An integrated semi-automated approach of Object Based Image Analysis (OBIA)–Synthetic Aperture Radar (SAR) for landslide delineation

Abstract

Landslide delineation is a critical task in geospatial analysis and disaster management, it requires precise and dynamic methodologies for accurate detection and risk assessment. Traditional Object-Based Image Analysis (OBIA) techniques have leveraged spectral characteristics, geometry, and textural features in the optical datasets. Separately, Synthetic Aperture Radar (SAR) based methods have been employed to detect landslides through the analysis of phase changes in temporal images captured at different angles by radar satellites. This study presents a method of integration of SAR subsidence data with OBIA methodologies to enhance landslide delineation accuracy. By incorporating subsidence values as a dynamic parameter alongside the static parameters used in OBIA, this study demonstrates advancement in landslide detection. The integration aims to leverage the strengths of both methods: the detailed spatial analysis capability of OBIA and the temporal, movement-sensitive detection capacity of Differential Interferometry Synthetic Aperture Radar (DInSAR). The methodology involves a systematic combination of SAR-derived subsidence data with OBIA's spatial analysis to identify areas that are potential landslides more accurately and comprehensively. The study's findings reveal that this integrated approach improves the Recall score by 10.6%, indicating a higher success rate in identifying true landslide events. However, it also results in a 4.2% decrease in the Precision score, highlighting an increase in false positives. Despite this trade-off, the incorporation of dynamic subsidence data into the OBIA framework presents a more nuanced assessment of landslide, contributing to a better understanding of potential future changes and enhancing emergency preparedness and mitigation strategies.