Search terms and languages
We relied on different reviews, meta-analyses, books, reports or scientific articles (see for example [8, 13,14,15, 19, 27,28,29,30,31]) to develop a list of ISGs used in various research domains (nature conservation, ecosystem functionality, biodiversity indicator or ecosystem services). We identified twenty-two ISGs: flora, mammals, birds, reptiles, amphibians, spiders, bees, parasitoid wasps (ichneumonids and the braconids), orthopterans, butterflies, beetles (carabids, coccinellids and staphylinids), syrphids, lacewings, ants, slugs, snails, annelids, nematodes, soil mites, springtails, millipedes, and centipedes. These ISGs cover a wide range of ecological niches and trophic levels (Fig. 1). To collect the literature corresponding to this preliminary list we developed a set of ninety-three search terms (Additional file 3, see Population in Fig. 2). We acknowledge that soil microorganisms (i.e., fungi, bacteria, or archaea) are also an important component of biodiversity in agriculture (also in term of ecosystem services), but we finally decided not to include them as ISG due to the too large number of articles that were collected for this group (between 20′000 and 50′000 additional references depending on the keywords).
AMPs in European arable systems are spatially and temporally diverse and it is thus challenging to group them into broad but meaningful categories. Here, we considered ten main AMPs categories: soil preparation, fertilization, sowing, irrigation, crop protection, harvesting, mowing, grazing, intermediate cropping, and ecological infrastructure implementation. We built a set of fifty-five search terms to effectively gather literature corresponding to these main AMPs (Additional file 2; see Intervention in Fig. 2).
To record the effects of AMPs on ISGs, we are interested in studies reporting a difference or a change in the diversity, abundance, or survival of the latter. We combined eleven search terms (see Outcome in Fig. 2) to include taxonomic, structural, and functional diversity indices. Each of these terms was then associated with the word “species” (i.e., “species richness”) or with one of the ISG search term (i.e., “spider richness”). When the search platform offered the possibility of using proximity operators, they were combined with the Boolean operator “NEAR/3” (in Web of Science) to find records where both terms are within three words of each other (i.e., “richness of spiders”), otherwise they were combined with “AND”.
To restrict the literature search to European agricultural environments, we defined two additional sets of keywords. Six “Environment” keywords aimed at focusing on agricultural landscapes (crop and grassland), and forty-nine “Location” ones restricted the search to European countries (see Environment and Location in Fig. 2). The geographical range of the study includes most continental Europe (Additional file 5), apart from Russia and Turkey (and countries further east of the latter), and islands as that are commonly known to have different conditions from the continent (i.e., species guilds, types of agriculture or weather conditions).
To account for alternative spelling or hyphenation, search terms were truncated and a wild card (*) added to include alternate forms of the word (i.e., fertili* to account for fertilizer, fertilizers, fertilization, fertilizing, and associated British spelling) or a dollar sign ($) to allow the inclusion of an additional character (i.e., bird$ to account for bird or birds, but not for birding or birdwatch; only available on Web of Science platform). Quotation marks were used to limit the search to exact phrases (i.e., “pest control”). All literature searches will be performed with terms in English.
Finally, as this systematic map is conducted as part of a project aiming at measuring and optimizing the environmental impact of farms in Switzerland (Indicate project; [25]), we focused on the type of agricultural fields present in the Swiss lowland agricultural environment (Additional file 6). These fields were grouped in four main categories: annual crops, perennial crops, grasslands, and ecological infrastructures (Additional file 4). They also correspond to the main field types cultivated in Europe (except for olive groves, citrus, nuts, and some fruits [26]), which ensures that the data collected in this study and the conclusions we will be able to draw from it will be useful and applicable to all European countries. We did not define specific keywords to select for field types during the literature search phase, but we will use them as inclusion/exclusion criteria during the screening process (see Eligibility criteria).
Search strings
Search terms within categories will be combined using the Boolean operator “OR”, and between categories using the Boolean operator “AND” (Fig. 2). This implies that studies must include at least one term of each of the five categories to be retained. When possible, the search will be restricted to the article title, abstract and keywords, except for the “Location” search terms that will be screened across the full text. Due to keyword number limitations in the searches, we conducted a literature search separately for each ISG. This means that the keywords for the categories Intervention, Outcome, Environment and Location were combined to each ISG keywords successively (see Fig. 2 for an example).
Publication databases
We will search for relevant literature on Web of Science Core Collection and CABI platforms.
Internet searches
We will conduct an internet search on the Google Scholar website using a simplified search string (Additional file 7) at the whole text level (as it is not possible to restrict the search fields in Google Scholar). The first 500 results will be exported in Excel format and screened. To reduce the algorithm biases associated to previous internet searches, browser history and cookies will be disabled during the internet search and the “private” navigation mode used.
Supplementary searches
A search for grey literature at the European scale would go beyond the scope and resources of this systematic map. So, supplementary searches will be carried out for Switzerland only, to specifically collect relevant literature that has to be considered in the context of the Indicate project. We will search for grey literature in English, French, German, and Italian on Swiss specialized websites (Additional file 7). In addition, we will ask for additional Swiss grey literature through the professional networks of the research team. The final analyses will be conducted with and without the grey literature to assess its impact and avoid any grey-literature-based bias in the conclusions.
