We thank the reviewers for their critical readings that served to improve our paper. Below, we reply to their comments and describe our subsequent modifications. Please note that Dr. Maier identified six issues, each of which he numbered. Our replies to these issues are numbered accordingly. With respect to Dr. Marreiros, his comments are included in bold text and are followed by our responses.
 

Andreas Maier’s comments

Concerning the introduction paragraph, we wish to clarify that the aim is not to test the hypothesis that the Garonne River Valley marks a border between two cultural trajectories. Rather, we test the hypothesis that the different cultural trajectories on either side of the Garonne River Valley are related to the exploitation of different environmental conditions (i.e., ecological niches). We have clarified this point and associated initial description of the research approach in order to eliminate confusion (l. 133 to 136).
 

1. As correctly pointed out, no two niches characterized using two sets of occurrences that are geographically differentiated will be identical. However, by taking into account the background areas that are used to estimate those niches, one can reliably evaluate to what degree their similarities or differences depart from what would be expected by chance. In our case, the niches have a strong degree of geographical similarity, but when compared to the distribution of similarity values of a thousand random background models, we observe that the empirical niche estimations are more different from one another than would be expected by chance. The similarity observed, although not significant, is not surprising considering the theory of niche conservatism and the fact that the niches pertain to closely related populations.
    

2. Occurrence data certainly influence resulting niche estimations, and it makes sense to assume that comparisons between niches will be influenced by one occurrence data set being twice the size of the other, especially with respect to predictive architectures that do not extrapolate or only minimally so. This is a potential limitation that is stated at the end of the discussion where we have added a sentence explaining why this possibility cannot be entirely excluded (l. 586 to 590). Taking into account this comment, we have modified slightly our modeling design in order to reduce such a potential bias by removing two feature classes (hinge and threshold), because they are complex response variables. Increasing model complexity while having a low number of occurrences points can result in model overfitting and increase model variability, thus rendering the comparisons less reliable (section 2.3.2). By removing complex variable responses, the variability between models has indeed decreased and therefore our comparisons appear to be even more robust. (Figure 5) The fact that the Pyrenees occurrences are distributed within a smaller area and are more clustered does not necessarily mean that their niche should be smaller. In fact, one could also expect the Pyrenees niche to be broader in some dimensions since this region is characterized by a higher degree of altitudinal variability (generally indirectly related to temperature).
   Concerning the removal of duplicate occurrences, the objective is to eliminate pseudoreplication caused by duplicated signal of the same pixel to the algorithm. If we do not clean duplicate occurrences, we overfit ecological niche models. We did an additional thinning with the objective of reducing potential spatial autocorrelation between occurrences that would result in oversampling in some regions as opposed to others. We thus thinned the dataset so that each occurrence point is separated from another by at least two pixels. The northern middle Gravettian dataset was more impacted by the thinning than the Pyrenees Noaillian dataset because of the clustering of sites in the Northern Aquitaine region (related to research history). (section 2.3.1)
   Finally, the comment that the grid size is too coarse to capture the ecological diversity of the Pyrenees would be relevant if we were reconstructing niches at a more local geographic scale. Obtaining reliable environmental layers at a higher resolution from the kind of climate simulation that we used is not feasible. Downscaling the layers further could increase environmental variability, but at the same time, uncertainty would increase. As thinning results in conserving most geographically extreme records, the risk of missing extreme environmental values (which will determine how broad niches are) is also, somewhat, reduced.
   With respect to the calibration area M, we consider that the hypothesis proposed is not problematic as the region selected could have been accessible to the target populations of interest. The background points required to create Maxent models are drawn from this area (M) and help to measure how similar background environments are those known to have been occupied.
    

3. The fact that the niche is smaller and less broad while associated with a higher faunal diversity is not necessarily counter-intuitive. We found that the Pyrenees Noaillian niche is smaller than the Northern Middle Gravettian niche. Their multidimensional morphology in terms of width and breadth are non-extrapolative with respect to fauna. The niches might overlap with those of multiple other species, but by themselves, niches are not proxies for species richness. We did not attempt to discuss biodiversity patterns with our current data since that is largely beyond the scope of this paper. While the possibility exists that there could be an equal degree of faunal diversity in the north, we unfortunately do not possess data with which to evaluate such diversity―one would need paleontological sites where animal accumulations are not biased by human predation. We agree that this was not explained clearly enough, so we have added a sentence in the discussion paragraph to clarify this point (l. 496 to 497).
    

