Geographic, Taxonomic and Metric Gaps in Biodiversity Research Limit Evidence-Based Conservation in Agricultural Landscapes: An Umbrella Review

Abstract

Agriculture is fundamentally dependent on biodiversity, yet unsustainable management practices increasingly threaten various organisms and ecosystem services. Confronting the global crisis of biodiversity loss requires a thorough understanding of the gaps, clusters and biases in existing knowledge across various management practices, spatial scales, and taxonomic groups. We undertook a comprehensive literature review, synthesising secondary data from 200 meta-analyses on agricultural management impacts on biodiversity in croplands. Our systematic map covers 1885 comparisons (mean effect sizes), from over 9000 primary studies. In the latter, seven high-income countries prevail (notably the USA, China and Brazil), with particular focus on fertiliser use, phytosanitary interventions and crop diversification. This emphasis on individual practices overshadows research at the farm and landscape levels. In secondary evidence, arthropods and microorganisms are most frequently studied, while annelids, vertebrates and plants are less represented. Evidence predominantly stems from averaged abundance data, revealing substantial gaps in studies on functional and phylogenetic diversity. Our findings highlight the need to analyse combinations of multiple practices to accurately reflect real-world farming contexts, and covering a wider range of taxa, biodiversity metrics and spatial levels, to enable evidence-based conservation strategies in agriculture. Given the uneven evidence on agricultural impacts, caution is required when applying meta-analytical findings to public policies and global assessments.

Keywords: agricultural management practices; associated biodiversity; farmland; indicators; synthesis research; systematic map.

FIGURE 1

A conceptual representation of (a) our umbrella review and (b) the multi‐level nature of the evidence base. At the lowest level (level 1), the individual primary studies report one or several paired data comparisons on an outcome of interest. At level 2, meta‐analyses report one or several mean estimates (i.e., effect size) of an intervention across several primary studies. Level 3 is a descriptive state of the art (i.e., a systematic map) of 200 meta‐analyses quantifying a measure of impact of an agricultural practice on biodiversity. These meta‐analyses presented 1885 effect sizes from 69,850 paired data reported in > 15,000 total primary studies, of which > 9000 primary studies are unique (i.e., one primary study could be used in several meta‐analyses).

FIGURE 2

Locations of the primary studies included in the 200 meta‐analyses and the top 10 agricultural interventions analysed by world region. Some meta‐analyses did not provide the references of their original studies, and the full text or locations of some primary studies were not available, thus not accounted for. Among the 9080 unique primary studies, 5600 provided enough information to code the location. In the barplots, yellow bars: Field‐level practices; green bars: Farm‐level practices; orange bars: Landscape‐level practices (see Table 1 for definitions).

FIGURE 3

Alluvial diagram showing the distribution of effect sizes across agricultural interventions (left bar), taxa at the kingdom level (middle bar), and biodiversity metrics (right bar) for the 200 meta‐analyses studying the effects of agricultural interventions on biodiversity. Each bar represents the 1885 effect sizes of our database, subdivided into categories whose width is proportional to the number of data it contains. The width of the flows between the categories is proportional to the number of primary studies examining each combination of agricultural interventions, taxa, and biodiversity metrics. For example, 1345 effect sizes focus on ‘individual practices,’ from which 572 examined impacts on Animalia, 589 on microorganisms, and 173 on Plantae. For clarity, on the middle bar: ‘Archaea’ (n = 3) were grouped within ‘Bacteria’, and ‘Chromista’ (n = 3) were grouped within ‘Fungi’. On the right bar: ‘Other metrics’ refers to ‘Phylogenetic diversity’ (n = 3) or ‘Trait‐based’ (n = 4) metrics. ‘Multiple metrics’ indicates effect sizes aggregating multiple or unspecified biodiversity metrics.

FIGURE 4

Taxonomic resolution for each major Kingdom and for the total 1885 effect size data. The total number of effect sizes per Kingdom is depicted at the bottom of each plot, while the available information at lower taxonomic ranks is represented as a percentage of the total effect sizes.

FIGURE 5

Evidence map of the 1885 effect sizes extracted from the 200 meta‐analyses: (a) heatmap representing the number of effect sizes by agricultural management practice (y‐axis, top) and biodiversity metric (y‐axis, bottom), and for each taxonomic group (x‐axis); (b) stacked bar chart (in %) representing the direction of the effect sizes by agricultural management practices and biodiversity metrics (y‐axis, as in panel a), for all taxonomic groups (see also Figure S7 for min‐max range of the mean effect size values). In panel a, tile labels and colour intensity represent the number of effect sizes. Taxonomic groups are presented at Kingdom level, except animals at Phylum level (see Table 2). For clarity, Archaea are grouped with Bacteria, Chromista are grouped with Fungi, and effect sizes depicting phylogenetic and trait‐based metrics (n = 7 in total) are not shown. The number in parentheses on the x‐axis represents the total number of effect sizes for each management practice and biodiversity metric. Tile colours: Red for animals, blue for microorganisms, green for plants, orange for multiple or unspecified kingdoms.

FIGURE 6

Over‐ or under‐representation of taxonomic groups in syntheses on biodiversity in agroecosystems (our study) compared to the general biodiversity literature (data from Mammola et al. 2023). The y‐axis represents the ratio between the percentage of occurrence of each taxonomic group in both studies, the x‐axis represents the percentage of occurrence of taxonomic groups in Mammola et al. (2023) after harmonisation with respect to the Catalogue of Life classification. The four major groups (see Figure 3, Table 2) are presented with grey background boxes, and detail is given with white background boxes. Microorganisms are detailed at Kingdom level and animals at Phylum level. When no box is drawn, the group is absent from one of both studies, that is, Chromista is absent from Mammola et al. (2023), Mollusca and Platyhelminthes are absent from our study. For clarity, Protozoa and Algae are hidden behind Platyhelminthes and Mollusca, respectively. Raw data are given in Table S3.