Please see the attached PDF.

This paper focuses on the interannual variability of the cold water mass in Iyo-Nada, Japan, using both observations along a transect and three sensitivity runs of a hydrodynamic model (POM). The main conclusion of this study is that it is the heat transport during the warming season (May-July), rather than initial temperature before warming (Dec-Feb) that dominates the cold water mass’s interannual variability. In the end, this paper also compares with other coastal cold water masses using idealized equations to explain why the inter annual variability of the INCWM is not dominated by initial temperature in winter.

This is an interesting topic, however, the major conclusion is not successfully delivered through the three sensitivity runs. As Table 1 suggests, the authors increased local and remote air-sea heat fluxes by 10% during the winter (Dec-Feb) and warming (May-July) seasons, in order to evaluate whether it is initial temperature change in winter or heat transports in warming season that dominates the interannual variability of the cold water mass. Design of these sensitivity runs does not separate these two roles at all, thus not convincing at all. Their reasoning is like this, process A + process B will lead to phenomenon C. The authors changed 1.1A kept B the same, and it leads to 1.3C, but the combination of A and 1.1B leads to larger change in C, so B is the dominant role for C, instead of A, which is funny.

For example, authors finally found Case 2 - increase air-sea heat flux by 10% during May-July that brings the largest q and cr(T) values. In this case, the interannual signal in initial temperature is still kept in the model, it is not convincing to conclude as the authors suggested, only if the initial temperature signal on interannual time scale is totally removed. On the other hand, the increase value (10%) is tricky as well. Authors at least need other runs with different increase values to make sure this 10% is not sensitive and the conclusion is still valid. Not to say the results suggested by cr(A) are not supporting, which also need reasonable explanation.

Another confusing and disappointing point is, the authors do use a numerical model to study the inter-annual variability of this cold water mass, but none of heat budget terms from the model is analyzed, which is super crucial for determining which physical process that is mainly contributing. The authors write many texts for the last part (Section 4.2), which is actually not that interesting. Too many idealized assumptions, e.g. no interannual variabilities in rho, R, and H? And the final conclusion of this part is obvious without these texts.

Therefore, this paper is suggested to modify their reasoning method (three sensitivity runs) and the paper structure (too many uninteresting texts in discussion and few analysis about heat budgets). Other comments are listed below.

Major Comments:

1. Section 2.1 does not include the time information of the observational dataset. Is it only collected in Jan/Apr/July/Oct each year for all eight stations? It seems not true, because Figure 2b compares between observational datasets and model results on a time scale of multi-year monthly climatology. Also, How many obs. in each month are collected? Do you calculated monthly averages first?
2. Figure 2a. In the text, it says the R station has a depth of 75m, but the figure only shows the upper 50m. In the caption, it says the 18°C isotherm is considered to define the studied cold water mass, which is better to be illustrated in the main text, e.g. the method section.
3. Section 2.2 Model validation: This paper focuses on the interannual variability of the cold water mass, but only verified it on the multi-year average seasonal cycle (Figure 2b), and concluded that “Therefore, this model is suitable for examining the factors influencing the interannual variation in the INCWM via sensitivity experiments. ” More comparisons between obs. and model results are needed to confirm model validation, including episodic and interannual time scales, transections comparisons.
4. Does the CONTROL run only have seasonal climatology or actually it is an interannual hindcast of this region? It should be clearly stated in the main text. If it only has seasonal climatology, it is not reasonable to compare with the three sensitivity runs, and evaluate the contributions of air-sea heat flux. If it is actually a hindcast, why authors do not show any interannual comparisons with observations to validate the model? Without validation of this model on interannual time scale, it’s really hard to trust and interpret the interannual results from the sensitivity runs. On the other hand, if authors do have a hindcast of this model from 1994-2015 and it shows solid results comparing with observations, it is pretty interesting to quantify the cold water mass properties on interannual time scale using the model as well. Because the observational dataset is only present at 8 stations with very coarse spatial resolution, however, the model has a three dimensional distribution of the cold water mass.
5. Ln 136-141: How the f time series is calculated? Is it area-averaged for each domain? How about delta f? Do they have units? If so, please add them to Table 2. The q is the absolute value of the relative change of Temp/Area. Therefore, it is not sure if it is a decrease or an increase in Temp/Area in sensitivity runs. Also, in Table 2, the q value for each case is ground to one value. Is it a multi-year average? More details needed for the calculations.
6. Ln 193-194: April is selected as the initial month of the water mass formation, but figures only starts from April. To make the point, authors are suggested to show correlation panels at least for March as well. Ln95-97: the authors stated that “water temperature in April depended mainly on the cooling process, the initial temperature of the INCWM is likely associated with local air-sea heat flux from winter to early spring”, any reference for it?
7. Ln 198-199: “correlation coefficient below 10 m is larger in May-July than in April (Figs. 4b-d)” Actually it is not true based on Figure 4. Most of the markers show almost the same values, some even decreases, only two particular markers show the increase.
8. Figure 5. The blue dots does not match those in Figure 3. Some dots are not denoted with years. Please add their years in the figure.
9. Ln 245-247: This is a really confusing conclusion. cr(A) and cr(T) provide opposite results, which is obviously weird, as T and A are always strongly related, but clearly authors choose to trust cr(T) instead of cr(A). More explanations needed here.
10. Ln 253-256: Winds are weak during the warming season, how about cooling season, Dec-Feb? Case 1 should not be small, right? 0.006 and 0.06 are for case 2 and 3?
11. Section 4.2 Repeated texts for the different cold water masses. e.g. Ln 275-279 vs. Ln 323-329

