|Review for Haubner et al., TCD|
The paper submitted provides an evaluation of thickness and velocity evolution for Upernavik Isstrøm for two sets of simulations: (1) driven entirely by SMB; and (2) driven by SMB and prescribed changes in calving front location, thus circumventing the need for a calving law. The authors find that the present day velocities are generally replicated by the end of the simulation driven primarily by calving front change. If valid (see below), this is impressive given that the transient part of the model simulation time is ~160yr.
Having read through the previous version of the manuscript and sets of reviewer comments, the paper is greatly improved in terms of clarity regarding model initialisation. However, I have a few concerns regarding the interpretation of some of the model/observed comparison measures, in addition to a couple of questions for the authors that could be addressed by two/three more model simulations and allow for a better evaluation of their conclusions.
Prescription of calving in the model
Rather than implement a calving law, the authors prescribe terminus change based on their observational record which spans from 1849-2012. Their prescription of calving assumes that terminus change from the previous observation occurs either instantaneously or (for larger calving events) via staged retreat between observations evenly spread through time. Given that one of the author’s main conclusions is that terminus change strongly impacts dynamics, for me there is currently a large gap in the paper in testing the sensitivity of the final results to the manner and timing of how the retreat is prescribed. For me, this gap currently casts doubt on the author’s argument in the abstract that the results of their terminus change simulation could be used as a metric for evaluating the effect of calving laws.
To take an example, the terminus change between any two successive observations could have occurred extremely rapidly, or it could have occurred more gradually. If this was to be investigated by the model there are two extremes of how terminus change is prescribed: (1) instantaneously forcing retreat of the calving front irrespective of the distance of retreat (admittedly highly unrealistic); and (2) forcing retreat between every terminus observation in steps, similar to how the larger calving events are dealt with in the existing simulation. Within the second scenario there is also likely to be sensitivity to how often the calving front is updated (e.g. every timestep or every 6 model months say).
My understanding from the paper is that the authors have taken a compromise approach between scenarios 1 and 2 in order to provide a more ‘realistic’ simulation, and are able to identify 0.5-6 month accelerations that are coincident with prescribed retreat events. By investigating scenarios 1 and 2 outlined above it could be evaluated whether the dynamics and geometry after velocities have stabilised are conditioned by the terminus change events themselves, or just the terminus position. This is important to know if the results of this paper’s simulations are going to be compared to model results where calving laws have been implemented (i.e. for decadal-centennial timescales is it important to regularly update your calving front location, or will you get different/similar decadal timescale dynamics (and ice volume lost) by updating at intervals of every few years?).
Comparison of model results with observations
I’m not sure that the author’s statement on p11 L12 that “In 1985-2012 ISSM_PT simulates mass changes similar to observations” is fully justified/qualified (this is of course heavily dependent on what you mean by “similar”). Looking at Table 3, only 1985-2002/5 mass and dynamic changes represent a good match with the Kahn et al. (2013) values, and 2009-11 of the Larsen et al. (2016) values. The remainder of the values quoted in some cases have substantial mismatches even when the full error range is taken into account. The authors need to be a bit more up front about this, but also critique the reasons why this may be the case (e.g. the length of observation versus the fact that these comparisons are being made for the end part of the transient run). Personally I think getting the good match between obs and modelled results for 1985-2002/5 represents a fantastic result in itself, though the reasons for the mismatch at other times needs to be explored and explained.
It’s also unclear at times in the manuscript what numbers quoted actually represent. For example, P9 L16 quotes the surface as lying within +/- 30% of the observed thickness, though it is not immediately obvious whether this is for the lower portion of the catchment affected dynamic drawdown or for the entire catchment. Given that a lot of the behaviour is driven by calving, a much more detailed view of the near-terminus region would be useful throughout the supplementary (especially though for figure S3) as it would provide a much better impression visually of what is described in the text in addition to where the mismatches are occurring.
P5 L6 – grammar could do with clarifying
P5 L10 – how was the ice surface interpolated and how was the post-relaxation 1849 model configuration compared with the reconstructed configuration
P6 L17 – would remove the link between air T’s and the likelihood of retreat unless it can be fully substantiated
P6 L19-20 – explain the timesteps of calving fronts that are used
P6 L22 – spell out CFL
P7 L5-6 – if simulations are run with stepwise/gradual prescribed calving retreat, is the average surface vel increase the same as the original simulation?
P7 L11 – replace “succeeding” with “following”
P9 L2-3 – I assume the observational data %s aren’t exactly the same as the modelled DIL values – the observed percentages are needed for comparison.
P9 L9-12 – in addition to the movies can you provide a figure showing time evolving evolution of the surface (similar to Jamieson et al., 2012, figure 2). Personally, I find it much easier to visualise and interpret glacier change this way rather than having to play a movie over and over.
P10 L11 – for which locations on UI1,UI2 and UI3 are these velocity increases measured at? Are they the same fixed point or a given distance upstream of the terminus?
P11 – make axes on figure 5 equal
P12 L26 – how does the model deal with advance? Is it mass conserving or is mass in effect added to the domain?
Jamieson SS, Vieli A, Livingstone SJ, Cofaigh CÓ, Stokes C, Hillenbrand CD, Dowdeswell JA. Ice-stream stability on a reverse bed slope. Nature Geoscience. 2012 Nov 1;5(11):799-802.