Home » Uncategories » Animal Population Time Series Data Sets Publicly Available

Animal Population Time Series Data Sets Publicly Available

Nat Commun. 2016; 7: 12747.

Wildlife population trends in protected areas predicted by national socio-economic metrics and torso size

Megan D. Barnes,^a, ^ane, ^2, ^* Ian D. Craigie,^a, ^iii, ^* Luke B. Harrison,⁴ Jonas Geldmann,^5, ^six Ben Collen,⁷ Sarah Whitmee,⁸ Andrew Balmford,⁶ Neil D. Burgess,^v, ⁹ Thomas Brooks,^ten, ^11, ¹² Marc Hockings,^1, ^9, ¹⁰ and Stephen Woodley ¹³

Megan D. Barnes

¹School of Geography Planning and Environmental Direction, the University of Queensland, St Lucia, Queensland 4067, Australia

²Australian Research Council Centre of Excellence for Ecology Decisions, the University of Queensland, St Lucia, Queensland 4072, Australia

Ian D. Craigie

³Australian Research Council Centre of Excellence for Coral Reef Studies, James Cook University, Townsville, Queensland 4811, Australia

Luke B. Harrison

⁴Redpath Museum, McGill University, 859 Sherbrooke Street West, Montreal, Quebec H3A 0C4, Canada

Jonas Geldmann

⁵Heart for Macroecology, Development and Climate, Natural History Museum of Denmark, University of Copenhagen, Universitetsparken 15, Copenhagen E 2100, Denmark

^half-dozenConservation Scientific discipline Group, Department of Zoology, University of Cambridge, Cambridge CB2 3EJ, UK

Ben Collen

^sevenCentre for Biodiversity and Environment Research, University Higher London, Gower Street, London WC1E 6BT, U.k.

Sarah Whitmee

^viiiIndicators and Assessment Unit, Institute of Zoology, Zoological Society of London, Regent'due south Park, London NW1 4RY, United kingdom of great britain and northern ireland

Andrew Balmford

^sixConservation Science Group, Department of Zoology, University of Cambridge, Cambridge CB2 3EJ, UK

Neil D. Burgess

⁵Eye for Macroecology, Development and Climate, Natural History Museum of Denmark, University of Copenhagen, Universitetsparken xv, Copenhagen East 2100, Denmark

⁹United Nations Environment Programme World Conservation Monitoring Centre (UNEP-WCMC), 219 Huntington Route, Cambridge CB3 0DL, UK

Thomas Brooks

¹⁰International Matrimony for Conservation of Nature, 28 rue Mauverney, Gland 1196, Switzerland

¹¹Earth Agroforestry Center (ICRAF), University of the Philippines Los Baños, Laguna 4031, Philippines

¹²Schoolhouse of Geography and Environmental Studies, University of Tasmania, Hobart, Tasmania TAS 7001, Australia

Marc Hockings

ⁱSchool of Geography Planning and Environmental Management, the University of Queensland, St Lucia, Queensland 4067, Commonwealth of australia

^ixUnited Nations Environment Programme World Conservation Monitoring Heart (UNEP-WCMC), 219 Huntington Road, Cambridge CB3 0DL, Great britain

¹⁰International Wedlock for Conservation of Nature, 28 rue Mauverney, Gland 1196, Switzerland

Stephen Woodley

¹³Woodley and Associates, Chelsea, Quebec J9B 1T3, Canada

Received 2015 Aug 28; Accepted 2016 Jul 29.

Supplementary Materials: Supplementary Information Supplementary Figures 1-half dozen, Supplementary Tables ane-6, Supplementary Note 1, Supplementary References

GUID: 9A54EF74-D0C5-43EE-ACD5-CD21399A8E71

Supplementary Dataset 1 Table listing all species included in the analysis

GUID: 8D6A9FE8-20BA-48C7-AE45-D9C19FD7FBCD

Supplementary Dataset 2 Table listing all protected areas included in the assay

GUID: CD9ABD58-EF06-4A60-99BC-E139A5A71378

Data Availability Statement: All relevant data are available from the authors upon asking. The population time serial are bachelor from the Living Planet Database to registered users: http://www.livingplanetindex.org/data_portal. Links and citations for all the freely available predictor data sets are available in Supplementary Table 3.

Abstract

Ensuring that protected areas (PAs) maintain the biodiversity inside their boundaries is central in achieving global conservation goals. Despite this objective, wild fauna affluence changes in PAs are patchily documented and poorly understood. Here, we utilize linear mixed effect models to explore correlates of population alter in one,902 populations of birds and mammals from 447 PAs globally. On an average, we observe PAs are maintaining populations of monitored birds and mammals within their boundaries. Wildlife population trends are more than positive in PAs located in countries with higher evolution scores, and for larger-bodied species. These results suggest that active management can consistently overcome disadvantages of lower reproductive rates and more severe threats experienced by larger species of birds and mammals. The link between wildlife trends and national evolution shows that the social and economical conditions supporting PAs are critical for the successful maintenance of their wild fauna populations.

Biodiversity is in crisis¹ ^,2. A fundamental response to global biodiversity declines³ and the associated threatening processes^iv, has been the establishment of protected areas (PAs). PAs underpin most global and national conservation strategies^v, covering at to the lowest degree xv.4% of global land surface area⁶. The importance of PAs is set to increase further given the latest Convention on Biological Diversity targets to increment global land coverage of PAs to 17% by 2022 (ref. ⁷). Ensuring that biodiversity is maintained within-PA boundaries is consequently fundamental to achieving global conservation goals.

Despite a primal objective of PAs being to conserve wild animals populations within their boundaries, many PAs are experiencing undesirable wildlife population declines^viii ^,9. Worldwide, wildlife population changes in PAs are patchily documented¹⁰, unquantified and poorly understood. While a usually held formulation is that PAs are constructive at maintaining wild animals populations within their borders, this assumption has not been widely tested. Conversely, the perception that some PAs are declining, or at to the lowest degree performing inadequately, has precipitated calls to radically change both conservation controlling and PA management¹¹, and emphasized the need to ensure that PAs are effectively managed in the long term¹².

