A global inventory of animal diversity measured in different grazing treatments
Synthesis and data extractionData were collected using a literature search of Web of Science for peer-reviewed journal articles published between 1970 and November 2019. We conducted two sets of searches to capture grazing with discrete comparisons (e.g., grazed/ungrazed, moderate vs. heavy intensity grazing) and a range of grazing intensities. The search terms used for each were as follows 1) (graz* OR livestock) AND (exclosure* OR exclusion OR exclude* OR ungrazed OR retire* OR fallow* OR fence* OR paddock*), 2) (“grazing intensity” OR “grazing gradient” OR “stocking rate” OR “rotation*grazing”). Our synthesis includes domesticated and wild grazer species, with the latter defined as an undomesticated species naturally occurring in the study area during the study. Wild grazers are typically native species to the region (e.g., the American bison in Western North America) but can include non-native species that are naturalized in the area (e.g., feral horses on Sable Island).We excluded any study that did not test the effect of grazing animals. A grazer was defined using the definition provided by the authors of the respective study to account for the proportion of forage types in a herbivore’s diet that varies between seasons and habitats. For example, we included animals where their diet is assumed to come from all (e.g., cattle, sheep), most (e.g., wapiti, kangaroos), or some (e.g., deer species) grass species. However, within the included studies, these animals were classified as grazers as most of their diet was grass for the duration of the study. For added clarity about the herbivore composition in each study, we extracted a list of any herbivores listed in the paper regardless of foraging type or if any data was provided.We only included studies that measured animal diversity or abundance as a response variable and included data we could extract or contact the author to obtain9. We included any study with a grazing treatment and included observations within these studies of any grazed and ungrazed sites. All studies with grazing included a comparison to either ungrazed sites, different grazing practices (e.g., cattle vs. sheep), and/or differences in intensities (e.g., heavy/light, extensive/intensive). Studies that only measured plants or soil biota were excluded because syntheses of grazing effects on these groups have already been conducted7,11,12, and our goal was to provide a robust inventory of animal diversity. However, if a study included plants, lichens, or fungi in addition to animals, we included this data. Studies discussing marine grazing or aquatic systems were also excluded. From these preliminary filters, we identified 3,489 published manuscripts. We reviewed these 3,489 published articles and found 245 studies that surveyed animals in grazed sites. In total, we extracted 16,105 observations for over 1,200 species.We extracted 28 variables that focus on management systems, assemblages of grazer species, ecosystem characteristics, and survey type (Table 1). The latitudes, longitudes, and elevations of each study were included when provided for use with geospatial data. In addition, we included variables about the study site’s disturbance history, including last time grazed, if a flood event or fire had occurred, if fertilization was used, if the area was open or fenced off, and if the area was publicly or privately owned. Furthermore, the timeline for the study (i.e., the years the authors initiated and completed the study) was also provided. Study initiation was described by the authors and could include when the grazing treatment started, another treatment was applied, and/or animal surveys began. These timeline columns can be useful in identifying long-term studies and differentiating single grazing events or multi-year experiments. Finally, we generalized the characteristics of the ecosystem of the sites used in each study based on the climate and dominant vegetation.Table 1 The attributes and description of the metadata.csv file that lists the general characteristics of each study.Full size tableWithin the grazing data, we included information about the grazer when provided, including any measurement of the intensity of grazing (e.g., animals per hectare, the height of residual vegetation). We also provided two columns that detailed whether the study tested grazing effects using a discrete comparison or gradient of intensities (Table 2). The value for the target specimens extracted may represent either a single observation or a summarized statistic (e.g., mean animals per site). We identify unique observations as “count” and summarized statistics by the metric used, such as mean, median, standard deviation (column stat in grazingData.csv). When possible, we also included any record of other grazers that co-occurred with the observed grazer species. The data for these variables were extracted from the papers by a single researcher who read through each paper and filled in available data on the mentioned variables.Table 2 The attributes and description of the grazingData.csv file that has the extracted data from each study.Full size tableWe extracted information about the target specimen, site, year, experimental replicate, and response estimate (Table 2). We included multiple categorizations of the target species to assist future users in synthesizing similar taxa (Table 2). When a species name or genus was provided, we conducted a search query (see detailedTaxa.r) through the global biodiversity information facility (GBIF.org) to determine the taxonomic classification of the species, including kingdom, phylum, order, class, and family. When a species name was not included, we provided the lowest taxonomic resolution available. We also included a broader classification of ‘higherTaxon’ to distinguish plants, fungi, vertebrates, and invertebrates. These columns may help group similar species together for community-level analyses. Lastly, we included the characteristic of the plant community (i.e., planted or self-assembled, tilled, and its vegetation class) when plant data was reported.Patterns among studiesMost of the studies took place in the United States (26%), Australia (9%), and the United Kingdom (7%) (Fig. 1). As expected, most studies were conducted in grasslands (n = 206), followed by forests (n = 92) and shrublands (n = 82) (Fig. 2). We included publications from the entire range of years (i.e., 1970–2019), but most were published after 2000 (76%). The number of sites in a study and the study duration showed a bimodal distribution with a long tail (Fig. 3). Most studies included one to eight different sites, and few were conducted longer than five years (Fig. 3). A few studies were highly replicated, while many were limited in their replication (Fig. 3).Fig. 1The locations of studies that measured the response of animals to domestic or wild grazing.Full size imageFig. 2The number of grazing studies conducted in ecosystems around the world. We generalized the characteristics of the ecosystem of the sites used in each study based on the climate and dominant vegetation community. We separated grassland communities into those that were (a) semi-natural without recent cultivation or seeding (self-assembled), (b) recently cultivated or had supplemental seeding (planted/cultivated), and (c) a combination of both. In most grasslands, the cultivation history was unclear.Full size imageFig. 3The number of independent sites surveyed and the duration of each study. Most studies were conducted at either a single site or with some replication (e.g., 6–8 sites). Similarly, most studies were either conducted in one year ( >30%) or over a few years (e.g., 3–6 years). Very few studies (32) or lasted longer than 15 years.Full size imageSite and management data were not reported in all studies, as found in other reviews of grazing impacts on ecological processes10. Of the studies that mentioned the ownership status of the land used, 46% were on private land, 42% were on public land, and 12% had a history of both public and private ownership. Most studies included binary comparisons (56%) of grazed vs. ungrazed plots or sites, though some also included a discrete (22%) or a continuous estimate of grazing intensities (18%).Of the studies that reported plant community origin, 76% were self-assembled, 17% were planted communities, and the remaining included sites were a combination of the two. Domesticated grazers as the focal herbivore made up 67% of the studies, with 12% of the studies having wild grazers as the focal herbivore, and 21% having both present. Domesticated livestock were the most frequently surveyed grazers including cattle (n = 164), sheep (n = 83), and horses (n = 21), but studies are included that examined wild grazers, such as kangaroos (n = 6), elephants (n = 5), and pronghorn (n = 5) (Fig. 4).Fig. 4The frequency in which a study reported herbivores. We included any mention of herbivores regardless of being a grazer, browser, granivore, or other class. This list was obtained by the text within the manuscript and is different than the representation of species in the database (i.e., the measured species).Full size image More
