Overview:

The paper quantifies the meteorological and synoptic conditions during the filling of a lake in the northwestern Sahara using reanalysis data and satellite images. The paper describes the synoptic conditions leading to the lake filling and shows that cyclones and Sahara moisture recycling are important prerequisites for a lake filling event. They propose a shift in the understanding of the mid-Holocene green Sahara phenomenon, emphasizing the frequency and intensity of events rather than a large-scale northward shift of the rain belts.

I found the paper very interesting and innovative, and one that will have an important impact on the thinking of the greening of the Sahara. I think the paper is suitable for publication once a few minor, mostly structural and clarification comments, are addressed.

Minor comments:

I found the structure of the results somewhat confusing. The paper would benefit greatly, if there is a better separation between the observations and interpretations and a clearer explanation of what the main points of the figures are. For example, in fig 7, instead of “Overview of synoptic-scale conditions during HPE5.1” use “An example of the formation of a southward moving deep cyclone during an HPE event”. Or fig 5, instead of “Overview of..” use: “Duration, intensity and runoff coefficients of HPE and LFE events”. And perhaps add commentary to what the reader should understand from the caption “showing an agreement of intensity and duration between ERA5 and IMERG data” etc. This is true for all figure captions. Please help the reader understand what the main point of the figure is.  

In the results, evolution of HPE5 is given as an example for all HPE’s. I think this is reasonable for the synoptic evolution. However, using the HPE5 as an example for the moisture sources isn’t clear and the discussion about the proposed “domino effect” is misplaced. The discussion of the “domino effect” should be linked with fig 13, where it is clearest. In the way it is written, the mechanism is proposed before the full results are presented, which makes the claim weaker during the reading and stronger only after the full results are presented. I suggest moving section 4.2.2 to the discussion. In this way, you can lay out the full argument for the “domino effect” in one place, with all the key observations required to understand the mechanism already laid out.

Duration of the events seems to be the key component of the LFE, if the system persists for long enough there will be a LFE (Fig. 5a shows this clearly). This aspect of the system was not discussed much. What causes the system to persist for 5 days over 1 day?

Runoff coefficient is a very important aspect of the LFE’s (as explained by the authors). However, it is unclear what generates the large Runoff coefficients. It would be beneficial to discuss this, even if there isn’t a concreate answer to this problem.

Line comments:

L11. Should be: eventuated

L116. Should be: spelled

L116. Add coordinates

Figure 1. Add location of cities for orientation: Bechar, Kerzaz, El Menia, etc.  

Figure 1b. From the decrepitation and figure the properties of the lake are unclear. There is no need for the inset on the left, it is redundant with fig. 1a. Would be better to remove it and add a bigger image of the sebkha itself, and perhaps an image showing a time the lake was dry/drier, and a topographic transect showing the water level.

L123. How deep does the water get? How deep does the water need to be for the lake to overflow?

L126. How do you know what the geology is? You didn’t cite a geological map.

L128. Please add the location of these dams to figure 1, otherwise, these location names are meaningless. Do you take these dames into account when calculating the size of the catchment?

L147. Should be e.g., precipitation and evaporation? Evaporation of precipitation?

L215. Why not use model estimates of evaporation? e.g., the method used by: Zhou et al., NC (https://doi.org/10.1038/s41467-022-31125-6).

L263. Where does the actual dq data come from? AR5? IMERG?

L320. 100 mm where? In how much time?

L364. Should be: predominantly

L375. It’s not clear to me why you invoke this domino mechanism. If there is subsidence which prevents uplifting of the moisture, can’t the moisture be caught in mid-levels and be pushed eastwards, which wouldn’t require precipitation and evaporation? and the long residence time is because the moisture never reaches condensation levels?

It is also not clear why this is a prerequisite for the HPE event? If the source of the moisture is in the Atlantic and all that is happening is recycling within the troposphere, why do you suddenly get a downpour? I think the mechanism and its link to the HPE needs to be explained more clearly.   

L384 – 389 and Figure 8. At what height are you measuring RH? It is interesting the RH increases only when rainfall reaches the ground 22/11 at 3:00. If there is recycling of moisture before this, shouldn’t RH go up already during the enhanced moisture input?

Figure 10. What are the units of the x axis in a and b? If this is time, it should be noted in the figure or caption. Also, I’m not sure how this figure helps with depicting the domino effect (from lines 390-393).

Figure 13. This figure is essential to your arguments of the “domino effect”, I think it should be presented earlier in the text. Is this based on the MSD analysis? Please explain how this data is constructed.

L482. What part of the tropics? How is this moisture getting to the catchment? How does it relate with the moisture coming in from the Atlantic? This statement seems to be new and requires a bit more clarification.

L608 – 614. I think the emphasis here is not precise. If during the mid-Holocene the lakes are higher, or more persistent, there are more LFE’s and most likely an increase in the multiyear rainfall mean. So, the distinction you are making is not about the mean rainfall but about the type of system that causes the LFE. I think what you would like to say and are not saying clearly enough, is that given the modern conditions, a large scale northward shift of the African monsoon during the mid Holocene is not necessary. Rather, a higher tendency of westward cyclones could produce the same paleo archive. The question you should be raising is: given that models find it hard to move the monsoon this far north under the moderate forcing of the mid-Holocene, is it easier to produce more frequent cyclones under these conditions? If so, this would reduce the mismatch between paleo data and models.  

The paper "Meteorological ingredients of heavy precipitation and subsequent lake filling episodes in the northwestern Sahara" by Rieder et al. uses satellite observations and ERA5 output to to identify lake filling events in the northern Sahara and characterize the large-scale meteorological conditions associated with them. The manuscript is well written and addresses the important topic of heavy precipitation events in a complex part of the world and has interesting implications for paleoclimate. However, clarification of some key points are needed, so I recommend this paper be accepted with revisions.

You've shown very clearly that the high precipitation events (HPEs) you've identified are associated with large-scale circulation patterns and moisture convergence, which makes sense. Are there conditions where the pattern of extratropical cyclones and and upper level PV anomalies *don't* result in these events? How frequently do these patterns coincide? It would be interesting to see how what composites of these large-scale patterns that do not produce HPEs look like: if they don't result in moisture convergence this would be a very strong support for your claim. Or if you showed cases of the low-level cyclone without the PV anomaly, that might show moisture transport but not the ascent. This is probably beyond the scope of your paper, but should be considered. 

I don't fully understand why this moisture recycling "domino-effect" is required for these HPEs. I think this is an interesting hypothesis which should be explored further, but there isn't fully evidence for it. I would suggest moving this to a discussion instead of continually emphasizing it as something that is definitely happening. Could this possibly be due to choice of reanalysis? I always worry when I see a study that relies only on a single reanalysis, especially in regions that have sparse observations. If you used MERRA2, which might represent below cloud evaporation differently than ERA5, would you see the same regions of moisture supply? 

One part of your methodology that confused me was the use of both LANDSAT and MODIS. Why don't you just use your algorithm on MODIS, instead of just using it for visual confirmation of these events? Some clarification on how you are using both of these and why would be greatly appreciated.