Large vertebrate responses to forest cover and hunting pressure in communal landholdings and protected areas of the Yucatan Peninsula, Mexico

Abstract

Although tropical forest vertebrate populations are often heavily impacted by both habitat loss and subsistence hunting, in many human-dominated landscapes their fate depends on the wider habitat matrix outside protected areas. This study used line-transect surveys to examine patterns of vertebrate species richness and encounter rates (ER) throughout Mexico's Yucatan Peninsula across a broad spectrum of anthropogenic forest disturbances. Censuses were carried out at eight forest sites, including two localities within the Sian Ka'an Biosphere Reserve, a communally managed protected area, a private forest reserve and three communal-tenure landholding units. Species richness increased with increasing forest cover in the surrounding landscape, but was not affected by the degree of hunting pressure. Responses to different forms of disturbance varied markedly across species and functional groups. Aggregate ER of old-growth specialists were positively correlated with forest cover, but unrelated to hunting pressure. Hunting pressure was strongly related to aggregate ER of game species, particularly of preferred prey, and a meta-analysis of species-specific correlations showed a significant effect of hunting pressure on the ER across individual game species. Sampling within old-growth forest of Sian Ka'an yielded relatively high ER for almost all species. In comparison, low ER were obtained for several species at Sian Ka'an sites consisting of a higher proportion of open habitat. Landholding units showed low relative abundances, but species ER varied depending on disturbance type. The communally managed reserve exhibited high ER for both game and non-game species. Synthesis of these results with comparable data from 19 additional Mesoamerican sites shows that species-specific ER found across the Yucatan Peninsula are within the ranges observed elsewhere, with a consistent pattern of lower abundances in hunted sites. It is critical that private, communal and public-sector reserves are integrated into a coordinated management approach across the wider landscape in this biodiversity hotspot and throughout the Mesoamerican Biological Corridor.

Introduction

Habitat loss and fragmentation are recognized as the most serious threats to forest biodiversity (Lindenmayer & Fischer, 2006). At least 70% of the resident vertebrate species in tropical forests depend on closed-canopy environments (Johns, 1992), although many species are not obligate forest specialists (Emmons & Feer, 1997). In addition, subsistence and market hunting can lead to population declines or local extirpations, and dramatically alter the species composition and size structure of residual faunal communities (Peres, 2001). Creation of strictly protected areas and analogous reserves has been central to contemporary conservation strategy worldwide. However, even if all reserves were effectively protected, their limited spatial extent cannot ensure the long-term persistence of many native plant and animal populations in the wider countryside (Putz et al., 2001). Assessments of the biodiversity-conservation value of human-managed ecosystems, and particularly of patterns of species persistence and abundance under different land-use scenarios have thus become a research priority in conservation ecology (Daily et al., 2003; Bowen et al., 2007). Yet, landscape-scale studies on forest fauna are sorely lacking; data on species distribution, status of wildlife populations, species habitat requirements and availability of optimal and marginal habitats are insufficient in many tropical countries, including Mexico (Valdez et al., 2006). Although well-replicated studies have related species incidence to forest cover and anthropogenic disturbance (e.g. Urquiza-Haas, Peres & Dolman, 2009), few previous studies have estimated population densities or encounter rates (ER) of forest wildlife at replicate sites across landscape-scale disturbance gradients.

This study examined patterns of vertebrate species richness and ER across the Yucatan Peninsula, where most local people still rely heavily on subsistence activities such as slash-and-burn agriculture and hunting (Urquiza-Haas et al., 2009). Moreover, the region is undergoing exceptionally rapid infrastructure development and population growth driven largely by the tourism industry, especially along the coast (Klepeis, 2003). On the positive side, the region still retains one of the last major blocks of tropical forest remaining in Mesoamerica, which exhibits low deforestation rates in a number of communally managed land-tenure units (Bray et al., 2004; Dalle et al., 2006). Under this land-use scenario, the international Mesoamerican Biological Corridor faces the challenge of integrating sustainable development while maintaining forest cover (Miller, Chang & Johnson, 2001). Although much attention has been given to land cover status, very little is known about the degree of threat faced by wildlife species and their current population status. Key study objectives were to (1) compare large-bodied mammal and bird species diversity and ER across eight study sites differing in the degree of hunting pressure and forest cover; (2) discuss the current role of several land-management units for wildlife conservation; (3) compare survey results with those reported for other Mesoamerican forest sites.

