This manuscript presents a protocol for conducting a systematic review on a very timely issue of practical and theoretical importance. The aim is to collate the evidence on the impacts of polycultural cropping on crop yields and biodiversity. The manuscript is clearly well-researched and beautifully written. However, there is a range of potential improvements and clarifications that should be considered by the authors. It is a god start (and I highly value the idea of publishing a protocol) but it needs more work to be usable and publishable, specifically:

1. The introduction does not articulate clearly why the two research questions (effects on crop yield vs. wildlife biodiversity) have to be reviewed together in a single systematic review rather than two separate ones. Typically, only closely related outcomes are reviewed together because it makes for a more focused question and easier searching for evidence. Is there a reason to examine these two together? Do the empirical studies commonly report both types of outcomes? If you aim to answer both questions together then only include studies that test both effects together. This will also require a multivariate meta-analysis approach if you get to that stage.

2. Table 2: Components of RQ2 – Outcome states” “Change in field-scale biodiversity, e.g. change in the abundance of one or more species…” – looking at the change of abundance of one species does not seem like a logical indicator of the changes at the biodiversity level. For example, if the polycultural cropping resulted in an increase of a single weed species, the effect is on this species abundance in positive, but it will likely also have a negative effect on other non-competing species, reducing the overall biodiversity, so the effect would be really a negative one. If the aim of the synthesis is to look at the effect on biodiversity, then only established indices of biodiversity should be used (and explicitly listed as such). This is partially matching with the more detailed description in Table 3, but this description also does not specifically exclude single-species effects. Also, it seems like beneficial invertebrates and fungi are considered as potential taxonomic groups of interest, but how about other taxonomic groups like vertebrates or plants – would they be included as well?

3. Search strategy: the authors only plan to use two academic databases (WoS, CAB) and one online search engine (Google Scholar). I recommend using Scopus as another broad-coverage main search source and moving Google Scholar to supplementary searches. Google Scholar is based on a “black box” non-reproducible search engine algorithm which spits out thousands of records and is not suitable as a main search source. You should also consider using another source for searching for grey literature specifically targeting theses.

4. Search strategy: please provide a full search string exactly as it will be used in a given database (ideally for all databases/search platforms – they differ a lot in the search string syntax and available filters) – at the moment the WoS example is just a Boolean group of keywords without field identifiers, so it is not clear if you are going to apply these search terms to titles, abstracts, author or database keywords, full texts, ontologies etc. time limits should be included too.

5. Please check if your search strings actually retrieve your benchmark articles and how many (you do not need 100% for each separate search engine). This assessment should be part of the work done to prepare the workable protocol, not the actual work after the protocol, because the testing results will almost certainly require changes to your search strings and final search strings should be in the protocol.

6. Please also report how many records will each of your search string retrieve from each of your planned search sources (except personal contacts and snowballing) – this will provide an estimate of how much time screening will take and whether it is doable.

7. Please get an Information Specialist involved in designing your search or a researcher with extensive experience in planning and conducting systematic reviews.

8. Duplication removal of the records collated from different sources is part of the typical process but is not mentioned in the manuscript. Also, duplicated studies and datasets are possible across published articles and grey literature – please describe how you will handle this.

9. Screening – will all records be screened independently by at least 2 reviewers during both screening stages?

10. Piloting is critical for preparing a comprehensive and realistic systematic review protocol. Please pilot all stages of the review. You should use the results of the piloting (e.g., starting from a set of 100 randomly drawn references from your final search strings) to estimate the expected numbers of records/studies at each stage of a systematic review process. Please report these numbers in the report. Also, based on these estimates you may decide to adjust your review strategy and report the final one (as well as the whole process getting there) in the protocol.

11. Line 255: “Replicates are not truly independent” – what would be a fully independent replication for this type of primary study? – e.g., would having different fields within the same farm sufficient? Or they need to be on different farms/regions/countries? How would this depend on the crops or taxa studied?

12. Lines 263-264: “If none of the above factors are identified, the studies’ susceptibility to bias will be defined as ‘low’, and its validity will be defined as ‘high’.” – how about other methodological factors, such as sizes of the plots used, randomisation procedures or blinding of the measurements?

13. Lines 264-266: “all reviewers will critically appraise a random subset of articles (approximately 5% of the total number of articles selected for inclusion)” – maybe go for a fixed number first, like 10 studies, and then do additional rounds of similar sets until you get acceptable agreement rates? In your plan, it is unclear what will happen after the initial round? E.g., if there are many inconsistencies (very likely) what this will mean for the apprising the remaining studies? Also, does this mean that the remaining studies will be assessed by a single reviewer only?

14. Lines 273-274: “Study data will be extracted and recorded in an MS Excel data sheet, the design of which will be tested and agreed upon by the reviewing team in advance of the review commencing.” – why it has not been tested yet? It is highly recommended to find out as much as possible before finalising the protocol to save time and frustration later on. Please use a subset of randomly selected studies to pilot your data extraction independently by at least 2 reviewers, refine your plans and report the process in the protocol.

15. Line 276: “Where sufficient, good-quality data exists, it will be included in meta-analyses” – please define what will be considered “sufficient” and “good-quality” in this context.

16. Line 315: “Where sufficient sample sizes of studies are available, meta-analyses will be performed” – please define what will be considered “sufficient sample sizes of studies”.

17. Lines 172-193: Data extraction: for each variable define data type: numerical, categorical, free text, etc. this should be in a detailed table (could be a supplementary table). Such a table should contain the full final list of variables to be extracted and full meta-data for each variable, so the data extraction can be replicated even by a person outside the project team. Small deviations from that plan can be then acknowledged in the final report. At the moment it is a very vague description of categories of data that the authors would like to collect, not a concrete plan. It would be extremely important to pilot data extraction sheet and then revise the meta-data accordingly, with detailed descriptions on how to do the extractions. The pilot-extracted data could be used for testing the plans to conduct a meta-analysis.

18. Lines 297-298: Which digital tool will be used to extract data from the figures?

19. Lines 299-302, for a complex data extraction, like in the proposed project, the initial independent extraction of a small portion of the papers will not ensure high data quality of the whole data set. If the majority of the data is extracted by a single person, the whole data set should be then cross-checked carefully and independently by another experienced researcher.

20. Lines 319-321: “The risk of publication bias in the literature sample (e.g., due to a tendency for only ‘significantly positive’ results to be published, or due to evidence gaps in the published literature) will be assessed using funnel plots.” – this is not sufficient as standard funnel plots do not perform well fir highly heterogeneous data and small sample sizes. Also, more than one approach should be used for testing publication bias.

21. Line 321: “All analyses will be conducted in R (R Core Team, 2020).” – which specific packages and functions? How will you deal with missing data and data non-independence? Have you piloted this sort of analyses on a small set of pilot-extracted data or a simulated data set? I highly recommend doing so or not promising to do meta-analysis within this project. I suspect that data will be too messy anyway and will require carefully identifying the subsets of studies and effect sizes that can be combined together to get something biologically and statistically meaningful. This could be done in a follow-up study, with the current one acting as a systematic map of available evidence. Alternatively, consider only meta-analysis for only the yield studies which might be more homogeneous in their measures than the biodiversity studies.

22. Data synthesis: for meta-analysis there is not enough details provided, even the general type of the model is currently not specified.

23. In Discussion, provide limitations of your approach. For example, only English-language searches and studies will be used, so you cannot claim it is a truly comprehensive or global systematic review.

Thank you for addressing all my comments.