Main comments
This manuscript evaluates how species evenness may mediate the effects of species richness on forest biomass and productivity using a comprehensive plot database. The results presented in this manuscript show that the impacts of species richness depend on species evenness, but the extent to which they do so varies regionally. The topic addressed in the manuscript is very timely, as it seeks to reveal how two complementary components of community structure (i.e. the spècies abundance distribution) lead to saturating or linear increases (or decreases) in both biomass and productivity. Additionally, the regional level analysis adds valuable to the global analysis by revealing shifts in the importance of species richness and evenness for biomass and productivity. Despite these considerable strengths, I have concerns about the methods (and the clarity with which they are explained) and, more importantly, the issue of how spatial resolution and extent may influence the results and their interpretation.

i) Spatial scale: There are two ways in which spatial scale influence the results of this study: a) variation in resolution (plot size; 0.02 - 1.5 ha) and b) variation in extent (# of plots per ecoregion). Both components of spatial scale have been demonstrated to influence the relative importance of meta-community processes such as niche or neutral processes (Viana and Chase 2019), with increasing spatial extent increasing the importance of niche-based processes and increasing spatial resolution increasing the importance of neutral processes. These are important considerations when comparing results between regions that where high evenness has differential impacts on biomass and productivity. However, spatial resolution of plots included in the analysis varies on a similar scale at which Chisholm et al. (2013) found variation in diversity-biomass relationships. On a related note, the authors do not present an analysis of how sensitive their measures of species richness and evenness are to plot size, which could also confound their results. While the authors justify their selection of Hill's evennes due to its consistency with Shannon's entropy and richness, there is no mention of its sensitivity to spatial resolution. One metric of evenness whose robustness to spatial resolution - Hurlbert's probability of interspecific encounter (PIE; see Chase and Knight 2016) - is known, could be a viable alternative. Similarly, the authors could try alternative methods for estimating species richness that are robust to number of individuals, e.g. rarefaction and extrapolation (Chao et al. 2014). With regards to spatial extent, the number of plots per ecoregion is very asymmetric, with more plots in temperate and boreal ecosystems than in tropical ecoregions. The greater spatial extent in temperate and boreal ecoregions than in tropical ecoregions may therefore confound direct comparisons among ecoregions.

Chao, A., Gotelli, N. J., Hsieh, T. C., Sander, E. L., Ma, K. H., Colwell, R. K., & Ellison, A. M. (2014). Rarefaction and extrapolation with Hill numbers: A framework for sampling and estimation in species diversity studies. Ecological Monographs, 84(1), 45–67. https://doi.org/10.1890/13-0133.1

Chase, J. M., & Knight, T. M. (2013). Scale-dependent effect sizes of ecological drivers on biodiversity: Why standardised sampling is not enough. Ecology Letters, 16, 17–26. https://doi.org/10.1111/ele.12112

Viana, D. S., & Chase, J. M. (2019). Spatial scale modulates the inference of metacommunity assembly processes. Ecology, 100(2), e02576. https://doi.org/10.1002/ecy.2576

ii) Methods: I had a difficult time understanding how plot biomass was actually calculated and if and how ecoregions are "represented proportionally" in the regression models. In the case of biomass, it is not clear how the authors selected the 422 allometric equations and if all were used to estimated biomass, or if they were further reduced to create ecoregion-level equations. The resulting estimate of biomass accumulation is also problematic, in that it assumes that biomass increases linearly with time and is not truly a measure of productivity. I detail an alternative method in the line-by-line comments below. Secondly, given that there is uncertainty in the estimation of tree-level biomass, and this error propagates when plot-level biomass is estimated, it would be good to account for this uncertainty when estimating plot-level biomass (Rejou-Mechain et al. 2017).

Rejou-Mechain, M., Tanguy, A., Piponiot, C., Chave, J., & Herault, B. (2017). BIOMASS : an R package for estimating above-ground biomass and its uncertainty in tropical forests. Methods in Ecology and Evolution, 8(9). https://doi.org/10.1111/2041-210X.12753

Minor comments/suggestions
i) I would suggest adding Fig. S5 to the main text. The variance partitioning in Fig. S5 is a more honest representation of the results, as it in includes the relative importance of environmental variables and human impacts. Together these variables in many cases explain more variation than the comination of species richness and evenness and their interaction. The variance partitioning in Fig. 4 gives the impression that all models (global and regional) only include species richness and evenness and their interaction, which doesn't seem to be the case (see L864-866). So, I would suggest replacing the variance partitioning in Fig. 4 with that of Fig. S5b, and moving Fig. S5a to the main text. Fig.S5a does a very nice job of showing how the relative importance of species richness and evenness and their interaction, human impacts, and individual environmental variables change across regions.

Line-by-line comments
l573 change "property" to component

L576-578 re-word to active voice

L580-581 This has been tested empirically, on multiple occasions (e.g., Wilsey and Potvin 2000, Wilsey and Polley 2004, Yan et al. 2021).

