Antarctic Ice Mass Controversies

Like many others, I was interested in the recent controversy arising from findings of Zwally et al 2015 that there had been ice mass gain gain of ~112±61 Gt/year over 1992-2001 and ~82±25 Gt/year over 2003-2008.  Zwally’s findings obviously contradict a widely held contrary belief, expressed, for example, in IPCC AR5’s assertion there was “high confidence” that the Antarctic Ice Sheet had been losing mass for the prior two decades and that the rate of loss had “likely increased” to ~147±75 GT/year over 2002-2011 or in NASA’s widely cited statement that “the continent of Antarctica has been losing about 134 billion metric tons of ice per year since 2002”.
I had no prior interest in the literature, but was intrigued by the dramatic contrast between Zwally and IPCC on such a widely covered topic.  This quickly led into a voluminous technical literature, which is the subject of today’s post.   The issues were not only about interpretation of satellite data, but quickly led into thorny interpretations of the history of the entire Holocene.
Warning: the following post is very lengthy, but I think that the details are worth paying attention to.
An important and extremely interesting subplot to the Zwally controversy is the remarkable difference between trends in East Antarctica and West Antarctica (plus Antarctica Peninsula).   Most of Antarctica (East Antarctica, which makes up over 80% of the area, in particular) has been gaining ice mass.  On the other hand, there is quite dramatic ice mass loss on the Antarctic Peninsula and localized (but highly publicized) areas of West Antarctica, especially the Pine Island and Thwaites glaciers,  the solid blue areas to the bottom left of Zwally et al 2015 Figure 6b shown below.

Figure 1.  Zwally et al 2015 Figure 6b, showing IceSat dH/dt for 2003-2008.
In my opinion, the most (and only) convincing explanation for this peculiar localization takes one back to the Last Glacial Maximum and changes throughout the Holocene – interesting paleoclimate topics.
Understanding the controversy also leads quickly into complicated problems of glacial isostatic adjustments to gravity surveys (GRACE), as the size of these adjustments turns out to be more or less equivalent to the size of the mass loss itself – hardly a desirable property of the method.  Over the past decade, the size of generally accepted glacial isostatic adjustments has fallen quite dramatically, with estimates of mass loss falling in conjunction.   In retrospect, IPCC AR5 can be seen to have adopted mass loss estimates that were far larger than up-to-date technical literature.
At the end of the day, it is up to specialists to determine whether ice mass gain in East Antarctica exceeds the large ice mass losses in localized areas of West Antarctica or not, but the inconsistency between the two areas is something that ought to interest non-specialist readers.  Behind that topic,  I think that readers ought to wonder what proportion of the large but localized ice mass loss in West Antarctica is a present manifestation of a long-term Holocene pattern, a phenomenon that is attractive to alarmists but one that will continue to take place regardless of climate policy.
IPCC AR4 
AR5 took a far more aggressive line on Antarctic ice mass loss than AR4, which, in retrospect, was rather cautious.   AR4 reported that there had been attempts to measure Antarctic ice mass balance using a variety of techniques: radar and laser altimetry,  the GRACE gravity surveys and input-output balances (using climate models to estimate accumulations and satellite data to measure glacier flow over their grounding lines. In its summary, it was unable to reach a conclusion as to whether the mass balance was positive or negative:

Assessment of the data and techniques suggests overall Antarctic Ice Sheet mass balance ranging from growth of 50 Gt yr-1 to shrinkage of 200 Gt yr-1 from 1993 to 2003.

AR4 was considerably more candid than AR5 about the continuing impact of the end of the last Ice Age.  In a section entitled “4.6.3.2 Ongoing Dynamic Ice Sheet Response to Past Forcing”, AR4 contained an interesting comment that “retreat of the West Antarctic grounding line in response to the end of the last ice age” was estimated to contribute “about 90 Gt yr-1” to ice mass loss:

Because some portions of ice sheets respond only slowly to climate changes (decades to thousands of years or longer), past forcing may be influencing ongoing changes (Box 4.1). Some geologic data support recent and perhaps ongoing antarctic mass loss (e. g., Stone et al., 2003). A comprehensive attempt to discern such long-term trends contributing to recently measured imbalances was made by Huybrechts (2002) and Huybrechts et al. (2004).[1] They found little long-term trend in volume of the Greenland Ice Sheet, but a trend in antarctic shrinkage of about 90 Gt yr-1, primarily because of retreat of the West Antarctic grounding line in response to the end of the last ice age.

This is not an incidental number as it is almost exactly equal to the total Antarctic estimate in AR5.  However, discussion of post LGM long-term effects over the Holocene vanished entirely in the AR5 (chapter 4) discussion of the cryosphere, which used the word “Holocene” only once in passing.
AR4 to 2011
Subsequent to AR4,  interest in Antarctica mass loss increased tremendously, not least because of the availability of new satellite data measuring gravity, altimetry and grounding lines.  The GRACE gravity surveys had become available in mid-2002 and have continued to the present. The high resolution IceSat laser altimetry satellite became available in 2003 and operated until 2009, but, prior to Zwally et al 2015, does not appear to have resulted in a technically published estimate of Antarctic mass loss (though estimates were noted in the Supplementary Information of Shepherd et al 2012, as discussed below).
In contrast, between 2006 and 2011, there were numerous articles estimating Antarctic mass loss using GRACE data. Articles in this period used “early” models for glacial isostatic adjustments (Peltier 2004; Ivins and James 2005), estimates which were dramatically reduced in 2011-12.  Although not evident in most GRACE articles, the size of the glacial isostatic adjustment was the same order of magnitude as the ice mass loss itself (and even larger).  Velicogna and Wahr 2006 (somewhat anomalously) reported both items in their calculation, showing that the “uncorrected” GRACE trend was actually positive and that the estimated mass loss of 152 km3/year (~139 Gt/year) arose from the glacial isostatic adjustment of 192 km3/year (~176 Gt/year):

We subtracted this PGR [post-glacial rebound: the viscoelastic response of the solid Earth to glacial unloading over the past several thousand years] contribution from the GRACE-minus-leakage ice mass estimates … The PGR contribution (192 ± 79 km3/year) is much larger than the uncorrected GRACE trend (39 ± 14 km3/year)…  The best-fitting linear trend, and our final estimate of the decrease in total Antarctic mass between the summers of 2002 and 2005, is 152 ± 80 km3/year

Velicogna and Wahr 2006, almost uniquely, contained a figure (excerpted below) showing the difference between unadjusted and adjusted mass in the figure shown below. Before glaciostatic adjustment, the GRACE data had a slight positive trend (blue circles); after glaciostatic adjustment, there was a dramatic decline (red-orange crosses):
Figure 2. Original Caption: GRACE monthly mass solutions for the Antarctic ice sheet for April 2002 to August 2005. Blue circles show results after removing the hydrology leakage. Red crosses show results after also removing the PGR signal. The latter represent our best estimates of the mass variability. The error bars include only the contributions from uncertainties in the GRACE gravity fields and represent 68.3% confidence intervals (13). Also shown is the linear trend that best fits the red crosses.
Velicogna et al (2009) extended the estimate of mass loss to the period from April 2002 to February 2009, this time emphasizing the “acceleration in mass loss” which was said to have increased dramatically in a short period:

the mass loss increased from 137 Gt/yr in 2002–2003 to 286 Gt/yr in 2007–2009

There were numerous other contemporary estimates using GRACE data, most of which were later quoted in IPCC, but I’ve shown Velicogna’s because she was subsequently an IPCC author and results from Velicogna et al 2009 closely anticipate IPCC AR5’s subsequent “assessment” estimate. They showed a fit using a quadratic, which they claimed to be “statistically significant
at a 99% level.

