Thank you for your detailed response. I read this new version, and the message is much clearer, I especially like the use of effect sizes to describe the magnitude of changes. The dataset is impressive but complex, and while some observed patterns are still challenging to interpret, the main messages are clearly stated and I have no further comments on the text.

Best,

Ignasi Bartomeus.

The authors have thoroughly addressed all comments, conducting additional analyses when needed, and providing important clarification on methods. I only have minor comments to improve readability because the analyses remain complex and sometimes difficult to follow. Overall, I suggest that the authors standardize terms throughout the whole ms, add a synthesis of the results at the beginning of the discussion section, and use more specific wording to avoid any confusion, in particular in the discussion. Below, I am providing detailed comments to illustrate these general suggestions. 

L34 Replace “that with “, which”

L51 Replace “are observable” with “have been observed”

L52 Add “temporal” before “trend”

L56 Replace “by taxonomic diversity” with “using taxonomic approaches”

L58 Replace “dimension” with “approaches”

L85 Replace “the temporal dimension of functional inter-specific trait variations” with “temporal changes in species trait distribution”

L90 Since the authors use many traits and associated adjectives, it would be useful to specify the link between traits and adjectives the first time they are being used, e.g. more acquisitive species means species with a higher SLA.

L103 Use the same wording as above “species richness, trait composition and ecological strategies”

L111 When possible, use the same wording throughout the ms, e.g. here “soil moisture” instead of “water resource”, to help readers.

L170 Add “and ecological requirements”

L215 Try to use the same wording, e.g. here you call CWV “divergence”

L262 Use the words “significant”, “strong”, “non-significant” or “weak” rather than “clear” and “uncertain”. The word “uncertain” may suggest that the statistical power is not sufficient or that confidence intervals are too large to conclude. It would be great if the authors can clarify this point.

L265 Since this study only focuses on French data, delete “In France,”

L280 Fig 3 would be easier to read if legends for y-axes are indicated along the axis rather than above

L296 As suggested above, it would be useful to specify at least once the link between adjectives and traits (here conservative and acquisitive)

L332 Specify “temporal models” on what?

L333 “Effect sizes” rather than “regressions”?

L339 and 342 Replace “community changes” with “community trait composition”

L349 Explicit the meaning of CWM in the legend.

L363 The way this sentence is constructed suggests that you ran a model on a data subset for “frequently managed margins of the MZ”.

L366 What does “seasonal effects” means? Date of observation?

L371-380 This whole section explains Grime’s theory and belongs to the method section rather than the beginning of the discussion. I believe a summary of the results would be more useful at the beginning of the discussion section.

L390 What is “functional variety”? Does it refer to “trait divergence”?

L394-396 It would be good to start the discussion with this kind of sentence on the main results.

L403 Reword “dry and heat stress”

L417 The authors suggest that “plant communities are also responding to past landscape diversification”. However, landscape simplification has been ongoing for more than 70 years in most agricultural landscapes in France (except in the Mediterranean region). Do they refer to a heritage effect from past landscape diversity rather than diversification?

L419 The authors mentions “arable lands decrease” whereas it has increased in most agricultural landscapes in France.

L423 The authors did not study “functional diversity”

L427 Farmers will certainly adapt to climate change so this sentence sounds a bit naïve.

L429 Replace “diversity and species assembly” with “field margin plant communities”

L435 Replace “can only have” with “are likely to have only”

L437 I do not understand what “functionally similar subset of communities without herbicide application” means

L438 Add “some” in front of “traits”

L444 Replace “this” with “fertilization”

L446-447 Delete “high nitrogen levels can favour” and “, which”. Replace “, explaining that” with “, which explains why”

L450 Specify which vegetation

L463 Change to “and implications for communities response to ongoing global changes”

L467 Are trade-offs evident due to the scale of the size of the species pool considered?

L468 Change to “occur”

L470 Specify “field-margin plant communities”

L479 Replace “space” with “spatial analyses” and “identify” with “imagine”

L483 Replace “specialized to their habitat” with “habitat specialists”

L485 Which “restricted areas”?

L491 Add “to understand the temporal dynamics of field margin plant communities” or something similar at the end of this sentence.

L497 The authors mention “at the national level” but they did not study field margin plant communities at that level.

