Quality Assessment of GEBCO 08, Smith and Sandwell and SRTM30 Plus Grids in the Arctic
Several global compilations of gridded bathymetric data are available to the scientific community often leaving users wondering which data set is best for a given application. The choice of a bathymetry grid is complicated by the fact that there is little supplementary information provided on the accuracy and quality of the gridded products. The IBCAO grid is considered to be the most authoritative representation of Arctic bathymetry, even though in some areas it is based on digitized contours from published maps. Recently Smith and Sandwell (S&S) extended their bathymetric model based on acoustic soundings combined with gravity data as far as 81°N, presenting an alternative to the IBCAO bathymetry model in the Arctic. This study is focused on an assessment of differences between the GEBCO_08 (a derivative of the IBCAO grid), S&S and the SRTM30_Plus global compilations in the Arctic. Differences between the grids were evaluated in terms of quality metrics, which were defined as important when choosing a bathymetry grid, in particular: depth accuracy, internal consistency and interpolation accuracy. Independent, high-resolution and accuracy multibeam sonar (MBES) derived grids, provided by the Geological Institute Russian Academy of Sciences, were used as a ground truth against which the global compilations were compared. The independent MBES grids occupied six sub-regions with distinct differences in morphology as well as accuracy and distribution of source data used for the construction of the global compilation grids; this allowed addressing regional differences between the grids. In terms of accuracy, GEBCO_08 performs better than S&S and SRTM30_Plus over the three out of six sub-regions mainly due to better source data coverage and proper sound speed corrections applied to the source data. S&S and SRTM30_Plus grids perform similar to or slightly better than GEBCO_08 in deep areas. Poor performance of S&S and SRTM30_Plus was observed within shelf areas with the standard deviation of differences with MBES grids exceeding 25% of the water depth. The main reasons for the low accuracy of the S&S and SRTM30_Plus grids on the shelf include the scarcity of source data in the region and, in general, the poor performance of gravity prediction in shallow areas and at high latitudes. In terms of internal consistency, all three grids have artifacts present in the bathymetry, but artifacts in the S&S and SRTM30_Plus grids are more pronounced when visualized. A quantitative assessment of consistency performed for the grids revealed that artifacts in the S&S grid can be as deep (or shallow) as 140% of the water depth on the shelf. These artifacts create wavelengths in the bathymetry shorter than 12.5 km (in areas of no MBES source data) that are not represented by satellite-derived gravity with a stated spatial resolution of 20 - 160 km. Similar performance of interpolation was observed for the S&S and GEBCO_08 grids due to the interpolation algorithm used for construction of both grids, namely a continuous curvature spline in tension.
Journal or Conference Title
Fall Meeting, American Geophysical Union (AGU)
Dec 3 - Dec 7, 2012
San Francisco, CA, United States
American Geophysical Union Publications
Abramova, A., Monahan, D., Mayer, L.A., Lippman, T.C., Calder, B.R., Quality Assessment of GEBCO_08, Smith and Sandwell and SRTM30_Plus Grids in the Arctic, Abstract OS51D-1905 presented at 2012 Fall Meeting, AGU, San Francisco, Calif., 3-7 Dec. 2012.