A framework of freshwater and saline lake typology classification through leveraging hydroclimate, spectral, and literature evidence

Abstract

Separating saline and freshwater lakes is of immense importance for monitoring lentic ecosystems and assessing surface freshwater availability. However, conventional lake typology often relies on in situ salinity measurements, thus limiting the feasibility of a global application. To enable an effective workaround, we here propose and test a framework that classifies saline and freshwater lakes through systematic, integrated leveraging of hydrological, climatological, spectral, and literature evidence. In principle, saline lakes were sequentially identified if a lake is (1) located at the drainage terminal where salinity tends to build up as a result of lacking surface outflow (hydrological evidence), (2) distributed in an arid/semiarid climate with lacustrine evaporites visible from a high-resolution multispectral image (climatological and spectral evidence), or (3) documented to be (sub)saline or have measured salinity exceeding the maximum freshwater threshold (literature evidence). We applied the HydroSHEDS hydrography dataset, the Köppen-Geiger climate classification, and a large volume of spectral satellite images and literature to implement this framework. This framework was tested on the Tibetan Plateau using the Landsat-based, 30-m-resolution UCLA Circa-2015 global lake mask, and the result was benchmarked against the Chinese national lake survey. Our collected literature shows a 95% agreement between the lake types classified using hydrological, climatological, and spectral evidence alone and those documented in the literature. After further corrections based on the literature evidence, our final classification reached an overall accuracy of nearly 70% in terms of lake count and 94% in lake area, implying that the efficacy of our method tends to favor relatively large lakes. Given such promising accuracies, we further applied our framework to Australia, showcasing its effectiveness for lake typology classification without additional in situ surveys across the world’s most saline continent. While our method, in theory, results in a conservative recognition of saline lakes and should not be considered a complete substitute for in situ surveys, the proposed framework demonstrates potential for performing global-scale freshwater and saline lake typology classification with reasonable accuracies, using only remote sensing images, open-source hydrography and climate data, and existing literature.