4 Corresponding author; Present address: Planeación Ambiental y Conservación, Centro de Investigaciones Biológicas del Noroeste, LA Paz BCS 23090 México; e-mail: email@example.com
A widely accepted biodiversity crisis in the tropics has been recently challenged by claims that secondary forests will gradually restore biodiversity losses. This prediction was examined for the herpetofauna in Quintana Roo, Mexico. Quantitative sampling (108 transects) of reptiles was undertaken monthly (January–September 2004) along a vegetation gradient covering induced grasslands, and regrowth and primary rain forests. A total of 35 species was found, 14 being present in and five showing dependence on mature forests. Lizards contributed > 90 per cent of the individuals observed. Reptile abundance and snake species richness was highest in primary forests, even though the lower abundance and richness did not differ between regrowth forest and induced grasslands. Multivariate ordinations and ANOSIM tests displayed clear differences in assemblage structure among vegetation types, mainly caused by contrasting abundances of lizard species having distinctive arboreal or terrestrial habits. There was no evidence that snake assemblages differed between secondary forests and induced grasslands. Microhabitat availability had a key role in shaping species composition through the vegetation gradient. Our results dismiss the hypothesis that secondary forests can act as reservoirs of primary forest reptile diversity on the basis that many taxa depend largely on habitat quality and have specialized life-history traits, and that biological succession does not guarantee the recovery of assemblage complexity.
A long and growing body of evidence has established that tropical areas currently face a massive extinction of species (Brooks et al. 2002, 2006; Brashaw et al. in press). However, the regeneration of rain forests through secondary succession has recently been advocated as a potential mechanism that will counteract current rates of deforestation and biodiversity loss in the tropics in the next few decades, as human populations increasingly migrate to urban areas therefore reducing their impacts on forest ecosystems (Wright & Muller-Landau 2006). A necessary prerequisite for such a biological recovery must consist in that secondary forests can gain over time the habitat complexity of more mature environments and thus host or become open to specialist taxa. Tropical reptiles are not an exception among vertebrates in displaying diverse and complex communities (Pianka 1986). This is accentuated in rain forests (Rohde 1992, Duellmann 2005, Luiselli 2006a), where factors affecting the return of native reptile assemblages to perturbed forests have been suggested to be habitat structure, connectivity with source populations, and the biogeographical context of species. All of these factors must be contemplated in restoration projects (Kanowski et al. 2006). In this sort of study, data are needed to quantify habitat selection of interacting species in both well-established and successional communities, species turnover among forest types, as well as proportions of taxa that rely upon old-growth vegetation cover for persistence (Kanowski et al. 2006, Gardner et al. 2006), before secondary forests can be proved as refuges to tropical forest biodiversity.
Both biological and environmental factors participate in the succession of reptile communities through spatial or temporal patches of forest of changing habitat quality. Community structure is influenced by microhabitat heterogeneity over large-scale ecological processes (Doan & Arriaga 2002, James & M’Closkey 2002) to the extent that pools of taxa may be completely wiped out locally if certain microhabitats are eliminated by agricultural or forestry practices (Glor et al. 2001). Furthermore, species turnover may be affected by the size of primary forest fragments resulting from perturbations (Bell & Donnelly 2006), while competition among ecologically and taxonomically closely related species intensifies along gradients of increasing human disturbance (Luiselli 2006b). Such patterns consistently support that reptiles are good ecological indicators of habitat quality and change; hence basic inventory data of abundance or even presence–absence of species stand to facilitate conservation and management efforts in highly fragmented rain forests (see Urbina-Cardona et al. 2006) even after restoration measures have already been put in place (Kanowski et al. 2006).
The Mesoamerican biodiversity hotspot is second top ranked for total species richness and endemism among the 25 hotspots worldwide, and scores the highest number of reptile species (Mittermeier et al. 1998). This region incorporates the southernmost states of Mexico (Chiapas, Tabasco, Campeche, Yucatan, and Quintana Roo) alongside the complete territory of the seven further countries of Central America. Community patterns of the herpetofauna have been described in some detail in rain forests at relatively pristine sites of both Central and South America (Doan & Arriaga 2002, Losos et al. 2003, Duellmann 2005, Gardner et al. 2007). In south Mexico, autoecological data as well as general life-history features of particular reptile species have been recorded in target areas of the Yucatan Peninsula (Lee 1996, 2000; Campbell 1998; Köhler 2003), but little effort has been made to describe species assemblages in relation to environmental gradients in comparison to elsewhere. Based on the arguments presented above, the goal of this study was three-fold, namely: (1) to describe the diversity of reptile assemblages (lizards, snakes, and tortoises) through a gradient of vegetation types covering primary and secondary forests, and induced grasslands, in a poorly known area of south Mexico, (2) to document the occurrence of taxa by their dependence on forest state and microhabitat features, and (3) to assess Wright & Muller-Landau's (2006) hypothesis using reptiles as case organisms. This new information is directly relevant to the ongoing debate about current and future biodiversity prospects in the tropics (Brook et al. 2006, Laurence 2007, Barlow et al. 2007a, Bradshaw et al. in press), which directly compromises the herpetofauna (Gardner et al. 2006).