L498 Change to “Our study suggests that species selected by…”

L500 Specify which agricultural practices

L510 Replace “some perspectives” with “other important research questions”

The manuscript presents an impressive dataset to ask how ruderal plants have changed in the last decade across space in France. The challenge is to explain such complex patterns effectively. While the manuscript is overall sound, my main concern is that some details or decisions are unclear or not justified. Most of my comments are in this direction. 

First, given the several analysis conducted, I would recommend focusing the discussion on the effect sizes of the different models and avoiding using p-values. The authors are aware of that and already report p-values at alpha 0.01, and highlight the R2 of the models, but I miss a clear interpretation of the models in terms of the effect observed. How many e.g. species on average were lost over years? If the authors can translate the results into clear units, the paper will be much easier to follow. A clear example is Fig 4, which is a nice summary of "important" variables, but it tells nothing about how big were each of the effects. 

Second, while I like the distinction between temporal and spatial patterns, I am not sure if using a single modeling technique would help explain better the story. First, temporal correlations are shown, but those treat all site-year combinations as independent. I would expect "site" to be modeled e.g. through a random structure as done for other analyses. Second, spatial trends are done by pooling across years, but I am not sure if this pooling can mask some interesting effects. Finally, a more detailed temporal analysis is done, this time considering "site" as random (this is in part redundant with the first analysis?). Moreover, all these models are run for different subsets of data. Would it be possible to tackle all questions within a single analytical framework? For example, by modeling the interaction between time and environment e.g. CWM ~ Time * Region would avoid doing a test for each region. I am not suggesting that all analyses need to be re-done, but if you can simplify the number of tests done, without losing insights the paper would gain clarity and robustness. 

Finally, while I appreciate the context of Grime's triangle, this appears only in the discussion, and for me was very hard to relate the analysis performed to this framework. For example line 333 says your analysis shows changes in stress-tolerant species, but you did not previously relate your analyzed variables with the CSR strategies. This can be easily fixed by introducing this earlier on, or alternatively, this comparison can be deleted, as the manuscript is already quite complex. I have to say I did not fully understand Fig 5. Calculating CWM on CSR scores, and averaging across space is oversimplifying the story, and the two data points (a mean of means) are very close to each other precisely because this averaging masks more subtle trends, but the variation across species might explain a better story. In any case, this should be integrated into methods and in your modeling framework. Otherwise, sentences such as 417-418 are not convincing.

In-line comments: 

Line 21-22: This justification is poor, and mixes a technique (meta-analysis), with an index (taxonomic diversity). I think it can be safely deleted. The abstract can be enhanced for clarity.

Line 79: I think there is an extra s in traits.

Line 111: You mention 4 crop types, but mention only three.

Line 123: By removing rare species your analysis focuses on changes in common species. This is fine but should be discussed and interpreted as such.

Table 1. The underlines, italics etc... are really hard to keep in mind, especially as afterward you refer to e.g. "explanatory variables" without mentioning them explicitly again. I am not sure which is the best solution, but maybe adding a column to indicate predictors, explanatory variables, etc... and mentioning them again when you specify the models might help.

Line 177: It's unclear if you always use this "richness detrended" CWM index, or if you also use the raw index in your models. I think both are valuable and tell you different things. In fact, the species richness detrended CWM is hard to interpret, and if this is the one you report in the analysis, this should be made very explicit and help the reader interpret it with phrases such as "smaller CWM than expected given a community of this richness".

Line 179: Why perform the PCA on CWM values, and not in the pool of species values first, and then calculate metrics by communities (e.g. as done by Villager, Laliberte et al. in the FD package you use to calculate the metrics)? My concern is that CWM is losing information by focusing on the mean (you lose the variance), but maybe there is a rationale to do this. Explaining it would help the reader. 

Line 195: Adding the number of observers does not fix observer bias (differences among observers). You do this later on by using observe id as random, but I don't think here is accomplished. Maybe I am losing something.

Line 263: Here is an example of the complexity for the reader to follow up on all the details. By using Ellemberg T, the reader needs to track what this variable meant. I would prefer to be redundant and explain what variables should be interpreted every now and then.