Wilsey, B. J., & Polley, H. W. (2004). Realistically Low Species Evenness Does Not Alter Grassland Species-Richness-Productivity Relationships. Ecology, 85(10), 2693–2700. https://doi.org/10.2307/3450427

Wilsey, B. J., & Potvin, C. (2000). Biodiversity and Ecosystem Functioning: Importance of Species Evenness in an Old Field. Ecology, 81(4), 887–892. https://doi.org/10.1890/0012-9658(2000)081[0887:BAEFIO]2.0.CO;2

Yan, Y., Connolly, J., Liang, M., Jiang, L., & Wang, S. (n.d.). Mechanistic links between biodiversity effects on ecosystem functioning and stability in a multi-site grassland experiment. Journal of Ecology, n/a(n/a). https://doi.org/10.1111/1365-2745.13725

L585-586 re-word "at the global scale" to "globally", as it is not clear if scale here refers to spatial extent or resolution.

L587 delete "at a global scale"

L622-624 re-focus opening lines, focusing on the science rather than on the popularity/impact of the sub-discipline.

L632 delete "global", insert "globally" after "forests"

L635 these references are not necessary here. i would suggest deleting them.

L652 This can be very scale dependent, see Antao et al (2021).

Antão, L. H., Magurran, A. E., & Dornelas, M. (2021). The Shape of Species Abundance Distributions Across Spatial Scales. Frontiers in Ecology and Evolution, 9. https://doi.org/10.3389/fevo.2021.626730

L655 see previous referenced studies on the effects of evenness.

L661 change "apparent" to "manifested"

L671 delete "using statistical models" - not necessary.

L669-679 Very well structured paragraph presenting the main objective and hypotheses.

L699 insert "at least" before "once". Presumably some plots have been measured more than once. it would also be helpful to indicate the mean number of times that plots were measured.

L713 I am a little sceptical of this approach, as it is does not capture net change in biomass but rather biomass accumulation rates. Also, it assumes that biomass accumulation is linear with age, which is a questionable assumption. Most plot-based measures of NPP are measured as net change in aboveground biomass, e.g. Chen and Luo (2015). This would imply limiting the analysis (for plot-based NPP, at least) to those plots with multiple measurements, which would be fine given the large sample size (367k plots).

Chen, H. Y. H., & Luo, Y. (2015). Net aboveground biomass declines of four major forest types with forest ageing and climate change in western Canada’s boreal forests. Global Change Biology, 21(10), 3675–3684. https://doi.org/10.1111/gcb.12994

L716 Please provide correlation between measures of NPP. I see now that you have done this (L825). It may be helpful to provide this information earlier (ie here) when introducing both measures of productivity.

L731 One key measure of evenness not included is the Hurlbert's probability of interspecific encounter. It has a number of appealing properties, primarily that is less scale dependent than other measures of evenness, is the complement of Simpson's diversity index,
complement of Simpson’s diversity index, and can be readily converted to effective number of species(Chase and Knight 2013). This metric would be particulary suited for this analysis given the variation in plot size, which is not tested for here.

Chase, J. M., & Knight, T. M. (2013). Scale-dependent effect sizes of ecological drivers on biodiversity: Why standardised sampling is not enough. Ecology Letters, 16, 17–26. https://doi.org/10.1111/ele.12112

L766-771 It is unclear if and how the 422 allometric equations are reduced to one per extra-tropical biome. Please explain.

L783-785 How comparable are biomass estimates that use wood density, compared to those that do not?

L790-793 Given the uncertainty related to the allometric models, it would make sense to ensure that these errors are propagated when estimated plot-level biomass (see Rejou-Mechain et al. 2017)

Rejou-Mechain, M., Tanguy, A., Piponiot, C., Chave, J., & Herault, B. (2017). BIOMASS : an R package for estimating above-ground biomass and its uncertainty in tropical forests. Methods in Ecology and Evolution, 8(9). https://doi.org/10.1111/2041-210X.12753

L806-807 Yes, this is the correct terminology.

L825 It would be good to acknowledge the limitations of not knowing how these two measures are correlated for non-temperate regions.

L825 delete "Yet,"

L845 It's a bit of a reach to say that it corrects for differences in plot size and successional. It would be preferable to just focus on its impacts on evenness and species richness.

L846 These are the same WWF eco-regions?

L854 please explain "proportionally represented". Do you mean that the number of plots included in the analysis is proportional to the area of the ecoregion? I would guess not, but this is not clear.

L914 Try to emphasize the magnitude of the slopes, rather than their significance, as the large N in this study is likely to result in significant effects regardless.

L929 delete "global", insert "globally" after "productivity"

L933, L936 etc. change "systems" to "ecosystems" (or "forests") for consistency.

L944 delete "global"

L943-L947 I would suggest deleting this, as the the major results are summarized in the following paragraph.

L951 should read "likely affects", and not "is likely to affect"

L953 delete "indirect,"

L954 insert period after "and 2)". Change "specifically" for "Therefore,"

L955 delete "necessarily"

L960 delete "those"

L961 change "and" for a comma, delete "accelarating and accentuating", and change "effect" to "effects"f

L962 change "is" to "are", delete "necessarily"

L969 delete "highly"

L977 change "on" to "over"

L981 delete "in terms of total productivity"

L983 delete "drastically"

L986 delete "local"

L988-990 delete "For example...clusters" This is not an accurate, or just an unclear, explanation of where biodiversity effects occur.

