Automatic calving front extraction from digital elevation model-derived data

Calving Front (CF) is an important parameter to analyse ice sheet dynamics and to measure glacier mass balance. DEM products covering polar regions are critical remote sensing data sources to provide fundamental terrain information for glaciological studies. It is necessary to provide accurate CFs for DEM specific applications, such as mass balance calculation or the substitution of water values with geoid values in the DEM editing process. However, much less attention is paid to automatically delineate CFs from DEM data. In this study, we propose a DEM-based Automatic CF Extraction method (DACE) to efficiently extract the CFs from DEM data. In DACE, a height-sensitive terrain feature is designed to enhance the contrast between the ice sheet and the ocean by combining elevation and roughness information. To improve the CF extraction performance, a two-category image classification based on game theory is proposed that considers the spatial consistency of the feature image created. For validation, DACE was applied to DEM products generated from the single-pass SAR interferometry mission TanDEM-X (TDM) at different posting sizes (12 and 90Â m) and to the optical photogrammetry-based Reference Elevation Model of Antarctica (REMA) with a 2Â m posting. The proposed algorithm can achieve a CF extraction accuracy of better than 14, 20, and 70Â m for the 2-m REMA, 12- and 90-m TDM DEMs, respectively, when compared with the manually delineated CFs. The experimental results demonstrate that the proposed algorithm can effectively extract the CFs from DEM data. DACE can be used to replace water values and edit the DEMs themselves. The CFs extracted from the DEM data can also be used for glacier mass balance calculation with the DEM-based geodetic method and for the temporal analysis of CF changes when multi-temporal DEM data are available