Methods

Study sites

This study was conducted between July 2003 and March 2004 across a wide forest disturbance gradient in the States of Quintana Roo and Yucatan. Main vegetation types in the study region are semi-evergreen low- and medium-statured forest (Pennington & Sarukhán, 1998; González-Iturbe & Tun, 2004). We compare large-bodied mammal and bird species diversity and abundance across a wide gradient of hunting pressure and integrity of forest cover. Although forest cover was the primary criterion for choosing study sites, land tenure and prior information on levels of hunting pressure therefore also influenced site selection (see Urquiza-Haas et al., 2009 and below). Forest sites surveyed included three ejidos or communal tenure landholding units, a private property, two small protected areas within a private landholding and an ejido, and two forest sites within the Sian Ka'an Biosphere Reserve (Table S1, Fig. 1). At the severe end of the disturbance gradient were ejidos, where the local economy is based on shifting agriculture, cattle ranching or mechanized agriculture (Tezoco Nuevo and Tierra Negra). Human population within these landholding units was high and subsistence hunting was practiced on a regular basis (Table S1). The most intact sites, in terms of forest cover and mean stand basal area, were Sian Ka'an-Uaymil, Las Palmas (private property) and X-Conha (a forestry ejido) (Urquiza-Haas, Dolman & Peres, 2007); however, these sites differed in their history of hunting pressure (Table S1). Intermediate in terms of forest cover and/or hunting pressure were the Sian Ka'an site near the reserve headquarters, and two recently created protected areas (El Zapotal, Yodzonot Laguna). Although the Otoch Ma'ax Yetel Kooh protected area was proposed in 1992, this reserve was not officially decreed until 2002 (Ramos-Fernández et al., 2005). El Zapotal Reserve was purchased in 2002 by a conservation organization to halt further agricultural expansion and protect a key wildlife habitat area adjacent to the Ría Lagartos Reserve (Faller-Menéndez, 2002). Further details of survey sites and their history of anthropogenic disturbance are provided in Table S1.

Figure 1.  Map of the study area in Mexico's Yucatan Peninsula, showing the location of line-transect survey sites. Inset maps of study sites are superimposed on classified Landsat images and show the vegetation mosaic within and outside 1 km transect buffers and the location of landholding units. Za: (A) El Zapotal Private Reserve; TeNu: (B) Tezoco Nuevo ejido; YL: Yodzonot Laguna, (C1) Otoch Ma'ax Yetel Kooh protected area, (C2) Valladolid ejido; SK: (D) Sian Ka'an Biosphere Reserve, near the reserve headquarters; XC: X-Conha ejido, (E1) forestry polygon, (E2) polygon designated for agricultural activities; Pa: (F) Las Palmas private property; SKU: Sian Ka'an-Uaymil, (G1) Sian Ka'an Biosphere Reserve; (G2) Uaymil protected area; (G3) Uninhabited private properties; TN: Tierra Negra ejido (H).

