Evaluation of methods for connecting InSAR to a terrestrial reference frame in the Latrobe Valley, Australia

Abstract

Deformation measurements from satellite-borne synthetic aperture radar interferometry (InSAR) are usually measured relative to an arbitrary reference point (RP) of assumed stability over time. For InSAR rates to be reliably interpreted as uplift or subsidence, they must be connected to a defined Earth-centred terrestrial reference frame (TRF), usually made through GNSS continuously operating reference stations (CORS). We adapt and compare three methods of TRF connection proposed by different studies which we term the single CORS RP (SCRP), plane-fit multiple CORS (PFMC), and the multiple CORS RP (MCRP). We generalise equations for these methods, and importantly, develop equations to propagate InSAR and GNSS uncertainties through the transformation process. This is significant, because it is important to not only estimate the InSAR uncertainties, but also to account for the uncertainties that are introduced when connecting to the CORS so as to better inform our interpretation of the deformation field and the limitation of the measurements. We then test these methods using Sentinel-1 data in the Latrobe Valley, Australia. These results indicate that differences among the three TRF connection methods may be greater than their estimated uncertainties. MCRP appears the most reliable method, although it may be limited in large study areas with sparse CORS due to long wavelength InSAR errors and that gaps and/or steps may appear at the spatial limit from the CORS. SCRP relies on the quality of the single CORS connection, but can be validated by unconnected CORS in the study area. The PFMC method is suited to larger areas undergoing slow, constant deformation covering large spatial extents where there are evenly distributed CORS across the study area. Selecting an optimal method of TRF connection is dependent on local site conditions, CORS network geometry and the characteristics of the deformation field. Hence, the choice of TRF connection method should be carefully considered, because different methods may result in significantly different transformed deformation rates. We confirm slow subsidence across the Latrobe Valley relative to the vertical component of the ITRF2014, with localised high subsidence rates near open cut mining activities. Subsidence of ~ -6 mm/year is observed in the adjacent coastal region which may exacerbate relative sea level rise along the coastline, increasing future risks of coastal inundation.