Earthworms contribute significantly to global food production

Earthworms are critical soil ecosystem engineers that support plant growth in numerous ways; however, their contribution to global agricultural production has not been quantified. We estimate the impacts of earthworms on global production of key crops by analyzing maps of earthworm abundance, soil properties, and crop yields together with earthworm-yield responses from the literature. Our findings indicate that earthworms contribute to roughly 6.5% of global grain (maize, rice, wheat, barley) production and 2.3% of legume production, equivalent to over 140 million metric tons annually. The earthworm contribution is especially notable in the global South, where earthworms contribute 10% of total grain production in Sub-Saharan Africa and 8% in Latin America and the Caribbean. Our findings suggest that earthworms are important drivers of global food production and that investment in agroecological policies and practices to support earthworm populations and overall soil biodiversity could contribute greatly to sustainable agricultural goals.

Overall I think more detail needs to be provided in the methods.
-It is not clear to me where the coefficients have come from in the van Groenigen meta-analysis.The data is not provided, and the code has not information on how those coefficients are derived (and therefore the weighting from the original meta-analysis).If these have been taken from the figures in the main analysis, this should be stated (potentially also including a table of the original coefficients, and the weight values applied by Fonte et al).
-In the underlying meta-analysis, some of the coefficients were not significantly different from zero (e.g., sandy soil did not significantly impact aboveground biomass.).Were non-significant coefficients still used in this synthesis analysis?Assuming so, is this appropriate?My assumption would be not.
-Line 142 "We then selected the most important drivers of the earthworm effect" -it is not clear how this was decided/done.The underlying meta-analysis provides many other coefficients that could have been used in this synthesis analysis (e.g., changes in climate, which are often found to be important).This should be expanded upon in terms of why those predictors were chosen and other not, and what implications this may have.
-Finally, how was the non-linear function for earthworm abundance calculated?In the underlying meta-analysis, there are three coefficients for earthworm abundance, so how robust is a function that has been created from only three data points?
As far as I can tell, the underlying meta-analysis tested each subgroup with a single model (e.g., one model on the effect of soil pH on yield, and another model on the effect of earthworm abundance on yield, etc. etc.).Fonte et al., have assumed that these coefficients (crop type, soil pH, soil texture, and crop-specific fertilizer N rates) can be multiplied together (i.e., a multiplicative effect, line 147).However, one could equally assume that these effects do not significantly interact with each other, and instead could also be causing an additive effect.Unfortunately, the van Groenigen meta-analysis can not be used to determine whether the equation should be multiplicative or additive, so I would suggest that the possibility is tested, and see how much the results change as a result of this assumption.
Minor point: Line 76-77: "…due *to* the vastly…" Reviewer #2: Remarks to the Author: This study is interesting as it shows clearly that earthworms contribute to food production at the global scale.They showed with a metanalysis that in the whole world the Impact of earthworm on global agriculture is 6.5 % for global grain 2.3 % of legume yield.It is true that up to now, there is not serious measurements of the earthworm potential contribution to global agricultural production.Most of the information comes from small trials and we don't have the idea of their impact at large scale.The results of the analysis are interesting, we see that they impact cereals and legumes differently.As well their impact is more important in the global south although they have impact globally.They suggest that if there is an improved management of soil biological communities, the impact of earthworm will be larger.I Like that they mention that it can be some biases in their analysis which can give some over or under estimations.The methods used are well explained although I am not an expert on this kind of analysis.The manuscript is well written and clear.I think this communication is important as it is one of the first work that gives the idea of the impact of earthworms and soil biodiversity can have in food production at a global scale, which is a proof of the importance of agro -and soil biodiversity for decision makers.So, I recommend the publication of this communication without any corrections, I hope that over time more data will be available to draw more robust and secure conclusions as their data comes from 54 studies (462) points.

