Review of Dolores-Tesillos et al “On the role of moist and dry processes for atmospheric blocking biases in the Euro-Atlantic region in CMIP6”.

This paper studies blocking biases in the wintertime Euro-Atlantic sector in CMIP6 models,. They argue that mean state biases and inadequate representation of moist processes within the models contribute to underestimation of Euro-Atlantic blocking frequency. This paper is well written and executed and the subject matter is consistent with the scope of WCD.  My comments are minor. I enjoyed reading this paper and think that it adds to the literature on this topic. Please find my comments below.

My primary concern is that the results presented are consistent with the hypotheses suggested (blocking biases due to lack of WCBs and overly zonal jets), but that their causal relation is more difficult to establish and will likely require carefully designed model experiments to test. For example, one could imagine that a lack of blocking can contribute to a southward shifted mean jet. That is, correlation does not equal causation. I therefore suggest some discussion of this nuance in the paper. The authors could even suggest experiments to test these ideas.

Minor comments:

Lines 93-99: Is there a reason that only models with an adequate representation of blocking are studied and do you expect these results to generalise to all models? This is quite a small subset of the full CMIP6 ensemble.

Lines 127-134: Is there a reason that meridional gradients of Z are used for the background flow rather than simply zonal wind? Could the authors explain this as surely these will be roughly equivalent under geostrophic balance?

Figure 1 caption – the blocking frequency for the ERA5 climatology seems quite low, only around 3% at most for blocking over northern Europe. I assume that this is a typo and should be contoured every 5% as typical blocking frequencies are more like 10-15% (e.g. Woollings et al 2018, figure 2)?

References

Woollings, T., Barriopedro, D., Methven, J., Son, S.W., Martius, O., Harvey, B., Sillmann, J., Lupo, A.R. and Seneviratne, S., 2018. Blocking and its response to climate change. Current climate change reports, 4, pp.287-300.

Review of egusphere-2024-2878: On the role of moist and dry processes for atmospheric blocking biases in the Euro-Atlantic region in CMIP6 by Edgar Dolores-Tesillos et al..

In the manuscript, the authors investigate the representation of atmospheric blocking in CMIP6 models and explore the role of moist and dry processes in the climate model biases. The authors consider the representation of the background flow, transient wave activity and warm conveyor belts (WCBs) to elucidate the climate model biases relating to both dry and moist processes. WCBs are identified using a machine learning technique as data is not available from the CMIP6 models to calculate the Lagrangian trajectories typically used to identify WCBs. It is shown that there are consistent biases in the climate models that together result in an underestimation of block frequency over Europe: the jet stream is located too far south in the models and extends further downstream; eddy activity peaks further equatorward in the models and act to accelerate the flow in the peak blocking region; and the models have a strong negative bias in WCB outflow across the North Atlantic/European region.

I very much enjoyed reading this paper: it is very well written, well structured and the figures are clear and support the main conclusions drawn by the authors. I believe the manuscript is a nice addition to the literature relating to climate model biases in atmospheric blocking and will be of interest to many readers of WCD. I have a couple of relatively major comment and some smaller comments that I would appreciate the authors response to. After addressing these comments I think the paper will be suitable for publication in WCD.

Major comments:

1. This comment relates to the identification of WCBs in the climate models.

The machine learning technique used to identify WCB inflow/outflow was built/verified using ERA interim, i.e. trained to identify WCB trajectories that have the same characteristics (such as total scent, ascent rate) as those in ERA interim. As I understand, we do not know if the technique can accurately identify WCBs in the climate models. For example, if the WCBs in climate models tend to ascend slightly slower or do not ascend by as much as in ERA interim then these will be missed and contribute to the large negative bias in WCB outflow in the CMIP6 models found here. Of course, this would still be a bias that is related to moist processes in the WCB and will contribute to the blocking biases but is nonetheless a different bias to just a gross underestimation of WCB outflow. I assume it is not possible to verify that the machine learning method identifies similar WCB characteristics to a trajectory tool such as LAGRANTO in climate model output, which would be the best option. So, I do not expect the authors to do this, but some more discussion about the potential caveats of using the machine learning method on the climate models is needed.

2. My second major comment relates to causality statements.

The main argument proposed by the authors is that the climate models have a too-zonal background flow and this drives a too far equatorward maximum in eddy activity which reduces the diffluent flow associated with blocking events in the main blocking region, as well as driving a reduced WCB outflow in the region. Could the background flow biases not be a symptom of too-zonally propagating eddies/cyclones and not necessarily the cause? The jet position may be biased in climate models for a variety of reasons and a southward/zonal bias in eddy-driven jet latitude would result in cyclones being steered across the Atlantic with the same bias. The background flow would then appear to have a southward and zonal bias, as it is made up of the average behaviour of the eddies. Some more discussion relating to this causality is needed I think.    

Minor comments:

1. L10: you write “we define the background flow as the most frequent value of the latitudinal gradient of the geopotential..”. I’m not sure what is meant by this. Do you mean the most frequent value of the maximum latitudinal gradient?
2. L59: some readers may not be familiar with “anomaly blocking indices" or even blocking indices in general, consider adding a brief description here.
3. L75-85: you mention how WCBs/diabatic processes are important for blocks in ERA5 and that this has not been assessed in climate models much yet (partly due to the difficulty in identifying WCBs in them). I wonder if this area has been studied in numerical weather prediction models and whether we should expect similar in climate models based on that. Some discussion around this could be interesting.
4. L94: it might be beneficial to briefly describe how Palmer et al. (2023) do their selection. To save the reader having to look up the paper themselves if they are unfamiliar.
5. L121: do you include all latitudes between 75S and 75N?
6. L138: is this a common way to compute the storm tracks? Does it differ much from approaches that track cyclones and then construct the storm tracks from the cyclone counts?
7. L220-225: how large is the spread in storm track/jet biases among the CMIP6 models? You indicate where the majority agree on sign of the bias but it would also be interesting to see how the magnitude/spatial structures vary among the models.
8. L280: could these Lagrangian WCB trajectories from Joos et al. (2023) be used to verify the machine learning algorithm works okay on climate model data.
9. Fig. 4 caption: what are the relative biases compared to?

Technical correction:

1. L93: sub-set —> subset

Thank you very much for this interesting manuscript. Please find my detailed comments in the pdf.