*Overview*
I am pleased to see that the authors have addressed one of my main concerns from the last review by extending the hindcast runs, and associated assessment, out to more than 1 year. Given that things are worse in the second winter than the first, I feel it would be better to have run the system for even longer, but 20 months is much better than what we had before.
However, I find myself still rather frustrated with the presentation of this manuscript. The wording used is still very informal/unscientific and the arguments are not laid out in a manner that is conducive to transfer information to the reader – i.e., detailing exactly what has been done, why certain decisions were made, etc.. Moreover, the figures are quite difficult to understand. Several of the newer figures use very small text and/or very small coloured dots, that make them almost illegible. Furthermore, nearly all of the figures have captions that are either wrong or don’t contain the required detail, and all figures suffer from a lack of annotation/labelling or panel subtitling, that would allow the reader to quickly ascertain exactly what is plotted where.
Finally, after reading this through again, I’m afraid to say that I now find myself even more confused about exactly what is done in the neXtSIM-F operational forecasting system (and why!). I don’t find Figure 1 very informative in this regard without the appropriate explanation. For example, I am not sure what the difference between the “initialization” and “assimilation” boxes – when most operational centres would actually class the assimilation as the initialisation (i.e., its only job is to get the forecast starting from the best possible initial conditions!). I am also left wondering why the assimilation seems to be being applied during the forecasts (P19 L29-31) and why it seems to make little difference either way. Given this last point, I am further left wondering what the analysis is about (i.e., what is analysed, how is it used, how does it differ from the 1-day forecasts that contain assimilation?).
I think this issue can be fairly easily fixed - the authors need to take some time to carefully explain the top-level structure of the forecasting system. I would recommend doing this right at the start of Section 3 where Fig 1 is introduced (i.e., before the model is discussed in Section 3.1).
Also I find that things are further confused by the dual discussion of MEB and BBM versions of neXtSIM but only the latter is evaluated here. In some places (e.g. Appendix A1) this feels like a CMEMS report about the operational system (in particular given that December 2020 is in the past)! Given that in this paper you are solely showing results from the BBM version of neXtSIM, much of the MEB stuff is not relevant - except to show the differences in the new (current?) system you are documenting here. So I feel it would make sense to tone the MEB descriptions down and just to focus on the new/current operational system.
*General comments*
You need to be careful when talking about other systems all using the VP rheology because many readers will likely ask “what about EVP?”. Many will not know how similar these are so it might be best to elaborate a little and say "based upon the viscous-plastic (E)VP formulation" or “the VP family of rheologies” or something. Also some might ask “what about the EAP” rheology? I think that RASM are using EAP for their coupled seasonal forecasts now(?). Maybe you could avoid that potential conflict by better quantifying the area you are talking about by stating "short-range analysis and forecasting system" instead of just "forecasting system"?
**NB. page and line numbers used below correspond to those used in the tracked-changes version of the document. I now notice that these are somewhat odd with some pages having some line repetition**
*Specific comments*
P2 L15: “while Hunke et al. (2020, Table 1) give a comprehensive list of modelling systems that include sea ice”: I would be a little careful here because the Hunke et al. paper is actually about the use of climate model sea ice modelling formulations for operational forecasting. Therefore, it does not aim to give an overview of all operational forecasting systems (certainly not all “modelling systems” as you claim!), instead only those using the classic AIDJEX continuum model formulation (CICE, LIM/SI3, MITgcm).
P2 L18: “[models/systems] do not vary in their numerical framework”: I would change the wording here because many would argue that there are lots of differences between models/systems - with some models having complicated thermodynamics schemes and others not having sea ice thermodynamics at all (zero-layer). If you mean the dynamics and the overall structural formulation (i.e., AIDJEX continuum dynamic-thermodynamic models) then you should say so.
P2 L24: I don’t quite get the RIPS->RIOPS paragraph here (aside from J-F requesting it!) because it almost invalidates your argument for running standalone neXtSIM. You need more discussion here, I think. For instance, why did ECCC move away from the basic/standalone RIPS forecast towards RIOPS? What physics/skill does neXtSIM-F likely miss out on by sticking with the "RIPS" style approach (i.e., standalone) rather than the full "RIOPS"?
P3 L8-10: again with the first forecasting system not to use VP you should be careful of RASM using EAP and to ensure that EVP is captured (as in General comments)
P4 L20: ocean forcing – please specify the frequency of forcing field update – daily? Hourly?
P4 L20: “TOPAZ near-surface (30m) velocity”: Is this exactly the velocity at 30m or the integrated velocity over the top 30m? You need to be more specific. Either way can you say you use this approach?
P5 L10: atmos forcing – please specify the frequency of forcing field update – daily? Hourly?
P5 L25: “(European western time)”: do you mean UTC? If so that would be easier to understand.
P6 L12-3: “(It is therefore an independent validation dataset for our forecasts.)”: I find this odd because the previous sentence – i.e., used for evaluation of free run and forecasts – does not make the data independent! Needs rewording.
P6 L19: “As part of our evaluation we sometimes apply a filter on the uncertainty…”: I don’t understand why we have sometimes here! Did you do this filtering here or not? If so remove the sometimes; if not why mention it?
P7 L11: “In order to compare neXtSIM drift…”: You don’t need to do this to compare drift though do you? You could use the model velocity fields? The point here is that you are trying to make a better comparison and “compare apples with apples” so why not say that?