It is thus vital to quantify how well PAs are conserving wild fauna¹³, and to identify enabling conditions and barriers to effective conservation. By identifying properties of those PAs more likely to maintain wild animals populations, it will be possible to ensure new PAs are established in a more spatially and financially efficient configuration and maximize biodiversity outcomes within resource constraints. Without a better understanding of those factors that contribute to wild fauna outcomes in PAs, so their selection, design and direction are likely to remain sub-optimal.

Wild fauna population alter is an important and useful metric for evaluating wildlife conservation outcomes in PAs. It is sensitive to long-term environmental alter¹⁴, ofttimes directly linked to PA objectives, and valuable in diagnosing extinction run a risk^fifteen. Critically, population trends can also quantify biodiversity change in a variety of habitat types, including savannah and other non-forested habitats, complementing studies of PA impacts on maintaining wood cover¹⁶. Nosotros compiled an extensive data set of 1,902 vertebrate population abundance time series from 447 terrestrial PAs, and calculated bird and mammal population trends for 556 species as a metric of PA effectiveness in meeting conservation goals. Directly counterfactual data¹⁷ from similar but unprotected populations were not available due to insufficient monitoring effort outside PAs. We conducted a broad-scale evaluation of within-PA wildlife abundance trends (alter over time) to place properties of PAs that contribute to variation in trends amid PAs. Learning from conservation outcomes for wildlife among PAs has the capacity to dramatically meliorate policy and direction for native wildlife in PAs by promoting the propagation of enabling practices globally.

We find that on average PAs are maintaining the abundance of populations of monitored birds and mammals within their boundaries, and that wildlife population trends are more than positive in countries with higher evolution scores, likewise as for larger bodied species. This trunk mass finding suggests active management tin can overcome disadvantages of more severe threats experienced by larger species¹⁸ ^,nineteen. It also suggests in that location is a need to manage smaller species direct, rather than assuming that management deportment targeted at conservation of iconic taxa volition lead to constructive conservation of all species. Our results too underscore the need to address social and economic conditions that support PA management—equally these announced to exist critical for maintaining wildlife populations within PA borders.

Results

Dataset

The total data set contains 1,902 population time series for 556 species in 447 PAs (Supplementary Table 1, Supplementary Data 1 and ii) across 72 countries ( Fig. i and Supplementary Fig.1), from time periods between 1970 and 2010. The species in our data ready are dominated past large mammalian herbivores and waterfowl (torso mass distributions—Supplementary Fig. 2), reflecting taxonomically uneven global monitoring efforts. Notation our information terminate before the recent poaching crisis that has significantly reduced populations of African elephants (Loxodonta africana) and also affected rhinoceros species (Diceros bicornis and Cerotherium simium)²⁰.

An external file that holds a picture, illustration, etc. Object name is ncomms12747-f1.jpg

Locations of our PA population time series.

Countries in grey are those included in the assay. (a) Proportionally sized pie charts indicate the number of bird (red) and mammal (blue) time series in each PA. (b) Mean per cent annual change in population abundances for each PA represented. Lighter (more yellowish) dots stand for greatest declines, and darker (more than blue) dots greatest increases.

Overall Trends

Overall, the mean per centum annual modify in population size within PAs was near nada (slightly positive: mean 0.52%, median 0.81%, s.d. 12.vii, Fig. 2 ). Overall, bird trends were marginally positive (hateful annual modify 1.72%, median i.71%, south.d. 12.45, Table 1 ), whereas mammal trends were slightly negative (mean −1.00%, median −0.62%, s.d. 12.45, Fig. 2a ). Trends in Europe were more than positive than those in Africa ( Fig. 2b and Table i ).

An external file that holds a picture, illustration, etc. Object name is ncomms12747-f2.jpg

Frequency distribution of wildlife abundance changes in protected areas.

(a) Changes past species type. (b) Changes by location. In a, grey shows all species, green shows all mammals and blue shows all birds. In b, grey shows all sample populations, green shows sampled populations in Africa and blue shows sampled populations in Europe.

Table 1

Annual percentage population change in each data set.

Information prepare	Mean	Median	s.d.
Global	0.52	0.81	12.72
Mammal	−ane.00	−0.62	12.45
Bird	i.72	i.71	12.45
Africa	−1.79	−1.67	13.68
Europe	ii.05	ii.15	12.05

Population trends showed substantial variation across species and PAs ( Fig. 1b ). To explore this variation, nosotros tested factors previously identified equally likely to influence biodiversity outcomes in PAs¹⁰, and collated data sets addressing those factors (Supplementary Tables two–4). We examined half dozen groups of possible influences: (1) PA design (for case, size, shape, IUCN management category), (2) socio-economic context of the region and country in which the PA is located (for example, wealth, corruption), (3) species' traits which might determine response (for case, body mass), (four) local human impacts (for example, route density, land-use modify), (five) biophysical context (for case, PA elevation) and (6) time series characteristics (for case, length). Nosotros used linear mixed outcome models to account for the hierarchically nested data construction, and in addition to a global model, produced dissever models for mammals and birds, and for the 2 well-nigh data-rich regions (Europe and Africa).

