A new global dataset of mountain glacier centerline and length

Abstract. Length is one of the key determinants of glacier geometry and is an important parameter of glacier inventory and modeling; glacier centerlines are crucial inputs for many glaciological applications. In this study, the centerlines and maximum lengths of global glaciers were extracted using an automatic extraction algorithm based on the latest global glacier inventory data, digital elevation data (DEM), and European allocation theory. The glacier polygons were reconstructed according to the geometric principle and an automatic checking algorithm for the global glacier outlines was designed to filter erroneous or unsupported glacier outlines. The DEMs of global glacier-covered regions were compiled using available DEMs. An updated automatic extraction tool was designed independently, and a parameterization scheme with empirical thresholds was applied for data production. The accuracy of the dataset was evaluated using random assessment with visible interpretation and comparative analysis with another dataset. The 10,764 erroneous glacier polygons, 7,174 ice caps, and 419 nominal glaciers from the Randolph Glacier Inventory (RGI) version 6.0 were identified and excluded, accounting for 8.25 % of the total. In total, 198,137 glacier centerlines were generated, accounting for 99.74 % of the total input glaciers and 91.52 % of the RGI v6.0. The accuracy of glacier centerlines was 89.68 %. The comparison between the dataset and previous datasets suggested that the majority of glacier centerlines were slightly longer than those in RGI v6.0. The extraction method of this study has a strong ability to obtain the maximum length of glaciers, meaning that the maximum lengths of some glaciers were likely underestimated in the past. The dataset constructed includes 14 sub-datasets, such as the global glacier centerline dataset, global glacier maximum length dataset, and global glacier DEM dataset, all of which can be found at link: https://doi.org/10.11922/sciencedb.01643 (Zhang and Zhang, 2022).