|The manuscript has improved significantly due to the changes addressing the issues raised by the reviewers and posted comments. |
The main goal of the manuscript is to introduce a new method for deriving multi-mission elevation changes and mass change time series of altimetry data. The new results are used for investigating climate and ice dynamics related changes of the Antarctic Ice Sheets. Part of the methodology is described in the supplementary information and some of the arguments of selecting certain models, eg., planar surface fit vs. higher order polynomial, are also explained in the supplement only. The supplement is well written and it is easy to follow. However, the main manuscript is fragmented and unclear in some places, making it difficult to follow the discussion as it moves from the main paper to the supplement and back. The manuscript should explain all main points of the study indicating where the reader should consult the supplement and references for details. Also, the grammar needs some improvement. For example, the words respective or respectively are used a total of more than 60 times (paper and supplement), in my view, often unnecessarily.
Following recommendations from the review of the original manuscript, the authors included a conversion of the elevation changes to mass changes. The applied method uses an ice density mask defined by a priori knowledge of the spatial distribution of changes due to ice dynamics and surface processes for the conversion. To my best knowledge, this technique has not been used before for long periods with significant changes in ice dynamics. There is ample evidence of changing Antarctic ice dynamics in the literature and in the manuscript itself. The authors describe the technique as “straightforward and robust”, but they don’t provide any assessment of its performance. Considering the changes in ice dynamics and the complex variations of firn-compaction and density during the 40 years spanned by the altimetry record, I expect large errors. These errors, only partially accounted in the error budget, could be responsible for some of the significant discrepancies between the mass balance derived from GRACE and altimetry in the study. Also, the manuscript and the references do not describe the method with sufficient details. Was the density an average density for a given time period? What was this time period and how did the authors account for the different length of the altimetry record? Was the firn-compaction removed? Also, the manuscript several times states that the FDM and SMB time series do not contain long-term changes. This is only partially true. Both FDM and SMB assumes a balance period. The inadequate knowledge or lack of such a balance period could result in an error in the FDM and SMB anomaly trends, but the remaining components of the long-term change will be correctly reconstructed. The cumulative SMB anomalies presented in the manuscript show several examples of these long-term trends (e.g., Fig 13).
Page 7. Figure 3. This figure is only referred after the reference to Figure 4. Either the figures order should be changed, or a reference to this figure should be included.
Page 9, line 8: I assume that all other corrections were performed and not “omitted”
Page 9, lines 8-14: I can only guess the meaning of “These "res" [residuals] represent the anomalies of typically a single satellite pass towards all respective parameters including the linear rate of elevation change.” It needs to be rephrased.
Page 11, lines 27-28. As it has been documented, significant changes occurred in ice dynamic during the 8 years used for the correction (e.g., Flament and Remy, 2012). The density of sampling would probably allow using a quadratic approximation to allow for non-linear dynamics. What was the reason for using the linear approximation and what could be the impact of the modeling error?
3.3.3 Merging different techniques:
A region with a 30 km diameter is large enough to exhibit variations in thickness change rates. Therefore, as the authors have stated, the residuals of the elevation changes within the region could reflect not only measurement errors but real variations. In my opinion, using a formal error propagation through the entire error assessment, i.e., instead of including equation (4) in Supplement C.4 could provide a better error estimate.
Page 12, line 9: What is the meaning of the “flow line of outlet glaciers”? Regions of fast-flowing outlet glaciers? Measurements along the flowlines of fast-flowing outlet glaciers?
Page 12: showing annual rates rather than differences relative to the reference time would make it easier to interpret the changes. Otherwise, the sign of the relative anomalies is different for dates preceding the reference dates and those after the reference data. Also, using differences relative to the reference time results in a scaling of the variation of elevation differences within the averaging area, as mentioned in the manuscript, and makes the interpretation of the error difficult.
Page 12, line 10: I suggest to refer to the area of averaging, rather than to the area of smoothing.
4.1 In situ observations. Briefly describe the measurement type and processing method for both types of observations (e.g., ground GNSS traverse using snowmobile, or airborne laser altimetry). Include the error of the in situ measurements based on literature. Elaborate on the reason of the difference between in situ measurements and new results in the flat interior region.
4.2 Firn model. Include details of the particular FDM use. Did it assume the same steady state period as in Ligtenberg et al., 2011? What was the temporal and spatial resolution? How were the modeled FDM changes interpolated to the altimetry grid? Finally, what is the error of the FDM model?
It is difficult to understand the comparison method applied between the FDM and the altimetry records. A figure, illustrating the approach would be very useful.
Detailed question: Were both time series detrended independently using linear approximations? In page 14, line 24: “RMS of these anomalies from the altimetry data”, while in the figure caption for Fig. 8. b): “RMS of the detrended anomalies of the 1992-2016 altimetry time series”. Do you refer to the same RMS in both places? Is it the RMS of the difference between the detrended FDM and detrended altimetry time series computed for each grid cell? Why is the RMS reflect “the magnitude of seasonal and interannual variations”? I assume that both the detrended FDM and the detrended monthly altimetry data reflect seasonal and interannual variations, so the RMS of their difference would be related to the errors and sampling of the two data sets and any short-term variations in ice dynamics.
Figure 8. Capital letters and drainage basins in a) are not explained and used in this section. The point density needs a color bar.
Results, 5.1 Surface elevation changes
Table S2. Include the projection system used for the X,Y coordinates and latitude, longitude locations. Expressing the multiyear changes as annual change rates would make it easier to interpret the results, e.g., relative elevation at a later date minus relative elevation at an earlier data divided by the number of years would always give negative values for thinning rates and the normalization would allow a direct comparison of the change rates during different time periods.
Page 15, line 19: elaborate on the connection between the stable grounding line position and observed elevation changes.
Figure 9. Since elevation changes are shown as relative values, a “floating” elevation change scale would work better than the current vertical axis. What is the difference between the black and colored line in the elevation change time series?
Page 16, line 3: how much longer are the new time series? Clearly articulate on the new findings that are allowed by the extended time period.
Page 17, lines 8-13: it is somewhat unclear if this paragraph discusses one event (resulted from two accumulation anomalies) or two distinct elevation increases (in line 10: these event – should be either this event or these events). Also, some of the time series show a continuous increase, while others show two periods of rapid surface elevation increase. This is a nice result, and an enlargement of the figure with a more detailed description would be helpful.
Page 17, line 23-30: I don’t understand the computation as described here.
Figure 10. Show the locations of the enlargements in Figure 9 to connect the two parts of the section.
Page 18, line 9: this part should be better connected to the changes presented on the Shirase Glacier earlier in this section.
Page 19, line 2: how much was the “huge” increase? Add a number!
Page 19, line 4: be more specific about the other regions.
5.2 Ice sheet mass time series See general remarks at the beginning of this review.
6.1 Multi-mission SEC time series
Page 23, line 9: use bias instead of offset to describe the difference between change derived from altimetry data and validation data
Page 23, lines 11-13: the >0.5 correlation values do not mean that “both time series agree very well”. It just means that the seasonal and interannual variations of the two time-series correlate well.
Page 23, lines 14-17: the good agreement between the detrended time series cannot establish the absolute accuracy of the SEC, especially not in a level of a few cm/yr.
Page 24, lines 1-2: These sentences appear to be incomplete: “This due to the mountain ranges just north of 72°S, which lead to many losses of lock of the measurements all the way across this part of the ice sheet. The same applies to the measurements at the APIS.”