Model Results

Correlates of species population trends in PAs were identified across four of the six groups of factors ( Figs three , 4 , 5 , Supplementary Fig. 3 and Supplementary Tables 5 and half dozen). In order of importance: outset, in most models, population trends were more positive in areas with higher national Human being Evolution Alphabetize scores (global, mammal, bird and Europe models), and greater Gini indices (that is, level of income inequality, in global and bird models; see also Supplementary Notation 1). 2nd, population trends increased with body mass (in all models) and differed among taxonomic classes (all models where tested, except Europe). Third, among our measures of anthropogenic impact, population trends were positively correlated with local road density (mammal and Africa models) and local Human Influence Index (Europe model). 4th, population trends in African PAs were more positive in subsequently years. Neither of the other 2 groups of factors, PA design or biophysical context, emerged every bit significant fixed effects in any models: in all models species and socio-economic factors were more of import. Models were found to exist robust to the effects of phylogenetic influence every bit reflected by taxonomy, and exhibited depression sensitivity to statistical outliers such as the positive outliers of African elephants and rhinoceroses. Note our abundance information pre-dates the post-2008 surge in illegal hunting of elephants and rhinoceroses²⁰ ^,21.

An external file that holds a picture, illustration, etc. Object name is ncomms12747-f3.jpg

Fractional-effects plots for variables in the most parsimonious global model.

Partial-effects plots showing fitted relationships betwixt change in population size and (a) body mass, (b) taxonomic class, (c) Gini index and (d) Human being Development Index (HDI). In a, c, and d, dashed lines are 95% credibility intervals. In b, the circles point the estimated partial effect size for each cistron level with apparent intervals displayed every bit mistake bars.

An external file that holds a picture, illustration, etc. Object name is ncomms12747-f4.jpg

Informative continuous fixed furnishings in preferred models for each subset.

Data are shown for (a) global, (b) mammals, (c) birds, (d) Africa, (e) Europe. Bar lengths signal the size of the parameter estimates and their explanatory ability in the model. Significance levels derived from highest posterior credibility intervals from Markov-Chain Monte Carlo methods: NS P⩾0.05;+P≤0.05; ++P≤0.01; +++P≤0.001. Colours show groups of potential explanatory factors: dark blue, species' traits; light blue, socio-economical context; light-green, local human impacts; and yellow, time series characteristics. Superscripts indicate power (that is, squared, cubed) of the variable for those variables which exhibit college guild relationships.

An external file that holds a picture, illustration, etc. Object name is ncomms12747-f5.jpg

Partial-effects plots for the most parsimonious model of each subset modelled.

Information are shown for (a) mammals, (b) birds, (c) Africa, (d) Europe. Dashed lines are 95% credibility intervals based on MCMC sampling with 10,000 samples. They evidence the relationship between each stock-still effect: (a) trunk mass, HDI, road density; (b) trunk mass, Gini index, HDI; (c) body mass, class, road density, mid year; and (d) torso mass, HDI, class, land-use change (Human Impact Index (HII) in 25 km), and population trends when all others are held constant. For chiselled variables (class) dots indicate the estimated partial effect sizes of each factor level, and error bars show 95% credible intervals. HDI, Human Development Index; MCMC, Markov-Chain Monte Carlo.

Discussion

Given the central part of PAs equally global conservation tools, information technology is reassuring that, on average, monitored populations inside PAs are stable or marginally positive (hateful 0.52% annual increase). The differences between birds and mammals (birds more positive), and between Europe and Africa (Europe more positive) are likely an effect of differing regional histories, as well as current pressures. Large-scale land conversion outside the tropics has led to broad-calibration historical extirpations²². More recent policy changes in Europe have led to widespread improvements in biodiversity direction²³, and the recovery and reintroduction of many wildlife populations, including birds in PAs²⁴. African wildlife populations are typically more intact in absolute terms, but are nether increasing anthropogenic pressure; causing abundance declines^eight.

Strikingly, larger-bodied species had more positive population trends in all models except Europe, indicating that PAs are more likely to maintain populations of larger-bodied wildlife than smaller bodied species. This finding is consequent beyond geographic realms, and taxonomic grade, so it was non driven past the difference in body mass between birds and mammals. In the mammal and Africa models the relationship of trunk mass to trend was u-shaped ( Fig. 5 ), suggesting perhaps that the smallest species are more than resilient because of their high reproductive rates²⁵; while intermediate sized species, lacking active management and having slower reproductive rates, are experiencing greatest decline. These findings accept not previously been quantitatively demonstrated, simply are supported by previous anecdotal evidence from Kenyan PAs²⁶, which posited a shift from elephant and rhinoceros to poaching smaller species in response to increased penalties, earlier the contempo ivory and horn poaching crisis.

Ane explanation for these findings is that threat processes of loftier severity (for example, hunting) are impacting intermediate bodied species particularly badly—a pattern previously detected among threat processes for mammals and birds²⁷ ^,28 ^,29. Management attempt and external project funding is usually prioritized towards large-bodied flagship and charismatic species³⁰, and larger species are key to tourist revenues and public/political priorities³¹. As a result, monitoring and management actions focus on the needs of these species and information technology is likely that smaller species practise non receive the same benefits from PAs. We item interactions betwixt torso mass, threatening processes and influential factors in a conceptual diagram (Supplementary Fig. 4). Larger species also tend to exist preferred for ecological written report and monitoring. Consequently, population declines are likely to be noticed sooner and their causes better understood, leading to more effective management responses. This effect has substantial implications for future allocation of conservation effort among species inside PAs.

Indicators of greater human wealth (gross domestic product) and evolution (Human Evolution Index) were associated with more positive population trends. Wealth and development have been shown to have a complex human relationship with conservation. Poor outcomes have been associated with both historical and accelerating threatening processes, while increasing wildlife trends tin can be associated with greater awareness and management chapters, that is, wealthier countries may have more resources bachelor for PA management, and spend more on conservation³² ^,33 ^,34. Moreover, man populations in wealthier areas take less need for resources directly extracted from PAs to back up livelihoods (for case, bushmeat, firewood). A final contributing gene to this pattern could be historical species loss (an extinction filter upshot)—with the biota of wealthier countries having been more extensively purged of species which are more sensitive to anthropogenic threats³⁵.