Wildlife surveys

Diurnal vertebrate surveys were conducted using standardized line-transect census (following Peres, 1999). Species were identified by direct sightings and/or acoustic cues, particularly alarm calls (see Table 1). Perpendicular distances of acoustic detections were often measured for terrestrial mammals by finding the tracks left by the animal before fleeing its location at first detection; tracks were also used to corroborate species identity. Observations outside surveys were noted, but were not considered in analysis. Surveys were carried out by one or two observers, typically T. U.-H. accompanied by a local field assistant. Owing to sampling at different study sites and the long duration of the study, we were unable to minimize observer bias by retaining the same field assistant. Nonetheless, all surveys were carried out with the aid of local hunters who were proficient in species identification based on direct or indirect observations, and had been trained previously in line-transect census techniques by T. U.-H. and Alejandro Tuz, a former hunter and field technician from ECOSUR, who participated in almost all surveys. Two or three transects of 3–4.5 km in length were walked once a day at each site (Fig. 1), amounting to a cumulative effort of 93.3–147.4 km per survey site, and a total of 927.2 km walked across all eight sites. Transects within the same survey site were placed at least 500 m apart, with more than 4 km between survey sites. Completely random transect placement was not feasible, however, and transect length depended on the spatial configuration of the landholding area, research permits and other logistical constrains. Given these restrictions, complete spatial independence between some transect segments could not be ensured. However, this is not a problem, as we do not treat different transects within surveys sites as spatially independent and for all but one or two species, the distances between the nearest segments of our transects placed within a site were sufficiently spaced to ensure that we sampled different animals in each population.

Table 1.   Latin and English names of large-bodied mammal (>600 g) and bird species (>200 g) surveyed in this study

Scientific name	Common name	Functional group
Species functional groups are listed according to the degree of arboreality, forest habitat use and prey choice by local hunters. ^a Although Ortalis vetula has been classified previously as a game species, it is a rarely sought after prey in our study area. A, arboreal; S, scansorial (climbing adapted); T, terrestrial; OGS, old-growth specialist species; SGT, second-growth tolerant species; HG, habitat generalist; G, game species; NG, non-game species.
Mammal species
Alouatta pigra	Yucatan black howler	A, OGS, NG
Ateles geoffroyi yucatanensis	Central American spider monkey	A, OGS, NG
Dasyprocta punctata	Central American agouti	T, SGT, G
Eira Barbara	Tayra	S, SGT, NG
Mazama spp.	Red and brown brocket deer	T, SGT, G
Nasua narica	White-nosed coati	S, SGT, G
Odocoileus virginianus	White-tailed deer	T, HG, G
Pecari tajacu	Collared peccary	T, SGT, G
Sciurus spp.	Yucatan and Deppe's squirrels	A,SGT, NG
Tamadua mexicana	Northern tamadua	S, SGT, NG
Tayassu pecari	White-lipped peccary	T, OGS, G
Urocyon cinereoargenteus	Grey fox	T, HG, NG
Bird species
Agriocharis ocellata	Ocellated turkey	SGT, G
Crax rubra	Great curassow	OGS, G
Crypturellus spp.	Thicket, Boucard's and little tinamou	SGT, G
Ortalis vetula	Plain chachalaca	SGT, NG^a
Pteroglossus torquatus	Collared aracari	SGT, NG
Ramphastos sulfuratus	Keel-billed toucan	OGS, NG

Landscape and local variables

Geographic information system-derived variables were processed using ArcView 3.2, supplemented by ESRI extensions. Four broad vegetation types were classified using a supervised image classification on three Landsat (ETM+) images (April 2000) based on training areas (>0.5 ha) selected on the basis of georeferenced control points obtained in situ for the following categories: young secondary growth (4–15 years; n=37), mid-successional forest (16–30 years; n=25); late-successional forest (>30 years; n=33); and non-forest land (crops, pastures and natural open vegetation areas; n=47). Water, clouds and shadows were excluded from calculations. Overall class accuracy was high (Kappa coefficient, κ>0.98 for all classified images). The most frequent misclassifications were young regrowth (>59% class accuracy) and mid-successional forests (>59%), while late-successional forest had a high accuracy (>95%) (see Urquiza-Haas et al., 2009). Percentages of different land cover classes were calculated within 1 km of each site (individual transects buffered and dissolved to form a single polygon); water, clouds and shadows were excluded from calculations (Fig. 1). In addition to the in situ plots selected for habitat classification training areas, at each survey site forest structure and composition were sampled within twelve 0.25 ha plots (10 × 250 m). One of the variables obtained was the aggregate basal area of all trees bearing fleshy fruits (for more details, see Urquiza-Haas et al., 2009).

