Review of «Solving dense shelf water cascading with a high-resolution ocean reanalysis» by Hos et al.

The authors investigated the dense shelf water cascading (DSWC) process at two submarine canyons in the Gulf of Lion (northwestern Mediterranean Sea). The aim of the study is to evaluate the performance of the reanalysis in representing the DSWC. To achieve this goal, the authors compare the reanalysis data with in-situ observations from two moorings over the period 1993-2021 for Lacaze-Duthiers Canyon and 2011-2021 for Cape de Creus Canyon. From the results, they conclude that the MedSea reanalysis is able to successfully reproduce the interannual variability of the DSWC, although the timing and intensity of the phenomena are biased at the daily scale.

I found the topic interesting because the reanlysis performance in deep layers has not been thoroughly evaluated for the Mediterranean Sea. Thus, testing reanlysis skills in representing an oceanic proccess such as the DSWC is a good excersise. The manuscript is generally well written and the results are relevant to the scientific community, especially the Mediterranean community. However, I found some points that should be improved before its publication in OS.

1) Abstract : I find the statement "...can be used to find other unreported cascading events elsewhere" somewhat problematic. While the approach is compelling, it's important to note that the reanalysis product used in this study is specifically designed for the Mediterranean. This does not necessarily imply that other reanalysis products (e.g. global) would perform equally well in capturing cascading processes in different regions. In addition, it may be premature to extend the conclusions to other cascading areas within the Mediterranean basin, such as the Adriatic or Aegean Seas, given the different water mass characteristics and potentially different local dynamics in these regions. It would be helpful to evaluate reanalysis performance in these areas before making broader generalizations. Please rephrase.

2) Introduction: Since the manuscript was previously submitted to another journal with strict space limitations, I understand why the introduction may have been kept concise. However, Ocean Science does not have such constraints, and I found the current introduction somewhat brief and missing some important background information. I would recommend that the authors provide a more detailed description of the state-of-the-art, including the criteria (e.g., temperature, salinity, potential density, current velocity) used in previous studies to characterize DWSC in the Gulf of Lions, as well as a more thorough geographical description of the study area. I also strongly recommend including references when referring to water mass characteristics (e.g., LIW, WMDW) and their typical depths in the water column. I suggest more clearly highlighting the novelty of this work in comparison to previous studies. Finally, a short paragraph at the end of the section introducing the content of the next section might also be nice.

3) Data and Methods: I suggest to include a small table summarizing the main information related to the reanalysis data and those related to the measurements from in situ observations. This would help the reader to easily follow the methods section. L68-71: It is not clear to me whether the criteria used to define DSWC events come from observations or from reanalysis. This should be clarified. Are you defining DSWC events from in-situ observations or are you moving to values (model reality) where reanalysis successfully reproduces DSWC?

4) Results: In general, I found this section to be very descriptive, I would recommend the authors to go a little bit more in depth in the analysis in order to provide answers to the open questions (such as, what mechanisms might be involved ?).

- L84: Why is 2019 not considered a DSWC event? It seems clear from the observations that there was an event in the CCC in 2019. However, this event is not captured by the reanalysis.

-L87: ‘Events tend to occur in paired years, where the second year is weaker and usually hardly detectable in the LDC.’. Any explanation for why DSWC occurs in paired years? Why is it more difficult to detect in LDC than in CCC?

-Figure 2: Please use the same ranges on the y-axis to make LDC and CCC easy to compare. Avoid using acronyms in the caption (figure captions should be understandable to the reader without having to read the entire manuscript).

-Figure 3: Why is the year 2018 missing in 3j?

- L123: Here, after finishing the analysis of Figure 3, you come back to the analysis of Figure 2e, which seems a bit strange. Please reorganize. These volumes are calculated for waters σθ ≥ 29.05 kg/m³? Clarify in the caption of Figure 2.

5) Discussion: Have you identified any EMT signals in the reanlysis that prevent deep cascading from 1988 to 1998?

6) Conclusions : As I noted in the abstract, I would caution against generalization of conclusions without detailed analysis in other regions. In this section I would suggest to highlight the biased performance of reanalysis at the daily scale and the main deficiencies found. In this way you can provide some perspectives to suggest improvements to the reanalysis developers.

Minor remarks:

-L13: As I understood through the manuscript, the monitoring of Cape de Creus started in 2005/2006 for winters and then permanently from 2011, not from 2003 as stated.

-L27: you may want to include the common names for these dry and cold northerly winds (Mistral and Tramontana).

L35: basin floor → bottom

-L38-40: Any study under CMIP5 or CMIP6 scenario projections?

-L50: (CMEMS, https://marine.copernicus.eu/)

-L60: from 1987 to 2021→ the reanalysis provides data until 05/31/2023, but you focus your analysis on 1987-2021.

