Water body distributions across scales: a comparison of three Arctic wetlands
Water bodies are ubiquitous features in Arctic wetlands, ranging from very small polygonal ponds to very large thermokarst lakes. Ponds, i.e. waters with a surface area smaller than 1 ha, have been recognized as hotspots of biological activity and greenhouse gas emissions. Regional and global models, however, cannot resolve ponds due to the coarse resolution. The aims of this study were to identify common characteristics of Arctic wetlands regarding (1) water body size and abundance, and (2) Landsat subpixel fraction of water cover. We mapped water bodies in three Arctic wetlands, i.e. Polar Bear Pass on Bathurst Island in the Canadian High Arctic, Samoylov Island in the Lena River Delta in Siberia, Russia, and Barrow Peninsula on the Alaska Coastal Plain. High-resolution (0.3 to 4 m) water body maps were overlain on to Landsat albedo maps to extract the proportion of open water within a Landsat mixed pixel. At all three sites ponds occupied 95% of the total number of surface waters. Surface waters smaller than 0.1 ha, which cannot be detected with Landsat data, still contributed 60% and higher to the total number. All study areas showed similar rates of decline in water body abundance with increasing water surface area (Fig. 1). Previous studies have fitted abundance-size distributions of water bodies to the Pareto distribution, which appears linear on a log-log plot. Our data, however, shows paretian behavior only in the upper tail of the distribution so that the Pareto distribution strongly overestimates small water bodies. Landsat albedo increased with decreasing proportion of open water cover per Landsat pixel. Linear regressions for albedo values with a subpixel water cover between 100% and less than 5% showed r-square values larger than 0.8, which constitutes a better performance than other more complex unmixing methods. In conclusion, all three wetlands showed similar properties regarding size-abundance data of water bodies, scaling errors, and retrieval of subpixel water cover via Landsat albedo. Common scaling procedures regarding surface waters can therefore be applied to similar wetland regions across the Arctic for implementation in regional and global ecosystem and climate models.