Glacier surge monitoring from temporally dense elevation time series: Application to an ASTER dataset over the Karakoram region

Abstract

Glacier surges are spectacular events that lead to surface elevation changes of tens of metres in a period of a few months to a few years, with different patterns of mass transport. Existing methods to derive elevation change associated with surges, and subsequent quantification of the transported mass, rely on differencing pairs of digital elevation models (DEMs) that may not be acquired regularly in time. In this study, we propose a workflow to filter and interpolate a dense time series of DEMs specifically for the study of surge events. We test this workflow on a global 20-year dataset of DEMs from the optical satellite sensor Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER). The multistep procedure includes linear non-parametric locally weighted regression and smoothing scatterplots (LOWESS) filtering and approximation by localized penalized splines (ALPS) interpolation. We run the workflow over the Karakoram region (High Mountain Asia). We compare the produced dataset to previous studies for four selected surge events, on the Hispar, Khurdopin, Kyagar, and Yazghil glaciers. We demonstrate that our workflow captures thickness changes on a monthly scale with detailed patterns of mass transportation. Such patterns include surge front propagation and dynamic balance line changes, among others. Our results allow a remarkably detailed description of glacier surges at the scale of a large region. The workflow preserves most of the elevation change signal, with underestimation or smoothing in a limited number of surge cases.