| 15 Sep 2026
Detecting glacier-scale mass-balance heterogeneity using ASTER DEM differencing
Abstract. Glaciers in the Southern Andes show substantial spatial variability in mass balance, reflecting the combined effects of spatial variations in climate forcing and differences in glacier geometry. This study analyses elevation change patterns and climatic drivers for two contrasting glacier systems in southern Chile: the Queulat Glacier Complex and the stratovolcano Melimoyu. Elevation changes from 2000 to 2026 were derived from multi-temporal digital elevation models and analysed using a hypsometric approach to resolve elevation-dependent mass balance.
The results present a pronounced contrast between the two glacier systems. The Queulat Glacier Complex experiences consistently negative mass balance throughout the study period, with a rate of mass loss increasing from approximately −0.23 ± 0.39 to −0.63 ± 0.19 m w.e./yr. In contrast, Melimoyu maintained positive or near-balanced conditions in the early 2000s, followed by a transition to slightly negative mass balance in recent years. Elevation-dependent analyses show that mass loss in Queulat is predominantly associated with extensive low-elevation glacier tongues, while Melimoyu is characterised by a top-heavy hypsometry with a large accumulation area at high elevations.
Examination of downscaled ERA5 reanalysis data shows a consistent regional warming and a decrease in solid precipitation over the study period, while differences in total precipitation between the two sites remain small. This suggests that the contrasting glacier behaviour is not primarily driven by differences in climatic forcing, but rather by differences in glacier geometry and elevation distribution, which modulate the sensitivity to warming.
These results underline the key role of hypsometry and spatial glacier structure in determining individual glacier responses to climate change. The study indicates that glaciers under similar climatic conditions can exhibit substantially different mass-balance trajectories due to their geometry.
This paper presents a time-series analysis of two glaciers in the Andes using ASTER DEMs and climatological data. The authors justify the work by pointing out that existing global compilations (e.g., Hugonnet et al., 2021) are relatively coarse and lack recent updates, which is a reasonable premise for a regional study.
However, there are major methodological and literature-related limitations that need to be addressed:
DEM Generation and Comparison: The study relies on standard AST14DEM products, which are well-documented to be unsuitable for detailed elevation change studies. As established by Girod et al. (2017, https://doi.org/10.3390/rs9070704), the automated SilcAst algorithm used to generate standard AST14DEMs lacks key corrections for satellite attitude jitter and uses a poor image correlation implementation. This routinely results in systematic cross- and along-track undulations adding to up to 20m in amplitude, severe vertical blunders ("salt-and-pepper"), and widespread matching failures over low-contrast snow and ice surfaces. Attempting to filter out these biases and artifacts post-hoc is a flawed strategy when dedicated pipelines—such as MMASTER (Girod et al., 2017), which was designed specifically to correct payload jitter and optimize matching over glaciated terrain—are publicly available and were used to create the input ASTER DEMs for Hugonnet et al. (2021).
Furthermore, comparing heavily post-filtered AST14DEMs against the global Hugonnet dataset—which derived its ASTER inputs from the same raw scenes processed with a jitter-corrected photogrammetric pipeline—is a methodologically flawed comparison. It essentially evaluates the success of post-hoc filtering against a dataset that solved the underlying photogrammetric errors at the source, but subsampled and time-interpolated its output for the sake of global homogeneity.
Given that the study focuses on only two glaciers, the number of ASTER scenes involved is likely quite small and manageable (the exact scene IDs or even the scene count are missing from the paper). The authors should re-process the raw ASTER L1A scenes using a tailored photogrammetric workflow rather than relying on uncorrected standard AST14DEM data. Complete documentation and setup instructions for the MMASTER workflow are freely available online (https://mmaster-workflows.readthedocs.io/en/latest/setup.html). If the authors encounter technical hurdles setting up the processing pipeline, I would be happy to point them in the right direction to help get their dataset properly processed.
Gap-Filling Methodology and Literature: While extensive effort appears to have been placed on gap filling, the specific interpolation algorithms used lack formal justification against established methodological literature. Most notably, the paper omits reference to key foundational work evaluating DEM gap-filling techniques, error propagation, and interpolation biases in glaciated terrain—specifically McNabb et al. (2019, https://doi.org/10.5194/tc-13-895-2019). While regional ASTER studies like Brun et al. (2017) demonstrate the necessity of gap filling, McNabb et al. (2019) provides the standard benchmark for evaluating which interpolation methods (e.g., hypsometric, local spatial, or temporal) minimize systematic elevation bias. The authors should explicitly define their gap-filling workflow and contextualize their chosen methodology against these established standard practices.
Addressing these core methodological and literature gaps will significantly strengthen the manuscript before the remainder of the analysis can be fully evaluated.