The F-test show that the improvement obtained with the quadratic fit is statistical significant at a very high confidence level (99%).

In the diagram below (first blink), I’ve plotted the linear fit as well as the quadratic fit.  Using a standard anova test on digitization of the diagram, their confidence claims cannot be replicated.
Velicogna et al 2009 did not show their results prior to GIS adjustment. They stated that their GIA adjustment was “removed… as described by Velicogna and Wahr [2006a, 2006b]” and reported a GIA correction of 176 Gt/yr, somewhat reduced from the 190 Gt/year in their earlier article.   In the diagram below (second blink), I’ve added back the GIA adjustment to reverse engineer the unadjusted GRACE data (green), which, as before, shows a slight increase prior to GIA adjustment.  I’ve also shown a quadratic fit to the data with the (linear) GIA adjustment added back. The R2 of the cumulative mass loss after GIA adjustment arises almost entirely because of the linear GIA adjustment.
Unadjusted mass loss is (somewhat surprisingly) greater in the early part of the data. As a result, a quadratic fit to unadjusted data is slightly concave downwards.  When this slight concavity is combined with the cumulative linear GIA adjustment (which is negative sloped down), one gets a concave downward quadratic, which looks like “accelerated” ice mass loss.  However, this “acceleration” is an artifact of the peculiar method.

There were numerous other contemporary estimates of Antarctic ice mass loss using GRACE data with some even higher results (e.g. Cazenave et al 2009’s estimate of 198 Gt/year for 2003.0-2008.0; Chen et al 2009’s estimate of 178 Gt/year from 2002.3 to 2009.1).
Rignot et al 2008 carried out fresh input-output calculations over much (but not all) of the Antarctic coastline, but including all the locations of rapid ice flow, extrapolating from the observed to the total – a calculation subsequently criticized by Zwally.  They used satellite data on outflow and estimated snowfall using climate models.  They estimated mass loss of 112 ± 91 Gt/yr in 1996, increasing to 196 ± 92 Gt/yr in 2006, with East Antarctica being in balance, with large and increasing mass loss in West Antarctica and the Antarctic Peninsula, with most of the mass loss coming from the highly publicized Thwaites and Pine Island glaciers in West Antarctica.
Prior to 2012, the new laser altimetry data does not appear to have been used to estimate overall Antarctic ice mass loss, though it was used in some local studies.
Mackintosh et al 2011
Mackintosh et al 2011, unfortunately not cited by AR5, was a thorough re-evaluation of evidence on changes in the Antarctic ice sheet through the Holocene, incorporating the considerable accumulation of information on ice sheet recession.  They confirmed the remarkable (in my opinion) scenario of an extremely high West Antarctica in the LGM, with massive reductions of up to 3 km ongoing through the Holocene to the present, but negligible change in East Antarctica – if anything, slight accumulation (see an excerpt from their Figure 5 below).   According to their Figure 5b, ice mass loss in parts of West Antarctica has been up to 3 km over 14 kyr, i.e. an average of 21 cm/year over the 14,000 years, though the changes were heavily front-loaded as shown in their Figure 5a, which shows that nearly all the Antarctic melt since the LGM has occurred in West Antarctica, with contributions continuing almost to the present (at reduced levels).

Figure ^. Excerpt from Mackintosh et al 2011 Figure 5.
 
Zwally et al 2011
In 2011, Zwally and coauthors attempted reconcile the various estimates, with Zwally et al 2011 being, in effect, their opposing answer to the question framed by Bamber in his realclimate article; they described their purpose as:

a critical assessment of estimates of the AIS mass balance in IPCC07, along with more recent estimates as listed in Table 2. Our purposes are to find areas of agreement among the methods, to identify outliers, and to provide a rationale for a narrowed range of estimates and a preferred estimate

First, they discussed three different estimates from radar altimetry estimates:  Wingham et al 2006 had estimated mass gain of +27 Gt/year for the Antarctic ice sheet for 1992-2003, whereas Zwally et al 2005 had estimated a mass loss of 31 Gt for 1992-2001.   Davis et al 2005 had estimated a mass gain of 45 Gt/year in East Antarctica, 29 Gt/year more than Zwally, and, if extended to the AIS, would have been about break-even overall.    Zwally et al attributed the differences to different assumptions on the density of the snow/firn/ice as the altitude changed, noting that the Zwally estimates were based on density of 900 kg/m^3 (the density of ice) whereas the Davis and Wingham estimates were based on density of 350 kg/m^3 (the density of snow).   Bamber argued that Zwally et al 2015 differed from prior studies on this issue, but, in reality, the issue had been clearly articulated in Zwally et al 2011 and, in this article, calculations using the higher density value led to more negative results, undermining Bamber’s line of criticism.
Second, Zwally et al 2011 critically analysed input-out analyses from Rignot and coauthors (including Bamber).  They levelled a number of technical criticisms, of which they singled out extrapolation as an area of largest concern.  Rignot’s observations covered a large fraction of the coast, but not all.  Much of Antarctic glacier discharge occurs in relatively fast-flowing channels; Rignot et al stated that their observations included all fast-moving channels.  Zwally et al argued that such parameters would over-estimate discharge in the non-observed (NOBS) portion in which there were no fast-moving channels. Zwally argued that the differences were substantial, enough to potentially tip the balance from net mass loss to net mass gain, rendering the method inconclusive:

Compared to R08, area scaling reduces the NOBS [not observed]  outflow from 624 Gt/year to 410 Gt/year (columns m and o) and the total outflow (O+) by about 10% from 2191 Gt/year to 1977 Gt/year, which would change the net mass balance from a loss of 136 Gt/year of R08 to a gain of 78 Gt/year…

we believe there is insufficient evidence in the IOM [input-output method] results for either an acceleration or deceleration.

I have no views on the merits of the respective sides of this dispute. AR4 had said that there were very large uncertainties attached to input-output estimates;  Zwally does not appear to have withdrawn his criticisms; one would have hoped that the issue would have been assessed in AR5, but they did not confront it, a topic that I’ll return to below.
Zwally et al itemized a long list of major uncertainties in the GRACE calculations, leading with the GIA adjustments, but including numerous other issues, some of which are also very large:

uncertainties in the GIA correction from model and density errors, uncertainties in the various terms in the spherical-harmonic expression of the Earth’s gravity field, uncertainties in the corrections for variations in the atmospheric mass, errors in the calculations of the scaling factor (”kernel”) to account for the effective size of the measurement region or limited spatial resolution of GRACE satellites, and ”leakage” of signals from mass changes outside the measurement region.

Again, one would have hoped that these points would have also been addressed in AR5.
New GIA Estimates: Whitehouse et al 2012, Ivins et al 2013
New calculations (Ivins et al 2013; Whitehouse et al 2012) dramatically reduced estimates of Antarctic GIA adjustments from previous methods.  The impact was illustrated in the IMBIE SI Figure S8 which showed ICE-5G GIA adjustments of ~170 Gt/year at plausible upper mantle viscosities and Ivins and James 2005 GIA adjustments of ~100 Gt/year at plausible viscosities.   The newer estimates of Ivins et al 2011 ((IJ05_R2) and Whitehouse et al 2012 (W12a) lowered estimates to ~50 Gt/year (!), down 120 (140) Gt/year from assumptions of Velicogna et al 2009 (Velicogna and Wahr 2006).

