An improved dataset of ASTER elevation time series in High Mountain Asia to study surge dynamics

Abstract

Some glaciers display flow instabilities, among which surge events particularly stand out. Surges are quasi-periodic flow perturbations with an abnormally fast flow over a few months to years. It can result in surface elevation changes of more than 100 m in a few months. The estimation of the mass transfer and the flow variation can be inferred from the glacier surface elevation and velocities. It is critical data to better understand the dynamics of a surge. While satellite-based DEMs provide useful information for studying surges, their use in previous studies was generally limited to a few DEM differences extending over periods of several years. To date, very few studies have leveraged the full time series of elevation data available since ~2000 which could help quantify the variations of mass transfer during the very short surge phases. Here, we exploited the high temporal and spatial coverage of the ASTER optical satellite sensor to compute a dense time series of elevation suited for the study of surges. Our case study area is the Karakoram range, in High Mountain Asia. We used non-filtered ASTER digital elevation models (DEMs) of 100 m resolution from Hugonnet et al. (2021). The time series range from about 2001 to 2019, with a median of 56 observations per on-glacier pixel over the whole period. We developed a specific method for filtering the elevation time series that preserves surge signals, as opposed to the original method that tends to reject this behaviour as outliers. A LOWESS method – locally weighted polynomial regression (Derkacheva et al., 2020; Cleveland, 1979) is at the core of this workflow. Then, we predicted the elevation over a regular temporal and spatial grid from filtered data, with the B-spline method ALPS-REML (Shekhar et al., 2021). In this presentation, we will present the results of this method applied to more than 1000 DEMs covering the Karakoram region to derive elevation time series at 100 m resolution. The filter and the prediction performances will be discussed. The results will be compared with those of other studies, in terms of surge onset and end dates, location or volume transported. Finally, the  elevation data set will be analysed with regard to velocities extracted from ITS_LIVE (Gardner et al., 2024) to validate the approach and highlight the complementarity of both types of observations.