Many thanks to the authors for their thoughtful replies to my comments and those of the reviewers. I am perfectly satisfied with the new analyses, and in particular I find the results much clearer and easier to interpret. The two reviewers also agree that the manuscript has substantially improved and only one of them (Scott Wilson) has still some minor remarks that I recommend the authors take into consideration.

From my side, I also have a few remarks that can be addressed through a minor revision.

The main one is that I think there is a margin for improvement in the discussion. As context to my comments: the study design is one where for three bird guilds (9 forest edge species; 3 grassland species; 2 reedbed species), the authors investigate how the abundance of species is affected by the area of a type of semi-natural habitat considered to be their primary habitat (respectively: woodlands; grasslands; wetlands) and by the area of a type of field margin that may correspond to a substitute habitat (respectively: hedgerows; grass strips; reed strips) as a way of testing whether field margins can act as suitable substitute habitats. The expectation was thus that the results should confirm the value of the semi-natural areas as primary habitat for each of the species; and then the relative value of the respective type of field margin in relation to the semi-natural areas would let us know if it is or not an adequate substitute habitat, thus allowing for a test of the habitat compensation hypothesis.

The results are (inevitably) not as simple. Regarding the forest edge species, only 1/9 responds positively to woodland area (great tit), and only 1/9 (greenfinch) responds positively to the area of hedgerows. Regarding grassland species, only 1/3 (corn bunting) responds positively to grassland area, and it responds even more strongly to grass strips. Regarding wetland species, only 1/2 (reed warbler) responds positively to wetland area, and it responds even more strongly to reedbed strips. What these results are telling me is that this turned out not to be such a good model system for testing the habitat compensation hypothesis after all, because for most of the species analyzed (11/14) there is no evidence that the areas of semi-natural habitat studied correspond indeed to areas of “primary habitat”. For the 3/14 species for which there is such evidence: for one (the great tit) there is positive but weak evidence that field margins are a substitute habitat (positive but no significant effect; for two of them, the substitute habitat turns out to be even better than the primary, so arguably it is not so much a “substitute” but a preferred habitat in its own right. 

This is the beauty of studies based on real data, and the fact the results did not go always in the expected direction does not detract from the quality of the study. Nonetheless, the discussion needs to acknowledge it more explicitly. Currently the discussion focuses strongly on the specifics of the three individual species for which the result shows that they prefer field margins, and on the adverse effects of some types of field margins to some of the species. This is all very interesting, but what is missing at the beginning of the discussion is a broad level analysis of what the results tell us (or cannot tell us) specifically regarding the habitat compensation hypothesis, which was the main question of the study (as reflected in the title), including perhaps a discussion of how the hypothesis could be better tested (with a different study design?)

I also find that currently the conclusions and the abstract are not well supported by the results (abstract lines 41-48: “Our study confirms that bird guilds are favored by the area of their primary habitat but are also influenced by the area of field margins. Reedbed birds are favored by the area of wetlands and reed strips and are negatively impacted by grassland cover. Grassland birds are favored by grassland and wetland areas and negatively impacted by woodland and hedgerow areas. Finally, forest edge birds are favored by hedgerows and negatively impacted by reed strips. These results suggest that field margins may represent substitute habitats for some bird species and highlight their importance for biodiversity conservation in wetland agricultural landscapes”; discussion lines 375-376: “In conclusion, our results highlight that field margins are valuable landscape components to improve biodiversity conservation while keeping a sufficient area dedicated to food production in rice paddy landscapes”). Indeed, in my interpretation the study does not add strong evidence to the importance of field margins as either a substitute habitat (the main question of the study; only weakly supported for 1/14 species tested – great tit) or as a habitat (valuable landscapes) in its own right (only for 2/14 species tested – corn bunting and reed bunting). 

Other minor points:  

-          Line 64: I recommend to add “focus on maintaining and increasing the capacity” 

-          Lines 80-81: recommend being more precise as “the habitat compensation hypothesis has been investigated in the context of farmland abandonment and in dry agricultural areas”

-          Lines 106-108: this sentence is ambiguous; clarify what the “the greatest rate of decline” means (in relation to what?)

-          Line 128: A reference from 1994 is not sufficient to support the statement that “the area… is now stable”

-          Line 135: “fields” rather than “field”

-          Lines 154-155: when you say that the area of each type of field margin and semi-natural habitat was measured “within a 500 meter buffer around the centroid of each crop field”, it would be useful to clarify that the fields are much smaller than this (so that the field margins, where the bird counts were made are well within this buffer). You currently provide the size of the fields in the results (line 237), but it may make more sense to give earlier (in line 135, or in the legend of Figure 1). Also, in the abstract (line 37), saying “a 500 m buffer around each sampled crop” gives the impression it is a 500 m buffer around the boundaries of each field; please clarify it is around the centroid of the field.

