Abstract
Sea ice concentrations based on AMSR-E 89GHz data are unprecedented in combining data timeliness (about 6 hours after overflight), horizontal resolution (about 5km) and daily global coverage. Here the geoloclation of the AMSR-E Level 1 data (required to use due to the time constraints) is corrected and the sea ice concentrations are validated. The geolocation adjusts the cone angle and scan angle of AMSR-E's conical scanning scheme based on the comparisons of the jump of the AMSR-E brightness temperature at the global coastlines with a global landmask. The average residual error increasing from 250m for the 89GHz channels to 1425m in the 6GHz channels. The ice concentrations are based on the ARTIST (Arctic Radiation and Turbulence Interaction STudy) Sea Ice (ASI) retrieval algorithm which is an enhancement of the Svendsen 85GHz algorithm. Here we review the results of four types of comparisons of the ASI/AMSR-E ice concentrations, namely with (1) Arctic ship based bridge observations of RV Polarstern, (2) optical images of the multispectral imager ETM+ operating on Landsat-7, (3) Envisat and Radarsat-1 SAR images and (4) two other AMSR-E sea ice concentration algorithms (Bootstrap and NASA Team 2) which use the 19/37GHz channels. In spite of the different sensor types, wavelengths and interaction principles of the electromagnetic radiation the four comparisons yield a rather consistent picture. On average the ASI ice concentrations range between those from Landsat and SAR. Both the bias intervals (-2.9...2.6%) and the rms errors are slightly higher than those of the NT2 algorithm, applied to the same scenes. In the hemispherical (Arctic and Antarctic) comparisons of the ASI results with the widely used NASA Team 2 and Bootstrap concentrations, the biases do not exceed 2%, the rms error ranges between 7 and 11% ice concentration.
