Abstract

Tropospheric delays (TDs) limit the accurate detection of slow slope motion using interferometric synthetic aperture radar (InSAR), especially in subtropical coastal regions prone to frequent changes in humidity. Although TDs can be estimated through external weather data, their spatiotemporal resolution and data availability are greatly limited, which is not applicable for individual slopes. This paper presents new TD correction methods for slopes with both small dimensions and small elevation changes. We simultaneously estimated and eliminated the TD signal from a line-of-sight (LOS) time series through a blind source separation (i.e., independent component analysis). The stratified TD sources were isolated according to a spatially elevation-linked and temporally periodic independent-component (IC), which was determined via a correlation test and power spectrum analysis. Hence, the TD was corrected without the use of any external weather products/meteorological data and had unprecedented spatiotemporal details equivalent to the synthetic aperture radar (SAR) images. The proposed method was verified using CosmoSkyMed (CSK) and Sentinel-1 (SNT-1) images covering a slope in Tai O, Lantau Island, Hong Kong, and validated using a series of geodetic, meteorological, and hydrological data. Up to 3–4-cm relative TDs were measured in the LOS directions of CSK and SNT-1. The relative TD exhibited a slower increment rate than the slope elevation and was largely affected by specific air conditions (e.g., temperature and humidity) on the SAR image-acquisition days. The analysis of InSAR data yielded reasonably good estimates of millimeter-scale downslope slips (due to increases in pore-water pressure in the wet season) and upslope rebound (due to soil shrinkage in the dry season). It was found that soil swelling and shrinkage of the slope (also known as seasonal ratcheting) and the reclamation were likely regulated by rainfall and sea levels, respectively. Although the slope motion in Tai O was determined to be small (i.e., seasonal variations of 10 mm), the TD correction reduced the root-mean-square error by 42.3%, such that InSAR time-series measurements with millimeter-level accuracy (potentially 1–3 mm) were obtained.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call