An intensity triplet for the prediction of systematic InSAR closure phases

Abstract

Synthetic Aperture Radar (SAR), a microwave-based active remote sensing technique, has had a rich and contemporary history. Because such platforms can measure both the phase and intensity of the reflected signal, interferometric SAR (InSAR) has proliferated and allowed geodesists to measure topography and millimeter-to-centimeter scale deformations of the Earth's surface from space. Applications of InSAR range from measuring the inflation of volcanoes caused by magma movement to measuring the subsidence in permafrost environments caused by the thawing of ground ice. Advancements in InSAR time series algorithms and speckle models have allowed us to image such movements at increasingly high precision. However, analysis of closure phases (or phase triplets), a quantification of inconsistencies thought to be caused by speckle, reveal systematic behaviors across many environments. Systematic closure phases have been linked to changes in the dielectric constant of the soil (generally thought to be a result of soil moisture changes), but existing models require strong constraints on structure and sensitivity to moisture content. To overcome this obstacle and decompose the closure phase into a systematic and stochastic part, we present a data-driven approach based on the SAR intensities. Intensity observations are also sensitive to surface dielectric changes. Thus, we have constructed an intensity triplet that mimics the algebraic structure of the closure phase. A regression between such triplets allows us to predict the systematic part of the closure phase, which is associated with dielectric changes. We estimate the corresponding phase errors using a minimum-norm inversion of the systematic closure phases to inspect the impact of such systematic closure phases on deformation measurements. Correction of these systematic closure phases that correlate with our intensity triplet can account for millimeter-scale fluctuations of the deformation time series. In permafrost environments, they can also account for displacement rate biases up to a millimeter a month. In semi-arid environments, these differences are generally an order of magnitude smaller and are less likely to lead to displacement rate biases. From nearby meteorological stations, we attribute these errors to snowfall, freeze-thaw, as well as seasonal moisture trends. This kind of analysis shows great potential for correcting the temporal inconsistencies in InSAR phases related to dielectric changes and enabling even finer deformation measurements, particularly in permafrost tundra.