Vertical Accuracy Evaluation Of Digital Elevation Models For Mapping Of Glacial And Periglacial Environments On Keller Peninsula, Antarctica Marine

This paper analyzes the vertical accuracy and performance of Digital Elevation Models (DEMs) TanDEM-X (TDX) and Reference Elevation Model for Antarctica - REMA2 and REMA8 - and their potential for glacial and periglacial environments mapping using the Keller Peninsula (KP) as case study, located in King George Island (KGI), maritime Antarctica. The vertical accuracy of these models was evaluated using a DEM generated from photogrammetry by calculating the Root Mean Square Error (RMSE) for the entire area of the Peninsula and for three surface classes: ice-free, lakes and glaciers areas. The Pearson linear correlation coefficient (R) and error matrix from RMSE values were calculated by statistical analysis. The DEMs potential for geomorphological analysis was evaluated applying the Geomorphons method for automatic classification of glacial relief elements. With a RMSE of 9.85 m, the REMA8 shows the best results. The TDX underestimates the overall elevations and the REMA2 overestimates the highest elevations. The TDX produces the best RMSE values for ice-free areas class (10.4 m) when the lake coverage class is disconsidered. In general, all models show a high correlation coefficient. The Geomorphons classification by REMA2 shows the best results in relief forms mapping located on paraglacial environments, for identifying moraines, scarps and terraces. The macroforms from glacial erosion were better identified in classification of REMA8 and TDX DEM. REMA8 allowed the classification of raised beaches features in almost all coastal extension of the Peninsula and its use is recommended for the coastline paleolevel identification. The vertical accuracy differences among models were influenced by terrain surface characteristics and relief elements configurations.