|This paper presents an analysis of sea ice vorticity in the Atlantic sector of the Southern Ocean. In particular, it examines 4 datasets of sea ice velocity determined from feature tracking. Three of the datasets are from single sensors, while the fourth is a merged product. |
The main conclusions seem to be that i) cyclonic ice drift features are more common, intense, and variable than anticyclonic features in this region; ii) the merged product contains more intense cyclonic features than the other products; iii) an increase in cyclonic features occurred after the well-known decline in ice extent in this region around 2014.
I believe these results are interesting and important and do deserve publication. However, I have a few issues with the paper as it stands. I was not shown any evidence that the metric used here is an essential climate variable (cyclonic ice drift driven by atmospheric forcing). I also feel that each of the three conclusions I have enumerated above are shown but not explained. So this left me as a reader both confused about exactly what is shown and also wondering why those features have occurred. I think my main problem stems from the fact that all of the figures shown are purely statistical, and do not invoke the physics of sea ice in this region.
As I understand it, the key metric that is being plotted is the grid-scale vorticity averaged over overlapping circles with a 450km radius. The paper frequently refers to atmospheric cyclones/anticyclones and seems to assume throughout that all vorticity detected by this metric is associated with cyclones/anticyclones. However, I don’t see that demonstrated anywhere. I can imagine many other things that would create vorticity in these data, e.g. coastal currents, ice edge currents, ocean eddies, tidal effects, artefacts in the data from missing data, interpolation or edge effects from different satellite swaths, etc. One piece of evidence that non-atmospheric-cyclone vorticity is important is that the standard deviation of vorticity within the circles is large. Does that mean there are strong sub-circle features, which means smaller scale, which means not cyclones? So, what is the basis for claiming that the vorticity captured by this metric is from atmospheric cyclones/anticyclones? I think this may be particularly important when just considering the 95th percentile. Perhaps those are the relatively few circles that have strong sub-circle features, e.g. ocean eddies or floes at the ice edge or satellite processing artefacts? I think that demonstrating where the vorticity comes from in a general sense is crucial to the conclusions, and I feel it is missing in the current manuscript.
I think the paper needs to show some maps of vorticity. At present I don’t have a good feel for what area is under consideration, or what its vorticity looks like. I haven’t studied vorticity plots elsewhere in the literature and so I have no intuition here – maybe I missed something? I think the reader needs to know what the metric is, and I don’t think that can be shown by starting with considering higher statistics. I think maps could be constructed such as mean vorticity, mean cyclonic-only vorticity, snapshots of interesting cases (e.g. high cyclonic vorticity), etc.
Figure 1 etc. Why are the cyclonic features more intense on average than the anticyclonic features? I think the explanation here really depends upon whether they are dictated by atmospheric forcing or other features. I feel that maps are particularly needed here, to show where the cyclonic/anticyclonic features are and what they look like. If the explanation lies in the atmospheric forcing (cyclones are more intense than anticyclones) then that could be shown and discussed and literature cited.
Line 225: The merged product has a higher frequency of high cyclonic anomalies than the single-sensor products. Why is this? It is sensitive to the quality flag, but reducing the threshold for that flag just includes data that are suspect and so that is not a good test, in my opinion. Is it because there are more data, e.g. near the ice edge? Does mapping the vorticity fields for the different sensors help here? Mapping the data availability?
None of the regressions plotted in figures 1 and 2 any statistical significance test. P values should be calculated and quoted on the plots, and the text should be altered wherever the trends are not significant.
Line 137: I understand that for a given time, the circles overlap in space. However, I do not understand how they could overlap in time? And how you could allow them to overlap in time but not space?
Equation (2): Is sigma_tr a displacement error, in units of metres?
Optional: The paper does not consider seasons at all. That seems strange to me as I would imagine that the ice vorticity is very different in the summer and winter, both through the ice mechanics and the atmospheric forcing. I do not think it is essential that the authors introduce seasons to the paper, but I would do that personally as I think it would add a lot of insight.
Comments have been prepared in the attached supplement.