Rapid glacier retreat and downwasting throughout the European Alps in the early 21st century

Mountain glaciers are known to be strongly affected by global climate change. Here we compute temporally consistent changes in glacier area, surface elevation and ice mass over the entire European Alps between 2000 and 2014. We apply remote sensing techniques on an extensive database of optical and radar imagery covering 93% of the total Alpine glacier volume. Our results reveal rapid glacier retreat across the Alps (−39 km² a−1) with regionally variable ice thickness changes (−0.5 to −0.9 m a−1). The strongest downwasting is observed in the Swiss Glarus and Lepontine Alps with specific mass change rates up to −1.03 m.w.e. a−1. For the entire Alps a mass loss of 1.3 ± 0.2 Gt a−1 (2000–2014) is estimated. Compared to previous studies, our estimated mass changes are similar for the central Alps, but less negative for the lower mountain ranges. These observations provide important information for future research on various socio-economic impacts like water resource management, risk assessments and tourism.

In general the manuscript is concisely written and the results are appropriately discussed in the context of previous literature. Considering the supplementary material the presentation of the results in figures and tables is substantial. Just the results section should be realigned for better comprehensibility. Sommer et al. provide sufficient methodological detail such that the experiments could be reproduced. Nevertheless, I recommend the authors to publish the glacier outlines as an openaccess data set along with the manuscript. Since a number of methods where applied to different optical or elevation data in the recent years, the set of outlines will provide an idea of uncertainties and differences of semi-automatic and manual mapping based on different data sources. Since no fundamental analysis has to be applied for revisions, I recommend to accept after consideration of the revisions suggested in the specific and detailed comments.

Specific comments
Despite the well elaborated presentation of the methods and the results, there are three main concerns to be considered for revisions.
1 Uncertainty of glacier elevation changes The presented uncertainties of mean elevation change rates (Tab. 1) appear to be rather low. For both periods they are less than 2% or a maximum of 0.21m for the total period. These are theoretical values regarding co-registration performance in stable terrain, but do not include the uncertainty caused by errors of the DEM in steep terrain, along steep slopes at glacier margins or glacier areas shaded by topography, which is important in particular for small to very small glaciers and low-lying glacier tongues. The presented uncertainty of the elevation change rates is distinctly lower than uncertainties presented in the publications referred to in line 290 (#17-19) For an additional uncertainty estimation, the authors should present the two-year elevation change rates between the two TanDEM-X DEM, as e.g. shown in Malz et al. 2018 (#17). Relating to Tab.1, mean elevation change rates of the regions are varying to be either more or less negative for the different periods. Accumulating the elevation change rates to total elevation changes of the periods and dividing the difference of the total changes by two years results in elevation change rates varying significantly for nearby regions between 2012 and 2014 (e.g. -0.94 ma-1 for region 01 and +0.18 ma-1 for region 02). In particular, for region 08, the two-year elevation change rate would be -0.85 ma-1 and, thus, more negative than the 12/14-year period mean. However, 2013 glacier mass balances have been comparatively less negative compared to mean mass balances since 2000 (see WGM mass balance data, e.g. Hintereisferner and Kesselwandferner). If those differences can be related to the median dates of TanDEM-X acquisitions showing differing snow accumulation, this has to be stated as a potential error source increasing uncertainty of elevation change rates nonetheless. I suggest to revise the uncertainty estimation of the elevation change (rates). If those uncertainties are found to be higher, uncertainties of (specific) mass change rates have to be corrected accordingly.

