Review of “Low confidence in multi-decadal trends of wind-driven upwelling across the Benguela upwelling system due to internal climate variability” by Bordbar et al.

The present study analyses the coastal upwelling regime at the eastern boundary of the South Atlantic. The focus is on multi-decadal variability of the wind forced total upwelling composed of coastal upwelling due to alongshore winds and wind-stress-curl-driven upwelling. The manuscript analyses a specific reanalysis dataset of the global climate, ERA5. This dataset is somehow validated against few other available datasets. Total upwelling and wind-curl-driven upwelling is found to be strongly related to the strength and position of the South Atlantic Anticyclone (SAA). Long-term trends identified in the ERA5 dataset are mostly weak, showing considerable heterogeneity, and have only a small signal-to-noise ratio due to enhanced interannual climate variability. My major concerns are:

1) ERA5 is a reanalysis dataset that includes information about uncertainties for all variables. I think it would be necessary to evaluate this uncertainty with regard to possible trends. Particularly trends in the wind forcing are still highly uncertain and a clear description what are the uncertainties provided with the dataset is necessary.

2) There is a validation of the ERA5 dataset with respect to the satellite-derived ASCAT wind dataset and to sea level pressure in St. Helena. This validation focuses on short-term or seasonal variability. However, to strengthen to topic of the paper it would be important to look at the comparison of multi-decadal variability. How good do ERA5 represent long-term changes identified in ASCAT or St. Helena sea level pressure?

3) Inherent to the discussion in the manuscript is that long-term changes are associated with human-induced global warming and shorter-term (interannual to decadal) variability to internal variations of the climate system. This is an assumption that cannot be proven with the current dataset. I would suggest to more carefully discuss multi-decadal, decadal and interannual variability pointing to possible mechanisms that could be associated with internal climate variability or global warming.

This is an overall well-written manuscript that contribute to our understanding of the wind-driven upwelling, its variability and its long-term changes. It might be acceptable after revising the raised major and specific point.

Specific points, comments and suggestions:

Abstract: In the abstract should be mentioned that the study is based on ERA5 reanalysis.

L11: "cross-shore integral of wind-driven coastal upwelling": What is this? Do you mean coastal upwelling driven by alongshore winds?  The different upwelling terms must be clearly defined and understandable. Also L15: “integrated wind stress curl-driven and total upwelling”: again term is not well defined and there should be a simpler way to introduce the three terms: coastal upwelling driven by alongshore winds, wind stress curl driven upwelling in the near coastal band and total upwelling as the sum of both.

L16: “more signatures” is unclear

L20: Sentence “However, …” implies that internal climate variability exists only on shorter timescales and that trends are necessarily due to other forcing. Would it be better to write “… can be obscured by interannual to decadal climate variability”?

L28: These modes result …

L28: “nonlinear climate dynamics” Why nonlinear? There are many linear climate feedbacks. Maybe: coupled ocean-atmosphere dynamics.

L31: at the eastern margin

L37: “Farther offshore, …” brackets are not logical: divergence is associated with upward, convergence with downward velocity. I would suggest to remove the terms in the brackets.

L48: “Based on …”. There is a rich literature on the local and remote forcing of Benguela Niños and Niñas. This sentence do not reflect our current understanding of these anomalous coastal warm and cold events.

L55: “equatorward part of the BUS” requires a definition, e.g. by a latitude range.

L58: “poleward portion”, same as before. Please provide a definition of the region.

L59-61: At least there is some uncertaintiy about the role of southward migration of the SAA. The southward movement is also suggested to change the upwelling in the northern portion of the nBUS (Jarre et al. 2015).

L106: In the Supplementary material the seasonal cycle etc. is analyzed. Is there any support from independent data that ERA5 reliably describe trends in the region? As the paper is dealing with long-term variability, the provided validation is not that meaningful. Please provide an additional comparison of trends and decadal variability between ERA5, ASCAT and SLP. It is well known that such long-term changes are particular uncertain and this study could contribute to assess these uncertainties.

L118: volumes transport per unit length (m²/s)

L140: remove comma after coast

L149-150: unclear: upwelled (or downwelled) … transported offshore (or onshore) ?

L151: maybe include: i.e., where the meridional wind stress is maximum

L154: I do not understand what the authors want to say. What is the reason for the simplifications? How good are the made assumptions? How large is the error, e.g., neglecting the onshore geostrophic transport? It would be good to give some more insight.

Table 1: “alongshore-driven” should be “alongshore-wind-driven”

L216: Last sentence is not necessary and distract.

L231: It would be good to provide separate correlations for correlations on interannual and decadal timescales. Which timescale dominate the correlation?

Fig. 2a-c: Please show the complete functions for positive and negative correlations and provide the significance.

L280: (used boxes are marked in Fig. 2e)

L292: “For example, the years 1997 and 2006 are characterized by persistent eastward SAA displacements”: I cannot see this behavior in Fig. 3b.

