Cloud-free imageries, acquired from Landsat 8 OLI during 2016 to 2018, were used to delineate the extents of the glacial lakes in the mountainous terrain of CPEC
1 Data input Because of the remote location of glacial lakes, the most accessible way to acquire its position and area data is remote sensing images. 14 Landsat 8 OLI Level 1T images during July to October from 2016 to 2018 were selected.Path/Row and date: (148/35 2016/09/24)(148/36 2017/10/29)(149/33 2018/07/19)(149/34 2016/10/01)(149/35 2016/10/01)(149/36 2017/10/04)(150/33 2017/09/09)(150/34 2017/09/09)(150/35 2017/09/09)(150/36 2017/10/27)(151/33 2018/07/17)(151/34 2016/10/15)(151/35 2016/10/15)(151/36 2016/10/15) In addition to the Landsat image, ASTER GDEM V2, with resolution up to 30 m, from USGS was also utilized to analyze slope conditions in this study. Moreover, to identify the different types of glacial lakes, Google Earth was widely applied. 2 Classification With the aim to provide a comprehensive and systematic glacial lake inventory to strengthen the preconditional estimation of GLOF in the CPEC, we categorized the in situ glacial lakes on the basis of their physical conditions and formation mechanisms. The major categoriesglacial lakes are defined as Blocked lake，Erosion lake，Supraglacial lake(SP) and Other lake(OL). And under Blocked lake there are End moraine-dammed lake (EM), Lateral moraine-dammed lake (LM), Other moraine-dammed lake (OM) and Ice blocked lake (IB). Under Erosion lake there are Cirque lake(CQ) and Other erosion lake(OE). 3 Glacial lake coding Consulting the coding system of glaciers formulated by National Snow and Ice Data Center (NSIDC), in this study, we developed a glacial lake coding criterion: “XXnnnnnnNmmmmmmmE” ( “XX” is the abbreviation of glacial lake types; “nnnnnn” is the centroid latitude value of glacial lakes multiplying 100000; Same as “nnnnnn”, “mmmmmmm” represents the longitude while with a “0” in the start if the longitude value is less than 100; “N” and “E” mean north and east). 4 Uncertainty In satellite imagery interpretation, previous studies reported that the interpretation of mixed-pixels can cause the uncertainty in glacial lakes delineation , . A distance of ± 0.5 pixels was regarded as a proper range to represent the uncertainty. So a buffer with a distance of a half-pixel outside the boundaries was adopted to appraise uncertainty in this inventory. As a result, the uncertainty of glacial lake mapping is less than 8%, which could ensure credibility and reliability.