the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 30 Jul 2025
Characteristics of ocean mesoscale eddies in the Canadian Basin from a high resolution pan-Arctic model
Abstract. Mesoscale eddies are ubiquitous in the Arctic Ocean and are expected to become more numerous and energetic as sea ice continues to decline. Yet, the spatio-temporal characteristics of these eddies are poorly documented. Here, we apply an eddy detection and tracking method to investigate mesoscale eddies in the Canadian Basin over the period 1995–2020 from the output of a high resolution (1/12°) regional model of the Arctic - North Atlantic. Over that period, about 6,250 eddies are detected per year and per depth level and are distributed about equally between cyclones and anticyclones. On average, these eddies last 10 days, travel 11 km and have a radius of 12.1 km. These statistics hide strong regional and temporal disparities within the eddy population studied. In the top 85 m, the seasonal, decadal and interannual variability in the number of eddies and in their mean characteristics follow that of the sea ice cover. In contrast, below the upper pycnocline, the eddy number and properties show a weakened seasonality. At all depths, eddy characteristics and generation rate show a strong asymmetry between the slope and the centre of the Canadian Basin. The upper 85 m show an increase in the number of eddies generated along the slope, while a net diminution of the number of eddies generated is visible within the pycnocline layer along the slope presumably due to the stabilizing effect of the slope. An increased number of eddies are generated in the vicinity of the cyclonic boundary current in the AW layer. The vast majority of eddies have no temperature signature with respect to their environment, although a significant portion of long-lived eddies, located along the Chukchi shelf break, have a non-negligible temperature anomaly and penetrate into the Beaufort Gyre, thus suggesting a mechanism for the penetration of heat into the gyre. The number of eddies generated within the upper 85 m increases by 34 % over the 25 year of simulation, with the largest increase occurring in the open ocean and marginal ice zone. The number of eddies between the upper and lower pycnoclines increases by 45 %, with a strong year-long increase in 2008, presumably in response to the Beaufort Gyre spin-up in 2007–2008. The number of eddies in the Atlantic Waters (AW) layer shows an overall increase of 41 % with little interannual variability. Finally, the analysis shows that the dominance of anticyclonic eddies within the Beaufort Gyre reported from measurements with Ice Tethered Profilers is partly due to a spatial sampling bias. This model-based eddy census can thus help interpret some of the discrepancies found between observational studies by providing a consistent spatio-temporal characterization of mesoscale eddies in the Canadian basin.
- Preprint
(7071 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 24 Sep 2025)
-
RC1: 'Comment on egusphere-2025-3527', Anonymous Referee #1, 26 Aug 2025
reply
This paper uses a 1/12 degree ocean general circulation model to studies the statistics of ocean mesoscale eddies in the Beaufort gyre of the Arctic Ocean. The authors focus on the period 1995-2020. The authors show strong regional and temporal variability. There is strong linkage with the sea ice cover, while the seasonality of the eddies is weaker during the pycnocline. The authors find, that except along the Chukchi shelf break, that most eddies have little to no temperature signal. The eddies in the upper layer increase with time over the simulation. The authors also suggest that the results from ITPs that suggest most Beaufort Gyre eddies are anti-cyclonic may be due to sampling bias.
This is an interesting paper, well worth publishing. And appropriate for Ocean Science. That said, there are ways the manuscript could be improved and I would suggest major revisions. Detailed justification is provided below.
Major points for further discussion or analysis:
Is a 1/12 degree model really high resolution? Many models of higher resolution now exist – for example there are papers that look at eddies at 1/60 degree in simulations of the Arctic Ocean, Nordic Seas and the Labrador Sea, for example. The advantage of a 1/12 degree simulation is the length of integration for the analysis. So, as long as the given resolution is able to resolve a significant part of the eddy spectrum at the given resolution, I can understand the authors using the configuration they did. But I’d like to see some more discussion of mesoscale eddies and understand how they are represented in CREG12. Given the model’s stratification, could the authors provide information on where the Rossby Radius is resolved and with how many grid cells, especially at the margins of the Beaufort Gyre. And if the motivation to use 1/12 degree is the longer timeseries, this needs to be discussed. As well, some discussion of what is found at higher resolution is needed to provide the reader with confidence with respect to their results at the given resolution.
I’m trying to understand what the authors’ number of eddies mean. The paper states about 6,250 eddies are detected per year and per depth level. That seems like a huge number. First off, I understand the authors point about the difficulty in determining the vertical coherence of the eddies. And the authors do show some changes with depth. But just saying per depth level makes it seem like there are significantly more eddies that there actually are – given most eddies are likely found through multiple levels. In might be better to average numbers over several levels and then state there are on average X eddies detected per year above pycnocline, Y in the pycnocline and Z below it (or using some other depth metric).