Human population density (people/km²) within the boundaries of protected areas and ejidos was derived from the INEGI (2002) national population census of 2000. As an additional measure of wild meat demand and potential human impacts from nearby sites, and because landholding boundaries rarely inhibited poachers, the total number of people of the settlements that intersected study site polygons was summed (hereafter neighbouring human population, Table S1). A 3 km buffer around settlements was applied as hunters reported exploiting catchment areas between 2.5 and 9 km of their settlements, but typically within 3–4 km.

Hunting activities were evaluated qualitatively and assigned to one of four broad categories of hunting pressure – low, moderate, high and very high – on the basis of (1) information from key informants on both present and past hunting practices; (2) semi-structured interviews with long-term local residents to obtain information on overall hunting frequency, frequency of group-hunting with dogs, use of traps (cable snares or traps placed at burrows), household meat consumption and prey selectivity; (3) direct evidence of hunting encountered during census work (shotgun shells, traps, hunting trails and waiting stations). However, it was only possible to conduct semi-structured interviews in the four communal tenure landholding units. Despite high levels of past hunting pressure at Las Palmas and El Zapotal, current hunting pressure at these sites was classed as moderate because poaching had been discontinued at least 2 years before the study, and human population densities around these sites had been consistently low (Table S1).

Data analysis

Despite the relatively large census effort, for most species too few sightings (<40) were obtained per site to enable fitting of distance-detection functions and calculation of reliable species-specific population density estimates using distance software (Buckland, 1993). Instead, abundance was compared among sites of varying levels of disturbance, using ER (individuals or groups per 10 km walked). Detection data were pooled for sympatric congeners that could not always be unequivocally identified in the field, and only partially or entirely diurnal species were included in the analysis of species richness and abundance (Table 1). Two additional game bird species potentially occurring in the study region (crested guan and great tinamou) were never encountered.

To investigate whether visibility conditions may vary across sites, mean perpendicular distances were compared among sites for five species, analysed in glim 4.09 assuming a γ error distribution. Perpendicular distance of visual detection events did not differ among sites for three of the five species examined (Sciurus sp.: F_7, 325=1.15, P=0.33; Crypturellus spp.: F_7, 70=1.54, P=0.17; Ramphastos sulfuratus: F_4, 81=1.54, P=0.19), but differed marginally for the remaining two (Nasua narica: F_7, 62=2.04, P=0.06; Ortalis vetula: F_7, 397=2.17, P=0.04;). Although this suggests broadly similar probabilities of detecting both arboreal and terrestrial species despite differences in vegetation structure among sites, detection probabilities might have varied slightly among habitats and sites, in part because animals can adapt behaviourally to stress factors, such as hunting. Shorter detection distances were observed for O. vetula at El Zapotal, which has a higher proportion of more densely vegetated habitats in comparison with X-Conha; thus, any bias towards lower detection in denser growth would make our results conservative. In addition, ER is a robust and conservative metric of abundance of social species (expressed as groups/km walked); group sizes did not differ significantly among sites for all but two of the species for which a minimum number of reliable group counts were obtained (n=7), and group size differences for these species (spider monkey, coati) were only significant between a few sites.

Species accumulation curves were calculated in EstimateS 8.2.0 to standardize samples for the comparison of species richness and diversity (Simpson D′ index) among survey sites. Two-tailed Spearman's rank correlations were computed between ER, species diversity and the disturbance variables considered a priori to be most important (proportion of mid- to late-successional forest cover and hunting pressure).

Finally, ER were compared among other hunted and non-hunted Mesoamerican and South American forest sites, when multiple surveys from different authors and years were available, most recent censuses were given preference and the highest value of abundance (ER) was considered for any given species. For social species (Alouatta spp., Ateles spp. Nasua spp., Pecari tajacu, Tayassu pecari), group ER were considered; when such data were not available values of individual ER were divided by mean group size based on estimates from the same or nearest sites.

