Introduction

Catalogues of species names and atlases of distributions are essential to describe and understand the patterns of biological diversity [1]. Unfortunately, almost-universal biases and incompleteness in the available distributional information on invertebrate species [23–4] hinder the reliability and usefulness of this information. In the case of distributional information, several methodological procedures seek to derive geographical representations of “real” or “potential” distributions from partial chorological data and different environmental predictors [5]. These so-called species distribution models (SDMs) or ecological niche models (ENM) have used the exponentially growing number of massive biological and environmental databases that are freely available on the World Wide Web [6, 7], aided by increasing computational and statistical capacity [8]. The modeling procedures use different conceptual and methodological approaches. We highlight those approaches that use the available information on species occurrence to determine the environmental conditions prevailing where the species are present (the species' Grinnellian niches [9]) in order to predict the suitable region (“potential distribution” [10]). Because these potential distributions are partial geographical representations of localities with climatic conditions similar to those where the species of interest are known to occur, these simulations are useful for designing future surveys to improve biogeographical and taxonomic knowledge [111213141516–17].

The genus Eurysternus Dalman, 1824 is the only genus of the tribe Eurysternini and one of the three Neotropical endemic dung beetle groups, together with the Eucraniini and Phanaeini [18]. The genus presents many morphological and ecological particularities within the subfamily Scarabaeinae. Eurysternus is distributed from central Mexico to southern Brazil. This genus occurs across a broad range of forest conditions and consumes a varied diet of carrion, faeces, and even leaves [1920–21]. The taxonomy of the genus has recently been revised [22], increasing the number of recognized species from 31 to 53. In this study, we compile and georeference the distributional information on the Eurysternus species available in an exhaustive number of natural history collections and bibliographic sources. The purposes of the study are: i) to examine the general macroclimatic determinants associated with the occurrence of these species, ii) to estimate the potential distributions of each of these species for the American continent, and iii) to distinguish those climatically suitable areas that have not been adequately surveyed to pinpoint the regions that need to be explored in the future.

Methods

Information from the Musée Canadien de la Nature (CMNC) database was the primary source of distributional and taxonomic information [22]. This information was complemented by data from other museum specimen labels (see Appendix 1) and bibliographic sources (see Appendix 2). All of this information was included in the MANTIS v. 2.0 [23] database system. The resulting database currently includes approximately 22,000 records belonging to the 53 currently recognized Eurysternus species [22]. The locations were georeferenced using freely available gazetteers at GeoNet Names (see  http://www.geonames.org/[24]). Due to differences in the geographical precision of the information sources, a cautious approach was applied through the use of relatively high resolution (0.08° cells; approximately 100 km², depending on the latitude) to fit both climatic and biological information. All of these data will be freely available in  www.biogeografia.org.

The potential distributions were estimated with a multidimensional envelope procedure according to the protocol recently proposed by Jiménez-Valverde et al. [10] and the former conceptual proposals [25, 26]. The aim of these methods is to obtain geographical representations of the set of locations with climatic conditions similar to those where the organisms of interest are known to be present. The goal is to geographically represent the set of abiotic requirements in which a species can maintain a net positive rate of population increase without immigration (i.e. a geographical representation of the species' Grinellian niche [10]). Presence data are the unique source of information used to determine potential distributions; this implies the assumption that the environmental conditions in these presence localities constitute a partial but reliable description of the whole climatic requirements of the species.

The 19 freely available bioclimatic variables provided by WorldClim (see  www.worldclim.org [27].) obtained by the interpolation of climate-station records from 1950 to 2000 (for details, see [7]) were used as predictors. As the number of used predictors may condition the obtained potential representation, most relevant predictors were selected by using Ecological Niche Factor Analysis (ENFA; see [28]), which is a method firmly rooted in the niche concept [29]. ENFA computes new uncorrelated factors by comparing climatic data in presence locations with conditions in the entire study area. This procedure serves to maximize both the marginality (the distance between the optimum recorded for the species and the average climatic conditions in the area) and the specialization values of the species (the ratio between the climatic variability in the study area and that existing at the points where the selected species occurs). The factors were retained based on their eigenvalues relative to a broken-stick distribution [28]. The original bioclimatic variables selected for each species are those that show the highest correlation values with these retained ENFA factors (absolute factor scores > 0.30). After the most relevant bioclimatic variables were selected individually for each species, we calculated the maximum and minimum scores (the extreme bioclimatic values) in all observed presence cells. All of the cells with climatic values falling within that range were selected as suitable to derive a binary potential distributional hypothesis. This procedure assumes that the recorded occurrences provide a reliable representation of the full spectrum of climatic conditions in which the species can survive and reproduce [10, 30].