Study area.— This study was carried out at the Caobas Ejido (68,000 ha, 150 m asl, 18°26′39″N, 89°06′15″W), east of the Yucatan Peninsula in the municipality of Othon P. Blanco (Quintana Roo, Mexico; Fig. 1). The climate is warm (26°C yearly on average) and subhumid, with heavy rains (1200 mm yearly on average) particularly in summer (June–November, Chan 2002). The influence of human activities, principally agriculture, on vegetation has been extensive and long lasting, currently casting a mosaic of forested and cultivated lands all over the Yucatan Peninsula. Despite primary forests being the predominant vegetation in the region, current deforestation rates have been set at 0.39 percent loss in forest cover yearly (Chowdhury & Schneider 2004). Moreover the peninsula is often whipped by catastrophic events like hurricanes that produce an exceptional increase in woody debris and so enhance the occurrence of fires during the dry season (Whigham et al. 1998).
Vegetation types.— (1) Mature forest (MF). MF (i.e., medium-saturated semievergreen forest; Miranda & Hernández 1963) covers the major part of the Quintana Roo state and, generally, thrives in areas with high precipitation levels (1000–1500 mm/year). Adult trees range between 15 and 25 m in height and a total of 25 percent of the species lose their leaves in the dry season. The most characteristic taxa are Manilkara zapota (L.) P. ROYEN, Brosimum alicastrum Sw., Talisia olivaeformis (KUNTH) RADLK, and palm species such as Thrinax radiata LODD. ex SCHULT & SCHULT and Coccothrinax readii QUERO (Sánchez-Sánchez 1986). MF makes up to 33 percent (32,500 ha) of the Caobas Ejido. (2) Regrowth (i.e., secondary) forest (RF). This sort of vegetation develops in abandoned lands after vegetation clearing or logging have taken place, but also pioneers those areas of forest that burn naturally after hurricanes. It is composed of fast-growing tree species (mainly Cecropia obtusifolia BERTOL, Lysiloma latisiliquum (L.) BENTH, Piscidia piscipula (L.) SARG, Guettarda combsii URB, Caesalpinia gaumeri GREENM, Metopium brownei (JACQ.) URB and Acacia gaumeri S.F. BLAKE), as well as many climbers and shrubs (Sánchez-Sánchez & Islebe 2002), and occupies some 20 percent of the Caobas Ejido (17,000 ha). (3) Induced grasslands (IG). This vegetation type is characterized by large tracts of grassland only punctuated by isolated trees (e.g., Brosimum alicastrum SW) and bushes. Grasslands originate from the severance of primary forests leading to cattle pastures. IG grows upon flat terrains, principally nearby human settlements, though often intermingles with patches of secondary forests forming a typical anthropogenic landscape. Currently, this vegetation blankets up to a further 20 percent of the Caobas Ejido. In this study, we sampled IG in the neighborhood of the villages Caobas and San José, which are the two main human settlements of the ejido. Other than the vegetation types under study, the remaining area of the Caobas Ejido (∼25%) is made up of water bodies, seasonally flooded lowlands, and cultivated terrains.
Field methods.— The distribution of the three target vegetation types (MF, RF, and IG) was identified from a 1997 land-use map (4 bands, 30 × 30 m per pixel) of the Caobas Ejido (Chowdhury & Schneider 2004). IG and RF occurred in the vicinity of the main human conurbations, whereas MF was mainly located in the southernmost part of the study area (Fig. 1). Sampling was done over a period of 9 mo from January to September 2004 by means of line transects 500 m long and 9 m wide (total surface = 0.45-ha per transect). Four transects were surveyed monthly in each vegetation type, with a total of 36 transects per vegetation type throughout the whole period of study (N= 108). The position of the 108 transects was chosen before the onset of the study so that, within each vegetation type and over the study period, transects were separated at least 500 m of one another and at least 1000 m apart from those of other vegetation types. Before monthly surveys, ground-truthing of both vegetation types and transect locations was undertaken. The start, middle, and end points of each transect were marked with small colored labels on top of wooden sticks, all of which were only removed after the whole study had finalized.
Each sampling day, from 0800 h to 1400 h, two people (in all cases VHL and a local guide) walked through the center of the transect width, some 3 m apart from each other, and recorded reptile individuals visually (Crump & Scott 1994). On-site records were complemented with focal searches in microhabitats that could act as platforms, refuges, or burrows. On each reptile sighting, we took note of the hour of the day, species, and microhabitat type where the animal was seen (see below). Sampling effort amounted to 5 person-hours per transect, with a total of 540 person-hours over the whole study.
