The manuscript presented by Lézine et al deal with the climate changes experienced in tropical West Africa during the last millennia. The manuscript first makes use of a number of paleo-records in the study area to define two new indices able to quantify the hydrological and vegetation context. The goodness of these paleo records is first evaluated by computing two multi-proxy indices which are validated against instrumental data for the period 1840-present with good results. Then, the authors make use of modelled data (from 850 to 1850) to characterise the precipitation in pre-instrumental period, subsequently discussing the relation between the modelled climate and the observed variability of the paleo-records, offering an interesting discussion and relevant results.

The paper is well written and in general it is clear. Personally, I find this work quite interesting, as it deals with a region still poorly characterised because of the sparsity of instrumental or even proxy records. Therefore, I recommend its publication in Climate of the Past.

I however have some concerns (general comments) that, if addressed, will probably improve the clarity of the manuscript:

The most important is related with the methodology used to homogenize the paleo data in section 2.1. As far as I understand, the method used is based on rescaling each individual paleo record (table 1) to a common 6-level scale. However, the details of this conversion are not explained in the manuscript making it impossible to know how this index is ultimately computed (this is essential at the time of evaluating the goodness of the original data or even to allow reproducibility). In my opinion, the authors should describe a little more the way this rescaling has been performed.

Another question is related to the reason why the authors have limited their study period to 850-1850. I’m not familiarised with the past1000 experiment data but ending in 1850 most probably indicates that the past1000 experiment was conceived to model the pre-industrial era. However, if possible, it will be extremely interesting that the modelled precipitation series were extended to present time. This would allow to compare the model results with the instrumental ASWI (figure 2 of the manuscript) and, providing the result is good, it would add a lot of confidence to the results. Anyway, I would like to stress that I find figure 2 very interesting as, beyond some indirect evidence, the humid period described by the ASWI between 1840-1890 had not be confirmed by independent data up to now.

Apart of these questions, there are some minor aspects that could help to clarify the text (specific comments):

Lines 31-32. The west African monsoon is not only driven by land-sea contrast, but it is also a consequence of the migration of the ITCZ (see for example Gagdil et al. https://doi.org/10.1007/s12040-017-0916-x).

Line 50. I consider that this manuscript is not a “review” but a “research”.

Table 1: Maybe expressing the latitude and longitude in sexagesimal form will be clearer.

Line 97: In my view, the validation performed is not indicating that the methodology is “realistic” but instead, it is testing the similarities between the paleo-data and the instrumental ASWI.

Figure 1. The blue arrows are a little difficult to see where the underlying colour is also blue.

Line 180. I believe that the way the past1000 index is constructed should be more explained.

Figure 4. I’m sure that presenting such amount of series in a single figure is not easy, but it is quite difficult to interpret some of the y-axis scales in this figure. For example (not the only case) in figure 4A “Jikaryia” the axis is scaled by not consecutive values (3, 1.5, 2 , 0.5). Please clarify.

Line 326. Please indicate the methodology used to compute statistical significance.

Line 352. The local term “Heug” could be unknown by readers not familiarised with the climate of this region. Please explain.

In this manuscript, Lezine and colleagues present a synthesis of palaeorecords representative of the WAM covering the period between 850 – 1850 CE. They compare these data with recent simulations over the same period. The paper is interesting, well-written, and summarises the region's state-of-the-art quite nicely. As such, I recommend its publication in Climate of the Past, providing some clarifications of some technical elements supporting the study (see below). Most data seem to be represented with some indices. How these are constructed is rarely described, making reading and interpreting the figures quite difficult. The authors mention a 6 levels scale, but many plots display decimal values (e.g. 2.5 for GeoB9501 in fig. 4). Overall, the manuscript would greatly benefit if all the technical details of this study were better described.

Specific comments

L76: Is this the mean resolution of 100 years? Or the highest time between two consecutive samples should be 100 years?

L78-81: I understand this part perfectly, but I do not like the use of the term ‘degraded’, which is biased towards human perception. Plants or animal species that live in ‘degraded’ (as defined here) environments would probably not call it that way. A more objective description, from most arid to humid conditions, seems more appropriate.

L97-113: I think a description of how the ASWI is calculated is warranted. Not in detail, because it has been published elsewhere, but with sufficient information to avoid checking the Gallego et al. 2015 ref.

L140: Why a reference to ENSO here? It doesn’t seem to be contributing to the rest of the study.

L141: Define the acronym AMV

Fig. 3: This may be a problem with the preprint, but the axes and colour scale labels are difficult to read. This comment also applies to most figures.

L173-175: Replace ‘slightly’ with a measure in distance or degrees. The northward expansion seems to be several degrees south of where it should theoretically go. Then discuss why this is acceptable.

L180: An index is calculated or derived, not performed. What does this index measures? What do you do to the first and last 9 samples when computing the moving average since they do not meet the criterion of 10 samples for the moving average? Are you reducing the length of the record? (No wrong answers here, but the methodology needs more clarity).

L179-189: A better description of how the index is calculated and what it means is warranted here.

Fig. 4: What do the blue shades indicate? And how were they determined? Also, what are the numbers of the y-axes? ASWI values?

Results hydrological records: I am struggling to find commonalities within the groups of hydrological records, except for the Sahel region, where a general trend toward drier conditions seems to be consistently reconstructed. Perhaps the authors could run some statistical analyses to extract the trends shared by the records and limit the impact of the ‘local’ signals.

Results pollen records: How are pollen records summarised to one single curve? They seem to be plotted against – broadly speaking – the same scale as the hydrological records. Did the authors convert every pollen sample, a high-dimension type of data, to one single value by hand? More details are definitely needed here.

In addition, hydrological and pollen records from the same site are, more often than not, quite different.

Fig 6: The signal obtained from the data seems consistent, despite my comments above. Maybe this suggests that their representation in Fig. 4 could be somehow improved.

Another figure similar to Fig. 6 but for the MCA would be pretty important here to see if the changes captured by the data represent the effect of MCA to LIA transition or if it is something else.

Fig. 7A: How are these ‘regional’ curves derived?  [I found the explanation later on lines 406-407. It would still be good to add it to the caption]