Future Scenarios
In contrast to existing model simulations of future glacier scenarios of the Alps, the analysis presented here does not consider dynamic adjustments, future scenarios of changing glacier mass balance and already observed processes like increasing debris coverage and glacier disintegration for the calculation of future glacier coverage. The presented scenario is also based on ice thickness data showing high relative deviations in ice thickness particularly for Alpine glaciers. This issues incorporates a high uncertainty of the presented results for 2050 and 2100 either in percentage of glacier coverage or time. The calculated future glacier volume also do not reflect the high variability of recent glacier changes and geometries in each of the regions. Model studies of e.g. Zekollari et al. 2019 (#4) present a large range of results with respect to different scenarios. To my opinion the results presented here are neither necessarily needed for the manuscript, which mainly aims in the presentation of a temporally and regionally consistent analysis of recent change rates, neither are they substantial considering all the assumptions and limitations. Thus, I suggest to remove this part of the results.
3 Results sections In general, the results (L70-105ff) are hard to read in the present form. I would suggest to realign this section by presenting at least one characteristic result for each region in order of the IDs (West to East) first and then to close with the main findings of the comparison (e.g. L72ff, L77ff). L72ff: This appears to be a finding which also can be highlighted in the abstract. L74: significantly: do you mean (statistically) significant? L77: Is this a statement or a result? If latter, this can also be an important point for the abstract L101: 'large' instead of larger L104: Similar to what? L140: add'-' before 0.10 L165: Those are areas of small glaciers and prone to higher uncertainties in analysis of surface elevation change (see specific comments) L172ff: With respect to the specific comments, I would suggest to remove this section L231ff: The climate classification is, as far as I can see, not further used in the analysis and may be removed. L238: Replace 'periods' by 'dates' L326: applying a 'conversion factor assuming a mean density' of… L413ff: With respect to the specific comments, I would suggest to remove this section !"#$%&'(%$")'"*$+,$-$*". In the revised version of their manuscript, Sommer and colleagues have addressed all specific and detailed comments accordingly. They have changed the title to more focus on the observed changes. The authors have retained the future scenario analysis, although the importance of this estimation was reduced. With respect to processes like dynamic adjustment and climate change impacts on glacier mass balance, which are not considered in the approach, the results appear to be less robust compared to more detailed model studies. Conversely, the observed volume changes include processes of glacier vanishing such as basal melt and high mass losses due to the present state of low glacier dynamics in particular for many of the small to medium-size glaciers. Thus, the presentation of the results in Line 118ff, also stating the limitations, appears to be appropriate.
Still, I'm concerned on the presented uncertainty of the surface elevation changes. Coregistration will eliminate a number of systematic errors of surface elevation changes (mean/median deviation, shifts). Nevertheless, the standard deviation of deviations between TandemX DEMs and plain surfaces was found to be in the order of several decimetres with RMSE in the order of meters (e.g.
Becek area. This is about 4 times to an order of magnitude higher compared to the uncertainties of surface elevation changes presented in Table 1. I agree that this simple calculation cannot match the results of the detailed analysis considering surface slopes. However, I recommend to present deZ IQS D EHUUHS SHRSQGVFLELNLUY QI VPFHSUDLPUY HTULODULQPT LP QPH QS LP D FQOELPDULQP QI UKH following ways: (i) In Figure S8, the authors present the normalized median absolute deviation (NMAD), which is an error measure in particular for non-normal error distributions (Höhle and Höhle, 2009). If the distribution of the elevation differences between coregistered SRTM and TandemX DEM in stable DSHDT LT PQP$PQSODN# UKLT ODY EH GLTFVTTHG% :I UKH GLTUSLEVULQP LT PQSODN# deZ FDP EH RSHTHPUHG instead of NMAD.
Apart from this point, Sommer and colleagues present a well elaborated study of a first regionwide consistent analysis of glacier surface elevation changes over the European Alps worth to be published in Nature Communications. Figures and tables of the supplementary material complete the clear presentation of the results in the main text.
In the recent version of their manuscript, Sommer and colleagues have revised the uncertainty calculation as suggested. They changed the uncertainty measure presented in Fig. S8 and added a table in the supplement showing the mean standard deviation between the coregistrated DEM. Thus, the presented uncertainties correspond to the applied methods and data used. My only minor comment is that in Tab. 1 almost all uncertainty values are updated except uncertainties in the column presenting the RGI6.0 based mass balance rate. These values should be revised, too. After revision of this single comment, I highly recommend this work for publication.