Results

Sampling by walked transects was of sufficient intensity at each site to provide robust asymptotic estimates of species richness (Fig. 2). Variation in species richness and diversity among the eight survey sites could be attributed to differences in proportional forest cover (>16 years old) but not hunting pressure (Table 3). Animal ER varied greatly across species and forest sites (Table 2). Squirrels and chachalacas were usually the most frequently encountered species across all sites, whereas howler monkeys, northern tamanduas and tayras were rarely detected, occurring in only three to four sites.

Figure 2.  Species accumulation curves (S_obs Mao Tau). Sites listed in the legend are ordered by a decreasing percentage of forest cover (>16 years old).

Table 3.   Spearman's rank correlation coefficients showing the relationships between key variables and either species-specific or assemblage-wide encounter rates

Scientific name	% Forest cover (>16 years old)	Hunting pressure
Numbers in parentheses indicate significance values. a Encountered only in the Sian Ka'an Reserve (SK). b ^b Preferred game species: brocket and white-tailed deer, white-lipped and collared peccary, ocellated turkey and great curassow. c ^c Standardized species richness (estimated using species accumulation curves) considering only direct encounters of species listed in this table.
Mammal species
Alouatta pigra	−0.10 (0.821)	−0.02 (0.96)
Ateles geoffroyi yucatanensis	0.34 (0.404)	−0.32 (0.442)
Dasyprocta punctata	−0.02 (0.955)	−0.47 (0.241)
Eira barbara	0.60 (0.119)	−0.18 (0.674)
Mazama spp.	−0.30 (0.471)	−0.47 (0.237)
Nasua narica	0.76 (0.028)	−0.36 (0.384)
Odocoileus virginianus	−0.57 (0.136)	−0.36 (0.38)
Pecari tajacu	−0.51 (0.194)	−0.34 (0.407)
Sciurus spp.	0.12 (0.779)	0.59 (0.121)
Tamadua mexicana	0.40 (0.328)	−0.10 (0.814)
Tayassu pecari^a	−0.08 (0.846)	−0.51 (0.193)
Urocyon cinereoargenteus	0.56 (0.151)	0.53 (0.174)
Bird species
Agriocharis ocellata	0.56 (0.146)	−0.48 (0.23)
Crax rubra	0.82 (0.012)	−0.48 (0.236)
Crypturellus spp.	0.38 (0.352)	−0.25 (0.555)
Ortalis vetula	−0.62 (0.102)	0.26 (0.535)
Pteroglossus torquatus	0.59 (0.122)	0.45 (0.26)
Ramphastos sulfuratus	0.71 (0.047)	0.10 (0.816)
Groups of species
Game species (n=9)	0.55 (0.16)	−0.72 (0.046)
Preferred game species (n=6)^b	0.02 (0.955)	−0.87 (0.006)
Non-game species (n=9)	−0.02 (0.955)	0.25 (0.555)
Old-growth specialist (n=5)	0.91 (0.002)	−0.32 (0.438)
Second growth tolerant species (n=11)	−0.41 (0.32)	0.30 (0.476)
All species (n=18)	0.31 (0.456)	−0.24 (0.576)
Species richness (n=18)	0.88 (0.004)	−0.09 (0.839)
Species diversity D′ (Simpson)^c	0.69 (0.058)	−0.03 (0.954)

Table 2.   Encounter rates of mammal and bird species at eight study sites in the Yucatan Peninsula, Mexico