Figure S8. The optimum IJ05_R2 and W12a model corrections vs. the past IJ05 and ICE-5G corrections employed for GRACE Antarctic ice mass balance analysis. Model GIA calculations of IJ05 and ICE-5G assume two lithospheric thicknesses (h = 120, 65 km) and two upper mantle viscosities; çUM = 1 x 1020 Pa s and çUM = 1 x 1021 Pa s. (See Fig. S6 for the differences in melt history). Error bars for the open circle and diamond are ± 4 Gt/yr, based upon scatter in the filtering methods that were tested in the course of the IMBIE. The blue star is placed at (41) recommended ‘stiff’ Earth structure model, and the red star, near the structure model advocated by (39) based upon global inversions. The two new optimum models are discussed in the text of this SOM. (Adapted from (32)).
Ivins et al 2013 also observed (once again) that the GRACE mass balance without correction was slightly positive over the 2003-2012 period:

They also noted that implementation of the revised GIA adjustments would dramatically reduce prior estimates of Antarctic ice mass loss.
Ivins et al 2013 similarly estimated positive (though considerably lower) unadjusted mass gain over 2002-2013, but its GIA adjustment was less than one-third of the estimate in Velicogna and Wahr, resulting in a dramatically lower estimate of ice mass loss than used in the NASA webpage or IPCC AR5:

GRACE monthly solutions from the Center for Space Research Release 04 (CSR-RL04) release time series from January 2003 to the beginning of January 2012, uncorrected for GIA, yield an ice mass rate of +2.9± 29 Gt/yr. The new GIA correction increases the solved-for ice mass imbalance of Antarctica to −57±34 Gt/yr.

In 2011, Thomas et al  (GRL) attempted to reconcile observed uplifts from then recent GPS information with uplifts predicted from GIS models and reported systemic errors. They concluded that the GIA models were “unreliable” and GRACE estimates of ice mass loss were biased high except for a few regions:

part from a few regions where large ice mass loss is occurring, the spatial pattern of secular ice mass change derived from Gravity Recovery and Climate Experiment (GRACE) data and GIA models may be unreliable, and that several recent secular Antarctic ice mass loss estimates are systematically biased, mainly too high

Whitehouse et al 2012 (Geophys. J. Int.) attempted to reconcile these inconsistencies through estimating paleoclimate glaciation through the Holocene and forward modeling to present uplift.
Implementation of their GIA model in King et al 2012 which dramatically reduced prior GIA estimates and, with them, estimates of ice mass loss from GRACE data:

After applying the model to 99 months of GRACE data spanning Aug 2002 to Dec 2010, we estimate a continent-wide ice mass change of -69±18 Gt yr-1 (+0.19±0.05 mm yr-1 sea-level equivalent). This is 36-48% of the most recently published GRACE estimates2,5 that cover a similar time period but are based on older GIA models. …

They reported that West Antarctic ice loss was limited to basins along the Amundsen Sea (and the Antarctic Peninsula) and that the rest of West Antarctica was more or less in balance, with gains in East Antarctica:

We resolve 26 independent drainage basins and find Antarctic mass loss, and its acceleration, is concentrated in basins along the Amundsen Sea Coast. Outside this region we find West Antarctica is nearly in balance (-10±7 Gt yr-1 in total) and East Antarctica is gaining substantial mass (60±13 Gt yr-1 in total).

The SI to King et al 2012 noted that the W12a GIS model contained serious inconsistencies along the Siple Coast of the Ross Sea (West Antarctica) where there was negligible observed uplift:

the relative sea-level change to be nearly zero (-0.3 mm yr-1) over our GRACE  data  period. This suggests substantial over-prediction of the magnitude of the major GIA uplift centre along this coast in W12a.

They observed that the inconsistency could result from Late Holocene (recent millenia) accumulations and attempted ad hoc patches.
The potential impact of recent (last two millennia) ice sheet advances and retreats is the topic of ongoing analysis by this group.  Nield’s thesis (entitled “The Effect of Late Holocene Ice-Mass Changes on Glacial Isostatic Adjustment in West Antarctica”), partly reported in Nield et al 2014 (Earth Plan Sci Letters), points to the likelihood that explanation of present-day uplift requires knowledge of late Holocene loading (e.g. Little Ice Age) loading.  Nield observed that mantle viscosity under West Antarctica was low, entailing rapid response to recent loading and the requirement to consider recent history for GIS adjustment in West Antarctica and the Antarctic Peninsula:

Conversely in West Antarctica, where mantle viscosity is lower, the Earth responds much more quickly to changes in surface loading so that present-day GIA is likely to be dominated by recent ice-mass changes rather than LGM ice history. Therefore currently unmodelled Late Holocene ice-mass changes are a potential source of large errors in GIA models of West Antarctica.

Similar issues arise in Bradley et al 2015 (another article by the same group) concerning GIA in the Weddell Sea area (which drains some West Antarctic ice), which reported that its observed uplift “rates are noticeably smaller than many current GIA model predictions for Antarctica (Peltier, 2004 and Whitehouse et al., 2012b)”, noting a mean bias of 3.4 mm/year.   Bradley et al considered a variety of ice histories, forward modeling them to estimate present uplift.  Whereas prior models, including Whitehouse et al 2012, had assumed that deglaciation was monotonic and had tapered off some time ago, they considered scenarios which included retreat and re-advance in the late Holocene (a pattern for which there is paleoclimatic support.) They improved model performance “by revising the Late Holocene deglaciation pattern within the Weddell Sea to include an early retreat behind the grounding line defined in the W12 model (at 6 kyr BP) and a relatively long stillstand followed by a short readvance that continues to present day”.  While they did not carry their analysis forward to its impact on GIA adjustments to ice mass loss, they noted:

Finally, the revised Holocene ice-loading history proposed in our study might have important implications for the GIA correction applied to the GRACE data, with a likely reduction in the GIA correction producing a smaller estimate of the present-day ice loss around the Weddell Sea than previously suggested (King et al., 2012).

For comparison to previous GRACE work that used older GIA models, replacing W12a with ICE-5G19 produced total rates of estimated Antarctic mass loss that were greater by 90 Gt yr-1
IMBIE (Shepherd et al 2012)
In preparation for AR5,  there was an attempt by specialists in the field in 2012 purporting to reconcile the wildly discrepant estimates: the Ice Sheet Mass Balance Exercise (IMBIE), reported on in Shepherd et al 2012 (pdf and especially its SI).  Co-authors included Zwally as well as IPCC Chapter 4 coauthors Velicogna and Rignot, and Chapter 4 Coordinating Lead Author David Vaughan.
Their Table S2 summarized their collation of estimates from different methods: note, in particular, the much reduced estimated showed mass loss of 57 ± 50 Gt/year from GRACE data.
Shepherd et al 2012 Table S2 (Antarctic Gt/year)

 
The GRACE estimate of 57 ± 50 Gt/year (2003.8-2009.0) is derived from six analyses of GRACE data commissioned for IMBIE, each carried out by one of six different groups of specialists.  Each of the various groups (including Velicogna’s) appears to have used the more recent (and much reduced) estimates of GIA adjustment, as these values are much lower than (for example) Velicogna et al 2009 or Chen et al 2009 (see above.)   The uncertainty of the above estimate appears to be the standard deviation of the six analyses, each using the more recent (and reduced) GIA adjustments and does not incorporate uncertainty/error from changing GIA adjustments, about which IMBIE noted uncertainty of “up to 130 Gt/year” (though this uncertainty was not carried forward into the mass loss uncertainties.)

the use of GIA models has in practice introduced considerable uncertainty (up to 130 Gt/year) into ice-sheet mass balance estimates derived from satellite gravimetry (31-33).

