Glacier-level and gridded mass change in the source rivers in the eastern Tibetan Plateau (ETPR) from 1970s to 2000
Abstract. The highly glacierised eastern part of the Tibetan Plateau is the key source region of the five major rivers Yangtze, Yellow, Lancang-Mekong, Nu-Salween and Brahmaputra rivers. These exotic rivers are vital freshwater resources for more than one billion people downstream for their daily life, irrigation, industrial use, and hydropower. However, the glaciers have been receding during the last decades and are projected to further decline which will profoundly impact the water availability of these larger river systems. Although few studies have investigated glacier mass changes in these river basins since the 1970s, they are site and temporal specific and limited by data availability. Hence, knowledge of glacier mass changes is especially lacking for years prior to 2000. We therefore applied digital elevation models (DEMs) derived from large scale topographic maps based on aerial photogrammetry from the 1970s and 1980s and compared them to the SRTM DEM to provide a complete picture of mass change of glaciers in the region. The mass changes are presented on individual glacier bases with a resolution of 30 m and the grided (0.1° and 0.5°). Our database consists of 17444 glaciers with a total area of 17426 ± 523 km2. The annual mean mass loss of glaciers is -0.28 ± 0.15 m w.e. in the whole region. This is larger than the previous site-specific findings, the surface thinning increases on average from west to east along the Himalayas-Hengduan mountains with the largest thinning in the Salween basin. Comparisons between the topographic map-based DEMs and DEMs generated based on Hexagon KH-9 metric camera data for parts in the Himalayas demonstrate that our dataset provides a robust estimation of glacier mass changes. However, the uncertainty is high in high altitudes due to the saturation of aerial photos over low contrast areas like snow surface on a steep terrain. The dataset is well suited for supporting more detailed climatical and hydrological analyses and are available at https://doi.org/10.11888/Cryos.tpdc.272884 (Liu et al., 2022).
This preprint has been withdrawn.
Yu Zhu et al.
Yu Zhu et al.
Glacier-level and gridded mass change in the source rivers in the eastern Tibetan Plateau (1970s-2000) https://doi.org/10.11888/Cryos.tpdc.272884
Yu Zhu et al.
Viewed (geographical distribution)
Review of Zhu et al., “Glacier-level and gridded mass change in the source rivers in the eastern Tibetan Plateau (ETPR) from 1970s to 2000” in discussion in ESSD, March 2023
by Romain Hugonnet, University of Washington
The study by Zhu et al. presents an elevation change dataset for glaciers in the Tibetan Plateau. This dataset is based on historical topographic maps compared to the SRTM.
While digging into historical archives is of great interest to the field of glaciology (and others), many processing steps in the study are flawed. My three most concerning issues are the radar penetration assumptions, the bias correction of elevation differences, and the spatial correlations used during error analysis.
Individually, each of these steps can impact the current estimates to a magnitude close to that of the signal (or error) estimated.
In addition, many statements are unclear, and many references in the text are out-of-date or irrelevant.
1/ General error reporting
The authors do not specify which confidence level is reported, so I assume it is 1-sigma (68% confidence) everywhere. To make robust statements, one should rather use 2-sigma (95% confidence), as +/-1-sigma is too narrow a confidence interval.
With 2-sigma, the current regional result of the authors is not statistically different from zero at -0.28 +/- 0.30 m yr-1. From basic error propagation theory, any result aggregated at a larger-scale has lower uncertainty than its smaller-scale components. This implies that all other results of the study (from pixel scale to subregions elevation change) must also not be statistically significant.
In short, if the errors provided by the authors are realistic, then the results are largely useless for the community. This is worrying, but might only be a question of whether the error estimation is reliable or not (some sources of error might be overestimated). However, looking into the details, several sources are actually highly underestimated in the current approach, as I further describe in my main comments 2/, 3/ and 4/.
In short, the authors should strive to improve their entire error analysis (systematic AND random) and paint a clear picture of the confidence in their dataset and result. However, even if they do, they might find that the data is not of good enough quality, and that the results are not statistically significant (which would not surprise me, given the structures of noise known to exist in some of the DEMs used, see comment 3/).