L306: “changes in the global heat budget”: Do you mean global warming or net radiative imbalance at the top of the atmosphere? Why are 1990s important the increase started earlier.

L309: “due to the enhanced radiative forcing”: what is meant with radiative forcing? of the ocean? Do you mean global warming?

L313: “positive trend is more prominent over higher latitudes”: Is this associated with poleward migration of the center of the SAA?

L365: “Based on equation 5, …” In equation 5, upwelling velocity depend on x. Do you mean maximum upwelling velocity calculated using R1 given in Fig. S5?

… corresponds to a maximum upwelling velocity …

L397: the spatial resolution

L408-413: this section must be improved: possible there are multidecadal changes or trends and there are interannual to decadal variability. It is not obvious from the analysed data what is due to global warming or what is internal climate variability. Used terms such as “historical changes” or “historical trends” are unclear. These conclusions must be formulated more carefully.

Review of "Low confidence in multi-decadal trends of wind-driven upwelling across the Benguela upwelling system due to internal climate variability" by M.H. Bordbar, V. Mohrholz, and M. Schmidt; submitted to Earth System Dynamics.

The paper discusses the relationship between long-term trends in the position and strength of the South Atlantic Anticyclone (SAA) and upwelling dynamics. The author uses ERA5 reanalyses and focuses on different aspects of the upwelling system, namely the windstress-driven and the curl-driven upwelling. Overall, the paper is well presented and well written. In particular, the methodology is clear and seems promising to disentangle different aspects of the complex and multiscale nature of an Eastern Bounduary Upwelling System (EBUS), i.e. from large-scale wind forcing to local-scale modifications of the wind by orography and/or SST/wind interactions.

In particular, the authors show that a robust change in the position and strength of the SAA together affects both curl-driven and wind-driven upwelling, with a pronounced increase in the curl-driven mechanism. However, this robust change cannot be attributed to Bakun's theory because it is difficult to attribute the SAA change to climate mode variability or to the difference in warming between the land and ocean. Strong spatial heterogeneity of a weak long-term trend in ERA5 is shown, making any strong conclusion difficult. However, the limitations of the study and the use of ERA5 are well presented in the last sections of the paper.                                                      

The paper could be suitable for publication after discussion of the following aspects: 

(1) ERA5 is a medium-resolution reanalysis with a relatively short time span. It remains difficult to relate anthropogenic global warming to the long-term trends present in the data set. Also, with an analysis period of 40 years, it is complicated to disentangle the anthropogenic signal from climate mode and decadal variability. I remain convinced that a lot can be done and said with ERA5, which remains a very useful dataset, albeit limited. For example, one can ask whether there is a signal in ERA5 for differential warming over land and ocean. If so, can this signal explain the long-term trend in SSA strength and position? How are climate modes represented in the ERA5 dataset, and what are the consequences for SAA position and strength? The paper gives some hints, but would benefit from a clearer and more thorough discussion of these two aspects. Please note that this comment does not necessarily require important new diagnostics.

(2) Several mechanisms are responsible for curl-driven cross-shore upwelling. (i) Offshore, winds blowing along a permanent or semi-permanent front adjust through the so-called downward momentum mixing mechanism, with a decrease (increase) of the surface wind over the cold (warm) flank of the front (see e.g. Chelton et al., 2004). Even with the relatively low resolution of the dataset, this mechanism is important in the dynamics of the ERA5 surface winds, especially in the EBUS where near-neutral conditions prevail. (ii) Near the coast, the coastal wind dynamics are mainly driven by the differential stress between the ocean and the continent and the role of the coastal orography. This mechanism usually implies a dropoff zone (as mentioned in the paper) and possibly a cape effect.

Both points mentioned here are driven by important structures in the curl that favor upwelling and are hidden in the Wcurl calculated in the paper.

The paper would benefit from a fuller discussion of these aspects. The orography-induced wind drop is undoubtedly not well characterized in ERA5 due to poor resolution (coastal points also suffer from the so-called Gibbs phenomena). The coastal orography varies

from the southern seasonal upwelling to the more permanent northern counterpart; inducing a different permanent wind curl signal at the coast.

at the coast. The wind/SST interaction is an important factor for the Ekman pumping in the BUS. Even though it is a coupled mechanism, some insight into the SST forcing used in ERA5 would be nice to discuss (possible cold biases; structures of the fronts, etc.). The paper, and with legitimate argument, examines only atmospheric variables, but the quality of the SST forcing seems crucial to discuss, especially with the role of the front in shaping the Ekman pumping. 

(3) The morphology of the continental shelf is important when comparing WSD and WSCD upwelling. As Marchesiello and    

Estrade (2010), the competition between reducing coastal upwelling and increasing Ekman pumping can be assessed by comparing the length scales of coastal upwelling (Lu) and orographically induced wind drop-off (Ld). If the two length scales are equal, there is no effect on the total upwelling water in the cross-shore direction. The continental shelf of the area, especially around the upwelling cells mentioned, could be discussed and drawn on one of the maps.