For the input-output method, they simply applied Rignot’s method to somewhat updated data, reporting an estimated mass loss of 142 Gt/year for a common period of 2003-2008,  negligibly different from the results of Rignot et al 2008 interpolated to this period.   For East Antarctica, they estimated a mass loss of 30 ± 76 Gt/year, whereas estimates of laser altimetry, radar altimetry and even gravity were all positive – especially laser altimetry.  As noted above, Zwally et al 2011 had severely criticized how Rignot and coauthors had extrapolated from observed areas (which included all fast flowing glaciers) to non-observed area (which necessarily had no fast flowing areas.)   Shepherd et al made no attempt whatever to rebut Zwally or otherwise refute his criticism.  I have not been able to figure out how they purported to calculate uncertainties for this method, but it is my strong surmise that there is no allowance whatever for potential issues in respect to extrapolation, since they did not rebut or comment on Zwally et al’s critique or explicitly include Zwally’s alternative calculations.
Although the IceSat satellite had operated from 2003-2009, at the time of the IMBIE intercomparison in 2012, there (remarkably) does not appear to have been any formal publication of IceSat laser altimetry results. Zwally et al 2015 appears to be long overdue in this respect.  For the IMBIE intercomparison, four centers (Louise Sørensen and René Forsberg (SF) at the Technical University of Denmark, Hamish Pritchard (HP) at the British Antarctic Survey, Donghui Yi and Jay Zwally (YZ) at NASA Goddard Space Flight Center, and Benjamin Smith (BS) at the University of Washington) were commissioned to estimate mass change from the IceSat data.  Their estimates of volume change were reported in Shepherd et al Table S7  and mass change in Table S8, both in the SI.  All four centers estimated volume gain in East Antarctica exceeding the volume loss in West Antarctica plus Antarctica Peninsula. The Zwally (YZ) estimates of volume increase was greater than, but similar to, volume changes from other groups, with Sorensen and Forsberg (SF) estimates being particularly similar. Both Zwally (YZ) and Sorensen-Forsberg got similar mass gains for the AIS as a whole, with subsector results by the other two groups being similar as shown in their Table S8, excerpted below.
Shepherd et al Table S8 (Antarctic):  Ice mass change 2003-2008 (Gt/year)

 
Zwally also reported the above IMBIE estimate at the SCAR ISMASS Workshop in Portland, OR on July 14, 2012 (added to NASA’s Technical Reports server on September 7th, 2012).  Zwally’s workshop presentation was noticed by Anthony Watts, who drew attention to it here, pointing out, unsurprisingly, the inconsistency with GRACE estimates of massive ice loss.   Part of Zwally’s presentation is online at WUWT.  In this presentation, he said that it had been his expectation that IceSat laser altimetry data would have put an end to the controversy (“we would launch IceSat and that would be it”, but acknowledged that it hadn’t, “at least some large negative values [are] going away”.
Zwally’s IMBIE estimate in (49 Gt/year mass gain for 2003-2008) is somewhat less than the later estimate of Zwally et al 2015 (82 Gt/year for 2003-2008), but both show mass gain.  In addition, though Bamber’s realclimate article attempted to isolate Zwally et al as a unique outlier, both the Sorensen-Forsberg group and Benjamin Smith (supplemented with a plausible Antarctic Peninsula estimate) also show mass gain, though less than Zwally.  So there appears to be a larger issue with the discrepancy between laser altimetry and gravity survey results that cannot be blamed on Zwally being an outlier.
This survey still left a very large range between the different methods, ranging from estimated mass loss of 142 ±  86 Gt/year from input-output calculations  (with only a single representative with unclear uncertainty calculations) to mass gain of 21 ± 71 Gt/year from laser altimetry.   Shepherd et al then appear to have simply averaged these wildly discrepant results, but do not report let alone show any attempt at actual reconciliation of the differences.  (The term “reconciliation” is an accounting term and requires that the differences be explained.)    This papered-over average is carried forward into Table 1 of the main article, where the mass balance estimates are evidently derived from the average of Table S2 (over varying time periods.)

In addition, although the differing methods revealed a very wide range of estimated mass change (both gain and loss), Shepherd et al showed a lower uncertainty (±  43 Gt/year) for a simple average than for any of the individual methods.   This seems highly unjustifiable to me.
In addition, their methodology gave disproportionate weighting to the single Rignot-style input-output calculation, which showed much larger mass loss than the other two methods. If Zwally’s alternative calculation had been given equal consideration, the average would have been close to zero, with an enormous uncertainty.  The resulting average of laser altimetry and gravity methods would have yielded a much lower mass loss of ~20 Gt/year over 2003-2008 with wider uncertainty.  So there is an enormous effective weight on the very uncertain and problematic IO calculation in the IMBIE composite.
While Shepherd et al entitled their IMBIE report as “A Reconciled Estimate of Ice-Sheet Mass Balance” and used the term “reconciled” repeatedly in the article,  the title, in my opinion, is a misrepresentation, since the article did not reconcile the discrepancies, but merely collated results of inconsistent approaches without confronting the discrepancies.
IPCC AR5
Despite Zwally’s optimistic belief that new data had resulted in “at least some large negative values going away”, IPCC AR5 did nothing of the sort, reviving some of the estimates that Zwally had believed to be discredited.  Two advocates of high negative values were AR5 Lead Author and Contributing Author: respectively, Eric Rignot, also a coauthor of the Copenhagen Diagnosis, and Isabel Velicogna, who was prominently featured in the SKS Denial series.
For the period 2005-2010, IPCC estimated mass loss  of 147 ± 74 Gt/year, almost double the corresponding IMBIE estimate of 81 ± 37 Gt/year:

Overall, there is high confidence that the Antarctic Ice Sheet is currently losing mass. The average ice mass change to Antarctica from the present assessment has been –97 [–135 to –58] Gt yr–1 … over the period 1993–2010, and –147 [–221 to –74] Gt yr–1… over the period 2005–2010. . …

In the Chapter 4 Executive Summary, IPCC stated:

The average rate of ice loss from Antarctica likely increased from 30 [-37 to 97] Gt yr-1… over the period 1992-2001, to 147 [72 to 221] Gt yr-1 over the period 2002-2011 … [4.4.3, Figures 4.16, 4.17]