Best,

Ignasi Bartomeus 

General comment

The manuscript entitled “Functional trade-offs: exploring the effects of climate change and agricultural practices as drivers of field margin plant communities” presents a very interesting study of the interactive effects of climate changes and land-use changes on plant diversity in agricultural landscapes. It develops a spatial and temporal analysis of a large national dataset, based on a functional approach (species traits and the conceptual framework of Grime’s CSR triangle), to assess potential trade-offs in the response of plant species to climate changes and land-use changes. The results of this study may contribute to better predictive models and improved conservation strategies for plant species in agricultural landscapes.

The manuscript is well written. The introduction convincingly argues for the need to use a functional approach to understand the interactive effects of climate changes and land-use changes on biodiversity and develops sound hypotheses based on the literature. The material and method section presents the complex dataset collected within the 500 ENI monitoring program and the set of statistical analyses used to test hypotheses. Results are explained clearly. The discussion is structured around three straightforward messages: the effect of climate, the effect of agricultural practices and their interactive effects on field margin plant communities.

I believe there is still a margin for improvement in terms of clarity of the study design and in terms of interpretation of the results. I therefore provide two sets of comments that will hopefully help the authors to improve their manuscript, as well as a third set of comments on other specific points.

Study design

This study is based on a national dataset collected within the 500 ENI monitoring program. Such programs tend to produce large but complex datasets. It seems to me like some key variables (based on the literature) were not considered in the study and some methodological choices were presented only briefly. Since these variables and methods have the potential to influence the outcomes of this study, it seems important to clarify their consequences (or lack therefore). Below, I develop four main comments regarding the study design.

1)      The herbaceous field margin represents the uncultivated vegetated area located between the cultivated strip and the adjacent habitat. The manuscript mentions that crop type varied within the ENI dataset, with a clear latitudinal gradient (wheat in the north, vineyard in the south; see Appendix A). Similarly, the type of adjacent habitat is likely to vary within the ENI dataset (e.g. crop, grassland, woodland, road, hedgerow…). Yet, these two variables were not considered. What is the likely influence of the type of crop and adjacent habitat on plant communities in the field margin? Could the authors explain why these effects do not create a bias in the study?

2)      Landscape structure is a major determinant of biodiversity. Andrade et al. (2021) actually explains that landscape variables were measured around each field within the ENI programme. Moreover, the effects of landscape variables on biodiversity have been shown to interact with the effects of agricultural practices (e.g. Ricci et al. 2019). Yet, landscape variables were not considered in the present study (unlike in another manuscript from the same authors on the same dataset, which is currently under review). There may be some valuable reasons for this omission, but I believe these reasons need to be provided and, here again, the absence of bias needs to be justified.

3)      The manuscript develops both a spatial and a temporal approach. However, it does not provide a clear description of the data available for these two approaches. It states L190 “To analyze the effects of spatial variations in agricultural practices, explanatory factors and abundances were averaged across years within sites having at least five years of data and five floristic observations, leaving a total of 349 sites”. This sentence suggests that some sites were not surveyed every year. Could the authors provide some statistics about the data available and explain why the absence of data for some years does not influence the results of the temporal analysis? Moreover, it is not clear how using the average value across years may influence the spatial analysis. It would certainly be useful to know more about the temporal variability of vegetation and practices at the field level.

4)      A subset of well-known species (142 species over 855 taxa) was used for this study. I fully understand that this selection was necessary to avoid observer biases. However, it is not clear whether this choice may have triggered other biases. For instance, this may have resulted in a stronger underestimation of species richness in species-rich field margins. It may also have resulted in the selection of a subset of species with specific traits. Could the authors present the traits of selected species compared to the whole species pool? Could the authors explain why using this subset have not influenced the results, on either species richness or traits?

Discussion of results and perspectives

This study focuses on field margin plant communities but also aims to discuss its broader implications regarding the interactive effects of climate changes and land-use changes on plant communities in agricultural landscapes. The authors discuss their results in a very straightforward way but I believe they could further discuss some unexpected results, the implications for other plant communities within agricultural landscapes and possible avenues to further explore the ENI dataset. Below, I develop four main comments regarding the discussion and perspectives.

1)      The authors state L416 “Our results show little effect of agricultural practices, likely because of their limited temporal trend and great range of spatial variability”. It is however hard for the reader to grasp this spatial variability. Could the authors provide more detailed information on the spatial variability of agricultural practices? and further discuss why this great range of spatial variability did not result in a stronger effect of agricultural practices in spatial analyses? The reader may also wonder to which extent this lack of effect of agricultural practices may result in underestimating the interactive effects of climate changes and land-use changes. This may be harder to assess but it may warrant a comment in the discussion.

2)      The authors conclude that agricultural practices have a weak effect, yet they claim that “the persistence of intensive agricultural practices and accelerating climate change would have critical consequences for the conservation of floristic diversity in agroecosystems”. If I play devil’s advocate, based on the lack of effect of in-field practices, I could argue that increasing fertilization and the number of herbicide treatments is likely to have a limited effect (or even no effect) on field margin plant communities, as long as the number of management actions in field margins remains low. This would correspond to a land sparing strategy. I certainly don’t have the full picture but it feels like these results could actually be discussed in a more nuanced way than the usual conclusion “intensive agriculture is a threat for biodiversity”.

3)      The authors provide a very valuable analysis of spatial and temporal variations based on a functional approach, which may allow more generic discussions based on Grime’s CSR triangle.  Are the functional trade-offs observed in this study likely to apply to other plant communities within agricultural landscapes, in particular weeds? Are the conclusions on the interactive effects of climate changes and land-use changes likely to apply to them too? At present, the authors tend to draw broad conclusions about plant communities in agricultural landscapes without discussing to which extend their results can be generalized. I believe adding such discussion would strengthen the overall conclusion of the manuscript.

4)      The 500 ENI monitoring program has produced (and keeps producing) a huge dataset on plant communities as well as other taxonomic groups. The present study is among the first ones to analyse part of this dataset. I believe this study opens the way for very exciting research questions and it would be a shame not to discuss these perspectives (non linearity, time-lag, trophic networks, etc..). It may also be useful to provide a feedback on this program (are all relevant data on agricultural practices available, e.g. dosage, toxicity, type of adjacent habitat? Is the distribution of fields and years sampled adequate? Etc…)

Other specific comments

L21 “However, these impacts are often studied using meta-analyses at large scales, or using taxonomic diversity” Why is the fact that they are studied “using meta-analyses at large scales” is a problem? It seems to me like the knowledge gap primarily comes from not considering functional traits in order to understand the interactive effects of land use changes and climate changes.

L30 “Changes in plant communities were noticeable, especially as they relate to climate change” This phrasing sounds ambiguous to me. It would be clearer to state whether changes are significant and what the key drivers of these changes are.

L58 The amount of pesticide use is not the only dimension of agricultural practices that influences biodiversity. The toxicity of pesticides has been shown to be a key aspect, as well as crop diversification, field margin/boundary management and landscape structure.

L98 It would be useful to clarify somewhere in your hypotheses (and not only in the discussion L374) that the impacts of in-field farming practices on vegetation in field margins are likely to be limited (more limited than for weeds).

L101 I believe there are only two zones considered: Mediterranean and Continental.

L108 How were fields selected within the ENI programme? Are they representative of practices in France or are practices less intensive on average because farmers are volunteering? This might also contribute to the lack of temporal trend in agricultural practices.

L149 It would be useful to make a clear link between Ellenberg codes and their meaning, probably both in Table 1 and in the text the first times they appear in a section, because it is not trivial for non-botanist.

L155 I may have missed it but I couldn’t find any information on where chorology data comes from.

L174-177 I find these two sentences hard to understand. It may be easier to follow if these sentences were moved to the relevant sub-section on temporal and spatial analyses.

L191 What is the difference between a “year of data” and a “floristic observation”?

L222 The term “elusive” does not seem appropriate here since you did observe some significant trends.

L242 It may also be useful to specify the meaning of Ellenberg-F in the legend

L251 The two PCA represent either sites or observations and not species. Please rephrase all sentences accordingly.

L288 “Regressions between Ellenberg values and climatic factors”. This sentence gives the impression that you conducted several regressions rather than a single one with all explanatory variables. Please reword to avoid any confusion.

L356 “These trait values (low SLA, high seed mass) are also known to be linked to less intensive agriculture” It is not clear whether the ENI dataset is consistent with this relationship?

L407 Change “varies” to “vary”

L419 Is a trend towards more intensive practices not as likely, if not more likely, than a trend towards more extensive practices? This hypothesis also contrasts with the following sentences focusing on agriculture intensification.

L426 “Mediterranean species could face severe agricultural intensification in northern France”. Did you really mean “northern France”?

In the data provided, the first row has shifted in files flora_temp and flora_spat

​​​​​