Processing Steps

A global ocean reference network was constructed from four years of Topex/Poseidon altimeter data (Repeat Cycles 11-158, 01 January 1993 to 31 December 1996). This network provides an ideal framework for estimating radial orbit errors and systematic intersatellite and intrasatellite biases. Ice sheet elevation differences were computed at the crossover points between the GEOSAT (GM and ERM) and SEASAT satellite tracks. The crossover data were initially screened to ensure that instrument and environmental corrections were consistently applied to surface elevations. An editing procedure was used to eliminate altimeter waveforms where the initial waveform samples were exceptionally large and noisy (Davis et al. 2001).

The altimeter waveforms were retracked using a threshold algorithm designed specifically for measuring ice sheet elevation change. The Topex/Poseidon orbit error and intersatellite bias corrections were applied to the ice sheet crossover data using a root-difference-square as a measure of the magnitude of the orbit-error reduction in ocean and ice sheet data.

Elevation change data were spatially averaged to 50 x 50 km grid cells. This coarse spatial resolution provides a representative estimate of ice sheet change, because the elevation change data are heavily biased toward the northern interior of the ice sheet. A spatial average was then computed by averaging the individual results for all available cells. The following map shows estimated surface elevation change (Davis et al. 2001):

This map is similar to the one published in Davis et al. (2000), with two important changes. First, Davis computed the average elevation change estimate for each 50 km grid cell after excluding individual elevation change measurements that exceeded two standard deviations (SD) from the primary Gaussian distribution (2 SD edit). This reduced the spatial variability of the dH/dt estimates between adjacent cells in a few areas with low numbers of measurements within a given cell. Second, Davis excluded dH/dt estimates for a given cell when the random error estimate exceeded ± 4 cm/yr after the 2 SD edit.

Many of the radar-based dH/dt estimates in the northwest part of the study area are negatively biased because of differences in surface conditions resulting from strong melt years in 1977 and 1978. Davis computed a linear regression between dH/dt estimates from 1993-1998 laser altimetry and 1978-1988 radar altimetry as a function of elevation. Radar dH/dt cells north of 70.5°N and west of 38°W were used in the regression along with radar dH/dt cells west of 43.5°W and 69.5 to 70.5°N (west of the radar dH/dt discontinuity). These areas show viable evidence that radar dH/dt measurements are erroneous (Davis et al. 2001).

Davis corrected the dH/dt values in these two areas using the average elevation within each radar dH/dt cell and a linear regression described in Davis et al. (2001). The following map shows the resulting corrected dH/dt estimates; the negative radar estimates in the northwest part of the study area are now slightly positive.

Error Sources

Refer to Davis et al. 2000 for details of orbit error analysis with SEASAT and GEOSAT. Comparison with laser dH/dt results and other in-situ data indicates that a typical error value for elevation change estimates in a 50 km cell is approximately ± 2 cm/yr (Davis et al. 2001).