Data analysis.— Alpha diversity was measured as the total number of individuals (N) and the total number of species (S) counted in each transect (Magurran 1988). The precision of average estimates for N and S was quantified through their standard errors; this was done on a descriptive basis so the normality assumption could be relaxed. Differences in N and S among vegetation types were assessed by means of PERMANOVA (Anderson 2001), a permutation-based analysis of variance. The use of Euclidean distances among transects allowed to contrast the Fisher's F statistic (Anderson 2004a) over a total of 9999 permutations in both global (overall differences among vegetation types) and pairwise (differences between each pair of vegetation types, note that the t statistic applies here) tests, allowing a minimum P= 0.01 (Manly 1997). Finally, sampling efficiency (i.e., species yield) was assessed by means of species per area curves, and estimates of total species richness (i.e., species seen + unseen) were obtained by bootstrapping, a method that makes no assumptions on taxa interactions or distributions, and gives better estimates than jackknife approaches when the number of samples is relatively high (Smith & Van Belle 1984).
Multivariate ordinations were used to characterize reptile species assemblages. The term assemblage was defined as a group of phylogenetically related species coexisting in the same geographical area but not necessarily exploiting the same resources (Fauth et al. 1996). A species × transect (35 taxa × 108 transects) data matrix was compiled, and assemblage structure was assessed by the Bray–Curtis similarity index (Clarke et al. 2006) in a transect × transect triangular matrix. This triangular matrix was employed in all multivariate analysis throughout (CAP, ANOSIM, SIMPER, see below). A Canonical Analysis of Principal Coordinates (CAP; Anderson & Robinson 2003, Anderson & Willis 2003) was performed. CAP is a mode of Discriminant Analysis and, as such, an eigenanalysis-based method that traces principal axes in the direction of maximizing the separation of groups (of transects) specified by an a priori hypothesis. Our working hypothesis was that reptile species assemblages differed among the three vegetation types. The main advantage of the CAP against classical Discriminant Analysis is that canonical axes are calculated in a PCoA space produced by a similarity measure of choice. Three CAP ordinations were carried out. Firstly, untransformed reptile species × transect abundance data allowed to assess species assemblages mainly by their numerically dominant taxa. Secondly, in order to balance the influence of dominant and rare taxa on ordination patterns, the relatively mild square root-transformation was deemed sufficient as densities per transect never exceeded 10 specimens among individual taxa. This transformation was necessary as the Bray–Curtis index is prone to emphasize dominant species (Clarke et al. 2006), a notable feature of the data collected. Thirdly, ordinations from presence–absence of species accounted for compositional patterns of reptile taxa irrespective of their relative abundances. The working hypothesis was tested by ANOSIM (Clarke 1993), a nonparametric MANOVA-like test whose statistic (R) has a frequency distribution obtained by permuting the ranks of the similarity values in a triangular matrix, here the same Bray–Curtis similarity matrix was used in the CAP. A total of 9999 permutations was carried out in global and pairwise tests, allowing a minimum α= 1% (Manly 1997). In both PERMANOVA and ANOSIM tests, statistical significance meant an effect size unlikely to occur by chance. Indicator species within vegetation types were detected by SIMPER analysis (Clarke 1993). This routine calculated average similarities (and their standard deviations) contributed by each taxa to between-transect similarities within each vegetation type. The SIMPER ratio for each taxon was the average to the standard deviation of similarities in each vegetation type. Indicator species were investigated further by multivariate logistic regression (Zuur et al. 2007). One model equation was constructed for each species, with the presence/absence of that species being the response variable and the three vegetation types in binary format as the three explanatory variables. A backward stepwise procedure was employed to select the set of vegetation types that maximized the fit at P= 0.05. A species was regarded dependent on those vegetation types that had slopes statistically different from zero in the best-fit model.
Finally, the frequency of reptile records by microhabitat type and by diel activity habits was also investigated. Microhabitat guilds sorted species by their association with particular habitat features defined as the main structural elements common to all three vegetation types, whether related to standing structures of living vegetation (i.e., canopy, branch and shoot of trees and bushes) or ground substrates (fallen trees, bare rock, stones, leaf trash or, though locally very rare, bare soil). We defined microhabitat as the specific location of an organism within its habitat in terms of those factors defining the internal structure of habitat variation within a community (Ehrlich & Roughgarden 1987). This term being associated with small-sized habitat features is well accepted in reptile studies (James & M’Closkey 2002, Pianka 1986). The five microhabitat guilds encompassed species that were observed (1) only upon trees, (2) only upon bushes, (3) moving back and forth between bushes and ground substrates, (4) only upon ground substrates, and (5) upon ground or fossorial. The latter classification (see Lee 1996, 2000; Campbell 1998) delimited a gradient from purely vegetation- to purely ground-dependent occurrence. On the other hand, diel activity guilds aggregated species by their diurnal or nocturnal habits according to Lee (1996).
