Performance Analysis of Open Source Time Series InSAR Methods for Deformation Monitoring over a Broader Mining Region

Abstract

Time Series Interferometric Synthetic Aperture Radar (TSInSAR) methods have been widely and successfully applied for spatiotemporal ground deformation monitoring. The main groups of methodological approaches are often referred to as Persistent Scatterer (PS), Small Baseline (SB), and hybrid approaches that incorporate PS and SB concepts. While TSInSAR techniques have long been able to provide accurate deformation rates for various applications, their corresponding performance in complex environments such as mining areas has to be investigated. This study focuses on comparing the performance of three open source TSInSAR toolboxes (Stamps, Giant, Mintpy) over an extended region that includes an active opencast coal mine. We present the deformation results of each TSInSAR method on a Sentinel-1 dataset of 125 acquisitions spanning around 2.5 years over the Ptolemaida-Florina coal mine site that is characterized by several environmental and surface deformation conditions. First, a cross-comparison analysis is presented over different land cover classes. The study shows that all TSInSAR methods are capable for generating similar ground deformation results when the area has stable ground scattering conditions and the dataset sufficient temporal sampling. The most controversial results between TSInSAR approaches were found in land cover classes that include medium to high vegetation. An external comparative analysis between the different results from TSInSAR methods and leveling measurements is also performed. Stamps approach presented the best agreement with the in-situ deformation rates. The Giant approach yielded the best cumulative deformation results due to our a priori knowledge of temporal behavior of deformation in the vicinity of the leveling locations. Finally, we discuss the main pros and cons of each TSInSAR approach and we highlight the importance of comparison analysis that can provide insights and can lead to better interpretation of the results.