Despite the repeated mantra that "correlation does not imply causation", ecological studies not amenable to experimental settings often rely on correlational patterns to infer the causes of observed patterns. In this context, it's of paramount importance to build a plausible hypothesis and take into account potential confounding factors. The paper by Aizen and collaborators (2023) is a beautiful example of how properly unveil the complexities of an intriguing pattern: The decline in yield of some crops over the last few decades. This is an outstanding question to solve given the need to feed a growing population without destroying the environment, for example by increasing the area under cultivation. Previous studies suggested that pollinator-dependent crops were more susceptible to suffering yield declines than non-pollinator-dependent crops (Garibaldi et al 2011). Given the actual population declines of some pollinators, especially in agricultural areas, this correlative evidence was quite appealing to be interpreted as a causal effect. However, as elegantly shown by Aizen and colleagues in this paper, this first analysis did not account for other alternative explanations, such as the effect of climate change on other plant life-history traits correlated with pollinator dependence. Plant life-history traits do not vary independently. For example, trees are more likely to be pollinator-dependent than herbs (Lanuza et al 2023), which can be an important confounding factor in the analysis. With an elegant analysis and an impressive global dataset, this paper shows that the declining trend in the yield of some crops is most likely associated with their life form than with their dependence on pollinators. This does not imply that pollinators are not important for crop yield, but that the decline in their populations is not leaving a clear imprint in the global yield production trends once accounted for the technological and agronomic improvements. All in all, this paper makes a key contribution to food security by elucidating the factors beyond declining yield trends, and is a brave example of how science can self-correct itself as new knowledge emerges.
Aizen, M.A., Gleiser, G., Kitzberger T. and Milla R. 2023. Being A Tree Crop Increases the Odds of Experiencing Yield Declines Irrespective of Pollinator Dependence. bioRxiv, 2023.04.27.538617, ver 2, peer-reviewed and recommended by PCI Ecology. https://doi.org/10.1101/2023.04.27.538617
Lanuza, J.B., Rader, R., Stavert, J., Kendall, L.K., Saunders, M.E. and Bartomeus, I. 2023. Covariation among reproductive traits in flowering plants shapes their interactions with pollinators. Functional Ecology 37: 2072-2084. https://doi.org/10.1111/1365-2435.14340
Garibaldi, L.A., Aizen, M.A., Klein, A.M., Cunningham, S.A. and Harder, L.D. 2011. Global growth and stability of agricultural yield decrease with pollinator dependence. Proceedings of the National Academy of Sciences, 108: 5909-5914. https://doi.org/10.1073/pnas.1012431108
DOI or URL of the preprint: https://doi.org/10.1101/2023.04.27.538617
Version of the preprint: 1
This is a very nice and compelling paper challenging the mechanisms behind previously reported trends showing that pollinator-dependent crops are declining in yield growth faster than non-dependent crops. Exploring confounding variables such as growth form is a clever and needed addition. As pointed out by Reviewer 1, global analyses are powerful, but data management is complex, and the devil is in the details. Hence, both reviewers make minor but fair questions about the choices taken to manipulate the original data. I think backing up some of their concerns with additional supporting analysis (when data quality and quantity permits) will ensure the results are robust.
In particular, I am concerned with the interpretation of growth rate. In the text is mentioned that “a negative growth rate can be taken as evidence of long-term yield decline”. In parts of the text is unclear what "yield decline" refers to. My understanding is that behind these trends, we can find changes in the area occupied by the crop, changes in management practices, and changes in production per hectare. I understand that this last mechanism (changes in production per area) is the most tightly linked to pollinator declines or climate change effects, which is the focus of the introduction/discussion. Is there any elegant way to get closer to analyzing trends in yield per cultivated area (i.e. is area cultivated per year and crop available)? Alternatively, I think this should be noted early in the introduction to avoid any misinterpretation of the results. Also, Fig 1 is intriguing, as it seems the extreme years are highly influential. Do you think starting in 1962 or cutting off in 2019 would change the results?
I also have a suggestion to reinforce the main analysis. I think it will be illustrative to add a variance partitioning plot. This can be done using a Venn diagram, where you calculate the proportion of variance explained by the model with only one variable (model 1a), the proportion of variance explained by the other one (model 1b), and the joint explained variance (model 2). I might miss some technical details here, but I think exploring this would help the story.
Lastly, I miss some discussions on recent advances in plant trait correlations and trade-offs, where the growth form and pollinator dependence are discussed along with other traits. e.g. Lanuza et al. 2023 (10.1111/1365-2435.14340), Friedman, 2020 (https:// doi.org/10.1146/annurev-ecolsys-110218-024638 ); Paterno et al., 2020 (https://doi.org/10.1073/pnas.19106 31117 ); Roddy et al., 2021 (https:// doi.org/10.1111/nph.16823 ).
I think the changes proposed by reviewers and myself should be viewed as a robustness check, but they are mostly minor concerns. I am looking forward to seeing the final version of this stimulating paper.