Logistic regression analyses were done in the SPSS version 11.0 software (SPSS 2007), bootstrapping was carried out in Estimates (Colwell 2006), diversity indices, species per area curves, ANOSIM tests and SIMPER were performed in the statistical package PRIMER version 5 (Clarke & Gorley 2001), whereas PERMANOVA and CAP analyses were implemented using the DOS computer programs by Anderson (2004a and 2004b, respectively).
Species richness.— After 108 transects, 756 individuals belonging to 35 species (15 lizards, 18 snakes, and 2 tortoises) were recorded. No reptiles were spotted in three transects from MF and one transect from RF. Only one specimen of each of the two tortoise taxa was found (Appendix 1), with no single encounter of the unique species of crocodile present in the area (Crocodylus moreletii DUMÉRIL & BIBRON 1851) though it is common in adjacent water bodies. The total and average number of species and of individuals tended to be larger in MF than in RF and IG (Table 1). PERMANOVA tests revealed that differences in species richness among vegetation types were not statistically significant either for the whole reptile assemblage (P= 0.15, N= 36) or for the lizards (F= 0.7, P= 0.51), but were highly significant for the snakes (P= 0.01, N= 36). Snake species richness did not vary between IG and RF (P= 0.64, N= 36), but was on average 4 and 2.6 times larger in MF than in RF (P= 0.01, N= 36) and IG (P= 0.03, N= 36), respectively. The highest reptile species richness was found in one transect through MF with 20 individuals from eight different species (6 lizard and 2 snake spp.). Fifteen of the 18 snake species recorded belonged to Colubridae, while Polychridae (four Anolis spp.) and Gekkonidae (3 spp.) were the most diverse lizard families. The three vegetation types shared seven lizard and three snake species, 12 reptile taxa were common to RF and IG or MF, while 14 species were exclusive to MF (Appendix 1).
Table 1. Number of species and of individuals of lizards and snakes (±SE) surveyed in three vegetation types at the Caobas Ejido (Quintana Roo, south Mexico). Percentage of species seen out of predicted total species richness was estimated by bootstrapping. The number of transects within vegetation types yielding reptile records are also presented (totals are %). A total of 36 transects per vegetation type were sampled (0.45-ha/transect). MF=Mature Forest, RF=Regrowth Forest, IG=Induced Grassland.
Number of individuals
8.7 ± 0.5
6.6 ± 0.3
5.7 ± 0.4
7.9 ± 0.4
6.4 ± 0.3
5.4 ± 0.4
0.8 ± 0.0
0.2 ± 0.0
0.3 ± 0.0
Number of species
3.8 ± 0.1
3.3 ± 0.1
3.1 ± 0.1
3.1 ± 0.0
3.1 ± 0.0
2.8 ± 0.1
0.7 ± 0.0
0.2 ± 0.0
0.3 ± 0.0
Percentage of total species seen (%)
Number of transects with reptile records
Snakes and lizards were observed in 31 percent and 96 percent of transects, respectively. The detection of species as a function of sampling effort leveled off in RF and IG for all species (Fig. 2A), and lizard and snake species (Figs. 2B, C), but kept peaking in MF as a result of the yield of snake species (Fig. 2C). This pattern indicates that local species richness was appropriately estimated for lizards and snakes in RF and IG, and for lizards in all three vegetation types. However, snake species richness was likely underestimated in MF. Bootstrapped estimates for species richness forecasted that over 80 percent of the species present in the area may have been surveyed in this study (Table 1), whereas those estimators were generally over 75 percent for all reptiles except those for lizards in IG (Table 1).
Abundance.— Total reptile (P= 0.01, N= 36), lizard (P= 0.01, N= 36), and snake (P= 0.01, N= 36) abundances varied significantly among vegetation types according to PERMANOVA tests. In all pairwise tests, differences in abundance did not differ between RF and IG (P= 0.20–0.64, N= 36), whereas such differences were statistically significant between RF and MF for snakes (P= 0.01, N= 36) and all reptile species considered together (P= 0.02, N= 36), and marginally so for lizards (P= 0.07, N= 36). Lizards and snakes differed markedly in density (Table 1). Lizards represented nearly 94 percent of the reptiles seen. They were the dominant group in the three vegetation types, hence six species as a whole and four species in each vegetation type provided with around 70 percent of the individuals counted. Dominant lizard species varied among vegetation types (Table 2), i.e., Anolis tropidonotus PETERS and Anolis lemurinus COPE were largely abundant in MF and RF, respectively, while Sceloporus chrysostictus COPE and Aspidoscelis angusticeps COPE were commoner in IG. Finally, Ameiva undulata (WIEGMANN) and Basiliscus vittatus WIEGMANN were conspicuous irrespective of the vegetation types. The highest densities found were 10 individuals of A. tropidonotus for lizards and two individuals of Coniophanes imperialis (KENNICOTT) for snakes, each in one different transect through MF.
