Scholarlink

<p>Landslides are a well-known risk to human life and infrastructure; therefore, failure time prediction for slopes exhibiting progressive rupture mechanics is of immense value for early warning and disaster risk reduction. However, beyond the challenges posed by complex rupture evolution and satellite data noise, while uncertainty estimation has been explored in some methods, conventional failure time prediction approaches often lack comprehensive uncertainty quantification and confidence intervals, limiting their practical application in early warning systems. In this study, a probabilistic model based on Persistent Scatterer Interferometric Synthetic Aperture Radar (PS-InSAR), time series cross-correlation (TSCC), and Bayesian inference is proposed for landslide early warning with uncertainty quantification. The method combines geometric reliability screening of the best PS-InSAR monitoring point selection, TSCC analysis for leader-lagger relation detection, and Bayesian inference for probabilistic failure time prediction with quantified confidence intervals. The appropriateness of the proposed method is tested using 159 Sentinel-1 acquisitions (2017–2022) across six active landslides on the China-Nepal Friendship Highway in the Himalayan region. The results suggest that the proposed framework performs well. All six actual failure events were within the predicted 95% highest-density intervals (widths 13.58–38.20&#xa0;days). Overall, this study provides a useful framework for landslide early warning with uncertainty bounds using PS-InSAR monitoring data.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Probabilistic landslide time-of-failure prediction using Persistent Scatterer Interferometric Synthetic Aperture Radar (PS-InSAR), time series cross-correlation, and Bayesian inference

  • 期刊论文

  • 发表日期：

    2026年6月3日

查看全文
  • Diwakar Khadka,
  • Bhim Kumar Dahal,
  • Jie Zhang,
  • Meng Lu,
  • Atma Sharma

摘要

Landslides are a well-known risk to human life and infrastructure; therefore, failure time prediction for slopes exhibiting progressive rupture mechanics is of immense value for early warning and disaster risk reduction. However, beyond the challenges posed by complex rupture evolution and satellite data noise, while uncertainty estimation has been explored in some methods, conventional failure time prediction approaches often lack comprehensive uncertainty quantification and confidence intervals, limiting their practical application in early warning systems. In this study, a probabilistic model based on Persistent Scatterer Interferometric Synthetic Aperture Radar (PS-InSAR), time series cross-correlation (TSCC), and Bayesian inference is proposed for landslide early warning with uncertainty quantification. The method combines geometric reliability screening of the best PS-InSAR monitoring point selection, TSCC analysis for leader-lagger relation detection, and Bayesian inference for probabilistic failure time prediction with quantified confidence intervals. The appropriateness of the proposed method is tested using 159 Sentinel-1 acquisitions (2017–2022) across six active landslides on the China-Nepal Friendship Highway in the Himalayan region. The results suggest that the proposed framework performs well. All six actual failure events were within the predicted 95% highest-density intervals (widths 13.58–38.20 days). Overall, this study provides a useful framework for landslide early warning with uncertainty bounds using PS-InSAR monitoring data.