A correlation-based approach to modeling interferometric radar observations of the Greenland ice sheet

Interferometric synthetic aperture radar (InSAR) phase observations have greatly increased our understanding of the topography and motion of ice sheets, but yield little information on the sub-surface structure, a needed description for mass-balance estimates. Inversion of a diffuse volume scatter model shows that InSAR correlation values, ρ, can be related to radiowave penetration depths, d, which depend on characteristics of the snow/ice volume. Application to European Research Satellite (ERS) images (VV, 5.6 cm, 23° incidence angle) of the Greenland ice sheet imply C-band d of 0 m along the rocky coast, 10–20 m in the bare ice zone, and 20–35 m in the percolation zone and dry snow zone, consistent with in situ results. Moreover, volume scattering reduces the ERS critical baseline from about 1100 m to 300 m.; Correlation and backscatter power (σ0) observations can be combined for further understanding of the snow/ice volume. In particular, ρ and σ0 data of 15 km-long, 50 m-high topographic undulations in the dry snow zone are minimum on the windward side and maximum on the lee side, with 1 to 3 dB variation typical. These spatial variations in the scattering medium appear to follow from differences in snow accumulation due to prevailing winds. Assuming that snow-grains are the dominant source of backscatter, the classical independent-scatterer model is physically implausible at firn densities; a second-order dense-medium radiative transfer model also is unable to explain both the observed d and σ0. A modified Born approach provides a better match to σ0 and ρ separately, but leads to different grain size solutions for each measurement type. A buried layer model based on the incoherent addition of echoes from hoar layer interfaces, in which scattering from a single layer is derived from small-perturbation methods, reconciles the ERS σ0 and ρ data, with variations in hoar layer spacing of 12–17 cm providing the needed structural fluctuations for the observed range of σ 0 and ρ. Translation of layer spacing into accumulation rates predicts a 40% variability in accumulation rate from the windward to lee side and, more importantly, addresses high-resolution mapping of continental accumulation rates.