Abstract:

Supraglacial lake is an important part of the ice sheet hydrological system, which can reflect the characteristics of ice sheet surface meltwater and influence the polar climate and environmental changes. In this paper, a fine extraction method of the supraglacial lake by spectral information and morphology information in northern Greenland is proposed based on WorldView-2 images. Firstly, the ice surface water was segmented from ice with index of water body suitable for ice environment NDWIice and threshold. Then, the ice surface water is divided into three types: Independent supraglacial lake, independent supraglacial runoff and mixed area, by morphological rules. Finally, Fourier descriptor-assisted watershed algorithm is introduced to extract the ice lake. In order to verify the feasibility of the method, WorldView-2 image sequences of nine periods in one ablation period were selected for experiment. Compared with other automated methods and manual digitalization results, the extraction accuracy and integrity of the proposed method are higher than 90%, and the average error rate is lower than 0.2%, which shows obvious advantages.

Key words:  supraglacial lake, normalized difference water index-ice, morphological rule, Fourier descriptor, watershed algorithm