Based on the ENFA analyses, the two most relevant bioclimatic variables were used to obtain further estimates. Specifically, the maximum and minimum scores for these variables in the observed presence cells were used to estimate the bioclimatic tolerance of species (maximum-minimum), as well as their averaged conditions (maximum-minimum/2 + minimum). The correlation between the tolerance and the average bioclimatic value was determined with a Spearman rank correlation coefficient (rs). Because the tolerance values depend on the scale measurements, both the temperature and the precipitation values were transformed to a 0-100 scale. Subsequently, minimum and maximum values in the observed localities for each one of the two previously selected variables were used to calculate the climatic area covered by each species, according to the rectangle defined by the ranges in temperature and precipitation values. These climatic-area values were used to estimate the percentage of climatic area overlapped among all possible pairwise species comparisons. Species with fewer than five 0.08° presence cells are excluded from these calculations (n = 6).

Multidimensional envelopes were obtained for each species. The potential distributions obtained from these analyses were overlaid to obtain a map of the potential species richness. This map showed a geographical picture that represented the capacity of a locality to harbor suitable climatic conditions without considering the limits imposed by contingent forces such as biotic factors or dispersal limitations. For the four species occurring in three or fewer 0.08° cells, multidimensional envelopes were constructed by increasing by 10% the minimum and maximum climatic values found at the occurrence points for the five most relevant variables (see Appendix 3). The location of the available database records was superimposed on this potential richness map to identify those potentially diverse regions that were not sufficiently surveyed. As we develop geographical representations of potential distributions, comparing the derived potential species richness map to current knowledge of realized distributions suggest that some climatically suitable areas may be uninhabited due to dispersal limitations (as in the case of Caribbean islands or northern zones above the Isthmus of Tehuantepec). However, when the relevance of these dispersal limitations decreases, as in the case of tropical South American areas, we assume that the comparison of the obtained potential species map with the distribution survey may provide useful information on the location of unrecorded places.

Results

The bioclimatic variables that most frequently showed statistically significant correlations with the retained ENFA factors are the annual mean temperature (in 96% of the species), the annual precipitation (in 43% of the species) and the mean diurnal range of temperature (in 51% of the species). Only two other bioclimatic variables appear in at least 20% of the species: the maximum temperature of the warmest month and the annual range of temperature (Appendix 3). ENFA analyses also indicate that the different factors selected generally explain more than 97% of the total species information. The marginality values range between 0.37 and 3.30 (mean ± standard deviation; 1.17 ± 0.70) and the vast majority of species show high values of specialization (Appendix 3).

The tolerances of Eurysternus species for the annual mean temperature and the annual precipitation, which are the most relevant variables, are significantly correlated (rs = 0.705; n = 47; P < 0.001). Tolerance values (Fig. 1) are negatively correlated with the average annual temperature (rs = −0.731; P < 0.001) but positively correlated with the average annual precipitation (rs = 0.682; P < 0.001).

Fig. 1.

Relationship between the averaged annual mean temperature and annual precipitation values of the 0.08º presence cells for each species (open circles) and the tolerance values (maximum-minimum) represented by whisker plots. Only those species with at least five occurrence cells are represented. The species (X-axes) are ordered according to their average temperature and precipitation values.

From the 1,081 possible pairwise estimates of the shared climatic area for the 47 species considered, in only 61 instances (5.6 %) does a pair of species share more than 50% of the total estimated climatic area. The average percentage of overlap was 15.4 ± 15.9% (mean ± standard deviation).

The sum of the individual potential distributions indicates that the highest values of potential species richness for Eurysternus species can be found in lowland areas below 1000 meters in altitude throughout the North America's east coast, South Florida, the Caribbean islands, the continent's southern Andean highland zone and the Amazonian region (Fig. 2).

Fig. 2.

(A) Potential species richness map for Eurysternus derived from the overlap of individual potential distributions. The records available in the database are represented by black dots with sizes corresponding linearly with the number of database records. (B) Areas with greater potential richness (upper quartile > 19 species) and location of the database records.

Discussion

Temperature-related variables appear to play the most important role in explaining American Eurysternus distributions. Because the marginality values range between 0 and 1 in 95% of the cases where the distribution is normal [28], approximately one-half of the species would have a climatic “optimum” similar to that of the averaged tropical American climatic conditions (marginality < 0.90); almost one-quarter would occur in climatic conditions relatively distant from the mean conditions (marginalities between 0.9 and 1.5); and one-quarter would occur under distinctive climatic conditions (marginality > 1.5).

