Document Type : Original Article
Authors
1
School of Surveying and Geospatial Engineering, College of Engineering, University of Tehran, Tehran, Iran.
2
School of Surveying and Geospatial Engineering, College of Engineering, University of Tehran, Tehran, Iran -
3
Department of Photogrammetry and Remote Sensing, School of Surveying and Geospatial Engineering, College of Engineering, University of Tehran, Tehran, Iran
10.22059/eoge.2026.414122.1223
Abstract
Rapid and reliable landslide detection in densely vegetated mountainous regions remains challenging, particularly under persistent cloud cover and adverse weather conditions where optical imagery becomes ineffective. This study proposes an integrated PolSAR and PolInSAR framework based on freely available Sentinel-1 dual-polarization SAR data for landslide detection following the 2024 Enga Province landslide event in Papua New Guinea. Polarimetric features, including entropy, alpha angle, radar vegetation index, cross-polarization ratio, and covariance matrix elements, were extracted to characterize changes in scattering mechanisms and vegetation structure. In addition, multi-temporal coherence features were generated from pre-event, post-event, and cross-event image pairs to capture decorrelation associated with surface displacement and structural changes. To evaluate the contribution of different feature groups, coherence-based, covariance-based, PolSAR, and integrated feature stacks were classified using machine learning approaches. Among the evaluated classifiers, random forest achieved the best overall performance with an overall accuracy of 0.88. Quantitative evaluation revealed that while the coherence-only and covariance-based stacks yielded accuracies of 0.63 and 0.82 respectively, the PolSAR feature stack provided strong discriminative capability with an accuracy of 0.87. The integration of polarimetric and interferometric features further improved classification performance, with the full feature stack achieving the highest accuracy (0.88) and F1-score of 0.88. These results demonstrate the robustness and effectiveness of utilizing dual-polarimetric Sentinel-1 data for rapid and accurate landslide mapping in complex, cloud-prone environments.
Keywords
Main Subjects
)