A global reptile assessment highlights shared conservation needs of tetrapods

We used the IUCN Red List criteria^34,35 and methods developed in other global status-assessment efforts^36,37 to assess 10,078 reptile species for extinction risk. We additionally include recommended Red List categories for 118 turtle species³⁸, for a total of 10,196 species covered, representing 89% of the 11,341 described reptile species as of August 2020³⁹.

Data compilation

We compiled assessment data primarily through regional in-person and remote (that is, through phone and email) workshops with species experts (9,536 species) and consultation with IUCN Species Survival Commission Specialist Groups and stand-alone Red List Authorities (442 species, primarily marine turtles, terrestrial and freshwater turtles, iguanas, sea snakes, mainland African chameleons and crocodiles). We conducted 48 workshops between 2004 and 2019 (Supplementary Table 1). Workshop participants provided information to complete the required species assessment fields (geographical distribution, population abundance and trends, habitat and ecological requirements, threats, use and trade, literature) and draw a distribution map. We then applied the Red List criteria³⁴ to this information to assign a Red List category: extinct, extinct in the wild, critically endangered, endangered, vulnerable, near threatened, least concern and data deficient. Threatened species are those categorized as critically endangered, endangered and vulnerable.

Taxonomy

We used The Reptile Database³⁹ as a taxonomic standard, diverging only to follow well-justified taxonomic standards from the IUCN Species Survival Commission⁴⁰. We could not revisit new descriptions for most regions after the end of the original assessment, so the final species list is not fully consistent with any single release of The Reptile Database.

Distribution maps

Where data allowed, we developed distribution maps in Esri shapefile format using the IUCN mapping guidelines⁴¹ (1,003 species). These maps are typically broad polygons that encompass all known localities, with provisions made to show obvious discontinuity in areas of unsuitable habitat. Each polygon is coded according to species’ presence (extant, possibly extant or extinct) and origin (native, introduced or reintroduced)⁴¹. For some regions covered in workshops (Caucasus, Southeast Asia, much of Africa, Australia and western South America), we collaborated with the Global Assessment of Reptile Distributions (GARD) (http://www.gardinitiative.org/) to provide contributing experts with a baseline species distribution map for review. Although refined maps were returned to the GARD team, not all of these maps have been incorporated into the GARD.

Habitat preferences

Where known, species habitats were coded using the IUCN Habitat Classification Scheme (v.3.1) (https://www.iucnredlist.org/resources/habitat-classification-scheme). Species were assigned to all habitat classes in which they are known to occur. Where possible, habitat suitability (suitable, marginal or unknown) and major importance (yes or no) was recorded. Habitat data were available for 9,484 reptile species.

Threats

All known historical, current and projected (within 10 years or 3 generations, whichever is the longest; generation time estimated, when not available, from related species for which it is known; generation time recorded for 76.3% of the 186 species categorized as threatened under Red List criteria A and C1, the only criteria using generation length) threats were coded using the IUCN Threats Classification Scheme v.3.2 (https://www.iucnredlist.org/resources/threat-classification-scheme), which follows a previously published study⁴². Where possible, the scope (whole (>90%), majority (50–90%), minority (<50%) of the population; unknown) and severity (causing very rapid (>30%), rapid (>20%), slow but notable (<20%) declines over 10 years or 3 generations, whichever is longer; negligible declines; unknown) of the threat was recorded. Threat data were available for 1,756 of the 1,829 threatened reptile species.

Assessment review

Each assessment underwent two reviews. First, a scientist familiar with the species but not involved in the assessment reviewed the account for biological accuracy and accurate application of the Red List criteria. Once the assessors revised the assessment satisfactorily, staff from the IUCN Red List Unit reviewed the assessment primarily for accurate application of the Red List criteria. The assessors revised the assessment again, if necessary, to satisfy any concerns of the IUCN Red List Unit before the assessment was finalized.

Data limitations

Although we made an extensive effort to complete assessments for all reptiles, some data gaps remain.

Missing species

As of December 2020, 1,145 reptile species, primarily snakes and lizards, were omitted from the present study, including the phylogenetic diversity analyses, because they were described recently and they were described after previous comprehensive assessments from the region. Geographically, they are primarily from tropical regions (as are assessed reptiles) with an underrepresentation of African species (distribution of omitted species: Asia, 41%; Africa, 8%; Australia, 7%; Europe, 3%; North/Central America, 20%; South America, 19%; Caribbean, 5%; Oceania, 4%; percentages sum to greater than 100% because some species occur in two regions). Because they are recently described, many are poorly known, may be rare or occur in a very restricted area, or in poorly surveyed areas that are often subject to high levels of human impacts. As such, recent descriptions are more likely to receive a data-deficient or threatened Red List category than be assigned of least concern⁴¹. The net effect on our analyses is a slight underestimate of the number of threatened snakes and lizards, and plausibly a slight overestimate of least concern species. With tetrapod species described in the future likely to be small-ranged, threatened lizards and amphibians^43,44, surrogacy levels may decline from those reported here.