Additionally, how many eddy exist per day, on average (i.e. a timeseries by day through the mean seasonal cycle)? I wonder if the large yearly number if because of the short duration of the eddies being a function of them being lost by the tracking software, then re-found and thus recounted as a new eddy. A median duration of 4 days is short! Or if the eddies are being damped quickly given the setting at the given model resolution? If I look at figure 3c, I don’t see that many eddies on the given day (with generally very short trajectories), so I am wanting more discussion of this, to help understand what that large eddy number the paper provides really means. The author’s do bring up the idea of turbulent soup, which I like – but I still feels this topic does need more to help the reader understand what the results mean.
The authors talk about no temperature signal – what about salinity? The eddies like play an important role in freshwater exchange into and out of the gyre. This would be good to further explore.
Strong increase in 2008 – BG spin-up and/or low ice year?
I have some technical questions about the model configuration and experiment. What does constraining the model to *about* 1.4 Sv at Bering Strait mean? A constant value with time? Constant annual mean? An annual mean of ~1.4 Sv with interannual variability? Additionally, this value seems a bit large compared to the observations. As well, does the model consider the increase in recent years suggested with the observations? As well, what are the heat and freshwater transports, and how do they compare with the observations?
The river runoff I think needs to better explained. I thought the Stadnyk et al. paper mentioned uses the AHYPE model? And wasn’t the output from that model used in Weiss-Gibbons et al., 2025? And in either case, I don’t believe the AHYPE output provided Greenland discharge? Are you sure that didn’t come from a different product?
Given the importance of the sea-ice in setting the seasonal cycle of the eddies, more on the model representation of sea-ice would be useful. What does comparable to that derived from satellite observations mean in practice? Especially given that the authors use ERA5 for forcing, which has a known warm bias in the Arctic. So I’m curious about what parameter set for SI3 was able to allow the simulation to get sea ice fields close to the satellite observations.
Given the discussion of the three main water masses, I would also be curious to see a time series of salinity (or freshwater content) in the model, compared to the Beaufort Gyre Experiment results. Especially since Rosenblum et al. has pointed out that many models under-estimate the freshwater content in the region. Also, why use fixed depth layers for the vertical splitting, instead of isopyncals (or isohalines) to define the different water masses?
Finally, given the authors look at changes with time, as the amount of sea ice is being reduced, can the authors speculate about what their results may mean for the future. And discuss the implications for those potential changes.
End the discussion with more discussion on the limitations of the model and the present study.
Smaller items
Line 19 – The increase in the AW layer is over what time period?
Figures 1, 3, 5, 6: Please use discrete color contour intervals to make the figures easier to read.
Figure 2. Could you explain in more detail how the anomalies are calculated, rather than just saying similarly from monthly mean anomalies.
Line 333. Is this small decrease significant?
L335: Might some of the increase with time be a function of changes in the winds and energy input with time?
L348. Half more eddies doesn’t read well.
Figure 7 caption. What does gradient of the SSH averaged over the CB mean?
Table 2. Remind people of the layer definitions in the table or caption.
Citation: https://doi.org/10.5194/egusphere-2025-3527-RC1 -
RC2: 'Comment on egusphere-2025-3527', Anonymous Referee #2, 05 Sep 2025
reply
This manuscript documents the characteristics of mesoscale eddies in the Canadian Basin simulated by a 1/12th deg regional model of the Arctic. While the work has the potential to be a useful contribution, the manuscript in its current form is largely descriptive, and the results presented are not particularly compelling. It is not entirely clear what new physical insights are gained from this study and also what the important implications of the results are.
Importantly, since the study is entirely model based, it is essential to demonstrate that the model simulates the eddy field in the Canadian Basin well by comparing with available observations. At present, the evaluation is very limited: only one figure (Figure 2) shows model-simulated EKE compared with mooring-based estimates at two depths, and the model-data comparison is addressed in a single sentence, which simply states that the agreement is good. This level of validation is insufficient. A more extensive and quantitative comparison with available observations is needed for the simulated eddy field. Without this, readers cannot be confident in the reliability of the subsequent analysis of eddy characteristics from the model output. In fact, Figure 2 shows that there are considerable discrepancies between the simulated and observed EKE.
The model stratification differs from climatology (Figure 1). What is the corresponding radius of deformation in the model? Can your model properly resolve it?
There are several other eddy detection methods, e.g., contours of SSH anomaly. Could briefly discuss why you choose a method based on the OW parameter and what the advantage of this method is over other existing eddy detection methods.
It would be of interest to many readers to see the spatial distribution of EKE in the three layers, as this would bring together information on eddy number and intensity.
All the trends reported need to be tested for statistical significance.
