Abstract
Time-series SAR interferometry (TS-InSAR) has been widely used to map the millimeter-scale interseismic displacements. Tropospheric noise is still a key error source that hinders further improvement of such measurement. In this article, a new spatiotemporal TS-InSAR tropospheric noise filtering is proposed to approach higher accuracy measurements. We first construct a 2-D arc network for all the data points, and based on that a temporal high-pass and spatial low-pass filtering is applied to estimate the tropospheric noise for all the points. To better filter out long-wavelength interseismic displacements in temporal high-pass filtering, we construct four-candidate time-series models to model the displacement histories for each arc. To avoid overfitting, the F hypothesis test is applied to select the most suitable model for all arcs. Notably, instead of the commonly used Delaunay network, the all-pairs-shortest-path algorithm in the graph theory is employed to reconstruct all arcs and to improve the applicability of the time-series model. Comprehensive tests using synthetic data and Sentinel-1 data covering Central San Andreas Fault (CSAF) creep section validate our tropospheric noise removal approach in measuring the interseismic velocity across the fault.
Keywords
All-pairs-shortest-path (APSP), Delays, filtering, F-test, graph theory, History, hypothesis testSAR interferometry (InSAR), Mathematical models, Spatiotemporal phenomena, Splines (mathematics), Time-series analysis, tropospheric noise removal