4. We cannot compare the Noaillian and the Rayssian because they are not defined in a similar manner: the Noaillian means “presence of Noailles burins” whereas the Rayssian means “presence of the Raysse method and eventually a conceptually similar blade reduction method”. Hence, they are not defined as technocomplexes but rather typo-technological “faciès”. This is why we chose to compare two regions that have different cultural histories, in the sense that they do not have the same facies across Greenland Stadial (GS) 5. To make this clearer, we added a note in the introduction explaining what we mean by “faciès” (note 1). The point of comparison that we chose in the discussion is armature type and associated chaînes opératoires. This was not clear in the initial submission, so we have added some text and rephrased some arguments in order to clarify (see l. 461 to 493).
   Concerning the comment, “… the Noaillian and Rayssian toolkit were both equally suited for the conditions in the northern niche”, we do not think that one can unequivocally make this statement because the Noaillian is not well described in the north. Therefore, we do not know whether the armatures associated with Noailles burins in the north are the same as in the Pyrenees. To date, it appears that Noailles burins can be associated with different industries (for example, those with and (almost) without Gravette points). This is paired with the fact that they seem to be used for different activities (see Calvo et al., 2019), which provides a heterogeneous vision of this typological category. Finally, we cannot say that the Raysse method was unsuccessful in the southern niche because it would suggest that it had been present in the south, at least shortly, which is not the case according to known data.
    

5. We cannot make the mechanisms explicit if we do not know what they are - hence the reason for doing the study. Furthermore, this study by itself may not necessarily help us identify specific mechanisms but it certainly provides us a piece of the larger puzzle. We have explicitly defined what we mean by mechanism in the introduction by referring to the definition provided by d’Errico and Banks (2013) (l. 80 to 82). (see also response to João Marreiros’ first comment)
   Concerning the other questions:
   a) In fact, the use of the northern area represents a broader niche compared to the south, but it does not imply any chronological relationship between the Noaillian and the Rayssian. We are not saying that the southern niche is the Noaillian niche, because the “Noaillian” is related to the presence of Noailles burins, so it is not a species and it does not have specific ecological requirements. It is more the niche of the groups using this kind of tool, which in the Pyrenees seems to be associated with other behaviors such as diverse backed blade and bladelet armatures, inferred high logistical mobility etc. The northern niche is the niche of groups having used both the Noaillian and the Rayssian across GS5. This northern niche is broader. The hypothesis that we propose to explain this difference is that the Raysse method was more successful in this context, probably because it was more suited to high mobility within geographically broad subsistence/settlement territories, which would lead to a broader ecological niche.
   b) The statement that the northern niche is significantly broader in environmental dimensions is based on the results of the comparison between the NicheA models. The northern niche is significantly broader because it captures a wider range of colder and drier conditions as can be seen in the figure. Moreover, the fact that there are less prey species at an archaeological site does not necessarily mean that there were fewer species present on the landscape; it could also be the result of these populations’ focusing on one species more than others.
   c) This sentence in the text was indeed confusing: we did not mean that reindeer were less predictable, but that other resources, such as good quality flint, could be less predictable or less available, hence a higher degree of curation in some assemblages located away from good raw material sources (Grotte du Renne). We have made this more explicit in the text (l. 621 and 624-625).

6. We agree that such a pattern would be interesting to explore, however addressing it would be beyond the scope of this paper, and we do not think that it is appropriate to speculate here about this phenomenon observed during later time periods. Furthermore, to examine the potential role of the sables des Landes desert and the Garonne valley in cultural geography would require analyses of geological data and considerations of how best to incorporate them into the approach that we employ―work which is well beyond this study’s objectives.
    

João Marreiros’ comments

Why is the authors advocate so? Do the authors mean by “diversity” the lithic techno-typological variability? Is the lithic variability the main characteristic that makes this a perfect case study?
We agree that the word “diversity” was not sufficiently precise in this context. We have replaced it with “typo-technological variability” (l. 31), which better describes what we mean. We think it is clear, following our descriptions of these two “faciès” and the approach that we employ, why this Middle Gravettian record, with its geographic variability and potential ecological variability, represents an excellent case study for identifying some of the factors implicated in this diversity. Identifying factors is a necessary first step if we are to identify mechanisms. (see also 5th response to Andreas Maier)
 

Introduction. “These two typo-technological facies appear to differ both chronologically and geographically.” According to the state-of-the art presented by the authors, this seems not to be the case. Both industries are overlapping chronologically and geographically. In fact, this is one of the aspects considered later in the results and discussion section. As an example, the poor absolute chronological dating of some of the sites is problematic, as discussed by the authors, and pointed as one of the main limitations of the study.
We agree that this statement seems to be in contradiction with the rest of the paragraph, and in the end, it is not necessary. We therefore removed it and broke the paragraph into two parts: one explaining the chronological aspect, and the other focusing on the geographic distribution (l. 98 to 130).  

