The manuscript describes analyses of ocean and other model output in the vicinity of the Canary Current Upwelling System.  Specifically, it examines three different metrics of upwelling system indices, the vertical temperature structure of nearshore and offshore water columns, and the patterns of temperature that are correlated with wintertime indices of the North Atlantic Oscillation and East Atlantic atmospheric patterns.  Oceanic temperature is obtained from the  Global Ocean Ensemble Physics Reanalysis dataset obtained from CMEMS. 

The authors find that (1) different upwelling indices have different values and seasonal cycles, (2) the isothermal depth of nearshore profiles during upwelling is less than that for offshore profiles, and (3) upwelling is most intense during the positive phase of the NAO and (4) especially that in combination with the negative phase of the EA.

In my opinion, the main advance of this paper is item (4) and this result is interesting and useful.  Analysis of the upwelling indices appear and vertical structure are to me less novel, though it might be argued that they raise interesting questions (e.g., about what is the best upwelling index to use) and provide useful context (e.g., typical and anomalous isothermal layer depths) for the remainder of the paper.  

Main Recommendation:

I think the manuscript would benefit from the authors choosing the best upwelling index to characterize the upwelling, presenting only that, and then expand on the NAO/EA parts of the paper. Is the comparison of UI important? Is the 5 month time-lag between UI_ERA and UI_SST important beyond showing that indices based on different data and approaches are different?  

It would be helpful if the authors would include a local map of winds associated with the NAO and EA patterns (to understand their impact on local upwelling), along with their time-series, showing highlighted periods of upwelling that were used for averaging the model.  At the moment, the method by which averaging is done is not clear.  How many days contributed to the NOA+, NAO-, NAO+EA-, and NAO-EA+ fields shown in Figure 6 and 7. How much uncertainty is there in the averages calculated? Perhaps time-series or pdfs of upwelling events associated with different climate conditions could further support the argument that statistics change in different climate conditions.

Other comments

Lines 115-124: The calculation of UI_ERA5 I think should rotate the winds to the alongshore direction and then calculate the wind stress, rather than the reverse as is done presently.

The results section (around lines 210), the authors claim that a change in the ILD of 1-2 m during upwelling events.  There is no error analysis to show significance of this, but even if there was, is a 10% deepening important?  And (as they point out), this is not a new result (Line 213).  This section might be dropped.

Similarly, What's to be interpreted as important in Figures 4 and 5.  They do show differences in coastal and offshore profiles, but the figures seems routine.  Why characterize the vertical profiles or representative sections?  As the authors point out, the description that they give are in agreement with other works (Line 243).

Line 178: The result that the trends in UI are small over 25 years is interesting and useful.

Line 271: Minor comment: I think the authors mean "observed down to ~50 m depth"?  Also Fig 6 has an error in listing 97.04 m twice.  I think the authors mean ~200 m in the bottom row?

Figure 7 is very interesting and compelling.

The authors describe an analysis of upwelling off the NW African coast associated with the Canary Current system. A 25 year period is considered using data from various sources. A particular focus of this study is the relationship between upwelling and variability that may be  driven by the North Atlantic Oscillation (NAO) and the East Atlantic (EA) teleconnection patterns. A new aspect of this study appears to be an analysis of the vertical structure of the water column associated with upwelling variability.

On the whole, the paper is well written and contains copious references to previous relevant literature. There are a few issues though that I would like to raise. I believe that this paper contains results that would be of interest to the community and once the issues mentioned below have been addressed, it should be appropriate for publication.