L1005 change make "forest" plural

L1005-1017 Spatial scale (ie plot size) and extent (total ha per bioregion) may impact these results, as niche or neutral processes may become stronger (or weaker) with changes with plot size. See Viana and Chase (2019). It would also be important to mention that niche-based processes are still present in tropical forests (see Shipley et al 2012), but their relative importance - again - depend on the spatial extent and grain.

Viana, D. S., & Chase, J. M. (2019). Spatial scale modulates the inference of metacommunity assembly processes. Ecology, 100(2), e02576. https://doi.org/10.1002/ecy.2576

Shipley, B., Paine, C. E. T., & Baraloto, C. (2012). Quantifying the importance of local niche-based and stochastic processes to tropical tree community assembly. Ecology, 93(4), 760–769. https://doi.org/10.1890/11-0944.1

L1016 insert "being" after "species"

L1054 It would be good to cite studies from experimental forests (e.g. Paquette et al. 2018).

Paquette, A., Hector, A., Castagneyrol, B., Vanhellemont, M., Koricheva, J., Scherer-Lorenzen, M., & Verheyen, K. (2018). A million and more trees for science. Nature Ecology & Evolution, 2(5), 763.

L1078-1093 this Conclusion is largely repetetive with the previous paragraph. I would suggest deleting L1062-1075.

Main comments

The authors moderately improved this version of the manuscript, in which they show that species evenness moderates the effects of species richness on productivity. While the authors acknowledged potential scale-dependent biases in their analysis, they failed to adequately explore them beyond mere correlations. I would strongly suggest that they do a more rigorous analysis of potential biases, by calculating the accuracy and precision of effect sizes (see Spake et al. 2021), which have been shown to remain even after accounting for plot size (as either a random or fixed effect). I remain skeptical that how biomass production is a reasonable way to quantify forest productivity, as it assumes that biomass accumulation is linear over time and that the estimation of forest age is accurate. For this reason, I would strongly suggest that the authors use the remotely-sensed NPP in the main text.

Literature
Spake, R., Mori, A.S., Beckmann, M., Martin, P.A., Christie, A.P., Duguid, M.C. & Doncaster, C.P. (2021) Implications of scale dependence for cross-study syntheses of biodiversity differences. Ecology Letters, 24, 374–390.

Line-by-line comments
L264 delete "indeed"
L267 change "tree" to "forest"
L268 change "limits" to "buffers"
L270 delete "highly" (very subjective)
L273 delete "global", "highly"
L275 insert comma after "richness", delete "levels"
L276 insert comma after "richness"
L280 change "at..levels" to "of speciose communities"
L281 delete "the"
L311 delete "(i.e. a greater... richness)"
L315 and L319 The Fraser paper is about correlation between productivity and diversity, and not about detecting effects of diversity on productivity
L330 unclear what "trends" are being referred to. Maybe clarify by adding "spatial"
L331 delete "highly"
L332 delete "relatively"
L334 delete "global"
L335 delete "relatively"
L335 delete "highly"
L344 delete "highly"; insert "a" before "few"
L353 delete "global"
L363 delete "global", "highly"
L365 delete "We use...forests"
L368-371 the wording of this hypothesis is unclear. It gives the impression that one is experimentally increasing the richness of certain communities (species-poor ones), which is not the case as this is an observational study.
L431-432 Please least the other indices, and a short explanation as to their selection (otherwise it sounds like a random selection, which is not the case).
L452 the null model needs at least a short explanation, as it is not clear that it is explained in the Supp. Information.
L491 The description of the biomass estimation does not seem to coincide with the explanation given in the response to reviewers. The authors should change the text accordingly.
L522 delete "fairly"
L587 change "does... on" to "is robust to"
L611 delete "the" before "tropical" and make "forest" plural
L617-619 it is not clear to me how the ratio of variance explained can be used for inference.
L627 delete "Strong", capitalize "interactions"
L629 insert "statistically" before "significant"
L644-663 This section seems quite redundant with the previous paragraph (the topic sentences of both are quite similar). If so, I would suggest deleting it. If not, clarify what new line of evidence is being described here.
L645 delete "strongly"
L650 delete "the" before "tropical", make "forest" plural
L652 change "systems" to "forests" (and in L653)
L660 delete "in our global forests"
L667 delete "community" before "productivity"
L697 change "this" to "the"
L701 This sentence is very confusing, as it sounds as if one can manipulate the relative abundance species, using data from an observational study.
L707 reword "do... clusters" to "are not spatially aggregated"
L708 Delete this sentence - aggregation of species was not measured in this study.
L723 delete "related.. strategies" , as it's redundant with "functional traits"
L724 delete "or even facilitation" (awkward wording)
L740 delete "the" before "boreal" and make "forest" plural