P7 L13: “…place Lagrangian drifters…”: I would be more explicit and say that you "…seed synthetic Lagrangian drifters into the model…".
P7 L15: Sentence starts with “We use this product” but you've not defined/identified a product yet! Unless you include the "CS2-SMOS" in the subtitle (which I don't). Needs rewording.
P7 L27-28: “A delay of one week for thickness would probably be acceptable for assimilation in a real time forecast in the future.” You need to be careful here because many others (including me) would be of the opposite opinion. I guess you mean that you can pretend that the thickness from 7 days ago is the thickness from today and assimilate it? Even so that doesn't sound ideal. If however you're talking about weekly thick ice from CS2 but still using daily SMOS data for thinner ice then perhaps that is ok. However you don't say these things! Either way I would reword to make things clearer. Many of us are pushing hard to get satellite SIT available for near-real-time assimilation and so any statements along the line of what you say here might derail that. So you need to be clear and careful.
P11 L4: “The mesh is generated with Unref (a component of Gmsh…” needs more information to explain what these are. Are they numerical packages or publicly available tools or something?
P11 L26-29: regarding satellite observations being interpolated to model grid - This is the opposite way around that most people do assimilation, and model-observations comparisons generally, where the model is always translated to the observational locations. Why do you do this the other way around? What impact might this have?
P12 L17: “(this is a kind of assimilation of extent)” given that you don’t use the concentration in the pack why do you even need to do the assimilation given that the ice edge would come through in the SMOS SIT initialisation anyhow?
P12 L3 (bottom of page!): “the heat flux out of the ocean increases and the ice freezes up again very fast”. Surely this is not always the case? Only if the atmospheric forcing thinks there should be ice? i.e., if there is ice in the ECMWF model used to provide the forcing then the atmospheric fields will be cold and conducive to ice regrowth but if, however, there is no ice in the ECMWF model then the near-surface atmosphere will be warmer?
P15 L4: Re Figs 2&3 - what is “mean concentration”? Is this the mean over the whole domain, only ocean points, only sea ice point, or what?
P17 L12-14: both “Jan-Feb” and “Nov-Dec” are discussed here in relation to Fig 4 but neither panels are present in Fig 4?!
P24 L4 (bottom page): “…with 1-day forecasts being launched in between so assimilation still performed daily.” Are these not the analyses?
P25 L11: (equation 7): why, when you say IIEE is analogous to RMSE, is the RMSE ratio squared for concentration (eq7a) but the IIEE ratio is not squared for extent (eq7b)?
P25 L12-15 (bottom page): this needs rewording. I have no idea what “straw man” means here – are you seeking feedback on a basic idea from others? or trying to scare birds away from your crops? Neither seem appropriate. Also “We have rough benchmarks from some other models however” doesn’t read well. If this relates to the next sentences then you might as well delete it. If not, what are these benchmarks?
P27 L12-13: “we don’t try to correct this as reducing the concentration in the pack causes serious problem with the drift and thickness”: this might work well for a standalone sea ice model but you could have trouble with heat fluxes if you coupled to the ocean and/or atmosphere. Can you say any more about this – particularly in light of the last line of the paper where you mention plans to couple with ocean and/or waves.
*Figures*
Figs 2 & 3: the text used for legend & axes is very small and hard to read comfortably
Fig 4 (caption): the caption should state at the start that these are “differences” between the free run and OSI-SAF
Fig 5: the text is far too small in this figure and the caption makes no mention of what the left and right panels show (I had to zoom in to 200% to see the dates to work out they are the 2 winters - 18-19 and 19-20)!
Figs 6 & 8: panel titles or annotation would make it much easier to see what is plotted in each column
Figs 9 & 12: The detail on this figure is tiny. Even at 200% zoom I struggle to see what's going on properly - particularly the tiny coloured dots.
Fig 9 (caption): the caption is all wrong and relates to 3 rows when there are only 2.
Figs 10 & 11: what do the columns show? panel titles or annotation would make it much easier to see what is plotted in each column
*Typos*
P1 L4 (and other locations): VP should be “viscous-plastic” not “viscoplastic”
P2 L12: “MOSAIC” should be “MOSAiC”
P2 L21; “operation” should be “operational”?
P3 L11: CMEMS = "Copernicus Marine Environment Monitoring Service". Your version with "Marine and Environmental" would be a very different beast rather than just doing the "Marine Environment"!
P4 L9: Arctic MFC = "Arctic Monitoring and Forecasting Centre" (see https://marine.copernicus.eu/about/producers/arctic-mfc)
P4 L12: (as mentioned in the last review) the reference for CICEv4.1 is "Hunke and Lipscomb, (2010)" not “Hunke et al.”. It looks like in the references you have changed the date to 2010 but not changed any other details of the reference - including the author list!
“Hunke, E. C. and Lipscomb, W. H.: CICE: the Los Alamos sea ice model. Documentation and software users manual, Version 4.1 (LA-CC-06-012).”
P4 L30: You should drop the “both” from “used both for assimilation” now that “and evaluation” is deleted
P5 L19: AMSR2 = “Advanced Microwave Scanning Radiometer 2”
P6 L17: “Cryosat-2” should be “CryoSat-2” |
- Printer-friendly version
- Supplement
