A refined mapping of Arctic lakes using Landsat imagery
Effective mapping of water bodies at regional scales is a challenge with respect to the description and monitoring of hydrological, climatic, and landscape processes. In a region as sensitive to climate change as the Arctic, inaccurate representation of lake cover has probably led to underestimation of the role of lakes as landscape constituents and thus of their contribution to biochemical cycles. To estimate lake cover reliably (and perhaps also its change through time), the scientific community necessitates techniques for mapping water bodies using satellite sensors that include rich historical data sets and have sufficiently fine spatial resolution. Here we applied a density-slicing detection technique to 617 cloud-free Landsat images for the summer months 2006–2011. We developed a comprehensive database of Arctic lakes with a detection accuracy of 80% and examined spatial patterns of lake distribution in relation to landscape properties. We mapped about 3,500,000 lakes; these cover nearly 6% of the Arctic land surface (about 400,000 km2) and are typically small (<0.1 km2). Lake density and lake fraction analyses show that lakes are most common in lowland permafrost areas with tundra vegetation. The method described here can also be used to map and monitor lake cover at regional to hemispheric scales and to monitor changes in lake cover over time.
