification against the July 2025 heavy rainfall event across multiple resolutions and spatial extents yielded Hit Rates of 0.984–1.000, while FAR (False Alarm Ratio) remained structurally high (0.607–0.648 for grids sharing the rainfall field with occurrence sites). These findings confirm that SWI serves as an effective regional-scale necessary condition indicator for landslide-triggering moisture, but FAR reduction requires integration with slope susceptibility information.