-          Line 190: recommend “birds that use urban areas” rather than “birds using” (the latter is a bit ambiguous in that it could suggest you assessed if birds were using urban areas as part of your study; the former makes it clearer that it is an information obtained from other sources)

This manuscript has been revised by two reviewers, both of which consider this work interesting and a valuable and topical contribution to the literature, and I fully agree with them. There is however in my view a margin for improvement in terms of clarity of presentation of the context of the study, and also in the statistical analyses. I add below some comments and recommendations, which are complementary to those by the two reviewers. 

I encourage the authors to consider these in a revised version of the manuscript.

1) Hypothesis tested

The study starts by discussing declines in farmland birds, then clarifying that “Patches of semi-natural habitats, such as woodlands, grasslands and wetlands, remaining within agricultural landscapes may provide permanent habitat for wildlife and host a large part of farmland biodiversity”. So here these semi-natural habitat patches are seen as part of the farmland landscape, contributing to farmland biodiversity. 

Then the study is framed in the context of the habitat compensation hypothesis (lines 72-73) which “states that species may compensate for the loss of their primary habitat by using agricultural habitats as a substitute”. In this context, the analysis and discussion treat semi-natural habitat patches (wetlands, woodlands and grasslands) as “primary habitat” and the field margins as “agricultural habitats”. 

But it could be argued that the field margins are not “agricultural habitats” but simply particular types of patches of “semi-natural habitat” within an agricultural landscape – and indeed in line 71 they are presented as such. 

All this to say that I am not sure that the study is actually testing the habitat compensation hypothesis (effect to the substitution of primary habitat by agricultural habitat). It seems to me that it is instead testing if field margins (particular types of semi-natural habitat) add valuable semi-natural habitat area to agricultural landscapes. 

To me this seems like a very pertinent question, given that field margins fulfil functions in agricultural systems (e.g. draining ditches, hedgerows as separation between properties, grass strips as access areas) and so are much more likely to be retained in agricultural landscapes than other types of semi-natural habitat patches that could be converted to arable land. If they are valuable semi-natural habitats, they can be a win-win between agriculture and conservation. 

In this light, to me one of the key conclusions of the study would be that field margins do have a habitat value for some of the species, but (for the same area) they are not as valuable as the habitat patches. For example, the effect of reed strips on reedbed birds is 0.58, compared to 0.79 for wetlands.

2) Landscape context of the study

Currently the Introduction and the Study Area section give the impression the paper is all about rice fields, with the field margins being those wide bands (>3m, to be waterproof) separating rice paddies. 

It is then a surprise when we find out in lines 144-145 that the study covers different types of crops. The reader will then have made a mental image of the field margins that does not quite match the reality of the field. 

I recommend that the introduction clarifies that different types of crops are cultivated in the Camargue, and also that the methods section presents a more precise definition of “field margin”. (Are they the darker lines in the map in Appendix A?)

3) Species’ guilds

I am not sure of the advantage of classifying a priory species into guilds, rather than letting the results speak for themselves. For example, the results for the common nightingale indicate that it has a quite different response to other species classified as a forest edge species (Fig 4). 

Furthermore, like reviewer 2 I am confused as how to interpret the results in Figures 2-4: I cannot tell if a negative contribution to a positive effect means that there was a negative effect (the species’ abundance declines when the area of a given landscape type increases), or just a slightly less positive effect (the abundance increases less than for other species in the guild). 

For these reasons, I would recommend that rather than classifying the species into guilds a priory, the authors include all the species in the same model (with species as random effects), and then present results on the effects of each of the landscape variables on each of the species (as in Table 2, but with one species per column). The interesting analysis then being to compare for each species the effect of type of habitat vs the effects of field margin type. For example, for species for which wetland area has a positive effect: how does the effect of the area reed strips compare (also positive? similar in magnitude?); and how does the effect of hedgerow area compare (negative? similar in magnitude?).

I predict that this would give clearer results as to the species for which field margins (and which types of field margin) can work as valuable habitat than the current discussion based on guilds. For example, I suspect that the current counter-intuitive result that forest edge birds do not respond to woodland area may be an artifact of having quite different types of species within the forest edge guild (including nightingales). Instead, I would recommend defining forest edge species a posteriori, as those for which woodland area has a significant positive effect. 

4) Interaction between habitat patches and field margins 

The analysis treats each landscape variable as independent, but it is plausible that the value of field margins depends on the presence of habitat patches. For example, reed strips may be more valuable if close to wetlands. This could perhaps be tested in the model using interaction terms between each type of field margin and the corresponding habitat patch (reed strips – wetlands; hedgerows – woodland areas; grass strips – grassland areas).

This is also related to the point rightfully presented in the discussion (lines 331-333) that field margins may be unsuitable breeding habitats even if species use them (and could even be ecological traps). Worth discussing there that field margins could also be useful adjacent habitats (e.g. used for feeding but not for nesting) in which case they add value to existing patches of semi-natural habitat even if they cannot replace them.