The manuscript is well written and easy to follow. However, I failed to understand the novelty of this study compared to that of Asbjornsen and Arthur (2023). Both use a decomposition of the AMOC to assess its decline under the SSP5-8.5 climate change scenario at 26.5°N using a similar number of CMIP6 models. Asbjorsen and Arthur (2023) do not use RAPID observations, but come to similar conclusions anyway. I do not doubt that both studies were done independently, but a clearer justification of what is new in this study and how its results and methods differ from Asbjornsen and Arthur (2023) is needed before recommending it for publication.

Possible typographical errors:

Line 77 and Table 2: Rapid instead of RAPID

Line 201: historrical instead of historical

The AMOC strength has been observed by the RAPID array in the subtropical Atlantic since 2004. These measurements have greatly increased our understanding of AMOC dynamics and variability. They are also a critical comparator for ocean and climate models. Typically, comparisons between observations and models are of basic metrics such as the maximum and variability of the overturning streamfunction and its vertical distribution. But the observations allow far more mechanistic comparisons to the models and that is what is done here for the latest group of CMIP6 models. This paper identifies how the circulation in observations may be compared to the that in models accounting for the fact models and the real world have different zonal structures due to model resolution particularly at the Florida Straits.

The AMOC metric is a zonally integrated view of the circulation. In this paper key mechanisms are identified: Florida Straits transport, Antilles Current and several key water mass layers. These are all mechanistically interesting being relatable to both wind forcing and thermohaline forcing, particularly the wind forcing of course.

For the period 2004 to 2014 the observations and models are compared. Based on insights gained from this comparison model trends 2015-2100 are then considered and discussed.

Some important results emerge:

1. The observational community needs to refine its data products to allow better comparison with the models. Specifically, careful consideration needs to be given to how the thermocline wind-driven circulation and the western boundary current (Florida Current and Antilles Current) are separated. This is because the coarse model resolutions do not resolve the narrow Florida Straits and their western boundary currents are the sum of Florida and Antilles Currents.
2. The observations and models when carefully compared in the period 2004 to 2014 have similar mean transports overall. There are a number of caveats relating to the depth structure of the North Atlantic deep-water layers, which are considered a result of model inadequacies in representing exchanges across the Greenland-Scotland Ridge.
3. The slowing of the AMOC in the 21^stcentury is quantified as a 30% reduction in the mean. Florida and Antilles current transport reduces by 11Sv. The wind-driven Sverdrup transport in the thermocline reduces by 4.4Sv so the wind-driven upper ocean circulation reduces by 6.6Sv. This reduction is driven by a 14% reduction of the wind-stress curl in the CMIP models. The additional necessary southward reduction in transport to achieve a mass balance across the section is 6.4Sv is in the NADW layer (in the model the reduction is mainly in upper North Atlantic Deep Water, whereas the reduction seen in observations in the early part of the 21^stcentury occurs in the lower North Atlantic Deep Water).

This rather nice third result diagnosing the buoyancy forced reduction in AMOC is discussed with reference to Baker et al (2023). Baker show that in CMIP models this reduction is supported by reductions in the Indo-Pacific upwelling of North Atlantic Deep Water.

While the AMOC functioning is important it is perhaps more important to additionally understand how heat transport is affected by the changing wind and buoyancy driven circulations. This is briefly mentioned in the discussion and suggests another detailed model observation comparison study on this would be valuable.

This is a nice, short, straightforward paper that has some long-needed analysis and hopefully points a fruitful way forward for more detailed comparisons between AMOC observations and CMIP models. The attempt to separate wind and buoyancy forced changes is nice.

Minor Comments

Lines 95-120: I think these definitions of water masses only apply to the model and not to the observations? It is not clear.

Line 95-102: What is the typical longitude of the eastern edge of the Antilles Current?

Line 104-105: Why was a temperature boundary chosen – is this to do with being able to compare the correct water masses between observations and models? Why 8°C which is 900m deep in the west and 800m deep in the east? The 7°C isotherm looks flatter and at ~950m deep would have more thermocline flow?

Overall a few more explanatory lines justifying the choices would be helpful.

1. Results

Table 2 is quite confusing, and it needs a much better caption to explain it and also state the units of Sv. Also note use of upper case in Table that is referenced by lower case in the text. 1. The Antilles Current transport of 5Sv is a result from Meinen et al. 2019 and should be referenced. Note also that the Meinen result for the Antilles Transport is ±4.7±7.5Sv not 5±10Sv as stated in the results; 2. Western Boundary Current (FS+AC) does equal 31.3+5=36.3Sv. But Thermocline Recirculation+AC -18.6+5=-13.6Sv does not equal -23.6Sv as in paper. The number you want in the table is the -23.6Sv but you need to read the results section to carefully understand your argument for separating the thermocline and Antilles transports. It needs an explanatory note in the caption or you need to write Recirculation-AC; The line Western Boundary Current+Ekman+Model TR should be written as AMOC= Western Boundary Current+Ekman+Model TR so it is obvious it can be compared directly to AMOC=FS+Ekman+IW+TR for the AMOC northward upper limb. The Deep Water part of the table is much simpler especially as the AMOC deep limb appears on the same line!

A recent paper discusses the Gulf Stream weakening which the authors might reference (Piecuch, C. G. & Beal, L. M. Robust Weakening of the Gulf Stream During the Past Four Decades Observed in the Florida Straits. Geophysical Research Letters 50 (2023). https://doi.org:10.1029/2023gl105170)

Review of Bryden et al. “Comparing observed and modelled components of the AMOC at 26ºN”

This paper presents a comparison of CMIP6 model output vs RAPID observations, with transports decomposed similar to RAPID. I found the paper easy to follow, had some nice results. It’s quite a short paper for OS but I appreciate it came about in an unusual way with the MSc student not pursueing it. 

The Figure 2 is very informative, especially l182–183: this is a very nice finding. Very striking WBC decline and the decomposition of this into wind-driven components is very interesting (l212). I think one more figure should be added: a figure that simplifies Fig 2 to show MOC decline broken into wind-driven and non-wind components. 

I confess I haven’t read Asbjornsen and Arthun but their abstracts sounds very similar to your paper. Can you clarify the differences and unique insights from this study?

Overall, I think it’s an interesting study that can be published following minor corrections. 

Minor comments: 

l36: add a line on importance of the global overturning circulation

l48: the latest RAPID data are available to 2022 at time of writing. Please explain why your analyses ends in 2018 and any issues that might arise because of this shortening. 

l52. If you’re not going to state the mean transport in the climate models, then you need to report the decline differently i.e. not a percentage. Weijer also reports the decline as 1 Sv/decade (I think). 

l64. I think you can bring in findings from a few more papers than just Weijer here: 

    Robson, Jon, et al. "The role of anthropogenic aerosol forcing in the 1850–1985 strengthening of the AMOC in CMIP6 historical simulations." Journal of Climate 35.20 (2022): 6843-6863.

    McCarthy, Gerard D., and Levke Caesar. "Can we trust projections of AMOC weakening based on climate models that cannot reproduce the past?." Philosophical Transactions of the Royal Society A 381.2262 (2023): 20220193.

l83. Why monthly?

l87. How was the one ensemble member chosen? Was it the first one?

l91. Can you show the range of the net transport through the section? There is surely quite a range around -1Sv.

l104–118. Why use isotherms rather than depth as RAPID does?

l120. Is Ekman recalculated in the models from their winds? Or taken as very near surface ageostrophic transport?

Figure 1: I’m surprised at the small spread of the climate models mean AMOC. Was there any selection of the models based on mean AMOC strength?