Scale-dependent landslide susceptibility assessment and model transfer: a multi-scale study in the Wenchuan–Maoxian area, China

Landslide susceptibility models are strongly influenced by the spatial scale at which conditioning factors and landslide inventories are represented. However, whether a regional-scale factor system can be directly transferred to finer-scale assessments remains insufficiently understood. This study investigates the scale dependence of landslide susceptibility assessment in the Wenchuan–Maoxian area, southwestern China, using Random Forest models at regional and segment scales. A regional model was developed using six conditioning factors derived from 30 m DEM and Landsat imagery, while a segment-scale model incorporated 12.5 m ALOS PALSAR DEM-derived geomorphic indices and GF-2 interpreted landslide polygons. In addition, a scale-mismatch transfer experiment was conducted by applying the trained regional model to high-resolution factors over the segment area without introducing segment-specific conditioning factors. The segment model achieved the best overall performance (AUC = 0.9375; F1-score = 0.8566), whereas the transfer model showed high precision (0.9167) but lower recall (0.6667). Its recall further decreased from 0.818 for large landslides to 0.477 for small landslides, indicating limited transferability of the regional factor system to fine-scale landslide identification. Feature-importance analysis showed that the dominant conditioning factors varied with assessment scale, with river-related factors being more influential at the regional scale and tectonic and local hydrological/topographic factors becoming more important at the segment scale. These results indicate that increasing spatial resolution alone cannot compensate for the loss of process-relevant information in a scale-inappropriate factor system. Landslide susceptibility assessment should therefore match the spatial resolution, conditioning factors, and geomorphic processes to the target assessment scale.

Authors

Institutions

Publication Details

Journal
Scientific Reports
Published
2026-11-15
DOI
https://doi.org/10.1038/s41598-026-78472-8
Primary Topic
Landslides and related hazards
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Scale-dependent landslide susceptibility assessment and model transfer: a multi-scale study in the Wenchuan–Maoxian area, China

Xinchun Yang, Yixiang Du, Jialun Cai, YI Shujian et al.
Scientific Reports
Landslides and related hazards
article

Scale-dependent landslide susceptibility assessment and model transfer: a multi-scale study in the Wenchuan–Maoxian area, China

Xinchun Yang, Yixiang Du, Jialun Cai, YI Shujian, Xianwen Liu, Chong Li, Wenjun Zhang, Wei Wang, Xuwei Liao
article en

Abstract

Landslide susceptibility models are strongly influenced by the spatial scale at which conditioning factors and landslide inventories are represented. However, whether a regional-scale factor system can be directly transferred to finer-scale assessments remains insufficiently understood. This study investigates the scale dependence of landslide susceptibility assessment in the Wenchuan–Maoxian area, southwestern China, using Random Forest models at regional and segment scales. A regional model was developed using six conditioning factors derived from 30 m DEM and Landsat imagery, while a segment-scale model incorporated 12.5 m ALOS PALSAR DEM-derived geomorphic indices and GF-2 interpreted landslide polygons. In addition, a scale-mismatch transfer experiment was conducted by applying the trained regional model to high-resolution factors over the segment area without introducing segment-specific conditioning factors. The segment model achieved the best overall performance (AUC = 0.9375; F1-score = 0.8566), whereas the transfer model showed high precision (0.9167) but lower recall (0.6667). Its recall further decreased from 0.818 for large landslides to 0.477 for small landslides, indicating limited transferability of the regional factor system to fine-scale landslide identification. Feature-importance analysis showed that the dominant conditioning factors varied with assessment scale, with river-related factors being more influential at the regional scale and tectonic and local hydrological/topographic factors becoming more important at the segment scale. These results indicate that increasing spatial resolution alone cannot compensate for the loss of process-relevant information in a scale-inappropriate factor system. Landslide susceptibility assessment should therefore match the spatial resolution, conditioning factors, and geomorphic processes to the target assessment scale.

Scientific Reports
Southwest University of Science and Technology (CN), Ministry of Natural Resources (CN)
Landslides and related hazards
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon