The paper is well-written, the abstract and the discussion sections are clearly based upon the results. The authors follow good practice by making available their datasets and r-scripts, the github repository is very well-organised.

The findings from this paper are interesting. I think the paper title could perhaps be changed to reflect that the authors only found significant changes in the variability of laying date.

I provide some line-by-line comments below but I wanted to summarise below the important point to consider.

Given that the effects are quite small, particularly for the lnCVR work where the difference in variability only becomes significant in your tri-variate analysis, I think some additional sensitivity analyses are required:
1) I don’t think it is appropriate to include your imputed SD results in the lnCVR analysis. This is basically imputing the dependent variable based on a regression model of lnSD and lnMean. You say that you got qualitatively similar results with and without the imputed values, but it makes sense that this would happen, because the imputed values are based on non-imputed values. I suggest you present the lnCVR results without the imputed SD values. I think it is fine to include the imputed values for lnRR analysis.
2) The coefficient of variation (and therefore lnCVR) assumes a slope of 1 between lnMean and lnSD, but the relationship is quite different in your data. Deviation from this slope of 1 can result in under- or over-correction for the mean-variance relationship. This is nicely explained in Doring et al 2015 (doi: 10.1016/j.fcr.2015.08.005). Looking at figure S2, it looks like the slopes could be approximately 0.5 for laying date, 0.7 for clutch size and 1.2 for number of fledglings. Please could you comment on this and consider presenting results from an additional arm-based analysis where the relationship between lnMean and lnSD is explicitly modelled (see Senior et al 2016 doi:10.1093/emph/eow020 and Nakagawa et al 2015 doi:10.1111/2041-210X.12309). This would be a useful sensitivity analysis to include here alongside the lnVR analysis.

There are a lot of complex analyses in this paper. I think it could be useful to give each meta-analytic model that you use a unique identifier that you use when you refer to the model in the methods section and in the results section, this could also link to where the model appears in your scripts on github (this could basically be a modification of the existing tables S3 and S5). This would allow readers to track more clearly which model is being used where and what particular assumptions are associated with it. Given the complexity of the methods used, I think a short summary paragraph at the start of the methodology would help readers follow the subsequent text.

Please see below for additional line by line comments and suggestions, I hope that the authors find these useful in refining the manuscript further.
L50 – I got a bit lost here (before having read the paper) because there were a lot of parts to the sentence. Differences in phenological variation, between populations, within breeding seasons. Consider rewriting.
L68-70 – can you include a ref here?
L85-86 – adaptation to local conditions could increase phenotypic variation –I think you need to make it clearer here that i) urban environments are generally more heterogeneous, so ii) adaptation to local conditions increases phenotypic variation between those different sets of local conditions in the heterogeneous urban environment. Perhaps spell this point about urban conditions being more heterogeneous earlier in the intro. You say this on L115-116 but this could come earlier as it is an important concept for readers to grasp.
L89-94 – this line helped me make sense of the text that had come before. Consider bringing this higher up in the introduction.
L108 – using ‘inter-annual’ or ‘intra-annual’ would help clarify which you mean here
L116-119 – this feels repetitive of L85-86, but explains it better
L128-130 – this would be really nice if you can use spatial scale at which the variance is measured as a moderator in your analysis
In general, I found the introduction perhaps too reliant on the Thompson et al 2022 reference. Can you consider including a little more detail here so that this present paper stands up better by itself?
L143-146 – this feels a bit strange in this paragraph, the rest was about what you predict you’ll find, but this sentence is about what you’ve been able to do. It might be sufficient to just remove ‘lastly’, starting the sentence with lastly makes it feel like this is very similar to the previous two sentences about your predictions.
L151 – given the complexity and length of the methods section I think this would manuscript could really benefit from a methods summary paragraph
L159 – it might be useful to explain to the reader in more detail here what ‘paired urban and non-urban populations’ refers to. Is there some minimum criteria by which you judge the populations to be paired? Are the populations within a certain distance of eachother for example? Without something like this, the comparison may be confounded by factors like climate or photoperiod. -> this is explained later in the inclusion criteria, so perhaps refer to that section here?
L152-189 – This literature review section is well-written and clear apart from the single comment above.
L194-197 – would be good to include a ref here
L197-198 – geographically close is a subjective term. Can you put a put this more objectively, perhaps explaining the maximum distance between populations that you included?
198-200 – I assume there were some other ‘good practice’ criteria that you looked out for too, that sampling effort was equal in the urban and non-urban environments for example?
200-204 – clear and sounds sensible, though now having read the rest of the paper, I wonder if keeping these gradient of urbanisation data points and running a separate analysis based upon them would help address some of the limitations of your landscape heterogeneity / urbanization index work?
206-212 – consider moving this to the supplementary information
212-213 – it’d be useful to know the slope of the relationship between lnMean and lnSD (see above) perhaps as an annotation to fig S2.
218-233 – these references might be included more usefully in a table format, in which could also provide information about the location(s) of each study, the number of effect sizes you used from it, the species studied etc.
L242 – change ‘these’ to ‘the’
L266-268 – I think this text needs amending slightly – were there 30 studies in the main analysis? I thought from text above that were 32, 42 or 44 studies? In table S3 and S5 which listed k for different random effect levels, is this a typo that there are c.151 studies? Please check and clarify what you mean by a study here and make sure it is consistent across the text.
283 – typo ‘ad’
290 – so you combined the three life-history traits in the same analysis. This will have introduced nonindependence. Please explain briefly here in this section how your random effects structure accounts for this, or how it is accounted for elsewhere (e.g. in the trivariate models).
292 – is ‘population identity’ the correct term here? I assume this is referring to a ‘pair of populations’ that are being compared in a given row of data?
348-355 – you include tables in the supplementary comparing difference vcv matrices, can you refer to them here?
397-403 – it seems a bit of shame that there were studies conducted on an urbanization gradient but you only picked the two extremes of land use from these studies, they would be well-suited for this analysis. Please comment why you didn’t do this and/or consider doing it. You describe the limitations to the current approach nicely in the discussion 609-626 and it seems using the gradient studies may have got around some of these limitations.
421-424 – ok but this is a very small effect and the CI overlaps zero, I think you need to include explicit mention of this in the text. I guess this is what you are saying when you use the word ‘tended’ but this could be more explicity (like you do on line 455).
436 – do you need to say ‘much earlier’ here, or just earlier?
462 – do you mean phenotypic variation here or phenology? the laying date variation does look quantitatively very similar, but the means for the other two traits are a different sign.
476-477 – suggest reword this, something like ‘however, the mean lnCVR estimate were very different between these models’ to emphasise the difference
509 – ‘phylogenetic’ or ‘phylogenetically controlled’?
529-532 – should this be ‘large variation among species’? In table S1 the heterogeneity in I2 seems to be a lot greater among-species rather than among-studies
539 – typo ‘increased’

Thanks to the authors for their clear responses to our comments. The authors have made substantial changes to the manuscript. I think these improve the manuscript. I just have a small number of typos that I noticed below.
265 – change ‘every study’ to ‘these studies’
271 – change ‘resulting into’ to ‘resulting into’
272 – change ‘weights in’ to ‘weights’
531 – typo ‘95%Cis’
560 – change ‘increase’ to ‘increase in’
564 – change ‘a many’ to ‘many’