“Although this situation is complicated by the fact that Noailles burins and the Raysse method are frequently found together within archaeological assemblages.” Is this really the case? This seems to be contradictory to what the authors advocate earlier, especially on what concerns the chronological and geographical distributions of these industries. Perhaps the issue here concerns the term “found together”. What is it meant here? Same site, same archaeological horizon?
We disagree that this contradicts the rest of the paragraph. We state that the Noaillian is usually described as being older than the Rayssian in stratified contexts (Pataud, Flageolet) and that they have partially overlapping territories (meaning that in some regions, there is no Rayssian and in others, there is no Noaillian). We have added text in the introduction to clarify the statement about the fact that they are “found together” in archaeological assemblages. In fact, they are often described as being associated within the same archaeological layer. At present, however, these associations cannot be considered to be cultural in nature, since studies evaluating their stratigraphic and historical contexts are sorely lacking. In fact, at many sites, they appear to be more the result of imprecise excavation methods or site formation and post-depositional processes (l. 102 to 106 and 109-110).  

“The application of Ecological Niche Modeling (ENM) methods to the archaeological record is one way to test this hypothesis”. In the previous paragraph the authors explain the key hypothesis that have being explored and explain the observed lithic variability. But, in this sentence is not clear to each of the hypothesis the authors are referring to.
We have moved and rephrased the research question to make the paragraph clearer (section 1.2).  

Materials and methods “Taking into account these potential limitations”. I think this is an overstatement. These are clear limitations which are not related to the quality or importance of this study presented here. In fact, these limitations are the reflex of the archaeological record, excavation methods, and study protocols used in the past. And these, independent of the methods applied here, could indeed constraint general interpretations.
We agree and removed the word “potential”.  

Discussion “The fact that the northern Middle Gravettian niche is significantly broader than that of the Pyrenees Noaillian suggests that the development of the Raysse method may be linked to the exploitation of a significantly expanded niche composed of colder, drier conditions and thus more open landscapes and associated large mammal prey species. Available archaeological data from the archaeological record support this hypothesis.” I agree with the authors, but I think here the archaeological support for this argument needs to be more elaborated.
We have added a paragraph about the Raysse method explaining why we think this method is well-adapted to the exploitation of large territories (l. 461 to 479 and also 509 to 521).  

Conclusion “These niche results further support the hypothesis that the Landes cold desert and Garonne River Valley corridor served to limit cultural interactions between the Pyrenees and regions to the north.” I think this statement should be taken with a “pinch of salt”. The study shows different technological adaptations to different ecological settings, but this does not necessarily represent cultural interaction, or in this case, the lack of it. From my perspective wither the authors explain what is meant by “cultural” processes in this context, or they should explain this scenario from a technological approach. We have clarified what we mean by “cultural interaction”, because it is correct that the term was too vague. We are referring to the homogeneization of technological traditions [i.e., a hypothetical (and not observed) adoption of the Raysse method by populations in the Pyrenees].  

N.B.: we made some minor modifications concerning the English language after one final English proofreading made by one of us (WB). We also slightly modified some figures so as to make them clearer (Lines connecting sites with site number on Figure 2; background similarity and identity test legends were made changed from “Overlap” to “Similarity/Identity” to avoid vocabulary confusion on Figure 5; NicheA ellipsoids colors were changed to make them more visible on Figure 6). We also added a reference: Delvigne et al., 2020, since it was cited as “in prep.” (note 6) in the previous version and has just been published.  

References

Calvo, A., Perales, U., García-Rojas, M., Normand, C., & Arrizabalaga, A. (2019). Just before sewing needles. A functional hypothesis for Gravettian Noailles-type burins from Isturitz cave (Basque Country, southwestern France). Journal of Archaeological Science: Reports, 25, 420–432. https://doi.org/10.1016/j.jasrep.2019.04.022
 

Delvigne, V., Fernandes, P., Tuffery, C., Raynal, J.-P., & Klaric, L. (2020). Taphonomic methods and a database to establish the origin of sedimentary silicified rocks from the Middle-recent Gravettian open-air site of La Picardie (Indre-et-Loire, France). Journal of Archaeological Science: Reports, 32, 102369. https://doi.org/10.1016/j.jasrep.2020.102369