2/ Penetration of X-band is a systematic error and not negligible
The penetration of X-band in glaciers of the Tibetan plateau is not negligible, especially given the period of year where SRTM was acquired. Many studies provide such estimates of X-band penetration, see for instance Li et al. (2021) or Zhou et al. (2022), to only point out recent work in HMA.
Following Li et al. (2021), about 2 meters of average X-band penetration would still be present. Such a systematic error for a period of ~25 years will affect thinning estimates by 0.08 m yr-1, which is 30% of the signal measured by the authors of -0.28 m yr-1.
In short, X-band penetration needs to be corrected, not neglected.
3/ Long-range spatial correlations of errors are not accounted for
All of the DEMs used by the authors contain long-range correlated noise (that can be seen on Figure 4, top panel): KH-9, topographic maps, SRTM-C and -X. All of these have especially strong structures of noises. For more visual examples, see Hugonnet et al. (2022), Fig. 1 and Supplementary Figs.
Not accounting for these patterns of errors can underestimate final mean elevation change uncertainties by a factor of 50 on high-res modern DEMs, and much more on historic ones. Right now the authors use a correlation length of 600 m, which is ridiculously small when it is known that KH-9 has errors correlated up to dozens of kilometers (Dehecq et al. (2020)), and similar for other DEMs as can be seen on the visual examples listed in the previous paragraph.
Hugonnet et al. (2022) developed a framework to account for these errors reliably for any DEM, with open-source tools to perform error propagation, used by other studies while still in development (Dehecq et al. (2020) for KH-9, Zhou et al. (2022) for SRTM).
It is also a bad habit to show elevation difference maps without stable terrain (Figure 10), as it helps visually assess the data quality. Authors should add this back.
4/ Elevation-dependent corrections introduce important systematic errors
Elevation-dependent biases generally arise from differences in high curvature terrain (peaks and cavities) rather than GCP alignment, as justified by the authors that apply such a correction. See Gardelle et al. (2013).
Problematically, curvature biases are specific to stable terrain (where the peaks are) but do not have an impact on glaciers. Applying a correction to all terrain will therefore introduce systematic errors for the entire elevation change maps, including glaciers. Those can have a magnitude in the order of meters.
In short, the authors should either better justify the relevance of such a correction, or not apply any.
There were too many unclear or exaggerated statements, so I stopped taking notes entirely.
I advise the authors to:
1/ Revise their statements carefully for scientific clarity and accuracy,
2/ Revise their references, as many are out-of-date or biased towards a group (too many Bolch papers) or simply irrelevant,
3/ Ask for English proofreading.
Dehecq, A., Gardner, A. S., Alexandrov, O., McMichael, S., Hugonnet, R., Shean, D., & Marty, M. (2020). Automated Processing of Declassified KH-9 Hexagon Satellite Images for Global Elevation Change Analysis Since the 1970s. Frontiers of Earth Science, 8, 516.
Gardelle, J., Berthier, E., & Arnaud, Y. (2012). Impact of resolution and radar penetration on glacier elevation changes computed from DEM differencing. Journal of Glaciology, 58(208), 419–422.
Hugonnet, R., Brun, F., Berthier, E., Dehecq, A., Mannerfelt, E. S., Eckert, N., & Farinotti, D. (2022). Uncertainty Analysis of Digital Elevation Models by Spatial Inference From Stable Terrain. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 15, 6456–6472.
Li, J., Li, Z.-W., Hu, J., Wu, L.-X., Li, X., Guo, L., Liu, Z., Miao, Z.-L., Wang, W., & Chen, J.-L. (2021). Investigating the bias of TanDEM-X digital elevation models of glaciers on the Tibetan Plateau: impacting factors and potential effects on geodetic mass-balance measurements. Journal of Glaciology, 67(264), 613–626.
Zhou, Y., Li, X., Zheng, D., & Li, Z. (2022). Evolution of geodetic mass balance over the largest lake-terminating glacier in the Tibetan Plateau with a revised radar penetration depth based on multi-source high-resolution satellite data. Remote Sensing of Environment, 275, 113029.