Table 2. SIMPER Bray–Curtis similarity contributions of reptile species to three vegetation types surveyed at the Caobas Ejido (Quintana Roo, south Mexico). A total of 36 transects per vegetation type were sampled (0.45-ha/transect). Square root-transformed abundances of all species (n = 35 including lizards, snakes and tortoises) and a 90 percent similarity cut-off were employed. Percent average abundance is the percentage of reptile individuals to total abundance per transect and averaged across all transects within a vegetation type.
Vegetation types and indicator species
Average abundance (%)
Between-transect Similarity (%)
Similarity contribution (%)
Species assemblages.— ANOSIM tests reflected very highly significant differences in reptile assemblage structure among vegetation types irrespective of the transformation being used (α= 0.01%, N= 36, for the three tests). Likewise, all ANOSIM pairwise tests between vegetation types were very highly significant (invariably α= 0.01%, N= 36). Ordinations by CAP clearly supported the ANOSIM results, with CAP 1 separating transects in IG + RF from those in MF, and CAP 2 separating transects from IG and those from RF, for both transformed and untransformed data (Figs. 4A–C). The proportion of correct reclassifications of transects to vegetation types amounted to over 85 percent from raw and square root-transformed species abundances and from presence–absence occurrences indistinctly (Table 3). Such an outcome strongly indicates that vegetation types were the main factor structuring herpetofaunal assemblages in the area.
Table 3. CAP re-classification success from an a priori hypothesis of differences in reptile assemblage structure among three vegetation types (36 transects per vegetation type, 0.45-ha/transect) considering all species, only lizards or only snakes surveyed at the Caobas Ejido (Quintana Roo, south Mexico). Bray–Curtis similarities were employed from raw species abundances that were not transformed (NT), transformed by simple square root (√), or converted to presence–absence (PA). CAP results include: m as the number of PCoA axes used by the Discriminant Analysis, and Correct reclassification percent for each vegetation type and in total. MF = Mature Forest, RF = Regrowth Forest, IG = Induced Grassland.
Taxa and guilds
Number of taxa
Number of transects
Correct reclassifications (%)
Differences in lizard assemblage structure among vegetation types were also detected by ANOSIM (α= 1%, N= 36), with all pairwise tests being significant (α= 1%, N= 36). In that line, the success in CAP reclassification rates to any given vegetation type was over 70 percent for lizard assemblages (72–94%) and much larger for them than for snakes (29–70%) (Table 3). The latter result was principally due to misclassifications of snake transects from RF into IG. Indeed, differences in snake assemblage structure among vegetation types were highly significant (α= 1%, N= 36), though pairwise tests displayed significant differences only between MF and either RF or IG (both tests scored α= 1%, N= 36) but could not discriminate the snake assemblages in RF and IG (α= 39%, N= 36). ANOSIM scores were identical from untransformed or transformed species abundances. Therefore, the assemblages of both groups of taxa clearly responded to the vegetation signal from mature to perturbed vegetation.
SIMPER average similarity between transects per vegetation types exceeded 40 percent for lizards and all reptiles, but it decreased to less than 30 percent when only snake species were considered (Tables 2 and 4). Those species contributing to largest similarities between transects in each vegetation type were the numerically dominant lizards (Table 2). On the other hand, the snakes C. imperialis and Drymarchon melanurus (DUMÉRIL, BIBRON, & DUMÉRIL) explained above 95 percent of the similarity in snake assemblages in MF and RF, respectively, whereas Stenorrhina freminvillii, Bothrops asper (GARMAN), and C. imperialis were the assemblage indicators in IG (Table 4). Nonetheless, the very low density of these species should be taken into account when interpreting these results. For both lizards and snakes, those taxa that could be regarded as indicators of any of the vegetation types were identical irrespective of whether data were transformed or not.
Table 4. SIMPER Bray–Curtis similarity contributions of snake species to three vegetation types surveyed at the Caobas Ejido (Quintana Roo, south Mexico). A total of 36 transects per vegetation type were sampled (0.45-ha/transect). √-transformed abundances of the snake species (n = 18) and a 90 percent similarity cut-off were employed. Percent average abundance is the percentage of snake individuals to total snake abundance per transect and averaged across all transects within a vegetation type.