Even though the IMBIE study had included leading authors in the field, including IPCC coauthors, and had been completed in time for the IPCC cycle (and was referred to by IPCC), IPCC AR5 presented its own calculations (described as the “present assessment”) instead.  The methodology of IPCC’s calculations was nowhere explained, but appears to be a combination of Rignot and Velicogna’s own prior results.
In their Appendix 4.A, they listed ten articles as sources, but do not describe how they went from the articles to the values presented in Figure 4-16 and the running text.  Eight of the ten listed studies were based on GRACE gravity surveys (and thus limited to mid-2002 on.) The average of the eight gravity survey studies in IPCC Appendix 4A is 144 Gt/year  (for varying coverage of 2002-2011), 2.5 times higher than the IMBIE gravity survey average of 57 Gt/year for 2003-2008.  All but one of the GRACE studies used older GIA adjustments and thus had much higher values than in IMBIE.  One of the GRACE studies was Velicogna et al 2009, results of which were stated in Appendix 4.A as having been “extended to 2012”.
For Antarctica, they asserted instead that there was “no significant difference” between their results and IMBIE results:

The recent IMBIE intercomparison (Shepherd et al., 2012) for Antarctica, where the GIA signal is less well known than in Greenland, used two new GIA models (an updated version of, Ivins and James, 2005; for details see, Shepherd et al., 2012; Whitehouse et al., 2012). These new models had the effect of reducing the estimates of East Antarctic ice mass loss from GRACE data, compared with some previous estimates.  For Antarctica, Shepherd et al. (2012) estimate an average change in mass for 1992–2011 of –71 ± 53 Gt yr–1… For the same period this present assessment estimates a loss of 88 ± 35 Gt yr–1 at the 90% confidence level. Averaging across technique ensembles in the present assessment, rather than individual estimates, yields no significant difference.

This claim was completely untrue. As noted above, the IPCC estimate of mass loss in the period 2005-2010 was nearly double the IMBIE estimate for the corresponding period. And for the GRACE data separately, the IPCC estimate was more than double the IMBIE GRACE estimate.  In fact, it would be more accurate to say that there was “no significant difference” with the Velicogna et al 2009 estimates using its obsolete and too high GIA adjustments.  This is shown in the annotated version of IPCC Figure 4.16 below, in which the IMBIE estimate (red) clearly diverges from the IPCC estimate, while the Velicogna et al 2009 estimate of ice mass loss over 2002-2009 closely matches the IPCC estimate.

Ironically, Velicogna appears to have submitted a lower estimate of ice mass loss to IMBIE, but IPCC appears to have reverted to the higher earlier estimate.  Velicogna does not appear to have ever published the extension as used in IPCC. Instead, the subsequent Velicogna et al 2014,  using recent GIA adjustments, reported much lower mass estimates than used in AR5. AR5 did concede that the size of the GIA adjustment was “similar in magnitude” to the estimate of ice mass loss and had very large (±80 Gt yr-1) uncertainty, but, as elsewhere, the GIA uncertainty was not carried forward into the uncertainty of the ice mass loss:

In Antarctica, the GIA signal is similar in magnitude to the ice-loss signal, with an uncertainty of ±80 Gt yr-1 (Velicogna and Wahr, 2006b; Riva et al., 2009; Velicogna, 2009).

Appendix 4.A listed one laser altimetry study as being considered in the IPCC calculation, but did not use Zwally’s IMBIE results which showed a slight mass gain. Instead, they used results from a Chinese language article (Shi et al., 2011), about which they had commented: “Methodology and error budget incompletely described.”
In addition, none of the IPCC numbers for mass loss over named periods reconcile precisely to the information in Figure 4.16, as illustrated here.
IPCC discussed IMBIE in both in its section on Greenland and in its section on Antarctica.  In their section on Greenland, they stated that there was “good agreement” between the different methods and the combination of methods “considerably improves the overall uncertainty”:

A reconciliation of apparent disparities between the different satellite methods was made by the Ice-sheet Mass Balance Intercomparison Experiment (IMBIE) (Shepherd et al., 2012). This intercomparison combined an ensemble of satellite altimetry, interferometry, airborne radio-echo sounding and airborne gravimetry data, and regional atmospheric climate model output products, for common geographical regions and for common time intervals. Good agreement was obtained between the estimates from the different methods and, while the uncertainties of any method are sometimes large, the combination of methods considerably improves the overall certainty.

While there may have been “good agreement” between methods for Greenland (I haven’t parsed these results), this was obviously not true for Antarctica.     While IPCC did not directly say that there was “good agreement” for Antarctica,  if “good agreement” was worth reporting for Greenland, then surely disagreement for Antarctica ought to have been reported, but it wasn’t.
IPCC Lead Author Eric Rignot was also an author of the Copenhagen Diagnosis, which, unsurprisingly reported the most lurid estimates of ice mass loss, without mentioning that specialists had begun the process of walking back these estimates.
Velicogna et al 2014
In her newest analysis, IPCC Contributing Author Velicogna dramatically reduced the estimated mass loss from the earlier Velicogna et al 2009 and AR5 estimates, but neglected to even mention the change.
The figure below shows their estimate of cumulative Antarctic mass loss (blue; smoothed as green), for which they diagnosed “acceleration”, which they discuss at length.  I’ve overprinted Velicogna et al (red) and IPCC AR5 (magenta), both displaced to line up in mid-2002.  The revised estimate of mass loss, as estimated in Velicogna et al 2014, is a fraction of the estimated mass loss published in AR5 the previous year.  One presumes that the Velicogna submission to IMBIE in 2012 was along the lines of Velicogna et al 2014, but, for some reason, Velicogna and IPCC coauthors reported much larger estimates of ice mass loss as the IPCC “present assessment”.
In this figure, I’ve also overprinted my estimates of the pre-adjusted GRACE measurements (black) – Velicogna et al 2014 did not report the quantity, but said that they used the adjustment of Ivins et al 2013, which, by difference, appears to be 60 Gt/year, which is added back as above.  The unadjusted mass increases somewhat in the first part of the period and decreases slightly in the second part, giving a very weak concavity.

Figure ^.  Annotation of Velicogna et al 2014 Figure 3a, showing reported ice mass loss after GIA adjustment (blue; smoothed in green); ice mass loss (smoothed) of Velicogna et al 2009 (red) and IPCC AR5 (magenta). Black – estimate of pre-GIA adjustment, showing linear and quadratic fit.
Zwally et al 2015
Although the IceSat laser altimetry data had provided unprecedented detail on height of the Antarctic ice sheet from 2003 to 2009, Zwally et al 2015 was the first fully published journal presentation of an estimate of Antarctic ice mass loss using this data.  (As noted above, it had been used for IMBIE contributions, but these only showed final results and otherwise had been used in some detailed glacier analyses, but not for a fully published Antarctic estimate.)  By the time that it was presented, it was criticized for being out of date – raising an obvious question of why there was such a delay between availability of data and its technical presentation.
Zwally et al 2015 contained a somewhat higher estimate of mass gain of 82 Gt/year, with losses of 29 Gt/year in the Antarctic Peninsula and 25 Gt/year in West Antarctica being offset by a gain of 136 Gt/year in East Antarctica. Zwally drew attention to an interesting aspect of West Antarctic mass change: while the mass loss in the Pine Island and Thwaites glacier area had attracted wide attention,  this was substantially offset by mass accumulation in the Kamb Ice Stream sector of West Antarctic, where slowdown of the ice stream had led to substantial accumulation (observed in the increase in height of the ice sheet in the Kamb sector).
In his 2012 workshop (and IMBIE contribution), Zwally et al had estimated slightly lower mass gain during the IceSat period of 2003-2008 of 86 Gt/year for AIS excluding the Peninsula (versus 111 Gt/year in the subsequent Zwally et al 2015), which was said to be similar to his revised results for 1992-2001 from radar altimetry.  This was a change of sign from the previously estimated loss of 31 Gt/year for 1992-2001 (Zwally et 2005, re-iterated in Zwally and Giovetti 2011).
The largest contribution to changes from the ice mass loss estimates of Zwally et al 2005 and Zwally and Giovetti 2011 came from implementation of the newer glacial isostatic adjustments, which also impact the estimates from altimetry, though the impact on altimetry estimates is less than 20% of the impact on gravity surveys, as shown in Zwally’s Table 8, an excerpt of which is shown below:

