Deriving high-accuracy lake water-level changes from multi-source satellite datasets

Abstract

ABSTRACT Satellite data have exhibited a great potential to track lake water-level changes. By integrating multi-source satellite datasets (i.e., Ice, Cloud and land Elevation Satellite-2 (ICESat-2) lidar datasets, Landsat imagery, and Global Surface Water Dataset (GSWD)), we proposed a method to track annual water-level changes of Lake Mead, USA over 1984–2018. Specifically, we first extracted ICESat-2 profiles as elevation information. Second, we used Landsat imagery to extract annual lake boundaries and calculate lake areas. Third, we constructed the relationship between elevations and inundation frequencies provided by the GSWD for ICESat-2 points, and applied it to extract annual water levels by using inundation frequency of Landsat derived lake boundaries. Finally, we created the relationship between lake water level and lake areas via a cubic polynomial model to refine the annual lake water levels. Our water level results agreed well with the in-situ data (R 2 = 0.99, p < 0.01, RMSE = 0.64 m). Our results showed that the ICESat-2 data performed well in providing accurate elevation information around the shore of lake. Using satellite datasets, our proposed method has the potential to derive water levels for water bodies worldwide in the future.