Scientific name	Encounters per 10 km of transect walked
	Za	YL	TiNe	XC	Pa	SK	SKU	TeNu
Survey sites are coded according to Fig. 1. Encounter rates based on acoustic cues are shown in parentheses. a Group encounter rates. b ^b Encountered outside census walks.
Mammal species
Arboreal
Alouatta pigra^a	0.20	0.00	0.00	0.09	0.00	0.00	0.09	0.00
Ateles geoffroyi^a	0.27	2.43 (0.18)	0.00	0.09	0.09	0.00	0.45 (0.09)	0.00
Sciurus spp.	3.93 (0.07)	5.32 (0.09)	4.26 (0.16)	5.70	3.51	1.48	3.00 (0.18)	1.72
Scansorial
Eira Barbara	0.00^b	0.18	0.00	0.09	0.26	0.17	0.00^b	0.00
Nasua narica^a	0.34	0.81	0.74	0.70 (0.09)	0.79 (0.09)	0.35	2.73 (0.27)	0.21
Tamandua mexicana	0.00^b	0.00	0.00	0.18	0.00	0.00	0.09	0.00
Terrestrial
Dasyprocta punctata	0.27 (0.14)	0.81 (0.27)	1.47 (0.65)	0.09	0.26 (0.09)	0.35	0.91 (0.64)	0.11
Mazama spp.	0.88 (0.54)	0.09 (0.09)	0.08 (0.08)	0.00	0.18	0.17	0.64 (0.18)	0.64
Odocoileus virginianus	0.34 (0.34)	1.08 (0.90)	0.25 (0.08)	0.09 (0.09)	0.09 (0.09)	0.52 (0.09)	0.18 (0.09)	0.32
Pecari tajacu^a	0.47 (0.20)	0.09 (0.09)	0.08	0.09	0.09	0.35	0.27	0.43
Tayassu pecari^a	0.00	0.00	0.00	0.00	0.00	0.09	0.00	0.00
Urocyon cinereoargenteus	0.00	0.09	0.08	0.53	0.26	0.00	0.00^b	0.00
Bird species
Agriocharis ocellata^a	0.00^b	0.54	0.08	0.35	0.18	0.43	0.36	0.00
Crax rubra^a	0.00	0.00	0.00	0.18	0.44	0.26	0.36	0.00
Crypturellus spp.	0.34	1.89	0.57	0.35	1.67	1.13	1.00	1.39
Ortalis vetula^a	4.14 (0.47)	1.53 (0.27)	9.98 (0.16)	2.19 (0.18)	2.02 (0.26)	9.57 (1.48)	1.90 (0.18)	7.08 (0.54)
Pteroglossus torquatus^a	0.00^b	0.00	0.00	1.67 (0.26)	0.18	0.00	0.00	0.00
Ramphastos sulfuratus^a	0.07	0.00	1.64	2.02 (0.44)	2.28 (0.18)	0.35 (0.17)	4.09 (0.91)	0.11

Percentage of mid- to late-successional forest cover was positively correlated with species richness (Fig. 2) and with ER of three species (coati, great curassow, keel-billed toucan), whereas positive trends towards significance were found for ocellated turkey, tayra and collared aracari, and negative trends for white-tailed deer and plain chachalaca (Table 3). Not surprisingly, aggregate ER of old-growth specialists increased with forest cover (Fig. 3), but this was not the case for species that tolerated young second-growth. Reduced forest cover translated into reduced resource availability, at least for frugivores, as there was a strong positive relationship between the mean basal area of trees bearing fleshy fruits and the percentage of forest cover (r_s=0.98, P<0.001). Although forest cover and hunting pressure were largely independent (r_s=−0.22, P=0.6), the percentage of cultivated land was positively correlated with hunting pressure (r_s=0.80, P=0.02).

Figure 3.  Relationship between (a) percentage of mid- to late-successional forest cover and encounter rates of old-growth specialist species (n=5); (b) second-growth tolerant species (n=11); (c) degree of hunting pressure ranked across all survey sites and encounter rates of game species (n=9); (d) non-game species (n=9).

Level of hunting pressure and ER were not significantly correlated for any single species. Nonetheless, pooled ER of game species were significantly higher in less hunted sites, and this negative correlation was even stronger for preferred game species (Table 3; Fig. 3). The strongest negative responses to hunting included some of the most preferred game species across the Yucatan Peninsula (agouti, brocket deer, ocellated turkey, great curassow, white-lipped peccary) (Table 3). Neighbouring human population was not significantly correlated with ER of all game species (r_s=−0.49, P=0.22), but it was strongly inversely correlated with ER of preferred game species (r_s=−0.87, P=0.005).

