Remote Sensing, Vol. 9, Pages 765: Correcting InSAR Topographically Correlated Tropospheric Delays Using a Power Law Model Based on ERA-Interim Reanalysis (Remote Sensing)

Tropospheric delay caused by spatiotemporal variations in pressure, temperature, and humidity in the lower troposphere remains one of the major challenges in Interferometric Synthetic Aperture Radar (InSAR) deformation monitoring applications. Acquiring an acceptable level of accuracy (millimeter-level) for small amplitude surface displacement is difficult without proper delay estimation. Tropospheric delay can be estimated from the InSAR phase itself using the spatiotemporal relationship between the phase and topography, but separating the deformation signal from the tropospheric delay is difficult when the deformation is topographically related. Approaches using external data such as ground GPS networks, space-borne spectrometers, and meteorological observations have been exploited with mixed success in the past. These methods are plagued, however, by low spatiotemporal resolution, unfavorable weather conditions or limited coverage. A phase-based power law method recently proposed by Bekaert et al. estimates the tropospheric delay by assuming the phase and topography following a power law relationship. This method can account for the spatial variation of the atmospheric properties and can be applied to interferograms containing topographically correlated deformation. However, the parameter estimates of this method are characterized by two limitations: one is that the power law coefficients are estimated using the sounding data, which are not always available in a study region; the other is that the scaled factor between band-filtered topography and phase is inverted by least squares regression, which is not outlier-resistant. To address these problems, we develop and test a power law model based on ERA-Interim (PLE). Our version estimates the power law coefficients by using ERA-Interim (ERA-I) reanalysis. A robust estimation technique was introduced in the PLE method to estimate the scaled factor, which is insensitive to outliers. We applied the PLE method to ENVISAT ASAR images collected over Southern California, US, and Tianshan, China. We compared tropospheric corrections estimated from using our PLE method with the corrections estimated using the linear method and ERA-I method. Accuracy was evaluated by using delay measurements from the Medium Resolution Imaging Spectrometer (MERIS) onboard the ENVISAT satellite. The PLE method consistently delivered greater standard deviation (STD) reduction after tropospheric corrections than both the linear method and ERA-I method and agreed well with the MERIS measurements.