Geographical coverage

Although we made extensive efforts to map the current known distribution for each species, this information is incomplete for some species. Where appropriate, and following expert guidance, we interpolated between known localities if the ecological conditions appeared appropriate. In addition, species occurrence is unlikely to be spread evenly or entirely throughout the area depicted in range maps, with gaps expected, for example, in patches of unsuitable habitat.

Data-deficient species

For species assessed to be data deficient (1,507 reptiles, 14.8%), there was inadequate information on the distribution, population status or threats (historical, current or projected future) of the species (both from published sources and expert knowledge) to make a direct, or indirect, assessment of the risk of extinction. All species were assessed according to their recognized taxonomic circumscription at the time of assessment. Taxonomic uncertainty therefore did not result in a data-deficient assignment, although some species were listed as data deficient because they are morphologically indistinguishable from another species and therefore estimates of distribution and abundance are not feasible.

Time span of assessments

The assessments were completed between 1996 and 2020, with 1,503 assessments completed before 2011. The IUCN Rules of Procedure (https://www.iucnredlist.org/resources/rules-of-procedure) recommend reassessment every 10 years and thus, as of 2020, 15% of the assessments can be considered outdated. Of the species assessed 1996–2010, slightly more species were threatened (23.0%) than the species assessed more recently (20.7%). This difference is largely explained by the greater percentages of crocodiles and turtles with outdated assessments (29% and 35%, respectively) compared with tuatara, lizards and snakes (0%, 12% and 17%, respectively) and the highly threatened nature of crocodiles and turtles (Supplementary Table 2). The continuing deterioration of biodiversity globally¹ suggests that the species with outdated assessments are more likely to be in higher threat categories today than when they were when last assessed, causing an underestimation of current reptile threat status.

Analyses

Percentage of species threatened with extinction

To estimate the percentage of species threatened with extinction (categories critically endangered, endangered and vulnerable), we used the following formula, which assumes that data-deficient species have the same proportion of threatened species as species that were not data deficient.

where Prop_threat is the best estimate of the proportion of species that are threatened; CR, EN, VU and DD are the number of species in each corresponding Red List category and N is the number of species assessed (excluding extinct and extinct in the wild species).

Data for amphibians, birds and mammals

For all analyses that included data for amphibians, birds and mammals, we used the 2020-1 version³ of the tabular and spatial data downloaded from the IUCN Red List website in May 2020.

Threats

Threats calculations were restricted to species in threatened Red List categories (critically endangered, endangered and vulnerable). Multiple threats can affect a single species. Summaries of threats are for the first level of the IUCN classification scheme. Threats thought to affect only a minority of the global population (<50% of the population) (coded as ‘minority’) were not included. In addition, we removed threats that were assessed to cause ‘no declines’ and ‘negligible declines’ from the analysis (as indicated by the severity coding). We considered all threats without scope or severity scored to be major threats and retained them in the analysis.

Habitat

Analyses of habitat use were restricted to the first level of the IUCN habitat-classification scheme. We excluded habitats for which the major importance to the species was scored ‘no’ and suitability was scored ‘marginal’ and considered all habitats without major importance or suitability scored to be suitable and of major importance and included them in the analyses. We did not consider artificial habitats in the analyses.

Only a small number of reptile species inhabits ‘caves/subterranean’ and ‘marine coastal’ habitats, so they were not included in Fig. 4 but their threat prevalence is summarized in Supplementary Table 4.

Statistics

Statistical tests were designed to avoid inclusion of multiple observations from the same species (because species can occur in multiple habitats and be threatened by multiple threats). To assess whether arid habitat or forest species were more likely to be threatened, we included only species that were restricted to one of these habitat types. For threats analyses, we compared species that occur in forests (including those that occur in forests and other habitats) to those that do not occur in forests. All tests were two-tailed Fisher’s exact tests.

Geographical patterns

The geographical patterns of threat and phylogenetic diversity shown in Fig. 2 are for only terrestrial species (so, for reptiles, excluding 87 species of marine turtles and sea snakes). Tetrapod classes vary widely in the numbers of pelagic marine species and in the methods used to map distributions. Restricting analyses to terrestrial species ensured more-consistent analyses and avoided wide variation in summary values caused by small numbers of species.

