Selecting the Best Image Pairs to Measure Slope Deformation

Wentao Yang

doi:10.3390/s20174721

Abstract

Optical remote sensing images can be used to monitor slope deformation in mountain regions. Abundant optical sensors onboard various platforms were designed to provide increasingly high spatial–temporal resolution images at low cost; however, finding the best image pairs to derive slope deformation remains difficult. By selecting a location in the east Tibetan Plateau, this work used the co-registration of optically sensed images and correlation (COSI-Corr) method to analyze 402 Sentinel-2 images from August 2015 to February 2020, to quantify temporal patterns of uncertainty in deriving slope deformation. By excluding 66% of the Sentinel-2 images that were contaminated by unfavorable weather, uncertainties were found to fluctuate annually, with the least uncertainty achieved in image pairs of similar dates in different years. Six image pairs with the least uncertainties were selected to derive ground displacement for a moving slope in the study area. Cross-checks among these image pairs showed consistent results, with uncertainties less than 1/10 pixels in length. The findings from this work could help in the selection of the best image pairs to derive reliable slope displacement from large numbers of optical images.

Highlights

Slope deformations are important precursors for imminent landslide hazards, which pose serious threats to mountain communities around the world [1,2,3]
By using multiyear dense optical images, this work showed that uncertainties fluctuate annually, and the least uncertainty can be achieved by using images of the same dates in different years
We found the Sentinel-2 images had a very short revisiting time of two/three days in this study area, but unfavorable weather conditions contaminated 65.7% of all observations

Summary

Introduction

Slope deformations are important precursors for imminent landslide hazards, which pose serious threats to mountain communities around the world [1,2,3]. With increases in population sizes and economic activities, detecting possible slope deformation in mountain regions is becoming increasingly important [4]. Optical remote sensing has been increasingly used to monitor slope deformations [5,6,7,8]. Image correlation is the most frequently used methods to derive slope deformation from optical images [5,6,8,9,10,11]. Compared to SAR images, optical remote sensing images are easier to process and are more reliable in the detection of large displacements [13,14,15,16,17,18,19]

Objectives

Methods

Discussion

Conclusion