Monitoring and classification of high Arctic lakes in the Svalbard Islands using remote sensing

Abstract

Arctic lakes are an essential element of the environment. This study inventoried and classified lakes on three of the Svalbard Islands (Spitsbergen, Barentsøya, and Edgeøya) using Sentinel-2 images. The Forel-Ule color index (FUI) was employed to measure the lake's color, and a new equation to calculate the concentration of suspended sediment using satellite images was developed and tested using measurements from coastal waters. These two measures, plus the normalized difference vegetation index for vegetation on the lakeshore, and the lakes' location were used for clustering and classification. More than 1000 natural lakes were found on the three Svalbard islands. Nearly 80 % of these were smaller than 0.01 km2 (1 ha), with the largest a few km2 in area. Lake distribution analysis identified six zones of increased lake density, five of which were located along the western coast of Spitsbergen. The FUI was thus demonstrated to be useful for the monitoring of Arctic lakes. The K-means clustering results and analysis of the spectral signatures divided the lakes into seven classes: clear water, moderately clear water, tundra, coastal tundra, meltwater, extreme meltwater, and supraglacial lakes. Meltwater and clear water lakes were the two most common types, accounting for about 80 % of the total lake area.