Responses to Referee Comments
Reviewer #1 (Remarks to the Author): In this manuscript, Fonte et al. use previously published data (a meta-analysis and global data layers) to understand how earthworms contribute to the yield of cereals and legume across the world.They find that on average earthworms have caused yields to be 5.4% higher than if no earthworms were present.The amount varied depending on the crop type, but also global region -due to differences in soil proper es and the prevalence of fer lisers.Soil biodiversity, including earthworms, are o en cited as having a large impact on food produc on, and it great to see a synthesis analysis that delves further into this ques on, and a empts to iden fy how much impact they have as well as which regions of the world are benefi ng the most from a healthy earthworm community.Although ar cles are being published looking at the contribu on of soil biodiversity to ecosystem services (such as food produc on), these are o en focused on microbial diversity, and so it is good to see this analysis focused on the an important member of the soil fauna community.RESPONSE: We thank the reviewer for their thorough review and encouraging remarks.
Overall I think more detail needs to be provided in the methods.
-It is not clear to me where the coefficients have come from in the van Groenigen metaanalysis.The data is not provided, and the code has not informa on on how those coefficients are derived (and therefore the weigh ng from the original meta-analysis).If these have been taken from the figures in the main analysis, this should be stated (poten ally also including a table of the original coefficients, and the weight values applied by Fonte et al).
RESPONSE: As the reviewer indicates, the coefficients used in our analysis are largely based on the main figures (effect sizes) presented by van Groeningen et al. (2014) and especially, the supplementary informa on from their paper, where this data is reported in greater detail.We now explain this be er in the methods sec on by adding the following: "Coefficients were based on the effect sizes reported by van Groeningen et al. 8 (Table 1) and weighted by the number of observa ons in each category, since sample size varied depending on the factor in ques on and availability of data from the original studies."To further address the reviewer's concern, we now include the relevant sample sizes, overall earthworm effects, and effect size associated with each category (as reported in the supplementary informa on van Groeningen et al. ( 2014)), to Table 1 in our manuscript.
-In the underlying meta-analysis, some of the coefficients were not significantly different from zero (e.g., sandy soil did not significantly impact aboveground biomass.).Were non-significant coefficients s ll used in this synthesis analysis?Assuming so, is this appropriate?My assump on would be not.RESPONSE: Yes, we recognize that some of the categories/coefficients did not show a significant influence (as is now indicated with zeros in the 'earthworm effect' column of Table 1).However, we felt it would more accurately reflect the data to include coefficients based on the mean values reported by van Groeningen et al. (2014) rather than to simply say there was no earthworm effect.We note that these coefficients are generally low (< 0.5) so should not have an outsized influence on our global es mates of the earthworm effect.Addi onally, we note that including a zero for any coefficient in our earthworm effect equa on, would essen ally nullify the es mated earthworm effect in categories with a non-significant effect (i.e., cells with sandy soils, legumes, or receiving > 30 kg N ha -1 yr -1 ).
-Line 142 "We then selected the most important drivers of the earthworm effect" -it is not clear how this was decided/done.The underlying meta-analysis provides many other coefficients that could have been used in this synthesis analysis (e.g., changes in climate, which are o en found to be important).This should be expanded upon in terms of why those predictors were chosen and other not, and what implica ons this may have.

RESPONSE:
We now provide addi onal detail about how we selected the variables included in the analysis.In the Methods sec on we explain that we "selected the most important drivers of the earthworm effect by examining the effect sizes reported by van Groeningen et al. 8 and considering plant and environmental factors where the categories showed clearly differing earthworm effects (e.g., no-effect vs. posi ve effect, as reported for high vs. low N applica on) or had non-overlapping confidence intervals for at least two of the categories (e.g., pH >7 vs. pH ≤ 7).These variables were then cross-referenced against available global data layers, such that we ignored poten al drivers for which there is no available data set (e.g., crop residue applica on rate).The factors thus selected included: crop type, soil pH, soil texture, nitrogen (N) applica on rate, and earthworm abundance (individuals m -2 )." We note factors such as soil organic ma er content and climate were excluded by our approach, since there were not large differences between the different levels (i.e., confidence intervals were overlapping; see van Groeningen et al. (2014) Supplementary Table 3).
-Finally, how was the non-linear func on for earthworm abundance calculated?In the underlying meta-analysis, there are three coefficients for earthworm abundance, so how robust is a func on that has been created from only three data points?RESPONSE: This is a good point, however, we note that there are actually four levels of earthworm abundance reported by van Groeningen et al. ( 2014): < 100, 100-200, 200-400, >400 ind.m -2 .We understand the concern here, as we tried several different approaches before we se led on using a con nuous non-linear func on.We originally generated coefficients using the same approach as for the other factors, but given that the vast majority of the cells in the earthworm density data layer contained <100 ind.m -2 , and the fact that the observed effect was unexpectedly high for this category, we felt that this might have exaggerated the earthworm effect.In choosing a con nuous func on to fit the four points we went with a power func on (see Table 1) that best addressed the higher than expected effect size for the <100 ind.m -2 category.We now feel that this is fairly-well explained in the Methods sec on: "Earthworm abundance, a, was calculated as a con nuous, non-linear func on that provided a best-fit for the available data (Table 1).We use this approach for earthworms rather than categories of abundance (as reported in the meta-analysis) since earthworm abundance for most cells in the global earthworm map was below the threshold for the 'low' earthworm density category (100 individuals m -2 ) reported by van Groeningen et al. 8 .We consider our approach somewhat conserva ve since it allows cells with just a few individuals per m 2 to have a near zero effect of earthworms."We are not sure what addi onal detail we could add to address the reviewer's concern here but are happy to provide more informa on if needed.
-As far as I can tell, the underlying meta-analysis tested each subgroup with a single model (e.g., one model on the effect of soil pH on yield, and another model on the effect of earthworm abundance on yield, etc. etc.).Fonte et al., have assumed that these coefficients (crop type, soil pH, soil texture, and crop-specific fer lizer N rates) can be mul plied together (i.e., a mul plica ve effect, line 147).However, one could equally assume that these effects do not significantly interact with each other, and instead could also be causing an addi ve effect.Unfortunately, the van Groenigen meta-analysis can not be used to determine whether the equa on should be mul plica ve or addi ve, so I would suggest that the possibility is tested, and see how much the results change as a result of this assump on.RESPONSE: We agree that there are several key limita ons with our approach and with nature of the underlying data provided by van Groeningen et al. (2014).We also feel that the reviewer touches on an important point here, but we are not sure that we completely understand what they mean by 'mul plica ve' vs. 'addi ve' effects.We note that by simply mul plying the coefficients together, our approach assumes that there are no interac ve effects between factors -i.e., the effect of one factor does not depend on the level of another factor.It should be noted that when embarking on this study we contacted van Groeningen et al. and obtained the original data set used for the meta-analysis, with the hopes of be er elucida ng some of the poten al interac ons between factors, but a er further explora on we concluded that there were simply not enough datapoints to tease apart interac ve effects and we opted for the approach used here.To address the Reviewer's concern, we have not added the following sentence to the Results and Discussion that touches upon this limita on: "Addi onally, we point out that our analysis assumed simple addi ve effects for the plant and environmental factors that influenced the earthworm benefit, as we were not able to parse out poten al interac ons between these drivers."This study is interes ng as it shows clearly that earthworms contribute to food produc on at the global scale.They showed with a metanalysis that in the whole world the Impact of earthworm on global agriculture is 6.5 % for global grain 2.3 % of legume yield.It is true that up to now, there is not serious measurements of the earthworm poten al contribu on to global agricultural produc on.Most of the informa on comes from small trials and we don't have the idea of their impact at large scale.The results of the analysis are interes ng, we see that they impact cereals and legumes differently.As well their impact is more important in the global south although they have impact globally.They suggest that if there is an improved management of soil biological communi es, the impact of earthworm will be larger.I Like that they men on that it can be some biases in their analysis which can give some over or under es ma ons.The methods used are well explained although I am not an expert on this kind of analysis.The manuscript is well wri en and clear.I think this communica on is important as it is one of the first work that gives the idea of the impact of earthworms and soil biodiversity can have in food produc on at a global scale, which is a proof of the importance of agro -and soil biodiversity for decision makers.So, I recommend the publica on of this communica on without any correc ons, I hope that over me more data will be available to draw more robust and secure conclusions as their data comes from 54 studies (462) points.Kind regards RESPONSE: We appreciate the reviewer's kind words and encouragement.