-L88: significant ‘anti-correlation’ → ‘negative correlation’

- L95: ‘temperature minma at CCC’ → and LDZ isn’t it?

-L96: (Sv; 1 Sv = 106 m³/s) → (Sv; 1 Sv = 10⁶ m³/s). Please revise the use of superscripts throughout the manuscript.

General comments:

The authors analyze the ability of a numerical modelling (The Copernicus MED-SEA physics reanalysis 2017-2023) to simulate dense water cascading in the Gulf of Lions during winter. The assessment is carried out by comparing simulations with independent observations in the Cap Creus Canyon (CCC) and the Lacaze-Duthier Canyon (LDC), located to the west of the Gulf of Lion. The results compare temperature and velocity near the bottom (30 m above the sea floor) at a depth of 1000 m along the axis of the canyons. Visual and statistical (Taylor diagram) are invoked to support assessment. This is followed by a brief description of the cascading processes modelled, focusing on the winter of 1999 and the role played by shelf water density during 33 years.

Although such modelling is not fully designed to simulate the cascading (relatively coarse resolution, poor representation of canyon bathymetry, z-coordinates, hydrostatic assumption, etc.), it is reasonably successful in triggering the cascade in line with observations. This is the main and interesting result of the paper.

Modelling appears to confirm a 6-7 years cycle of paired cascading events and suggests an inhibition of such processes during EMT. On this point, the paper does not seem complete to me. The authors should study the role of atmospheric forcing (in particular the buoyancy flux over the Gulf of Lion and the wind forcing) in pre-conditioning and triggering the cascading. This could be an interesting complement to previous cascading simulations (Dufaut et al. 2004, Ulses et al. 2008ab), which have focused solely on specific events, or to Hermann et al. (2008), which have dealt with climatological and more regional scales.

One would expect more deep interpretation of the modelled cascading free period (1988-1997). Is it related closely to the EMT as suggested - i.e. to the water masses characteristics due to a past and distant event - or to the NAO (or others Mediterranean Oscillation indexes) -i.e. to the current atmospheric forcing-?

Why the 2019 event is missing in simulation (figure 2). The cascading does not reach 1000m deep in the simulation?

As written in the paper, it seems suspicious as regard to the residence time in the Gulf of Lion to consider a connection between two consecutive events. Are the apparent ‘paired events’ triggered by the meteorological forcing?

In my opinion, the discussion remains too brief and some results require further interpretation, as numerical modelling makes it possible to study the dynamics (as the authors suggest in their conclusion). Otherwise, the preprint is limited to the assessment of the cascading process in a reanalysis in the Gulf of Lion, its scope becomes less ambitious and the title inappropriate. The current title is too generic. The reader is expecting more a numerical experiment rather than an assessment of a reanalysis in the western part of the Golf of Lions. Please consider rewording the title.

I'm suggesting a major revision, because I think that you need to go beyond simply assessing the modelling of cascading to the west of the Gulf of Lions. The success of this modelling calls for an interpretation that is only very briefly sketched out in this version of the paper. Alternatively, it is possible to focus on the assessment, but this also reduces drastically its relevance.

Detailed comments

Line 70. What means exactly detrended? Is there a global temperature trend in numerical modelling and or in situ data?

Line 70. The expression ‘through the sections in Figure 1a’ seems more appropriate than ‘along the water column’. Have you calculated the transport from the bottom to the surface by applying your salinity and density criteria or have you implicitly added a depth criterion?

Line 70.  The detection of IDSWC from in situ data is clear (T<=12.6 and downstream velocities >=0.1m/s at 30m from the floor), but the extraction of the same data in reanalysis need to be clarified. I guess the velocities and the temperatures are extracted from the deeper cell (i.e. at ~15m from the floor) in the modelled canyon axis.

Line 80-85. Why the 2019 event is missing in simulation (figure 2). The cascading does not reach 1000m deep in the simulation?

Line 90. In guess the correlations have been performed only on the winter (JFM + AM?) periods as quoted at the beginning of the next section (line98).

Figure 1. The figure 1b is a bit confusing and may be not useful (it is obviously not the model bathymetry and the colored boxes look like horizontal section while vertical cross-sections are defined in figure 1a).

Line 106-107. Neither figure 3i nor figure 2 show an overestimation of the duration of cascading. You should add the plots for winter 2018 to figures 3a-h to support your affirmation.

Line159-160. The expression ‘balance between’ (i.e. equilibrium) does not seem appropriate to me. You probably mean ‘contribution of’ or …

Line 231. The exact reference seems to be slightly different:  https://egusphere.copernicus.org/preprints/2025/egusphere-2025-1310/

Typos

Line 89. Probably “r=-0.40” instead of “r=0.40”

Line 95 in figure caption -> 1sv = 10⁶m³s^-1