Validation of supraglacial bathymetry models developed for optical sensors using high-resolution stereo-imagery: Implications for meltwater storage assessments across the ablation region of the Greenland ice sheet.

Tuesday, 16 December 2014
Mahsa S Moussavi1,2, Waleed Abdalati1,2, Allen Pope3 and Theodore A Scambos1,3, (1)University of Colorado, Boulder, CO, United States, (2)Cooperative Institute for Research in Environmental Sciences, Boulder, CO, United States, (3)National Snow and Ice Data Center, Boulder, CO, United States
Supraglacial lakes on the Greenland ice sheet have attracted a great deal of attention in the scientific community with respect to their role in seasonal enhancements of ice flow velocities. Large amounts of meltwater generated at the surface can be efficiently delivered to the base of the ice sheet through hydro-fractured pathways beneath supraglacial lakes, potentially increasing basal sliding velocities. Therefore, assessments of surface meltwater volumes stored in and transported from supraglacial lakes to the englacial and subglacial systems are crucial for better coupling models of ice sheet hydrology and dynamics, particularly in response to a warming climate.

Several physically-based and empirical passive remote sensing techniques based on MODIS, ASTER, Landsat measurements have been proposed to derive bathymetric information over supraglacial lakes. While unvalidated, most of the techniques have been calibrated against limited in-situ observations and yet have been applied across large regions of the ablation region. In this study, we investigate the validity of such techniques, specifically developed for MODIS, Landsat and WorldView-2 instruments. To make such an assessment, we calibrate the depth-retrieval models by using water-leaving radiances over lakes captured by a specific sensor early in the melt season, and depth measurements from a high resolution WV-2 DEM over the same lakes when devoid of water. Having applied calibrated models over lakes (validation dataset), we then compare the modeled depths against observations derived from the after-drainage DEM of the area. Our primary study site is located in a portion of the ablation region of the GrIS with its center situated at 67o 18’ N, 48o 55’ W at approximately 1200 m A.S.L[1]. Initial results from the study specifically point to the capability of WV-2 multispectral measurements in calculating lake depths with a high degree of accuracy (bias< 2% of mean depth) and precision (RMSE< 12% of mean depth) and an average volumetric error of 6%. Through analyzing MODIS and Landsat measurements over the same lake region, this work will improve the currently-used techniques by allowing calibration/validation over larger areas and will better quantify the errors and uncertainties in lake storage assessments made thus far.