Zwally’s Table 8 contained an instructive reconciliation of the impact of changing GIA adjustments on his calculations. Zwally et al 2005 (and 2011) had used an exceptionally high GIA adjustment, even higher than other contemporary estimates.  This had the effect of disguising the difference between Zwally’s results and other results.  Had Zwally et al 2005 simply excluded the Huybrechts (2002) GIA adjustment from his composite, he would have shown a mass increase of 16 Gt/year, rather than a mass loss of 31 Gt/year (as the GIA adjustment would have been 16.5 Gt/year, rather than 62 Gt/year.)  Zwally et al 2015 used the even lower Ivins et al 2013 GIA estimate, which, had it been used in the earlier articles, would have led to a somewhat higher estimate of mass gain.

 
Zwally converted changes in altitude of the ice sheet to mass through a model of changes from snow to firn.  Through information from ice cores, it is well known that density increases downhole, with snow being transformed to ice as it is subject to increasing pressure.  Zwally has being doing this sort of calculation for many years and his procedures do not appear to have been seriously challenged by other specialists.
Zwally observed that nearly all of the difference between his results and GRACE results occurs in East Antarctica. He noted that GIA model improvements had led to positive GRACE values in East Antarctica in the most recent studies, though still less than his East Antarctica estimate of 136 Gt/year:

GIA model improvements have been a primary cause of AIS mass estimates from GRACE becoming less negative, and those for EA becoming more positive. Recent GRACE estimates of dM/dt for EA are +35 Gt a-1 (Shepherd and others, 2012), +60±12.8 Gt a-1 (King and others, 2012) and 62.8±8.1 Gt a-1 (Luthcke and others, 2013). Additional convergence may occur as the full implications of long-term growth of EA are taken into account in the ice loading history.

Zwally observed that it would take only a small change (1.6 mm/year) in the estimate of East Antarctic GIA adjustment to reconcile the remaining difference:

a change in the modeled dB/dt for EA downward by 1.6 mm a-1 (from the small average uplift of +0.4 mm a-1 (Table 2) to -1.2 mm a-1) would bring the two GRACE-based dM/dt estimates of 60±12.8 Gt a-1 and 62.8±8.1 Gt a-1 in line with a corresponding adjustment of our ICESat value of 136±8 Gt a-1 at ~150 Gt a-1 for both methods.

Zwally also observed that present knowledge of ice unloading history supported the possibility of further revision of East Antarctic GIA adjustments in the indicated direction.  Zwally posited that there had been ice thickening in East Antarctica over the past 10,000 years which would result in negative, rather than positive, GIA in East Antarctica:

The small modeled uplift of +0.4 mm a-1 averaged over EA implies a history of ice unloading used in the GIA model. The ice-loading histories typically treated in the models are episodic ice unloadings during a glacial-interglacial transition, which cause relatively rapid isostatic adjustments, followed by a much slower residual response rate decaying over thousands of years. In contrast, our finding of a stable dynamic thickening of 1.59 cm a-1 over EA (Table 2), along with our interpretation as long-term dynamic thickening for 10 ka since the early Holocene, implies a slow long-term ice loading resulting in the addition of 159 m of ice averaged over EA in 10 ka. Using 6 for the ratio of the density of mantle rock to the density of ice implies an additional bedrock depression of 27 m continuing at a rate of -2.65 mm a-1 assuming full long-term isostatic adjustment. Therefore, the -1.6mm a-1 needed to bring the gravity and altimetry dM/dt estimates into agreement is only 60% of the full isostatic adjustment rate, and therefore within the range of what can be expected if the ice loading implied by the long-term dynamic thickening is accounted for in GIA models.

Zwally re-iterated their previous critique of Rignot’s input-output calculations, taking particular aim at his East Antarctic results, where Rignot had estimated a slight mass loss. Zwally once again observed that this estimate depended on a disproportionate extrapolation of ice mass loss to non-observed areas and that equally (or more plausible) extrapolations led to I/O estimates of East Antarctic mass gain consistent with his results:

The estimate in Rignot and others (2008) for EA was near-zero at -4.61 Gt a-1, which was based on an input estimate of 1131 Gt a-1, an observed output of 784 Gt a-1 and an extrapolated non-observed output of 349 Gt a-1. Therefore, a disproportionate 31% of Rignot and others’ total output from EA came from only 15% of the area that was non-observed. If the non-observed output had been scaled in proportion to the area as stated, then the non-observed output would be only 173 Gt a-1 (i. e. 15% from 15% of the area) and the net IOM balance would be +174 Gt a-1. Or if the non-observed slower-moving areas were only 70% as effective at discharging ice as the faster-moving observed areas, then the modified non-observed output would be +121 Gt a-1 and the net IOM balance estimate for EA would be +226 Gt a-1 (Zwally and Giovinetto, 2011).

As a reader who is familiar with reconciliation procedures in other fields, I was favorably impressed by Zwally’s effort to actually address and reconcile discrepant results,  an effort that seems far too unique within the field. On the other hand, the very long delay between availability of IceSat results (2003-2008) and Zwally’s presentation of results to IMBIE in 2012 and their eventual publication in late 2015 –  far longer than the usual cycle of publication of satellite data e.g. the numerous GRACE articles – is disquieting. Zwally was clearly worried that his results would “give fodder to the skeptics” and/or “dilute the message” (to borrow terminology from Mann’s correspondence about hiding the decline), worrying to Nature that “some of the climate deniers will jump on this”.  One hopes that this concern did not contribute to the overlong publication delay, but one fears that it might have.
Blog Reaction to Zwally et al 2015
Anthony Watts alertly noticed NASA’s subdued announcement of Zwally et al 2015 at WUWT here and, entirely reasonably, pointed back to his earlier cover of Zwally’s 2012 workshop presentation.
Predictably, Zwally’s results were repudiated by warmists.  Gavin Schmidt told VICE news that he would “pin more weight” on gravimetric data:

“data from a pair of satellites called GRACE, which measure gravity, actually points towards a net loss of ice on the Antarctic continent in more recent years. Schmidt said that there are two methods for measuring the mass of an ice sheet. The first measures gravity, and the second measures the elevation of the top of the ice sheet. Both methods need to take different variables into account to be accurate. The method used in this most recent study measured the ice sheet’s elevation, and the most recent time period it considered ended in 2008. “I would pin more weight to the GRACE data than to this latest paper,” Schmidt told VICE News

Since the impact of problematic GIA adjustments is nearly six times larger on GRACE estimates than altimetry estimates, it is not clear why Schmidt believed that it was appropriate to “pin more weight” on GRACE estimates, other than he liked the answer better.
Michael Mann, whose own comparisons of observations to models notoriously ended in 2005 (see ^), told the Guardian that  Zwally’s results were already obsolete and did not deal with the “acceleration” observed since 2008 (on this issue, see above analysis on the “acceleration” ):

According to climatologist Michael Mann, who was not involved in either study, the use of older satellite data could be the cause for the disconnect.  “It sounds to me as if the key issue here is that the claims are based on seven-year-old data, and so cannot address the finding that Antarctic ice loss has accelerated in more recent years,” he told AFP.

