TetraDENSITY: A database of population density estimates in terrestrial vertebrates

Motivation: Population density is a key demographic parameter influencing many ecological processes, and macroecology has described both intra- and interspecific patterns of variation. Population density data are expensive to collect and contain many forms of noise and potential bias; these factors have impeded investigation of macroecological patterns, and many hypotheses remain largely unexplored. Population density also represents fundamental information for conservation, because it underlies population dynamics and, ultimately, extinction risk. Here we present TetraDENSITY, an extensive dataset with > 18,000 records of density estimates for terrestrial vertebrates, in order to facilitate new research on this topic. Main types of variable contained: The dataset includes taxonomic information on species, population density estimate, year of data collection, season, coordinates of the locality, locality name, habitat, sampling method and sampling area. Spatial location and grain: Global. Spatial accuracy varies across studies; conservatively, it can be considered at 1 8 , but for many data it is much finer. Time period and grain: From 1926 to 2017. Temporal accuracy is yearly in most cases, but studies with higher temporal resolution (season, month) are also present. Amphibians in terrestrial phase, reptiles, birds and mammals. Estimates derive from multiple methods, reflecting the study taxon, location and techniques available at the time of density estimation. of in animals is a challenge of


| I NTR OD U CTI ON
Population density has been widely investigated in macroecology and is fundamental to conservation because it is a direct proxy for extinction risk (Brown, Mehlman, & Stevens, 1995;Currie & Fritz, 1993;Sanderson, 2006). Population density varies enormously among species but is also extremely variable within species, both in space and in time (McGill, 2008). Many macroecological studies essentially focus on presence/absence data. Species abundance and density data can be much more informative, but their application is generally limited by the lack of such data. Understanding the temporal, spatial and life-history drivers of population density in animals is a major challenge of macroecology.
Much research has already focused on these questions, but the noisy and sparse nature of data has led to several unclear findings. For example, it is widely known that body mass scales inversely with population density, presumably because it is the primary determinant of metabolism and resource use (Blackburn et al., 1993;Currie & Fritz, 1993;Damuth, 1981;Silva & Downing, 1995 and four orders of magnitude at any given body mass (Silva, Brimacombe, & Downing, 2001). Confounding factors, such as sampling area (Blackburn & Gaston, 1996), may alter the shape of the density-body size relationship, as may resource partitioning among sympatric species (Pacala & Roughgarden, 1982), biases in the published literature toward high density estimates (Lawton, 1990;White, Ernest, Kerkhoff, & Enquist, 2007), and the spatial extent of studies (Blackburn & Gaston, 1997).
In macroecology, population density has mostly been explored in terms of interspecific variation, yet there is substantial variation in the density of populations within species (McGill, 2008). The environmental context, including climatic conditions, resource availability and partitioning and direct biological competition, certainly plays a fundamental role in determining local population abundance (Currie & Fritz, 1993;Pettorelli, Bro-Jørgensen, Durant, Blackburn, & Carbone, 2009). For an investigation of such patterns, spatial information is required. Yet, these data are generally lacking in global datasets of life-history traits, which are largely based on average estimates (e.g. Jones et al., 2009). Clearly, the more data are available, the better we will be able to explore such questions.
Better data on population density can contribute to conservation biology by identifying conditions and traits that allow species to attain a larger population size within a given area. For example, a common assumption in biogeography and conservation is that abundance is high at the centre of the geographical range and decreases toward the edges (Brown, 1984), but this assumption is controversial and probably does not hold for many, perhaps most, species. Nevertheless, this notion of an 'abundant centre' has proved influential in a variety of areas in conservation biology, such as where reserves should be placed, where extinction risks are high, and around the dynamics of gene flow across broad areas (Sagarin & Gaines, 2002). Improving our understanding of how population density varies across time, space and species will ultimately contribute to more informed conservation decisions. Changes in population density are significant for purposes of biodiversity monitoring (Collen et al., 2009;Santini et al., 2017). Building a large dataset on population density estimates on a wide range of organisms becomes pivotal to building a solid theory that can contribute to conservation efforts.
In this data paper we present TetraDENSITY, a global dataset of population density estimates for terrestrial vertebrates, which can prompt new investigations on this fundamental aspect of animal ecology. From each paper, we recorded the species name, the density estimate, the year of data collection, the coordinates of the locality, the locality name, the season when applicable, habitat and sampling method (Table   1). In a few instances, data were extracted from figures using WebPlot-Digitizer 3.10 Desktop (http://arohatgi.info/WebPlotDigitizer/; Rohatgi, 2016). In some cases, the coordinates were reported in the papers (with no specified precision), whereas in others only the locality name was reported. As a consequence, coordinates can be more or less precise depending on how coordinates were reported, how small the locality was and how large the study area. All coordinates were transformed to latitude-longitude coordinates (in decimal degrees). Most of the references are published in English, and a minority were in Spanish, French, German and Italian.