Vegetation types and indicator species
Average abundance (%)
Between-transect similarity (%)
Similarity contribution (%)
Occurrence by vegetation and microhabitat types, and diel activity.— Our logistic regression models revealed that, among the 15 species of lizards, four were significantly influenced by the presence of primary forests alone, namely Thecadactylus rapicaudus (HOUTTUYN) (β= 3.4, P < 0.01), A. tropidonotus (β= 2.2, P < 0.01), Coleonyx elegans GRAY (β= 1.3, P < 0.05), and Hemidactylus frenatus SCHLEGEL (β= 1.3, P < 0.05). Only one species (A. lemurinus) was affected positively by secondary forests (β= 1.5, P < 0.01). Among the snake species, uniquely C. imperialis was positively related to MF (β= 1.7, P < 0.01).
Nearly half of the lizard species found in this study has been documented to live in trees and bushes, that is above the ground, whereas 11 of the 15 taxa are diurnal. In contrast, all the snake species are chiefly associated with ground living, and 80 percent of them (12 of 18) have nocturnal activity. The dominant lizard species in the study area are all diurnal and dwellers upon ground substrates and/or bush branches, and they accounted for the fact that most of the reptile records were made in these microhabitats throughout (Appendix 1). By vegetation type (Fig. 3), the ground was the most common reptile microhabitat in IG (77% of seen individuals), and the second most common habitat in RF (42.3%) and MF (31.9%). Bushes were the predominant microhabitat for reptiles in MF and RF (>80% of seen individuals). The four arboreal species were only found in MF (5.1%), and the five fossorial taxa occurred at very low abundances in MF and IG but were absent in RF. Finally, 10–20 percent of the individuals observed can move to and fro between ground substrates and bush stands and were observed in all of the vegetation types (Fig. 3; Appendix 1).
The density of lizards and snakes varied through the gradient of increased human intervention represented by the three vegetation types surveyed in the Caobas Ejido. Indeed, mature forests exceeded secondary forests and induced grasslands by (on average) one and two lizard individuals, respectively, and over 50 percent snake densities per hectare. This pattern is widespread where pristine versus perturbed rain forest surveys have been undertaken, such as in Madagascar (Lehtinen et al. 2003), Nigeria (Akani et al. 1999, Luiselli & Akani 2002), Peru (Duellmann 2005), and Mexico (Urbina-Cardona et al. 2006). On the other hand, lizards were markedly more abundant than snakes in the study area, an obvious trend given that local lizards are of smaller size than snakes and forage on preys low in the trophic chain thus supporting relatively higher abundant populations. On the contrary, the bulk of snake species predates on high trophic levels, quite often encompassing lizards in their diet, and so they are necessarily rarer in the environment (Barbault 1971, 1987, 1991; Fitch 1987; Greene 1997; Luiselli 2006a, c).
The establishment of both groups in recovered primary forest may well be reliant on the colonization of their prey, and so take place over longer time scales for snakes feeding on larger items. Furthermore, secondary forest and induced grassland are interspersed in a matrix of mature forests in the study area, hence natural corridors and ample ecotones seem secured for reptile migration to take place naturally, although some management might be necessitated as shown in Australian rain forests (Kanowski et al. 2006). For instance, fossorial habitats for specialists like the rare Yucatan banded gecko (Coleonyx elegans), absent in degraded environments in the Yucatan Peninsula, may need to be facilitated. The same may be true for the tortoise species (Luiselli 2003). It is however unclear whether secondary forest regeneration may actually be useful for increasing the population sizes of those taxa that specialize in living in mature forests, a key question to consider particularly in tropical hotspots (Laurence 2007). In fact, among the four lizard and one snake species that showed dependence on mature forests in the study area, there was no one species whose density was also positively influenced by regrowth (secondary) forest. Moreover, there was only one single lizard species (i.e., just 3% of the total number of species analyzed) for which occurrence was related to regrowth forest. Similar evidence has been presented for butterflies (Barlow et al. 2007b), birds (Barlow et al. 2007c), and amphibians and reptiles (Gardner et al. 2007) in the Brazilian Amazon.
Further research is necessitated to evaluate to what extent tropical secondary forests could recreate over time the habitat properties and ecological processes that shape the assemblages of terrestrial ectotherms characteristic of primary tropical forests. As suggested below, data on resource partitioning, habitat selection, and interspecific competition could shed light on assemblage patterning, and those factors could manifest over complex cascade effects through environmental and structural habitat gradients. Although no environmental variables were measured here in order to characterize the effect of habitat alteration on reptile community structure, the fact that the anthropogenic gradient was fairly strong may well be used to highlight some noteworthy trends. Indeed, distinct distribution of individuals among species and different species sets resulted in clear differences in lizard and snake assemblage structure through the anthropogenic gradient of vegetation, as obtained from multivariate ordinations. As to the lizards, a total of 40 percent of the taxa were unique to mature forests and this is clearly related to both structural and environmental gradients. On one hand, the three-dimensional habitat conferred by bushes and trees must enhance the recruitment of perching species and individuals. On the other hand, many rain forest reptiles respond positively to the high relative humidity maintained under closed and stratified canopies, while temperature regimes heavily condition species occurrence in open habitats such as pastures (Sartorius et al. 1999, Luiselli 2005, Urbina-Cardona et al. 2006). Those ecological requirements indicate that a potential transition from secondary to primary forest lizard assemblages should guarantee the recovery of not only the tree component but also the understory vegetation. Albeit species-specific ecological requirements are incompletely known, some of them are collated in the following for the most conspicuous species.