While one can regret Zwally’s failure to publish his laser altimetry results more promptly,  Mann’s comment is merely a talking point and does not refute the large difference between Zwally and IPCC results in the overlapping period.
SKS (e.g. here) has long asserted that claims that Antarctica was gaining ice as a skeptic “myth”.  They conceded that there had been gains in sea ice, but argued that satellite data showed that Antarctica is “losing land ice at an accelerating rate”, and that it was “cherry picking” to focus on one index, rather than the other. This argument was reiterated in their 2015 online lecture.  They stated that “skeptic arguments that Antarctica is gaining ice frequently hinge on an error of omission, namely ignoring the difference between land ice and sea ice [SKS bold].”  They have posted up a caveat saying that “that until 2008 there might have been a bigger increase in ice on East Antarctica than there is a decrease in the west, meaning that total Antarctic land ice is increasing”, adding that “the study disagrees with many other techniques”. They promise to return to the issue when further technical analyses become available, in the mean time pointing to one-sided comments at other warmist blogs.
Media Matters focused on Zwally’s concern that his results would (in Mann’s phraseology) give “fodder to the skeptics” and headlined its article: “NASA Scientist Warned Deniers Would Distort His Antarctic Ice Study — That’s Exactly What They Did”.  While Anthony Watts had certainly taken satisfaction in Zwally’s results,  Media Matters did not identify any actual distortion: it was Zwally’s results that undermined the narrative, not the supposed distortion of the results.
Like Mann, Phil Plait criticized criticized the study for ending in 2008 (when IceSat failed) and claimed that we “know” that Antarctica is losing “more than 130 billion tons of ice per year” and that mass loss “accelerated” after 2008 :

here’s more. They looked at data going from 1992–2008. Starting right around that time, mass loss due to melting ice in Antarctica (mostly in the west) has accelerated. It’s actually been speeding up for some time, but in recent years it’s really kicked in. Every year, about 6 billion more tons of ice are lost than the year before. In the past two decades, the loss rate has doubled. This is enough to easily outpace the mass gained by snowfall over East Antarctica. Using data taken by the Grace satellites (which measure how mass underneath them changes over time), we know that overall, Antarctica is currently losing more than 130 billion tons of ice per year, and again, that number is increasing every year. Since 2002 it’s lost about 2 trillion tons of ice.

As noted above, GRACE proponents have dramatically walked back their earlier estimates of ice mass loss, though these obsolete high estimates continue in use by warmists.  We don’t “know” that Antarctica is losing more than 130 Gt/year; on the contrary, current GRACE specialist estimates are much lower than that and, as Zwally observed, only slight changes in GIA adjustments would make even GRACE estimates positive.
Plait also fretted that the conversion of height into mass was “a little bit problematic” and a “little worrisome”:

There’s one step in the new study I found interesting, and a little worrisome: They measure how the height of snow increased in East Antarctica, but it’s a little bit problematic converting that to how much mass of ice was gained. As the snow falls, it gets compressed over time and turns into ice. That process isn’t completely understood, as Robin Bell (who studies Antarctic ice) points out in a quote in an article about this on Vice

While the process isn’t “completely understood”,  there are dozens of ice cores in which density has been measured downhole, so the process is well understood.  There is no evidence that the uncertainties in this process are anywhere close to the uncertainties in GIA adjustment, where specialists have already made dramatic changes to the adjustments.
Greg Laden agreed with Mann’s talking point about the 2008 ending:

The fact that the study being reported uses older data could explain why it conflicts with everything else the science is telling us. Michael Mann, quoted in The Guardian, notes, “…the claims are based on seven-year-old data, and so cannot address the finding that Antarctic ice loss has accelerated in more recent years.”

Laden also introduced a concern that the laser “observations are wrong”.  Laden pointed out that the satellite stopped working in 2008 and posited that there might have an instrumentation problem prior to that time:

If the satellite method is just a little off, this could cause a problem. (By the way, the data end in 2008 because the instrumentation on the satellite stopped working then.) This study’s main contribution may, in the end, to be to point out a problem with the instrumentation prior to that time.

While concerns over the possibility of instrumentation error must be considered, no specialist in the field has challenged Zwally on this point. Laden also worried that errors in calculations of the “response of the bedrock” underneath the ice sheet could also cause problems:

The second possible reason seems more likely. Part of the process of determining that Antarctica has a positive mass balance (more ice over time rather than less) involves assumptions (and some measurements) about the response of the bedrock underneath the very thick ice sheets. If that is wrong, then that is a problem.

Concern about glacial isostatic response is obviously a legitimate concern, but Laden was obviously incorrect in using this as a stick against Zwally: as noted above, the impact of GIA adjustments is six times higher on GRACE gravity estimates than on altimetry estimates.  To the extent that Laden is worried about this factor, he should “pin” that much less weight on the GRACE estimates which he likes.
Laden and other readers even wondered why Zwally would publish an article showing ice mass gain in Antarctica and how it passed peer review. Greg Laden speculated on its present appearance as follows:

Sometimes to close out a grant you need to submit a promised peer review publication. This looks to me like a paper that might have been hanging around a while.

Clearly, the paper had been “hanging around a while” – the reasons for this remain unexplained.
Sou purported to shrug off the difference as an uninteresting dispute among scientists:

So one group of scientists find that ice has been on balance increasing, while others find that ice has been on balance decreasing.

Had IPCC taken the same position, the topic would be of little interest.
Bamber at Realclimate
The most coherent warmist response came from Jonathan Bamber, an opposing specialist, who, writing at realclimate here, posed the question as follows:

why does it contradict a plethora of previous studies that suggest Antarctica has been losing mass over the same time period?

Unfortunately, Bamber did not rebut or even analyze Zwally’s thoughtful and detailed commentary on GRACE gravity estimates or Rignot’s input-output estimates, a commentary which provides a plausible explanation of how Zwally arrived at different answers than others.
Bamber instead challenged Zwally’s conversion of altimetry data on height to estimation of mass. Bamber pointed out that “small biases in the laser and radar data can have a big effect when the signal is only on the order of 1 cm/year”.  This is true enough, but the same criticism is even more valid against highly negative GRACE gravity estimates, on which Bamber had been silent.
As support for this criticism, Bamber pointed out that Zwally’s 2012 IMBIE calculations from the same satellite data were “only about half” of his most recent estimates.  Bamber challenged whether the new calculations were an improvement, editorializing that it “is probably more accurate to say that it is just different, rather than better”.   Bamber didn’t mention that GRACE estimates had changed far more dramatically – as shown above for the change between Velicogna et al 2009 and Velicogna et al 2014. Further, the change in East Antarctica, the main issue, was only ~22%.  Because the overall mass balance is an offset between East Antarctic mass gain and West Antarctic/Peninsula mass loss, the percentage change on the overall balance looks larger than the change in the component.
Bamber also observed that the laser altimetry data used in Zwally et al 2015 came from the IceSat satellite, which had a life span of 2003-2009 and that results from the most recent satellite equipped with laser altimetry (Cryosat 2) “may yield a different picture yet again”,  noting that intercomparisons thus far are “not very encouraging”.
Bamber also stated that there is “another way to measure the changing mass of the ice sheet and that is to weigh it using the Gravity Recovery and Climate Experiment (GRACE) twin satellites”.  As an editorial comment, the term “weigh” in this context is a huge exaggeration.  The GRACE satellites measure gravity as they pass over.  The estimation of the “weight” of the ice sheet results from a very complicated by-difference calculation that appears to have just as many (if not more) problems and assumptions as the competing altimetry calculations.  Bamber conceded that the satellite cannot distinguish changes in gravity due to changes in ice thickness from changes in gravity due to relative change in height of continental mass arising from “changes in ice loading over thousands of years”. Bamber conceded that “some critical variables” are not well known and that there was “very limited information” for the (very large) East Antarctica interior:

Geophysicists have attempted to model how the solid Earth responds to changes in ice loading and, although the physics is fairly well known, some critical variables such as the viscosity of the lower mantle and the history of ice load changes are not…in the interior of East Antarctica we have very limited information on how big the bedrock signal is.

