Benchmarking and inter-comparison of Sentinel-1 InSAR velocities and time series

Different InSAR algorithms and methods produce velocities and times series that are not identical, even using the same data for the same area. This inconsistency can cause confusion and be a barrier to uptake and widespread use of the data in the commercial sector. With the widespread availability of Sentinel-1 SAR data and a suite of new algorithms in the commercial and academic sectors, it is timely to develop a method for comparison of different results. In this study, we focus on developing and testing an independent and robust methodology for assessment of different InSAR processing results. Our proposed method is adapted from the Terrafirma Process Validation project; we compare geocoded line-of-sight velocities and time series, density and coverage, as well as some qualitative metrics. We use Sentinel-1 data from an area in Glasgow (UK) processed using 4 different approaches modified RapidSAR, SqueeSAR, GAMMA-IPTA and conventional StaMPS. The main areas of ground motion are detected using all approaches, with the average standard deviation of velocity differences for all inter-comparison pairs in all polygons equal to 1.1 mm/yr. Sentinel-1 InSAR therefore provides comparable results that are independent of processing approaches. However, there are considerable differences in some aspects of the results, in particular in their density and coverage. We discuss the reasons for these differences and suggest a framework for validation that could be used in future national or pan-national ground motion services.