While the derived climatic optimum of Eurysternus species is relatively variable, the obtained high specialization values suggest that these species generally live under highly restricted climatic conditions. In all, fifteen species show specialization values between 6 and 10, twenty-four species show values greater than 10, and seven species (E. cavatus, E. cambeforti, E. fallaciosus, E. gilli, E. truncus, E. calligrammus and E. uniformis) inhabit extremely specialized climatic conditions. The remaining twelve species have specialization values of approximately 5 or less, showing that they occupy a range of climatic conditions approximately five times smaller than the climatic range of the entire continent. The specialization character of Eurysternus species is related to low average percentage of overlap in the climatic conditions between all pairwise comparisons. Thus, although most species are typically associated with semitropical or tropical conditions, each appears to occur under relatively specific climatic conditions.

The correlation between species tolerance values for the two most important climatic variables suggests that the occurrence of these species in a broad range of temperature conditions is associated with their occurrence in a broad range of precipitation values. Therefore, we infer that a broader climatic niche in Eurysternus species implies at the same time a tolerance for different temperature and precipitation values. As was to be expected, the more temperature-tolerant species are those capable of living under colder conditions, whereas the most precipitation-tolerant species generally tend to occur under rainier conditions. As a consequence, the Eurysternus species with wider climatic niches would primarily inhabit the rainiest and coldest suitable American localities, which are mainly located in the Andean mountains (Fig. 3). Subsequent analyses should try to determine the ancestral or derived phylogenetic position of these tolerant Andean species.

As in other groups [31], dispersal constraints and historical factors seem to have played a determining role in the conformation of current American Eurysternus distributions, because these species are currently absent from certain. priori favorable climatic regions. Both the whole potential species richness map (Fig. 2) and individual predictions (Figs. 4 to 53) indicate that southern North America, Central America, the Caribbean islands and the southwestern portion of South America are frequently climatically suitable but uninhabited by many species. However, the role of dispersal limitation and historical factors should be minimized when the connectivity of the potential localities increases. Under these conditions it is more probable that the available records of Eurysternus species are incomplete and biased such that “real” distributions are wider than observed ones. How to detect these unsurveyed localities?

Fig. 3.

(A) Relationship between the averaged annual mean temperature and annual precipitation values in the 0.08º presence cells for each species. Histograms on top and right-hand side show the frequency of these variables. (B) The same relationship plotted with circles whose sizes are linearly related to the climatic area covered by each species according to the range of temperature and precipitation values associated with that species. The map represents those areas suitable for any Eurysternus species with mean annual temperatures lower than 20ºC and annual precipitation values higher than 2000 mm.

Implications Implications for conservation

Distribution models can be applied in conservation when we are able to generate reliable predictions of realized distributions that can be coupled with reserve selection algorithms. Unfortunately, the lack of reliable absence information seriously hinders this approach [25]. When only partial representations of the potential distribution can be estimated (as in this case and in most exercises that use invertebrate data), the obtained simulations may help to delimit those areas that are potentially rich in species but not yet adequately surveyed [32]. In the future, new species may even be found in these areas [11, 12], allowing to generate more accurate distributional maps and a better taxonomical knowledge that can be subsequently used for conservation purposes.

According to our results, many of the areas with greater potential richness values (upper quartile > 19 species) are located in Central America, extending from the Yucatán Peninsula to the Panama Canal. As a substantial number of database records exist for this area, we suspect that most species that potentially inhabit this region are in fact absent due to dispersal limitation. In South America, the areas with the greatest potential richness generally correspond to lowland zones: inter-Andean valleys, Andean lowlands, the western, central and southern regions of the Amazon and lowlands of the Guyana Shield. However, only a few of these areas have a substantial number of records: the upper Amazon at the frontier between Perú and Bolivia, the Amazon region of Leticia in the north of Perú and the south of Colombia, as well as the Andean region in northern Ecuador. The remaining climatically favorable areas include few records.

These findings suggest that future collection surveys aimed at improving the present taxonomic and biogeographical knowledge of Neotropical Eurysternus should be conducted in the upper Brazilian Amazon (Jurua and Rio Branco basins), the lower Amazon region between Manaus and Santarem, Amapá and the Guyanas, Mato Grosso to Paraguay, and in Central America, in Honduras and Nicaragua.