Bamber then said that Zwally had “argue[d]” that GRACE GIA adjustments in East Antarctica were “wrong because they do not know the full ice loading history”:

Zwally argues that the GIA estimates that have been used for correcting GRACE data are wrong in East Antarctica because they do not know the full ice loading history over the last ~22,000 years.  Maybe. But GIA in the EAIS is poorly constrained and there is, in general, poor agreement among different estimates of GIA over the EAIS. [my bold]

This summary is both tendentious and incoherent.  GIA specialists (Ivins, Peltier, Whitehouse and others) have themselves taken the position that their earlier GIA estimates were incorrect in light of improved knowledge and additional data.  Zwally et al 2015 used up-to-date GIA estimates from Ivins et al 2013 – estimates that have also been widely used in more recent GRACE articles (e.g. Velicogna et al 2014).
Further, to the extent that Bamber is correct that there is “poor agreement among different estimates of GIA” over East Antarctica, this is a much larger problem for the GRACE gravity estimates, which are approximately six times more sensitive (as Gt/year relative to mm/year) than altimetry estimates.  If this is an issue for Bamber, he ought not to have been silent on the GRACE studies.
Bamber concluded that the difference between Zwally and other studies ultimately rested in Zwally making a “different set of assumptions” about East Antarctica and that these assumptions were “by their nature, subjective and difficult, without additional evidence, to corroborate”:

So what is really happening? One thing that Zwally’s study does highlight is how difficult it is to nail what is happening in East Antarctica because the signal is small and contaminated by unwanted effects that are as large or even larger. Zwally et al get a different result from previous studies because they make a different set of assumptions. Those assumptions are, by their nature, subjective and difficult, without additional evidence, to corroborate.

This is hardly an argument for disregarding Zwally’s conclusions.  The problem of the East Antarctic “signal” being “contaminated by unwanted effects that are as large or even larger” as the “signal” itself is much larger for the GRACE survey data than for Zwally’s laser altimetry data.  Bamber ought to have spoken out earlier.
Despite seeming to concede that it was impossible to get past a “subjective” interpretation of the Antarctic  satellite data, Bamber himself remained undeterred in his conviction of Antarctic mass loss, falling back on the difficulty of explaining observed sea level rise without an Antarctic contribution:

There are, however, other lines of evidence that suggest that Antarctica is unlikely to have been gaining mass in the last few decades. That would, for example, make closing the sea level budget a whole lot harder (that is, making the sum of the sinks and sources match the observed rate of sea level rise)

However, this is a very indirect argument. IPCC Table 13.1 attributed only 0.27 of 3.2 mm/year sea level rise from 1993-2010 to the Antarctic ice sheet (equivalent to about 90 Gt/year), rating multiple other factors (thermal expansion, non-icesheet glaciers, Greenland ice sheet, changes in water storage on land) as greater contributors.  Uncertainties for thermal expansion and glaciers were each greater than the contribution attributed to Antarctica. In addition, a recent study argued that groundwater extraction could have contributed up 0.8 mm/year, far higher than values contemplated by IPCC.  While Bamber is correct that mass gain in Antarctica would make closure of the sea level budget more difficult, there is clearly considerable play in the numbers.
Nor was this line of argument relied upon in the IPCC chapter on the cryosphere in which the IPCC reported “high confidence” in its estimates of large Antarctic mass loss.
Conclusions
While it is obviously up to specialists to try to ultimately figure out whether Antarctic ice mass was increasing in the periods 1992-2001 and 2003-2008 (per Zwally) or whether it was decreasing (as IPCC and others had previously asserted),  there does not appear to be any objective basis by which, for example, Gavin Schmidt could reasonably “pin more weight” to highly negative estimates from GRACE gravity data than to Zwally’s positive estimates from laser altimetry.
The size of the GIA adjustment for GRACE gravity estimates is the same order of magnitude as the estimate of ice mass loss and, in many cases, is larger. These GIA adjustments have been dramatically reduced by specialists over the past decade and have concurrently reduced estimates of ice mass loss.
Many popular (warmist) discussions of Antarctic ice mass loss continue to use obsolete (overly high) estimates of ice mass loss e.g. NASA’s estimate of “134 billion tons” per year.   Such estimates rely on GRACE estimates using obsolete GIA adjustments.
The estimates of mass loss in IPCC AR5 were highly questionable. They were much higher (nearly double) than contemporary specialist (IMBIE) estimates.  They appear to have been based on studies using GIA adjustments, already known to be obsolete.  It was separately highly questionable to attribute “high confidence” (and relatively narrow confidence intervals) to these very high estimates of mass loss.
Most of the Antarctic continent (especially East Antarctica) appears to be experiencing ice mass gain, with ice mass loss being localized to less than 5% of the continent:  parts of the Antarctic Peninsula and est Antarctica (especially Pine Island and Thwaites glaciers).  This peculiar localization requires its own explanation. Recent specialist literature has concluded that West Antarctica was up to 3 km higher in the LGM, while the height of East Antarctica has changed little and might even have increased slightly through the Holocene. West Antarctica has experienced dramatic ice mass loss through the Holocene, attenuating to the present.
AR4 had pointed out the possibility that localized ice mass loss in Antarctica was continued Holocene ice mass loss.   This possibility vanished in AR5 without discussion. In an inline comment to Bamber’s realclimate article, Eric Steig said that his opinion, and that of “50% of experts”, was that the connection of Antarctic glacier retreat to “anthropogenic global warming” was “weak” and that the localization of the glacier retreat to West Antarctica was “well understood” and something that he had written about “extensively”:

I think the evidence that the current retreat of Antarctic glaciers is owing to anthropogenic global warming is weak. The literature is mixed on this, about 50% of experts agree with me on this. So you’ll get no argument from me there.  Second, the localization in West Antarctica is well understood, and I’ve written about it extensively.

Elsewhere, Steig has attributed the West Antarctic glacier retreat to erosion of the grounding line of the glaciers by relatively warm Circumpolar Water, rather than to very slight warming of air temperatures above West Antarctica.  Given the continuous retreat of West Antarctic grounding lines over the Holocene, it seems implausible to attribute present grounding line erosion to a different cause than past grounding line erosion that has taken place over the Holocene.  Steig’s position on this point seems entirely reasonable.
However, it still seems like one of those too typical situations where the less alarming explanation is presented in specialist literature, but left unmentioned or unconfronted when retreat of West Antarctic glaciers is presented as a cause of alarm.

Source