This finding of an association with evolution is promising, as it shows that boosted capacity or decreased dependence on natural resource, can help ameliorate wildlife declines in PAs. Thus, economic development and the associated comeback in food security and governance may lead to effective conservation. In reality, outcomes are likely related to a residuum of multiple socio-economic factors. Regardless of which of these explanations is ascendant, actress effort will be required to retain species in developing regions. All the same, to avoid the extinction filter in developing regions resulting in wild fauna loss catching up with historic losses in developed countries, extra endeavor will be required to retain species in these countries. The marginal finding that higher national Gini indices predict more than positive population trends is unexpected. The result can be explained by a small number of countries (for example, South Africa) that take both very high Gini indices and relatively positive wildlife trends.

Counter-intuitively, more positive trends in PAs were correlated with increased anthropogenic mural changes, indicated past locally denser road networks in the mammals and Africa models, and land-employ change in the Europe model. Evolution of new roads may open upward areas and crusade wildlife and habitat declines³⁶, but in areas of historical road construction at that place may be a filter upshot with PAs currently experiencing a wildlife recovery, peculiarly for certain robust species³⁷. Wildlife populations surrounded by extensive historical land clearing and roads maybe experienced their declines before our data ready just are now stable, whereas places with fewer roads accept recently experienced development and clearing, driving declines³⁸. Conversely, higher road density may correlate with increased levels of resources for direction, improving PA outcomes: road density was plant to exist positively correlated with greater PA management effectiveness by Geldmann et al.³⁷

Several well-studied ecological factors that conservation theory predicts should be important determinants of wildlife trends, including PA size³⁹ and PA shape⁴⁰, have no explanatory power in our models. We do not interpret the lack of significance of these variables to mean they are not important. We suggest that over the timescales addressed by our data their influence is overwhelmed by either the priorities of managers, the more than general landscape scale recovery of big species²³, or the socio-economical context of PAs, only nosotros cannot not easily differentiate between these drivers. Over multi-decadal timescales the ecological drivers are still likely to be influential, simply our results show the importance of managing PAs for more than immediate threats. A substantial quantity of research effort examines optimal blueprint of PAs and PA networks with respect to ecological processes, but our findings suggest that greater conservation benefit would result from a focus on optimizing network pattern because management and man influences on PAs.

The ability of PAs to maintain species populations is disquisitional to global conservation. Our finding that abundances of monitored birds and mammals are maintained inside PA boundaries is encouraging. However, our results evidence that PAs do not work equally well for all species or in all circumstances. Moreover the recent poaching crisis in Africa shows that population gains can exist rapidly reversed, if the threatening processes grow too big to mitigate^twenty, emphasizing the need to scale management effort with threat intensity⁴¹ ^,42. In addition, the time series data for this analysis are from PAs that are older, larger, and farther removed from humans than near PAs (see Supplementary Fig. v). Hence, nosotros should not be complacent. These findings are likely to represent a best-example scenario for protected populations as they are sourced from PAs experiencing lower than average anthropogenic threat. If nosotros expect PAs to act every bit refuges for all species in perpetuity, and then a wider range of species must be targeted for direction, and a item focus on conserving medium-sized species may be required.

Further, it is clear that PAs do non be in a vacuum. Our results show that the social and economic atmospheric condition that support PA direction are critical for the maintenance of wildlife populations within PA boundaries. Anthropogenic drivers of wild animals abundance change appear to have more influence than those drivers affecting ecological processes such as PA size, at to the lowest degree over the period of a few decades. Much of the enquiry effort into PAs targets ecological processes but it seems greater render would come from focusing efforts more on anthropogenic drivers of PA performance.

Managing PAs' socio-political, rather than simply ecological, dimensions is pivotal to wild animals conservation in PAs. Human being dependence on PA resources in poorer countries must be addressed if existing PAs are to retain their contents equally these nations continue to develop. Finally, to understand the render on our investment in PAs in the long-term, all-time practise adaptive management and systematic monitoring of biological outcomes, including appropriate counterfactual monitoring, is essential⁴³. The tools to sympathize bear on and improve outcomes be—only we must strengthen the volition and capacity to implement them.

Methods

Wild fauna population time series data

A global database of population abundance fourth dimension series of all available time series for native birds and mammals in terrestrial PAs worldwide was compiled from sources including the Living Planet Database⁴⁴, PA agencies, published literature, grey literature and not-governmental organizations. Fourth dimension serial in the information set represent the majority of the data available globally to address this topic. Time series consisted of population abundance count estimates, or proxies of abundance such every bit nest density, mark-recapture or density estimates. Marine species, other than those with at least one critical life history stage on land (for example, breeding colonies), were excluded. We used population time series that met a number of criteria based on Collen et al.⁴⁴: (1) the technique used to measure abundance was comparable over the length of the time series; (2) the geographic location of the population was provided; (three) the majority (>l%) of the measured population was within a PA; and (4) population time series were a minimum of five years in length betwixt 1970 and 2010, with at least 3 measures of affluence inside that time flow (that is, an estimate was not required for every twelvemonth within a time series). Where data were bachelor for multiple fourth dimension points in a single year (for instance, wet flavor and dry flavour), data were standardised to obtain a single per annum abundance estimate for each population. Standardization was carried out using the near advisable and comparable method given information type and species environmental to obtain an estimate of species population changes most likely to remain consistent through time, and accurately represent population change, in accordance with established LPI database practice⁴⁴ ^,45. For case, monthly abundance counts were averaged (using mean estimates) for resident non-irruptive species.