Anoline lizards (four species) were among the main taxa discriminating reptile assemblages by vegetation types in the Caobas Ejido, most outstandingly A. lemurinus (very abundant in RF) and A. tropidonotus (very abundant in MF) together accounting for ∼40 percent of the reptiles seen. Anolis spp. are widespread and often abundant in Neotropical forests (Losos 1994, Cast et al. 2000, Schlaepfer & Gavin 2001). A total of 51 species of this genus occur in Mexico, 36 of which are endemic (Flores-Villela & Canseco-Márquez 2004). Anolines live upon bushes and trees, and species distributions are determined by vegetation structure, i.e., branch width, height, and inclination (Mattingly & Jayne 2004), alongside thermal needs and prey size (Schoener 1974). The four species recorded here perch upon the canopy of primary and secondary forests in Yucatan (Lee 1996, 2000; Campbell 1998) so their minor occurrence in induced grasslands can be clearly explained by the extremely low or null density of aerial, woody vegetation. The presence of A. tropidonotus and A. lemurinus in primary versus secondary forests displays different patterns in areas of Mexico (Pozo de la Tijera 2001) and Costa Rica (Savage 2000), but mechanisms controlling habitat selection appear to be lacking for this complex lizard group. Only in Cuba do sympatric anolines show very weak correlations between ecological and phylogenetic relatedness across species, suggesting that closely related species are no longer ecologically similar as a result of evolutionary divergence (Losos et al. 2003). This type of analysis would be worth undertaking on the diverse Yucatan anolines, and also on the Colubridae snakes encompassing a pool of up to 15 species found only from our survey in the Caobas Ejido. In contrast, the lizards S. chrysostictus and A. angusticeps were the only dominant species that were more abundant in induced grasslands than in forested areas. This could be expected from the fact that both species are terrestrial, active foragers, commonly observed in cultivated lands and rocky outcrops in Yucatan, and one of the pioneer invaders following forest logging (Penner 1973, Lee 1996). It is likely that dense vegetation canopy may impair the thermal requirements of these lizards. Lastly, only two lizards were detected at relatively high densities anywhere in the three vegetation types, those are the generalist taxa, A. undulata and Basiliscus vittatus. Indeed, while A. undulata distributes widely over Central and South America (Hower & Hedges 2003), Ameiva spp. are active foragers (Anolis spp. can be among their prey; see Simmons et al. 2005), not territorial, and may alternate between terrestrial and arboreal life (Simmons et al. 2005). Temperature constraints in closed-canopy environments have been reported for A. undulata (Sartorius et al. 1999), and in fact this species appears to invade forested areas by taking advantage of the canopy gaps opened by dead trees (Urbina-Cardona et al. 2006). On the other hand, B. vittatus is found in lowland habitats from coastal regions of central Mexico to Central America and Ecuador (Savage 2002, Köhler 2003), where it is mostly associated with riparian habitats and frequently seen on the ground or perching on bushes and the lower branches of trees in rain forests subjected to different degrees of alteration (Campbell 1998).
Snake assemblage structure in mature forests could as well be discriminated from perturbed vegetations in multivariate space, as a result of larger densities and species richness in mature forests and one-third of taxa being exclusive to this habitat in our study. However, in contrast to lizards, secondary forests and induced grasslands held snake assemblages that could not be differentiated by multivariate tests, so driving forces other than, or combined with, vegetation structure may control snake species assemblages there. Similar trends have been observed in tropical Africa (Andreone & Luiselli 2000, Luiselli 2006b), particularly in species-rich areas (Luiselli 2006a, 2008), and might well be attributed to niche segregation and ecological specialization in response to resource availability. Specialization indeed explains how some tropical snake species peak in density in the high variety of niches provided by mature forests while being rare or absent elsewhere (Spawls & Branch 1995, Luiselli 2006b). Though data on diet not being recorded in this study, the fact that the pool of snake taxa collected included species predating on mammals, amphibia, and lizards, suggests that some degree of food resource partitioning may occur (Luiselli 2006a) and it would be interesting to investigate this further. In fact, recent work suggests that snake species co-occurring within certain habitats usually partition their niches trophically (e.g., food availability) but not spatially (Luiselli 2006a), while lizard assemblages do the other way round, segregating through microhabitat types (e.g., substratum types and trees with different diameter, Luiselli et al. 2007) independent of trophic resources (Luiselli 2007, 2008). Additionally, snake assemblages may be organized through vegetation gradients at spatial scales wider than those of lizards, as reported in tropical Africa (Hughes 1983), and their study may require larger transect areas. An indication of that emerges from the fact that although average snake species richness was higher in mature forests than in perturbed vegetations from the Caobas Ejido, they were found in less than one-third of all transects undertaken in forests. The ensuing spatial clumpedness, much larger than that observed in lizards, could be accurately quantified through increased sampling effort and area, whereas data on vegetation structure and trophic interactions could unfold mechanisms by which different habitat scenarios may shape species composition patterns per unit area and their habitat-related variability.