Comprehensiveness of the search
To evaluate the comprehensiveness of the literature search, we will compare different search strings results with a test-list of articles considered to be relevant. To produce the test list, we first selected sixty articles that we considered to be appropriate and wanted to obtain in the literature search. Secondly, to ensure the diversity and representativeness of the test-list, we added pertinent literature cited in five key publications on biodiversity in agriculture: a review on the biodiversity in agricultural areas [12], a review of soil biodiversity [14], a European project on agricultural biodiversity [32], and two comprehensive research articles on Swiss biodiversity in agriculture [33, 34]. This resulted in a test-list of ninety articles (Additional file 8), published over a period of thirty years (from 1991 to 2021) in thirty-nine different scientific journals.
Article screening and study eligibility criteria
Screening process
First, as we expect the different search sources to report several times the same references, duplicates will be removed based on the DOI identifier, and on title for references without DOI. Then, the study screening process will successively be performed at the title, abstract and full-text levels. At each level, articles will be classified as relevant (included in the review) or irrelevant (excluded from the review), or uncertain. In the latter case, articles will be passed to the next level of screening and reevaluated (i.e., articles uncertain at the title screening level, will be passed to the abstract screening stage). For each rejected article, we will record the level (title, abstract, full text) and the reason (list of choices) of exclusion, which will be published with the systematic map. Full articles will be collected from the literature access of the Agroscope, the ETH Zürich and the University of Lausanne (UNIL).
To guide reviewers’ choices of including or excluding an article, we defined a set of criteria (see Eligibility criteria), and to assess the repeatability of the screening process we will compare the choices made by different reviewers. To do so, a subset of 150 articles will be assessed by two reviewers and their agreement evaluated using a Cohen’s kappa coefficient (k > 0.6 considered as consistent). This will be repeated at each screening level. In case of inconsistency (k < 0.6), the reviewers will discuss to resolve the reasons of inconsistency in their choices, adapt the criteria, and then screen a new subset of articles, until consistency is reached. Even when consistency will be sufficient (k > 0.6), reviewers will discuss and solve the remaining disagreements to ensure a high replicability.
Eligibility criteria
To be included in the systematic map, articles must fulfill nine conditions:
-
Eligible population: articles must include at least one of the ISGs.
-
Eligible intervention: articles must include at least one of the AMPs. One-time events will be excluded (i.e., unique pollution event), as well as studies which do not directly study the effect of AMPs on ISGs (i.e., bird population fluctuation through time in agricultural areas, without specific AMPs associated).
-
Eligible comparator: articles must compare ISGs before/after intervention, between an intervention and a control, or between different interventions. Studies reporting ISGs without further comparison will be excluded.
-
Eligible outcome: articles must report a measure of species diversity (i.e., richness, abundance, or evenness), otherwise they will be excluded.
-
Eligible environment: articles must report research conducted in lowland agricultural landscape. Studies conducted in non-lowland farming landscapes (i.e., forests, or alpine environments) will be discarded.
-
Eligible location: articles must report studies conducted on the European mainland. Studies conducted outside the geographical area under consideration will be discarded.
-
Eligible study design: articles must report and analyze monitoring or experimental field-data. Modelling papers, books, reviews, or meta-analyses will be excluded.
-
Eligible crop types: articles must report the effect of an AMP on an ISG in one of the crop types considered in the present study.
-
Eligible language: articles must be written in English.
We developed a list of inclusion/exclusion criteria to guide and standardize the literature screening process (Additional file 9).
Study validity assessment
We will not perform a critical appraisal of study validity, but rather extract study characteristics (see Data coding strategy) that we will combine to evaluate the relevance of the articles. To do this, we will collect study information (geographical extent, type of design, duration of the study, number of sampling sites, statistical analyses), ISG specifications (ISG focus, ISG monitoring method description, species list, number of diversity measures calculated) and AMP specifications (AMP focus, intervention description). These extracted study characteristics will then be combined to obtain a study validity score, which will represent the relevance of each article based on the goals of the present study. We therefore intend to attribute to each article a value of fit with the study question (external validity measure), rather than a risk of bias evaluation (internal validity).
Data coding strategy
The data extracted from the articles will include bibliographic information, study design and characteristics, ISGs and AMPs studied, and effects reported. The data coding strategy will consist in the combination of three tables, linked by a unique article identifier (see Additional file 10 for an overview of the data coding strategy). The first table, “Article References”, will contain the articles’ bibliographic information. The second table, “Study Characteristics”, will contain information about study location and design. The last table, “Species Diversity”, will summarize the effects of AMPs on ISGs measured outcome according to a four-level classification (positive, negative, neutral, or unclear).
To assess the repeatability of the data collection process, we will compare the data extracted by different reviewers on a subset of 50 articles. For each data to be extracted (Additional file 10), we will calculate the reviewer’s agreement as the percentage of fit. Cases of disagreement will be discussed to improve the collection repeatability (i.e., by clarifying the definition of a variable, reformulating the different categories of a variable in the case of multiple choices, or adding additional variables if necessary). We will report in the systematic map the repeatability of each of the extracted variables.
Study mapping and presentation
In the systematic map we will provide answers to the primary and secondary questions developed in the present protocol, in the form of descriptive texts supported by tables, figures and maps. As the tables described in the Data coding strategy might be further developed during full-text assessment, we will provide all final table structures and contents in the upcoming publication, as well as the final database produced in the frame of this review. To facilitate the use of these files in other research projects, we will provide detailed definitions of each column and coding schemes, as well as R codes [35] used to perform the queries and analyses needed for the review mapping. Detailed information about the articles included/excluded at the different stages of the screening process will be reported, in addition to any eventual modification to the present protocol. Finally, we will discuss strengths and gaps identified, and propose future research to improve our understanding of the effects of the agricultural management practices on biodiversity.