Estimation of wildlife population trends

Population trends were estimated by fitting a generalized linear regression model on time with a log-link office to each population fourth dimension series⁴⁶ ^,47. The trend was taken to exist the slope value from each regression. Fitting a log-link function assumes the response variable (population abundance) has a Poisson error distribution, and that the logarithm of its expected value can be modelled by a linear combination of unknown parameters, and is most appropriate for zero-inflated information, such every bit count and grab per-unit effort data⁴⁶. Leading and trailing cipher values (that is, zeros which occurred at the beginning or end of time series) were excluded from each time series earlier calculating trends; such zero values generally occur when populations are present at a level below that at which the sampling method is able detect individuals, rather than when a population has been extirpated or introduced. These zeros are therefore inaccurate estimates of abundance, yet they exert undue influence on the estimated slope values (population trends). Slope values exceeding ±0.5 on the natural log scale were excluded following inspection of the data distribution. Such values are equivalent to almanac rates of population growth and decline of more than l% per annum and biologically implausible over periods greater than 5 years (equivalent of a population of one,000 individuals increasing to 7,594 in v years or 57,665 in 10 years). Such values are likely the result of errors in surveys or data entry⁴⁸ ^,49 ^,50.

Explanatory variable pick and grooming

Variables were selected to correspond characteristics of PAs considered almost likely to be important determinants in maintaining wildlife populations in PAs for vertebrate species, based on both theoretical supposition and empirical ascertainment as identified via a comprehensive literature review¹⁰ and can be regarded as belonging to six groups of possible influences: PA pattern (for example, size, shape, IUCN management category), socio-economic context of the region and country in which the PA is located (for example, wealth, corruption), species' traits which might decide response (for instance, body mass), local human being impacts (for instance, road density, land-use alter), biophysical context (for case, PA elevation), and fourth dimension series characteristics (for example, length). Several variables suggested by the literature to exist important for determining PA outcomes at the site calibration (for example, PA-specific management budgets, threat intensity) were unavailable for virtually sites, and therefore could non be tested in this analysis. The key socio-economic variables used in the analysis were simply consistently available at the national calibration, not at finer spatial scales relative to the PAs. A table summarising the justification for each variable, a priori hypotheses, and references underpinning the option of each factor are provided in Supplementary Table two. For each explanatory variable, data was nerveless at a scale based on a combination of availability, standard practice and relevance for population dynamics. Generally, this meant that the finest resolution data available at a global scale was used. Descriptions of the explanatory variables and data sources and resolution are given in Supplementary Tabular array three.

Spatial data were analysed using ArcGIS 10.0 and R 2.15.0 (ref. ⁵¹) using the packages raster⁵², geosphere⁵³, maptools⁵⁴, rgeos⁵⁵, rgdal⁵⁶ and sp⁵⁷. PA boundaries were calculated using spatial information from the World Database of Protected Areas⁵⁸. Spatial data relating to PA context (for example, human population density) were calculated in buffers of three different sizes (5, 10 and 25 km) around each PA polygon. Multiple buffer sizes were used because information technology was not known a priori over what altitude potential correlates would be most likely to exist acting. PAs represented simply by indicate information in the WDPA were included in the analysis by creating appropriately sized round buffers using the ArcGIS buffer tool under an equal-surface area (Mollewide) projection⁵⁹. The size of the buffer was given past the area (kmⁱⁱ) recorded for the individual PAs in the WDPA. In cases where the projection of the PA data set differed from the explanatory information layer, the PAs were reprojected using ArcGIS. Details of the training of the individual explanatory variables are provided in Supplementary Tables 3 and 4. Several explanatory information sets were available as raster layers. The R raster parcel⁵² was used to overlay the PA polygon onto those cells and calculate metrics based on the underlying raster cell values. However, using the raster package, overlayed polygons must comprehend the heart of a raster cell to exist considered as inside the polygon. Pocket-size, spatially circuitous PA polygons may just cover a few raster cells, particularly when overlayed on raster layers with fibroid spatial resolution. The selected cells may therefore poorly stand for the overall cell area truly overlapped past the polygon. For this reason, nosotros disaggregated each raster layer by a factor of 10 (30 for the spatially coarse agricultural suitability layers) using the R raster package for modest PAs (<2,000 cells overlapped past the PA polygon). We also enforced a minimum overlapping area for very pocket-size PAs earlier performing all calculations (noted in Supplementary Table 3 in terms of the equivalent number of raster cells before disaggregation). For some variables (noted in Supplementary Table 3), PA polygons were clipped of adjoining marine areas in ArcGIS using the World Vector Shoreline Plus layer (http://shoreline.noaa.gov/data/datasheets/wvs.html).

Explanatory variables were transformed to normalize distributions, where necessary, and standardised using the R part 'scale' so all variables in the data set had equal ways and standard deviations merely different ranges. Normalized and standardised variables were evaluated to make up one's mind collinearity by visual inspection of the data and past computing Pearson's correlation coefficient (meet Supplementary Fig. 6).

Population trend modelling relative to explanatory variables

The slope of the generalized linear regressions for all populations was used as the response variable to address the question: what factors predict trend in abundance for bird and mammal species in terrestrial PAs? Nosotros practical a linear mixed-effects modelling approach to explore the fundamental correlates of wildlife population of birds and mammals through time in PAs. Explanatory variables were hierarchically spatially structured; at the species, PA and national levels. We used linear mixed-furnishings modelling to account for the data structure and investigate the relationship between population trends and the suite of potential explanatory variables. Mixed-effects models permit partitioning of the variance in population trends in a nested hierarchy⁶⁰ ^,61. In this analysis, the information were structured such that each population trend referred to a particular species within a given PA. Most sites contain several species, and most species occur at several sites so in that location are multiple observations of individual species across a suite of site subsets, such that individual population trends are non mutually independent. Further, populations were distributed non-randomly across continents and countries. All models were implemented in R 2.15.0 (ref. ⁵¹) using the packages lme4 (ref. ⁶²) and MuMin⁶³. Random effects were kept consequent in all models with species, site (PA), and country fitted every bit random effects in every model based on the a priori understanding of the data structure. In models where both birds and mammals were nowadays, taxonomic Class was included as a fixed effect in the models. Models were generated for the entire global data set and subsets for taxonomic form and geographic realm. Subset models were generated for: birds, mammals, Africa and Europe. Although some population data was available exterior Europe and Africa, other geographic realms had bereft sample sizes to construct subset models.