The habitat quality of the mature forests of the Caobas Ejido manifests by the occurrence of umbrella species like the globally endangered Baird's tapir (Tapirus bairdii (GILL) (Gill, 1865)), and several IUCN near threatened taxa like the ocellated turkey (Meleagris ocellata (CUVIER) (Cuvier, 1820)), the great curassow (Crax rubra L.), the puma (Puma concolor (L.)) or the jaguar (Panthera onca (L.)), the latter feline displaying higher abundances than in the Calakmul Biosphere Reserve (Ceballos et al. 2002). These data derive from the joint work of national universities and regional NGOs, which to date have not targeted reptiles as case species for conservation or applied research. The species yield of the reptiles recorded in this study (35 taxa in 48.6 ha of area surveyed) represents nearly 35 percent of the reptile biodiversity known to the Quintana Roo state (Lee 1996, Campbell 1998), widens the distribution range of the lizard T. rapicaudus, and the snake S. freminvillii (Luja & Calderón-Mandujano 2005, Luja 2006a), and caters for a new record of the snake Tantilla shistosa (BOCOURT) (Luja 2006b) to Quintana Roo. Furthermore, a total of 11 species found is listed in international (IUCN, CITES) and/or Mexican red lists (see Appendix 1), hence should merit major conservation efforts. Our data clearly show that the Caobas Ejido possesses a diverse and emblematic fauna of reptiles, and corroborates that sampling efforts have been to date modest for the region. In an area in which information on the biology and ecology of internationally and nationally protected reptiles is scarce, it is hoped that this study will prompt further research that could potentially become instrumental in the management and funding of conservation strategies in south Mexico and the Neotropics. Assemblage fidelity where consistent species assemblages can be found throughout vegetation gradient has already been suggested as a useful criterion for the selection of protected areas (Oliver et al. 1998), where the herpetofauna showed spatial species turnover rates comparable to insects and vascular plants and much higher than those of birds of vertebrates. Conservation-focused research in the region should regard the assemblage structure of the herpetofauna as a potential surrogate for habitat quality that can be endorsed to the selection of protected areas, while their study implies costs much lower than those devoted to birds and mammals.
In conclusion, our results play down the role of secondary forests in the recovery of the reptile diversity of primary forests. Many reptiles may largely depend on habitat quality and microhabitat and niche structure, which may not be directly provided by secondary forests, while biological succession does not guarantee the recovery of assemblage complexity. In particular, our study revealed a different ecological response by lizard versus snake assemblages to a vegetation gradient. Thus, snakes seem to form distinctive, well-structured assemblages only in primary forests. In contrast, lizards may be more resilient than snakes in recolonizing secondary forests after a perturbation has occurred and form there assemblages whose structure can be distinguished from severely modified forest environments like extensive grasslands. This may further imply that snakes can be more threatened than lizards if the recovery of pristine forests had to rely on secondary forests. Ecological specialization and location in the trophic web emerge as key aspects to consider in future studies addressing the recovery of mature forest reptiles, but currently the paucity of biological and ecological data on given taxa, reptiles in particular, further precludes the assessment of global tropical forest recovery prospects.
We are grateful to F. Candanedo, A. González, and particularly to S. Cornelio for their generous assistance during fieldwork, and to Elizabeth Pearse, Judy Foster-Smith and Corey Bradshaw and for their thorough revision of the final version of the manuscript. R. Calderón contributed much to species identification and A. González designed the map of the study area. Fieldwork was sponsored by ECOSUR, CONACyT, and Idea Wild (VHL), while a grant from the Ayuntamiento de Onda (Spain; SH-P) supported the completion of the manuscript. Author Contributions: Study design: VHL DGS. Data collection: VHL. Analysis: SHP. Write-up: SHP LL.
Table APPENDIX 1.. Species, family and order list, and total abundances of reptile taxa found in 108 transects (0.45-ha/transect) from three vegetation types (36 transects/vegetation type) at the Caobas Ejido (Quintana Roo, south Mexico). Species status according to Mexican legislation (NOM, 2001), CITES, and IUCN are attached. Microhabitat = fossorial, ground, bush, tree. Diel habits = diurnal, nocturnal. MF = Mature forest, RF = Regrowth forest, IG = Induced grassland.