An Effective Low-Cost Remote Sensing Approach to Reconstruct the Long-Term and Dense Time Series of Area and Storage Variations for Large Lakes.

Abstract

Inland lakes are essential components of hydrological and biogeochemical water cycles, as well as indispensable water resources for human beings. To derive the long-term and continuous trajectory of lake inundation area changes is increasingly significant. Since it helps to understand how they function in the global water cycle and how they are impacted by climate change and human activities. Employing optical satellite images, as an important means of lake mapping, has been widely used in the monitoring of lakes. It is well known that one of the obvious difficulties of traditional remote sensing-based mapping methods lies in the tremendous labor and computing costs for delineating the large lakes (e.g., Caspian Sea). In this study, a novel approach of reconstructing long-term and high-frequency time series of inundation areas of large lakes is proposed. The general idea of this method is to obtain the lake inundation area at any specific observation date by referring to the mapping relationship of the water occurrence frequency (WOF) of the selected shoreline segment at relatively slight terrains and lake areas based on the pre-established lookup table. The lookup table to map the links of the WOF and lake areas is derived from the Joint Research Centre (JRC)Global Surface Water (GSW) dataset accessed in Google Earth Engine (GEE). We select five large lakes worldwide to reconstruct their long time series (1984–2018) of inundation areas using this method. The time series of lake volume variation are analyzed, and the qualitative investigations of these lake changes are eventually discussed by referring to previous studies. The results based on the case of North Aral Sea show that the mean relative error between estimated area and actually mapped value is about 0.85%. The mean R2 of all the five lakes is 0.746, which indicates that the proposed method can produce the robust estimates of area time series for these large lakes. This research sheds new light on mapping large lakes at considerably deducted time and labor costs, and be effectively applicable in other large lakes in regional and global scales.