Modelling procedure

Model choice was made following an information-theoretic approach, using the corrected Akaike Information Criterion AICc⁴⁶ ^,64. Forrard and backward stepwise model selections were conducted to narrow the set of candidate models. Thereafter all possible subsets of candidate models were compared using the role dredge⁶³. This included testing all plausible interactions, and polynomials (orthogonal squares and cubes). Some variables could not be fitted simultaneously as they were highly collinear (Pearson'due south correlation coefficient >0.five). In these cases the variable with the greatest explanatory ability, as defined past the best AICc value, was plant by substitution. Substitution was conducted by exchanging variables in the model that were collinear and expected to explicate the aforementioned component of the variation to assess which of them provided the best fit to the information. Fit was assessed past change in AICc, and during substitution the rest of the model specification was held constant. Substitution was carried out both during initial data exploration and before concluding model selection for each data set.

For each model with a ΔAICc of less than four the fitted residuals were examined using qq-plots, Melt's altitude leverage plots and histograms. Models selection was tested for sensitivity to outliers through systematic removal and replacement of outlying information. Outliers were identified using qqplots and Cooks distance leverage plots of the model residuals. When outliers were removed the preferred models and their parameter estimates were remarkably consistent, and additionally exhibited depression sensitivity the removal of extremely large species such equally Africa Elephant (L. africana) and rhinoceroses (D. bicornis and C. simium). The elephant and rhino outliers were highly positive; if data including the recent years' poaching-related population declines of these species had been bachelor these species may not have been outliers. Fitting genus and order as random effects, to exam for phylogenetic influence as reflected by taxonomy, did not amend model fit significantly for whatsoever models. When they were fitted, body mass parameter estimates remained stable and significant.

Consequence sizes

Markov-Chain Monte Carlo Highest Posterior Density estimates with 10,000 samples were calculated to estimate effect sizes and 95% credibility intervals⁴⁶ for the most parsimonious models. Partial-furnishings sizes were calculated and plots produced for parameters of the best-fit models ( Fig. 5 ). Variable relative importance was calculated post-obit Zuur et al.⁶⁵ using Akaike weights and relative frequency standardised across all models with a ΔAICc <4 (Supplementary Fig. 3).

Data availability

All relevant data are available from the authors upon asking. The population time series are bachelor from the Living Planet Database to registered users: http://www.livingplanetindex.org/data_portal. Links and citations for all the freely bachelor predictor information sets are bachelor in Supplementary Table 3.

Additional information

How to cite this commodity: Barnes, M. D., Craigie, I. D. et al. Wildlife population trends in protected areas predicted by national socio-economic metrics and body size. Nat. Commun. vii:12747 doi: 10.1038/ncomms12747 (2016).

Supplementary Fabric

Supplementary Information:

Supplementary Figures one-6, Supplementary Tables ane-6, Supplementary Note 1, Supplementary References

Supplementary Dataset 1:

Tabular array listing all species included in the assay

Supplementary Dataset 2:

Table list all protected areas included in the assay

Acknowledgments

We are members of the IUCN WCPA-SSC Joint Taskforce on Biodiversity and Protected Areas, which supported this projection. We thank W.N. Venables for input on the blueprint of the statistical analysis, We thank other members of the taskforce who contributed initial thoughts and feedback for this project: S. Andelman, North. Dudley, Due east. Enkerlin, N. Lopoukhine, Yard. MacKinnon, Due south. Stuart, G. Redford and S. Stolton. We give thanks 50. McRae for her assistance with LPI data and J. Ringma for graphics assist. Nosotros thank WWF and ZSL who funded and compiled the living planet database and anybody who contributed to the database. Nosotros thank H. Possingham and C. Carbone for manuscript comments and suggestions, and E. McDonald Madden for advice. Funding was provided past IUCN'south Global Protected Areas Plan and Parks Canada (S. Woodley), European Wedlock's Development Fund through the Biodiversity and Protected Areas Direction (BIOPAMA) Plan, School of Geography Planning and Environmental Direction (M.D.B., M.H.) and the Centre of Excellence for Environmental Decisions at the University of Queensland (M.D.B.) and Natural Environmen1t Research Council studentship NER/Southward/A/2006/14094 with CASE support from UNEP-WCMC (I.D.C.).

Footnotes

Author contributions M.D.B., I.D.C., A.B., Due south. Woods. and G.H conceived the projection. I.D.C., 1000.D.B. designed assay. Grand.D.B., I.D.C., S. Wood. Thousand.H., 50.H. J.G. prepared the data. M.D.B., I.D.C., L.H. analysed the data Grand.D.B., I.D.C., 50.H. and S. Whit. displayed the items. M.D.B. and I.D.C. wrote the manuscript, with input from all authors. All authors contributed to the paper.