| R ESU L TS
We collected a total of 18,246 population density estimates from 949 references, covering a wide range of orders, families and genera across the four classes of terrestrial vertebrates (Table 1).
These estimates span over several orders of magnitude. Amphibian densities span between 24 and 9,140,000 individuals/km 2 , reptiles between 0.003 and 9,587,000 individuals/km 2 , birds between 0.002 and 9,587 individuals/km 2 , and mammals between 0.00003 and 24,700 individuals/km 2 (Figure 1a  of species sampled per location (Figure 2). A detailed description of the variables presented in the dataset is provided in Table 2.

| DI SCUS SION
TetraDENSITY is the largest ever assembled dataset of population density estimates in terrestrial vertebrates. It includes site-specific estimates that can vary up to two orders of magnitude within the same species. Amphibians and reptiles, for example, can show extremely high densities, which can refer to highly suitable microhabitats of limited extents, but also reflect the three-dimensional nature of the habitat in arboreal species. Additionally, they can represent temporary fluctuations of the populations. This collation will facilitate exploration of ecological theories such as species abundance distributions (McGill et al., 2007;Xiao, O'Dwyer, & White, 2015) and range size-abundance relationships (Gaston et al., 2000), in addition to large-scale intra-and interspecific geographical patterns in population density (e.g. Currie & Fritz, 1993;Sagarin & Gaines, 2002).
It is well known that different sampling methods can provide different density estimates. This complicates the comparison of density data gathered using different methods in different areas, and therefore combining densities from a range of sources in the same analysis is non-trivial. However, our dataset includes information on sampling methods, enabling users of our dataset to account for these issues and even perform methodological comparisons.
The population density records are biased toward certain taxa and geographical areas. These biases largely reflect known patterns in ecological research globally. However, our search covered only languages that use Latin script. Including data published in other writing systems would alter the perception of geographical bias, particularly with respect to China and Russia. Additionally, the estimation of population density in animals is not equally applicable to different habitats and Sampling method Sampling method used to estimate density pooled in broad categories: Incomplete counts (any incomplete count that is extrapolated to a larger area), censuses ('complete' counts, which assume full detection of individuals), distance sampling (including different algorithms and sampling design), home range extrapolation (derived from home range area estimation), mark-recapture (including different algorithms and capture approaches), trapping (removal methods, indicate the minimum number known to be alive)

15,454
Method information Additional details on the method 9,616 Notes Opportunistic additional notes on the density estimate or the study 3,521 species. The density of populations in more impenetrable habitats and of more rare/cryptic species are less likely to be estimated. Figure 1 provides the clearest view to date of where population density data are lacking.

ACKNOWLEDGMENTS
We thank Manuela Gonz alez-Su arez, Jeremy Kerr and two anonymous referees for providing constructive comments that improved the manuscript. L.S. was supported by the Brusarosco fellowship provided by the Italian Society of Ecology (SItE).