A new automatic approach for extracting glacier centerlines based on Euclidean allocation

Abstract

Abstract. Glacier centerlines are crucial input for many glaciological applications. From the morphological perspective, we proposed a new automatic method to derive glacier centerlines, which is based on the Euclidean allocation and the terrain characteristics of glacier surface. In the algorithm, all glaciers are logically classified as three types including simple glacier, simple compound glacier, and complex glacier, with corresponding process ranges from simple to complex. The process for extracting centerlines of glaciers introduces auxiliary reference lines and follows the setting of not passing through bare rock. The program of automatic extraction of glacier centerlines was implemented in Python and only required the glacier boundary and digital elevation model (DEM) as input. Application of this method to 48 571 glaciers in the second Chinese glacier inventory automatically yielded the corresponding glacier centerlines with an average computing time of 20.96 s, a success rate of 100 % and a comprehensive accuracy of 94.34 %. A comparison of the longest length of glaciers to the corresponding glaciers in the Randolph Glacier Inventory v6.0 revealed that our results were superior. Meanwhile, our final product provides more information about glacier length, such as the average length, and the longest length, the lengths in the accumulation and ablation regions of each glacier.