Land use change effects on ant diversity in Neotropics.xlsx

<i>Data collection</i>We followed the PRISMA-EcoEvo methodology (O'Dea et al., 2021). First, we conducted a search for studies on the effects of land use change on ants in Brazil using the Ants of Brazil Project database (Feitosa et al., 2022; Schmidt et al., 2022). We also performed a preliminary search (naïve search) for relevant studies in the “Web of Science databases - Main Collection (Clarivate Analytics)” (www.webofscience.com), “Scopus” (www.scopus.com), and “SciELO.ORG” (www.scielo.org). Initially, we carried out a preliminary search using pertinent keywords related to the topic (Grames et al., 2019). To achieve this, we categorized the keywords into four sections, following the PICO/PECO model (population, intervention/exposure, comparator, outcome; Haddaway et al., 2016). The 'Population' section pertained to the specific population under study (<i>e.g.,</i> Ants), the 'Intervention/Exposure' section referred to various exposure or environmental factors (<i>e.g</i>., Land use), the 'Comparator' section related to what the exposure would be compared to (<i>e.g</i>., Biomes), and the 'Outcome' encompassed the measured variables (<i>e.g</i>., Diversity). The naïve search conducted in November 2023. The titles, abstracts, and keywords of these studies were exported, and additional keywords were identified using the 'litsearchr' package (Grames et al., 2019), within the R software (R Core Team, 2021). In the analysis of the new keywords, we included the term 'species composition' as an outcome and performed a subsequent search. In addition to the search using English terms, we also conducted searches in Portuguese and Spanish in the same databases. The new search returned 3,488 studies in Web of Science, 1,573 in Scopus, and 16 in SciELO.ORG. We also included an additional 72 studies from the Ants of Brazil Project, bringing the total to 5,149.<i>Exclusion and inclusion criteria</i>We removed duplicate studies and conducted a screening of titles, abstracts, and full articles. Our inclusion criteria focused on studies that specifically assessed the impact of land use changes on ant assemblages in Brazil. We applied exclusion criteria, including: (1) studies conducted outside Brazil; (2) studies unrelated to ant communities or assemblages; (3) studies not addressing the effects of land use changes; (4) studies lacking a comparison between natural habitat and anthropogenic land use. We also excluded non-case studies (<i>e.g</i>., reviews) and studies dealing with anthropogenic disturbances not related to land use categorization (<i>e.g</i>., chronic disturbances like logging or non-timber forest product extraction). Furthermore, we omitted studies focusing on ecological succession (<i>e.g</i>., succession or restoration) and those where it was impossible to separate the impact on ant assemblages from other evaluated organisms (<i>e.g</i>., macrofauna). Out of the initial pool of 5,149 studies, 128 remained for the qualitative review, and 48 were eligible for the meta-analysis.<i>Qualitative review</i>From the 128 studies, we extracted qualitative data including the study ID (citation), publication year, language, journal, impact factor, biome, sampling season, vegetation type, anthropogenic land use, sampling methods, sampled strata (epigeic, leaf-litter, subterranean, and arboreal), and response variable, such as species richness and abundance. We are counting votes for qualitative data from the studies and assessed whether land use affects species compositions. Species composition refers to the differences in the identity of ant species between natural habitats and anthropogenic land uses, extracted from analyses such as PERMANOVA (Permutational Multivariate Analysis of Variance) and ANOSIM (Analysis of Similarities).<i>Extracting and calculating effect sizes</i>For the meta-analysis, we gathered data on the most frequently assessed response variable, including species richness and abundance. Species richness was reported in studies as either observed species richness or estimators (<i>e.g</i>., Chao1, Chao2, Jackknife1, individual-based rarefaction, and sample-based rarefaction). Abundance was reported in studies as number of ant workers and species frequency. For these response variables, we collected mean values, sample size, and dispersion metrics such as variance, standard deviation, standard error, and 95% confidence intervals. We extracted these data from the text, tables, and figures of the studies. When the data was solely presented in figures, we utilized WebPlotDigitizer version 4.6 to extract the mean and dispersion metrics (www.automeris.io/WebPlotDigitizer/). We standardized all the dispersion metrics into standard deviation. We also obtained data regarding vegetation type and anthropogenic land uses. Vegetation types were classified into various categories, encompassing forest types such as primary and secondary forests, riparian forests, and woodland savannah, as well as grassland, “restinga”, savannah, “vereda”, mangrove, and “canga”. The anthropogenic land use data were classified into agriculture and pasture categories when these two anthropogenic uses were sampled jointly, annual agriculture (<i>e.g</i>., soybean plantation), perennial agriculture (<i>e.g</i>., coffee plantation), mining, pasture, silviculture (<i>e.g</i>., <i>Eucalyptus </i>spp. plantation), and urbanization.