Large lake gauging using fractional imagery

Large floodplain lakes provide riparian habitat, are sediment and nutrient sinks, help control flow connectivity and flooding along rivers, and are both used by humans and strongly impacted by human activity. However, water level in many remote large floodplain lakes, especially in developing countries, is often monitored inconsistently or not at all. In this study, a novel method for estimating large lake water level using passive, optical remote sensing data combined with any digital elevation model (DEM) is presented. The method obtains water level estimates at 30 m² resolution using Landsat, in this case in conjunction with SRTM elevation data, nested within a 240 m² grid "fishnet". A probabilistic mean of elevation values for all water-designated pixels (between 5% and 95% filled within each grid) produces lake water levels often accurate to within ±50 cm of gauged reference data on Lake Curuai in the Amazon River and Tonle Sap Lake along the Mekong River. The method is relatively insensitive to cloud cover, especially as lake size increases. This study is the first to use solely passive optical remote sensing data for water level estimation and thus could be used to produce accurate, long-term estimations of water level in many large lakes globally. The use of optical sensors to obtain lake water level is both an important complement and potential alternative to methods that use active sensors.