Analyses of the distribution maps used polygons either with the following IUCN map code designations or with no codes indicated:

Presence = extant (code 1) and probably extant (code 2)

Origin = native (code 1), reintroduced (code 2) and introduced (code 6)

Seasonality = resident (code 1), breeding season (code 2), non-breeding season (code 3) and passage (code 4).

Ranges for species categorized as critically endangered (possibly extinct) are coded as possibly extinct (code 4) and excluded from the spatial analyses.

All spatial analyses were conducted on a global 0.5° by 0.5 ° latitude–longitude grid (approximately 50 km at the Equator). To explore the influence of spatial resolution, we repeated the surrogacy and phylogenetic diversity analyses at a 100-km resolution. We converted polygon range maps (tagged with the appropriate codes as described above) to these grids. We used a global equal-area pseudocylindrical projection, Goode homolosine.

We mapped the distribution of threatened species as a count of the number of species with ranges overlapping each grid cell.

Estimating the spatial distribution of disproportionate threat and phylogenetic diversity loss

We identified global areas in which each tetrapod class is disproportionately threatened compared with all other classes by comparing the species-richness-adjusted level of threat among the four tetrapod classes. First, for each grid cell, we identified the proportional threat level of each class by dividing the number of species in threatened Red List categories (vulnerable, endangered and critically endangered) by the total number of species for the class found in that cell. Second, for all grid cells in which at least five tetrapod species are present, we compared proportional threat values across the four classes and identified a grid cell as having a disproportionate threat level for a given class if: (1) the grid cell had a proportional level of threat equal to 10% or higher for the class; and (2) the grid cell had a proportional level of threat for the class at least twice as high as the proportional level of threat for the next class. We assessed the sensitivity of disproportionate threat patterns to our definition of disproportionate threat by varying the degree of difference in proportional threat level between the highest and second highest class. We identified the number of grid cells with disproportionate threat for each class when the class had a proportional threat level: (1) higher than any other class; (2) 25% or more higher than any other class; (3) 50% or more higher than any other class; (4) 100% or more higher than any other class; and (5) 200% or more higher than any other class. In the main text, we report results for the 100% or more threat level. Results for all thresholds are included in Extended Data Tables 1, 2.

Conservation strategies

We identified global conservation priorities for each tetrapod class using two alternative strategies: strategy 1 prioritized areas containing many threatened species with relatively highly restricted ranges, whereas strategy 2 prioritized areas core to the most range-restricted threatened species globally. We implemented both conservation strategies within the spatial conservation-planning software Zonation⁴⁵ and the R package zonator⁴⁶, using the additive benefit function and the core-area Zonation algorithms for strategies 1 and 2, respectively, at 50-km and 100-km resolutions for threatened reptiles.

The additive benefit function algorithm prioritizes areas by the sum of the proportion of the global range size of all species included in a given grid cell—a quantity similar to weighted species endemism (as defined previously⁴⁷) and endemism richness (as defined⁴⁸). On the basis of this algorithm, cells with many species occurring only in that cell or few other cells receive the highest priority. The core-area Zonation algorithm prioritizes areas by the maximum proportion of the global range size of all species included in a given grid cell: cells including the highest proportions of the ranges of the most range-restricted species are given the highest priority.

Therefore, comparing the two strategies, strategy 1 gives more importance to the number of species within grid cells (that is, more species = a higher summed proportion), potentially at the expense of the single most-range-restricted species globally, which are instead prioritized directly by strategy 2.

Because complementary representation problems such as these spatial prioritizations often have multiple solutions, we ran five iterations of each algorithm used and summarized variation across those.

Estimating surrogacy

To assess the degree to which conserving the diversity of threatened species of birds, mammals and amphibians (individually or combined) serves as a surrogate for conserving threatened reptile diversity, we calculated a species accumulation index (SAI) of surrogate effectiveness. The SAI is derived from the comparison of three curves: (1) the ‘optimal curve’ represents the accumulation of the diversity of threatened reptile species when conservation is planned using data for threatened reptiles directly; (2) the ‘surrogacy curve’ represents the accumulation of the diversity of threatened reptile species when conservation is planned using the diversity of threatened species diversity of a different class as a surrogate; and (3) the ‘random curve’ represents the accumulation of the diversity of threatened reptile species when conservation areas are selected at random. We estimated optimal, surrogate and random curves based on each reptile-surrogate combination (birds, mammals and amphibians individually and combined). Using 100 sets of approximately random terrestrial grid-cell sequences allowed us to generate 95% confidence intervals around a median ‘random curve’. In addition, because we ran five iterations of each spatial prioritization algorithm for each tetrapod class, optimal and surrogate curves were also summarized using the median and 95% confidence intervals across the five iterations.

