Integrated satellite interferometry: Tropospheric noise, GPS estimates and implications for interferometric synthetic aperture radar products

Interferometric synthetic aperture radar (INSAR), like other astronomic and space geodetic techniques, is limited by the spatially and temporally variable delay of electromagnetic waves propagating through the neutral atmosphere. Statistical analysis of these variations, from a wide variety of instruments, reveals a power law dependence on frequency that is characteristic of elementary (Kolmogorov) turbulence. A statistical model for a major component of the delay fluctuations, the “wet” component, has previously been developed by Treuhaft and Lanyi [1987] for very long baseline interferometry. A continuous Global Positioning System (GPS) network is now in place in southern California that allows estimation of, along with geodetic parameters, the total delay due to the atmosphere above each site on a subhourly basis. These measurements are shown to conform to the Treuhaft and Lanyi (TL) statistical model both temporally and spatially. The TL statistical model is applied to the problem of INSAR and used to produce the covariance between two points separated in time and/or space. The error, due to the atmospheric variations, for SAR products such as topography and surface deformation is calculated via propagation of errors. There are two methods commonly cited to reduce the effect of atmospheric distortion in products from SAR interferometry, stacking and calibration. Stacking involves averaging independent interferograms to reduce the noise. Calibration involves removing part (or all) of the delay using data from an independent source such as total zenith delay estimates from continuous GPS networks. Despite the relatively poor spatial density of surface measurements, calibration can be used to reduce noise if the measurements are sufficiently accurate. Reduction in tropospheric noise increases with increasing number of measurement points and increasing accuracy up to a maximum of √N, where N is the number of points. Stacking and calibration are shown to be complementary and can be used simultaneously to reduce the noise to below that achievable by either method alone.