UAV‐Based Evaluation of National and Global Digital Elevation Models for Chronic Landslides in the Indian Himalayas

Vollständiger Abstract

Worum geht es in dieser Arbeit?

ABSTRACT This study evaluates one national Digital Elevation Model (DEM) (CARTOSAT) and eight global DEMs (ALOS12.5, AW3D30, ASTER, COPDEM, FABDEM, GEDTM, NASADEM and SRTM) to represent the topography of 10 chronic landslides in the Garhwal Indian Himalayas, using high‐resolution Uncrewed Aerial Vehicle (UAV)‐derived DEMs as the reference. The majority of the chronic landslides considered in the study were activated/reactivated after the 2013 Kedarnath disaster in the Upper Ganga Catchment. Elevation, slope, aspect, and transverse profiles were compared across all landslides, followed by a vertical accuracy assessment through separate quantification of systematic elevation bias and random elevation variability using a variogram‐based estimate of terrain‐specific spatial autocorrelation. Results indicate that all coarser DEMs exhibit systematic vertical bias relative to the UAV‐derived DEMs. Among them, CARTOSAT shows the closest agreement with the reference higher‐resolution DEM for representing elevation and slope. In contrast, aspect derived from the national and global DEMs adequately represents the dominant slope direction but fail to capture the full range of slope orientations. Following bias correction, across the landslides considered in this study, the standard error ranged from 1.66 m to 17.74 m, with CARTOSAT consistently exhibiting the lowest standard error, followed by AW3D30, FABDEM and COPDEM. The comparative analysis was extended to four global landslides (Slumgullion and Oso, USA; Trisuli, Nepal; and Xinmo, China) using the available high‐resolution LiDAR‐derived DEMs as the reference. All global DEMs overestimated elevations and slope angles, but adequately represented the dominant aspect of medium‐ to high‐slope angle landslides. Results indicate FABDEM, COPDEM and GEDTM showed the most consistent vertical accuracy across the global landslide sites, with standard error ranging from 1.64 m to 17.65 m, demonstrating consistent performance. Overall, DEM performance was found to be site‐dependent, highlighting the need for systematic bias correction and site‐specific vertical accuracy assessment before applying coarser DEMs to regional landslide investigations.