Minor Comments:

1. Some sentences need to be improved in language., e.g. Ln 49-50: This sentence is suggested to be rephrased: “This study focuses on …”.
2. Ln 148: 1994-2012 or 1994-2015?
3. Ln 169: Figure 3 and caption do not match. It says authors used “18.23°C” as the average temperature, but the figure suggests 18°C for that year. Should it change to “18°C” to match the figure?
4. The color scheme of Figure 4 is too similar from 0.45 to 0.75, and the markers are too small. Values where p>0.05 are also suggested to show, in order to see the correlation patterns.

General comments

In this study, the determining factors of the intensity of cold bottom pools in shelf seas and bays are studied. The major study site is a semi-enclosed bay in Japan, but also generalisations to comparable bay are made. The major conclusion is that the strength of the cool pool sometimes depends on the previous winter SST (other studies) or on the air-sea buoyancy flux during the warming season (this study).

While these results are potentially interesting, I think that their value is limited here, since the methods used are not state of the art. The quantification of the cold pool strength depends on a highly site-specific empirical measure, rather than on energy considerations. Furthermore, the applied numerical model uses climatological forcing rather than realistic forcing. With this, a comparison between model results and field observations is not possible and interannual variability, a major focus of this study, cannot be assessed.  

For these reasons, I recommend to reject the manuscript at this stage and motivate resubmission of a manuscript that uses state-of-the-art methods.

Specific comments

32-34: What defines a BCWM to be strong or weak? The temperature of it will certainly increase during spring and summer, such you probably define strength thought some temperature differences? Please specify.

37: better “… and the Middle Atlantic Bight Cold …”

40: Isn’t it more simply and more directly the winter temperatures of the vertically well mixed shelf sea waters and then probably also the summer SST that determine the strength of the BCWM? In addition to the heat fluxes, also laterally advective exchanges could set the temperatures.  This is basically what you argue in lines 41-43.

74: make sure that you avoid double brackets “)(“.

94-96: This measure certainly depends on the position of the transect relative to the cold water pool. Therefore, it is not a suitable measure. Also temperature itself is not a good measure, because it is too site-specific. Temperatures differences between surface and bottom need to be involved. A better measure would for example be the thermal contribution of the potential energy anomaly, integrated over an entire bay. Or you could use the thermal contribution to the Available Potential Energy (APE) of the bay. This can easily be calculated by means of a numerical model.  Measurements can be used to reconstruct this as well, when some assumptions about the geometry of the cold pool are made. The measure could be converted into a mixing time scale by division by the kinetic energy supply through tides and wind (plus/minus surface buoyancy flux contributions).  

96: With the two indices you probably mean the transect area and the temperature.

100: This measure is highly empirical and not physically based, see above.

115-117: The model is forced by some kind of climatological wind, which leads to underestimation of the wind-energy input and mixing. The method for calculating the surface buoyancy flux is not mentioned. Since the wind is climatological, the reviewer can assume that also the buoyancy fluxes are idealised. Also, no information is given about open boundary conditions and riverine freshwater forcing. For an investigations like this one, is would be state of the art to apply a model with realistic forcing. Some information on the surface buoyancy forcing is given in lines 146-151, and it seems indeed that this forcing is climatological as well.

The following sections include some interesting discussions, but since the study is based on a highly site-specific empirical measure for the size of the cold pool and the numerical model is highly idealised, I propose that the authors do first improve their methods according to the above suggestions and then repeat the study.