We then derived the quantitative measure of surrogacy as SAI =  (s − r)/(o − r), where s is the area under the surrogate curve, r is the area under the random curve and o is the area under the optimal curve. SAI = 1 when the optimal and surrogate curves are the same (perfect surrogacy). It is between 1 and 0 when the surrogate curve lies above the random curve (positive surrogacy), zero when the surrogate and random curves coincide (no surrogacy) and negative when the surrogate curve lies below the random curve (negative surrogacy). We calculated the SAI using R code modified from a previous study⁴⁹. For each reptile–surrogate combination, we report median and 95% confidence intervals across all combinations of optimal, surrogate and random curves (5 target and surrogate curve iterations and 100 random curve iterations).

Although not strictly a measure of surrogacy²⁵, we also calculated the spatial congruence (Spearman’s rank correlation, analogous to a previously published approach⁹) of Zonation priorities for each conservation strategy and spatial resolution.

Coverage by protected areas

We overlayed protected areas (polygons, categorized as IUCN I–VI from the World Database of Protected Areas⁵⁰) over the ranges of all threatened tetrapods and classified species with ranges completely outside any protected area as unprotected.

Phylogenetic diversity

To calculate phylogenetic diversity¹⁵, we used published time trees of mammals⁵¹, birds⁵² and amphibians⁵³. For reptiles, we combined two time trees: a comprehensive squamate time tree containing 9,755 squamate species, including the species Sphenodon punctatus¹⁶, and a turtle and crocodilian tree containing 384 species¹⁷. The time trees contain some species lacking genetic data, added by taxonomic interpolation⁵⁴ to maximize taxonomic coverage. In total, we analysed 32,722 tetrapod species including 10,139 reptiles, 5,364 mammals, 9,879 birds and 7,239 amphibians. For squamates, and for turtles and crocodiles, 10,000 fully resolved trees were available. For each group, we randomly sampled 100 trees and combined them to obtain 100 fully resolved reptile time trees, to accommodate for uncertainty. Similarly, we randomly sampled 100 amphibian and 100 mammal time trees over the 10,000 available.

We thoroughly compared the species name mismatches between geographical and phylogenetic data to match synonyms and correct misspelled names. We also imputed species for which the genus (but not the species) was already present in the tree, for example newly described species (262 amphibian, 1,694 bird, 236 mammal and 777 reptile species). Imputed species were randomly attached to a node within the genus subtree. Because polytomies can result in an overestimation of the phylogenetic diversity, we randomly resolved all polytomies using a previously published method⁵⁴ implemented in R code. This procedure was performed 100 times for birds, and one time for each of the 100 amphibian, 100 mammal and 100 reptile time trees. We included 30,778 tetrapod species, each with geographical and phylogenetic data, in the phylogenetic diversity analyses. This total included 6,641 amphibians, 8,758 birds, 5,550 mammals and 9,829 reptiles. For each class, we estimated phylogenetic diversity¹⁴ for all species and after the removal of threatened species, at 50-km and 100-km resolution. To consider phylogenetic uncertainty (that is, the placement of interpolated species) in phylogenetic diversity calculation for each of the 100 fully resolved trees for each class, we conducted a sensitivity analysis using a previously described method⁵⁵. This method calculates an evolutionary distinctiveness score that (1) increases the total phylogenetic diversity of the clade when including interpolated species and (2) corrects the evolutionary distinctiveness score of species in genera with interpolated species (missing relatives). Following this method, we calculated evolutionary distinctiveness scores⁵⁶ for each cell from the subtree including all species present in the focal cell with the R package caper⁵⁷. For genera with interpolated species, the mean evolutionary distinctiveness score of non-interpolated species was assigned to interpolated species of that genus. For those genera, we computed a second evolutionary distinctiveness score corresponding to the mean evolutionary distinctiveness score of the focal genera (including interpolated species). For species belonging to genera with no interpolated species, the first and second evolutionary distinctiveness scores were identical. Next, we calculated the mean of the two evolutionary distinctiveness scores and reported this value as the evolutionary distinctiveness score of each species. Finally, we computed phylogenetic diversity as the sum of evolutionary distinctiveness scores. Therefore, phylogenetic diversity corresponds to Crozier’s version of phylogenetic diversity⁵⁸, that is, the sum of the branch lengths connecting all members of a species assemblage without the root. Next, we reported median phylogenetic diversity, computed over 100 fully resolved trees for each class. In the figures, cells with fewer than five species were excluded to avoid outliers.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.

Source: Ecology - nature.com