Reviewers' Comments:
Reviewer #1: Remarks to the Author: Thanks to Fonte et al for making the changes to their manuscript as well as clarifying points within the cover letter.
I am still happy with the manuscript, so I am just responding to a couple of previous points I raised.
-For the edits to the 'non-linear function for earthworm abundance', I would just add into the main text that it is a power function.This information can provide a lot of insight into the form of the shape, which will highlight the remaining of the information given about the non-linear relationship.
-I know see the issue for the additive vs. multiplicative effects!The main text now reflects your code, as from looking again I can see you did use an additive model.However, unless I am mistaken, the equation you give in the main text, is indicating a multiplicative model (a '*' between the predictor terms).A simple edit to change the '*' to a '+' and I would very much agree with everything you have written.
Again, it's a cool paper, and I am glad an analysis like this has been done!

Responses to Referee Comments
Reviewer #1 (Remarks to the Author): Thanks to Fonte et al for making the changes to their manuscript as well as clarifying points within the cover letter.I am still happy with the manuscript, so I am just responding to a couple of previous points I raised.
-For the edits to the 'non-linear function for earthworm abundance', I would just add into the main text that it is a power function.This information can provide a lot of insight into the form of the shape, which will highlight the remaining of the information given about the non-linear relationship.

RESPONSE:
We now clarify that we used a "con nuous, non-linear (power) func on" for the coefficient related to earthworm abundance.
-I know see the issue for the additive vs. multiplicative effects!The main text now reflects your code, as from looking again I can see you did use an additive model.However, unless I am mistaken, the equation you give in the main text, is indicating a multiplicative model (a '*' between the predictor terms).A simple edit to change the '*' to a '+' and I would very much agree with everything you have written.

RESPONSE:
We remain unclear about the reviewer's concern here.If we are not mistaken, our model is indeed addi ve because it does not consider interac ve effects.For example, the effect of fer lizer applica on does not depend on soil texture, but rather operates independently of the other factors.Regardless, we feel that our coefficients should be mul plied together.Based on the way we have constructed them, each coefficient acts as a scalar to either increase or decrease the earthworm effect based on the level of each factor considered.
To clarify the mul plica on steps in the code, we have added the comment, "All coefficient layers are mul plied to inves gate total earthworm effects" to line 92.Within each respec ve crop group block, this func on is wri en in lines 132-133 (grains) and lines 185-186 (legumes).The later summing steps (i.e., lines 146-151 in grains and lines 199-204 in legumes) are calcula ng absolute change in crop produc on values.These changes can be found in the same Dropbox files (h ps://www.dropbox.com/sh/icbv07bhm7djg7m/AAA2xCmFM5K3OqvBQ4PcVxHma?dl=0) and will also be publicly available in our Zenodo repository (10.5281/zenodo.8235224).
Again, it's a cool paper, and I am glad an analysis like this has been done!RESPONSE: We appreciate the suppor ve comments.