General comments:

(1) For this reviewer, one of the major weaknesses of this study is the very coarse horizontal resolution of the GREP ocean data set that is used for analysis of SST and vertical structure. The horiontal grid spacing of GREP is only 1 degree which means that in reality the effective resolution is probably more like 3 or 4 degrees. On the otherhand, the width of the upwelling region due to Ekman divergence at the coast may only be ~ the Rossby radius of deformation, which is probably ~40km for the 1st barocinic mode. This is much shorter than the resolution of the GREP data. If wind stress curl is an important factor in enhancing coastal upwelling in the region, the width of the upwelling zone may be larger, but again still less than the GREP resolution. Therefore, this study is really more a reflection of how upwelling varies in the model ensemble described by GREP rather than in nature. While coastal upwelling is clearly being captured by GREP, it is undoubtedly a highly distorted view compared to the real world. This should be discussed clearly in the manuscript, at the outset  and in the conclusions.

(2) Two different upwelling indices (UIs) based on the wind were used: one is the standard PFEL product while the other is one that the authors compute based on equation (5). The definition of the PFEL index is not given in the manuscript, so it is not clearly what the relationship is between UI_PFEL and UI_ERA5. Figure 3b shows that they vary consistently over time, so why consider both? Why not just use the accepted UI_PFEL? This needs further discussion and justification.

(3) Figure 3b shows that there is a lag in the SST response and the upwelling indices. This is mentioned in the manuscript, and has been noted by others, but this manuscript sheds no further light on this issue. Studies of coastal upwelling by Marchesiello and Estrade (2010, J. Mar. Res., 68, 37-62) and Jacox et al. (2014, GRL, 41, 3189-3196) have shown that coastal upwelling can be suppressed by onshore geostrophic flow leading to considerably less upwelling than might be expected based on the wind alone. I wonder if this might be the reason why the upwelling peaks later than the wind-based upwelling indices in the Canary Current system. This could perhaps be easily checked using GREP which presumably also contains the near-surface ocean current data.

Specific comments:

(1) Caption for figure 1: More information is needed here - all the acronyms and symbols should to be defined in the caption.

(2) Line 115 and equations (2) and (3): Non-standard notation is used here. I would suggest using u and v instead of W_x and W_y, and U and V instead of Q_x and Q_y.

(3) Equation (3):  Can you comment on the relative role of Ekman divergence at the coast and wind stress curl? It is likely that both contribute in a significant way to upwelling, as they do at other upwelling centres.

(4) Line 124: The upwelling threshold of 1.5 for UI_ERA5 needs some explanation/justification.

(5) Line 127: Please provide a full definition of the PFEL upwelling here (see comment above).

(6) Caption for Fig. 3: Please explain the format of the boxplots in panel (a) (i.e. what do the coloured boxes and various tick marks represent?).

(7) Line 178: What are you referring to here by "trend" - there are no plots presented that indicate a trend.

(8) Lines 187-189: Have you computed lagged correlation coefficients? What about the potential role of onshore geostrophic flow (see comment above)?.

(9) Lines 210-214: The changes in depth for the ILD discussed here seem very small. Can you discuss their significance?  The very low horizontal resolution of the GREP model data sets used must be an important limiting factor here.

(10)  Figure 4: It would help to show the location of these profiles in Fig. 2. Also, this figure is not easy to read.  Is there a better way of demonstrating how T and ρ vary with distance from the coast?

(11) Caption for Fig. 5: I don't think you mean "representative" here. According to the main text these are the  average temperature profiles based on several events that exceed a threshold based on UI_SST - is that correct?

(12) Line 247: I think that you mean "combination" rather than "coupling." The reason why you chose to consider these particular combinations of the NAO (+ -) and EA (+ -) in Fig. 6 should to be explained. 

(13) Line 250: Replace "not present anymore" with "largely absent"

(14) Figure 6: Please indicate more clearly the depths represented by each row of plots.

(15) Line 268: Replace "coupled" with "combined"

(16) Line 272: Rephase "...NAO and EA couplings..." as "...the combined influence of the NAO and EA in the ..."

(17) Lines 296-297: Why is the lag larger in your study? See comment above about possible role of onshore geostrophic flow.

(18) Section 4: It would be useful to discuss clearly what this study adds to the existing literature.