References

Butchart S. H. Grand. et al.. Global biodiversity: indicators of recent declines. Scientific discipline 328, 1164–1168 (2010). [PubMed] [Google Scholar]
Tittensor D. P. et al.. A mid-term analysis of progress toward international biodiversity targets. Science 346, 241–244 (2014). [PubMed] [Google Scholar]
Barnosky A. D. et al.. Has the Earth's sixth mass extinction already arrived? Nature 471, 51–57 (2011). [PubMed] [Google Scholar]
Geldmann J., Joppa 50. Due north. & Burgess Northward. D. Mapping modify in homo pressure level globally on land and inside protected areas. Conserv. Biol. 28, 1604–1616 (2014). [PubMed] [Google Scholar]
Secretariat of the Convention on Biological Diverseness. Global Biodiversity Outlook 4: A Mid-Term Assessment of Progress Towards the Implementation of the Strategic Plan for Biodiversity 2011-2022 (Secretariat of the Convention on Biological Diversity, Montreal, 2014).
Juffe-Bignoli D. et al.. Protected Planet Report 2014 UNEP-WCMC (2014). [Google Scholar]
Secretariat of the Convention on Biological Multifariousness. Strategic Programme for Biodiversity 2011-2022, Including Aichi Biodiversity Targets Secretariat to the Convention on Biological Diversity (2010). [Google Scholar]
Craigie I. D. et al.. Large mammal population declines in Africa's protected areas. Biol. Conserv. 143, 2221–2228 (2010). [Google Scholar]
Laurance Westward. F. et al.. Averting biodiversity collapse in tropical forest protected areas. Nature 489, 290–294 (2012). [PubMed] [Google Scholar]
Geldmann J. et al.. Effectiveness of terrestrial protected areas in reducing habitat loss and population declines. Biol. Conserv. 161, 230–238 (2013). [Google Scholar]
Fuller R. A. et al.. Replacing underperforming protected areas achieves better conservation outcomes. Nature 466, 365–367 (2010). [PubMed] [Google Scholar]
Watson J. Eastward. M., Dudley N., Segan D. B. & Hockings 1000. The functioning and potential of protected areas. Nature 515, 67–73 (2014). [PubMed] [Google Scholar]
McDonald-Madden E. et al.. ENVIRONMENT: 'True' Conservation Progress. Science 323, 43–44 (2009). [PubMed] [Google Scholar]
Buckland S. T., Magurran A. E., Greenish R. E. & Fewster R. M. Monitoring change in biodiversity through composite indices. Philos. Trans. R. Soc. Lond. B Biol. Sci. 360, 243–254 (2005). [PMC free article] [PubMed] [Google Scholar]
Hoffmann Thou. et al.. The difference conservation makes to extinction chance of the world'south ungulates. Conserv. Biol. 29, 1303–1313 (2015). [PubMed] [Google Scholar]
Andam 1000. S., Ferraro P. J., Pfaff A., Sanchez-Azofeifa G. A. & Robalino J. A. Measuring the effectiveness of protected area networks in reducing deforestation. Proc. Natl Acad. Sci. USA 105, 16089–16094 (2008). [PMC free article] [PubMed] [Google Scholar]
Ferraro P. J. & Pattanayak S. Thou. Coin for nothing? A call for empirical evaluation of biodiversity conservation investments. PLoS Biol. iv, e105 (2006). [PMC free article] [PubMed] [Google Scholar]
Cardillo K. et al.. Multiple causes of high extinction risk in large mammal species. Scientific discipline 309, 1239–1241 (2005). [PubMed] [Google Scholar]
Bennett P. M. & Owens I. P. F. Evolutionary Ecology of Birds – Life histories, Mating Systems and Extinction Oxford University Press (2002). [Google Scholar]
Wittemyer Thou. et al.. Illegal killing for ivory drives global turn down in African elephants. Proc. Natl Acad. Sci. USA 111, 13117–13121 (2014). [PMC costless article] [PubMed] [Google Scholar]
Biggs D., Courchamp F., Martin R. & Possingham H. P. Legal trade of Africa'due south Rhino horns. Science 339, 1038–1039 (2013). [PubMed] [Google Scholar]
Fritz South. A., Bininda-Emonds O. R. P. & Purvis A. Geographical variation in predictors of mammalian extinction risk: large is bad, but just in the tropics. Ecol. Lett. 12, 538–549 (2009). [PubMed] [Google Scholar]
Chapron G. et al.. Recovery of large carnivores in Europe's modern human-dominated landscapes. Scientific discipline 346, 1517–1519 (2014). [PubMed] [Google Scholar]
Eaton Thousand. A. et al.. The State of the U.k.'s Birds 2012 RSPB, BTO, WWT, CCW, NE, NIEA, SNH and JNCC, Sandy, Bedfordshire (2012). [Google Scholar]
Brook B. & Bowman D. J. S. I equation fits overkill: why allometry underpins both prehistoric and modern body size-biased extinctions. Popul. Ecol. 47, 137–141 (2005). [Google Scholar]
Gibson C. C. Politicians and Poachers: The Political Economy of Wild animals Policy in Africa Cambridge Academy Press (1999). [Google Scholar]
Owens I. P. F. & Bennett P. M. Quantifying biodiversity: a phenotypic perspective. Conserv. Biol. 14, 1014–1022 (2000). [Google Scholar]
Isaac N. J. B. & Cowlishaw G. How species respond to multiple extinction threats. Proc. R. Soc. Biol. Sci. Ser. B 271, 1471–2954 (2004). [PMC free article] [PubMed] [Google Scholar]
Brashares J. S. et al.. Bushmeat hunting, wild fauna declines, and fish supply in west Africa. Science 306, 1180–1183 (2004). [PubMed] [Google Scholar]
Smith R., Verissimo D., Isaac Due north. & Jones Thou. Identifying Cinderella species: uncovering mammals with conservation flagship appeal. Conserv. Lett. 5, 205–212 (2012). [Google Scholar]
Hilborn R. A. Y. et al.. Effective enforcement in a conservation area. Scientific discipline 314, 1266 (2006). [PubMed] [Google Scholar]
McCreless Due east., Visconti P., Carwardine J., Wilcox C. & Smith R. J. Cheap and nasty? The potential perils of using management costs to place global conservation priorities. PLoS ONE 8, e80893 (2013). [PMC free article] [PubMed] [Google Scholar]
