Extraction of Glacial Lake Outlines in Tibet Plateau Using Landsat 8 Imagery and Google Earth Engine

Abstract

Glacial lake outburst floods (GLOFs) are among the most serious natural hazards in high mountain regions in the last several decades. The recent global warming has caused dramatic glacial lake changes and increased potential GLOF risk, particularly in Tibet Plateau (TP). Thus there is a pressing need to understand area and spatial distribution of glacial lakes at a large scale. Current efforts about glacial lake mapping in TP region is limited by spurious detections in the heterogeneous backgrounds. The nonlocal active contour algorithm, which takes full consideration of the regional heterogeneity in image, has been effectively applied in the field of medical image segmentation, but has not been tested at large scale of glaciated area yet. Moreover, the improved radiometric resolution and geographic coverage from Landsat 8 provides an opportunity to map glacial lakes. This study evaluated the potential of Landsat 8 images on annual glacial lake mapping in TP region which was characterized by various complex water conditions. The Google Earth Engine based cloud computing effectively facilitated the processing of a complete time series of Landsat 8 imagery from 2015 (156 path/rows and approximately 3580 scenes). Characteristics of glacial lake distribution were analyzed from aspects of size classes and elevation. Our results demonstrate that these strategies and methods automatically produce highly reliable glacial lake extents across the entire TP region, and are potentially applicable to other large-scale glacial lake mapping projects.