The strong reduction in eddy number along the southern edge of the BG is attributed to the stabilizing effect of the continental slope. However, the slope (as indicated by the grey contours) appears to lie south of this region of low eddy number? In addition, why would the stabilizing effect of the slope result in stronger eddy intensity there?
More minor comments:
Line 154. “from R=10 km to 2*pi*R=60 km”. Is there a typo here?
Line 270. Explain why there is a dominance of anticyclones at the very surface within the Canada basin and cyclones within the Chukchi shelf break area in winter?
Line 280. Explain what the theoretical mean turnaround time is and how it is calculated.
Line 298. How is “generation rate” defined?
Line 329. …in our domain…
Line 335-337. This sentence is not clear, as the expansion of the ice-free area also contributes to increased energy input to the ocean.
Citation: https://doi.org/10.5194/egusphere-2025-3527-RC2 -
EC1: 'Comment on egusphere-2025-3527', Julian Mak, 06 Sep 2025
reply
I was waiting for the second referee to post (and an informal referee to comment) before I posted, and thankfully they corroborate what I was going to say already. I also think the paper can be published eventually, but quite a bit of elaboration and appropriate literature review will be required, and would recommend the authors to do replies and make revisions in due course.
To be blunt about this, my big issue with a lot of the eddy detection papers is "so what?", and to me this paper falls into that category. The polarity, eccentricities, depth extent, numbers etc by themselves are to me rather meaningless if they are not linked up explicitly to the appropriate physical mechanisms and/or possible consequences. A few of these might be:
* biases in polarity may be related to deviation from geostrophic theory
* vortices have a tendency to circularise, so highly eccentric vortices may indicate forcing mechanisms at play disrupting their shape
* deep/intermediate/shallow extent of the eddies may be related to the generation mechanism of the eddies
* numbers may indicate changes in the generation or dissipation mechanism (e.g. ice-ocean drag)At the moment some of the eddy metrics are presented and it is as if the reader is supposed to just know why these are important (I am not personally convinced they are to be honest). Anyhow, the fix is probably reasonably "easy" to do in principle: motivate the science and relationship with the metrics to be presented in a more comprehensive fashion in the introduction section (and may be recap accordingly in the conclusion), adding (probably quite a few) references of the literature as appropriate. Set up the scene on why the metrics to be discussed are going to be relevant, then go into detail as much as the authors would like. Note that the referees point out that there are some details that are not entirely sufficient at present (e.g. statistical significance), so do have a look at those and address those accordingly.
Some of my other selected scientific issues are:
* Eddy detection per vertical level, as noted by the referees also. This overblows the amount of eddies there actually are, and the vertical coherence and structure of the eddies are one of the most important aspects associated with geostrophic eddies partly because these probably relate to the generation mechanism and/or background turbulence characteristics (e.g. Eady-modes are deep while Charney-Green modes are more vertically confined; QG theory suggests inverse cascades and barotropisation but maybe there are things present in the system that does not allow that to proceed fully). This really should be fixed, or failing it being fixed completely in this article, some samples of cases showing vertical coherency should probably be shown somewhere to at least demonstrate it is doable. I am making the assumption that there are cases with vertical coherency because that's what theory would suggest; it would be more surprising if there wasn't (although I would then be inclined to think the choice of method has deficiencies instead).
* 1/12 degree is probably not completely sufficient to resolve eddies in the Arctic region and the eddies that are actually detected here might have been mis-represented / mangled already. If the authors go with "high res models are expensive and we have this model handy so we will do it on this as a first try" then sure, but the authors go with "this is a precursor to compare with observations", then a natural question that arises would be whether a comparison is even meaningful at all. The latter stance to me is a problematic one to take with the present dataset; the former stance is at least slightly more defensible.
* Detection methods based on Okubo-Weiss. I have my opinions on Okubo-Weiss as well as Eulerian methodologies but the thing that is sticking out the most is a lack of discussion with other eddy detection methodologies (as raised also by the referees). By all means make a choice, but that choice needs to be evaluated against other known methodologies, and that's not really done here at the moment.* Comparison with observations: There was some suggestion the work forms a precursor to comparison with observations, but then do the current results tell us more about where we should observe, what kind of temporal / horizontal / vertical resolution would be need, and whether that provides a logistical constraint on what is observable and what is not?
I would suggest the authors consider justifying this work in another way, because I think that stance is hindering rather than helping the paper by raising more questions that lead to not very convincing answers...Citation: https://doi.org/10.5194/egusphere-2025-3527-EC1
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 420 | 37 | 10 | 467 | 24 | 26 |
- HTML: 420
- PDF: 37
- XML: 10
- Total: 467
- BibTeX: 24
- EndNote: 26
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1