Balmford A., Gaston M. J., Blyth S., James A. & Kapos V. Global variation in terrestrial conservation costs, conservation benefits, and unmet conservation needs. Proc. Natl Acad. Sci. The states 100, 1046–1050 (2003). [PMC free article] [PubMed] [Google Scholar]
Waldron A. et al.. Targeting global conservation funding to limit firsthand biodiversity declines. Proc. Natl Acad. Sci. USA 110, 12144–12148 (2013). [PMC complimentary commodity] [PubMed] [Google Scholar]
Balmford A. Extinction filters and current resilience: the significance of past selection pressures for conservation biology. Trends Ecol. Evol. 11, 193–196 (1996). [PubMed] [Google Scholar]
Laurance W. F. et al.. A global strategy for road building. Nature 513, 229–232 (2014). [PubMed] [Google Scholar]
Geldmann J. et al.. Changes in protected area management effectiveness over fourth dimension: a global analysis. Biol. Conserv. 191, 692–699 (2015). [Google Scholar]
Laurance West. F. & Balmford A. State use: a global map for route building. Nature 495, 308–309 (2013). [PubMed] [Google Scholar]
Maiorano Fifty., Falcucci A. & Boitani 50. Size-dependent resistance of protected areas to land-use change. Proc. R. Soc. Biol. Sci. Ser. B 275, 1297–1304 (2008). [PMC gratuitous article] [PubMed] [Google Scholar]
Ewers R. M. & Didham R. K. The effect of fragment shape and species sensitivity to habitat edges on animal population size. Conserv. Biol. 21, 926–936 (2007). [PubMed] [Google Scholar]
Ihwagi F. W. et al.. Using poaching levels and elephant distribution to assess the conservation efficacy of private, communal and authorities country in Northern Republic of kenya. PLoS Ane 10, e0139079 (2015). [PMC free article] [PubMed] [Google Scholar]
Wasser S. Chiliad. et al.. Genetic assignment of large seizures of elephant ivory reveals Africa's major poaching hotspots. Science 349, 84–87 (2015). [PMC free commodity] [PubMed] [Google Scholar]
Craigie I. D., Barnes M. D., Geldmann J. & Woodley Due south. International funding agencies: potential leaders of impact evaluation in protected areas? Phil. Trans. R. Soc. B 370, 20140283 (2015). [PMC free article] [PubMed] [Google Scholar]
Collen B. et al.. Monitoring change in vertebrate abundance: the Living Planet Index. Conserv. Biol. 23, 317–327 (2009). [PubMed] [Google Scholar]
McLellan R., Iyengar L., Jeffries B. & Oerlemans N. Living planet report 2014: species and spaces, people and places http://wwf.panda.org/about_our_earth/all_publications/living_planet_report/ WWF International, Gland (2014). [Google Scholar]
Zuur A. F., Ieno East. N., Walker Due north. J., Saveliev A. A. & Smith G. M. Mixed Effects Models and Extentions in Environmental with R Springer Science (2009). [Google Scholar]
Zuur A., Ieno E. N. & Smith G. Thousand. Analysing ecological data Springer Science & Concern Media (2007). [Google Scholar]
Norton-Griffiths M. Counting Animals 2nd Edn African Wild animals Leadership Foundation (1978). [Google Scholar]
Thomas L. et al.. Distance software: design and analysis of altitude sampling surveys for estimating population size. J. Appl. Ecol. 47, 5–14 (2010). [PMC free article] [PubMed] [Google Scholar]
McCarthy M. A. et al.. The influence of abundance on detectability. Oikos 122, 717–726 (2013). [Google Scholar]
R Core Team. R: A Language and Environment for Statistical Calculating https://www.R-projection.org/R Foundation for Statistical Computing, Vienna, Republic of austria (2012). [Google Scholar]
Hijmans R. J. et al.. Raster: Geographic analysis and modeling with raster data. R package version ii.ii-31 (2013).
Hijmans R. J., Williams Due east. & Vennes C. Geosphere: Spherical Trigonometry. R packet version 1.ii-28 (2012).
Lewin-Koh N. J. et al.. Maptools: Tools for reading and handling spatial objects. R package version 0.eight-16 (2012).
Bivand R. & Rundel C. Rgeos: Interface to Geometry Engine-Open Source (GEOS). R package version 0.ii-7 (2012).
Keitt T. H., Bivand R., Pebesma Eastward. & Rowlingson B. Rgdal: Bindings for the Geospatial Information Brainchild Library. R packet version 0.7-20 (2012).
Pebesma Eastward. J. & Bivand R. S. Classes and methods for spatial data in R. R News v, http://cran.r-project.org/doctor/Rnews/ (2005). [Google Scholar]
WDPA. World Database of protected areas (WDPA) Release June 2012 (web download version). The WDPA is a articulation production of UNEP and IUCN, prepared past UNEP-WCMC, supported by IUCN WCPA and working with Governments, the Secretariats of MEAs and collaborating NGOs (2012).
UNEP-WCMC. Earth Database on Protected Areas User Manual i.1. http://www.unep-wcmc.org/policies/wdpa-data-licence UNEP-WCMC (2015). [Google Scholar]
Pinheiro J. C. & Bates D. M. Mixed-Effects Models in S and S-PLUS US Government Press Office (2000). [Google Scholar]
McCulloch C., Searle S. & Neuhaus J. Generalized, Linear, and Mixed Models http://wcmc.io/WDPA_Manual Wiley (2001). [Google Scholar]
Bates D. & Maechler Chiliad. R package 'lme4' (2010).
MuMIn: Multi-model inference (CRAN, 2012).
Burnham K. P. & Anderson D. R. Model Selection and Multimodel Inference: A Practical Data-Theoretic Approach Springer-Verlag (2002). [Google Scholar]
Zuur A., Ieno E. Northward., Walker N., Saveliev A. A. & Smith G. Chiliad. Mixed Effects Models and Extensions in Environmental with R Wiley International (2009). [Google Scholar]

Manufactures from Nature Communications are provided here courtesy of Nature Publishing Grouping