Discussion

Patterns of observed vertebrate ER were influenced by the amount of remaining forest cover and the degree of hunting pressure. Reduced species richness and low aggregate ER of old-growth specialists were associated with surrounding matrix areas containing more open habitats, regardless of whether low forest cover was a result of anthropogenic deforestation (Tierra Negra, Tezoco Nuevo) or underlying environmental factors (Sian Ka'an, El Zapotal) (Fig. 1, Table S1). These results are in stark contrast with trends observed in Amazonia, where some arboreal species were actually more common in more disturbed forest (Lopes & Ferrari, 2000; Michalski & Peres, 2007). Discrepancies may result from dissimilarities in the configuration of forest habitat, and a ‘crowding effect’ temporarily sustaining higher population densities of strictly arboreal species in small forest fragments due to their unwillingness to cross an inhospitable matrix (Estrada et al., 2002). However, like virtually all large vertebrate studies anywhere, and in the tropics in particular, our study is observational and may therefore fail to truly disentangle cause and effect (Romesburg, 1981). On the other hand, visual detection probability might have varied among species and habitats making our results conservative, although our data showed that ER were not likely to have been significantly influenced by intrinsic biases in detection probability.

The broad vertebrate diversity and abundance patterns uncovered here are consistent with observations elsewhere in the Neotropics. Species diversity was lower in more disturbed sites, where small-bodied or more adaptable species were more abundant, while others were negatively affected (c.f. Chiarello, 1999, 2000; Daily et al., 2003; Naughton-Treves et al., 2003). ER of great curassow and keel-billed toucan increased with landscape-scale forest cover, which is expected given that the distribution of these species is related to overall forest cover, older forest stands and higher abundance of trees bearing fleshy fruits (Martínez-Morales, 1999; Graham, 2001; Urquiza-Haas et al., 2009). Coatis, tayras and ocellated turkeys were also sensitive to reduced forest cover. Although these species can frequently forage in open habitats, they require forest for shelter, nesting sites or provision of food resources (Reid, 1997; Gonzalez, Quigley & Taylor, 1998). Conversely, second-growth tolerant species, such as brocket deer, collared peccary and agouti, were resilient to reduced forest cover. Previous studies have shown that these species can be very tolerant to habitat modification (Naughton-Treves et al., 2003), even in more heavily disturbed landscapes (Daily et al., 2003) than those examined here. Yet, these species can be driven to local extinction in small forest fragments, most likely through the combination of habitat area effects and overhunting (Chiarello, 1999; Peres, 2001). Furthermore, red brocket deer Mazama americana in southern Mexico has a strong preference for old-growth forest and avoids other vegetation types including second-growth (Weber, 2008); thus, this species may be more vulnerable to habitat modification in the Yucatan Peninsula than thought previously.

Levels of hunting pressure were strongly related to the aggregate ER of game species, particularly of preferred game. This is consistent with many studies in neotropical forests elsewhere documenting the reduced abundance of game species in hunted areas (e.g. Carrillo, Wong & Cuarón, 2000; Peres & Palacios, 2007; Parry, Barlow & Peres, 2009). Although correlations were not significant when species were tested individually, this is likely because of the small number of sites such that even substantial values of the correlation coefficient were deemed non-significant. All six of the mammal game species, and two of the three game bird species, had moderate to strong negative association between species-specific ER and hunting pressure. A straight-forward meta-analysis of our data (assuming a fixed effect, Field, 2001) provides an estimate of the mean correlation between ER and hunting pressure for the nine game species of r_s=−0.42 (significance based on Z score of effect: P=0.001), compared with r_s=0.16 (P=0.27) for the nine non-game species, providing strong support for the hypothesis that local densities of each individual game species are reduced in sites exposed to greater hunting pressure. Target species that were strongly related to our measures of hunting pressure included brocket deer, agouti, white-lipped peccary, ocellated turkey and great curassow (all with r_s≤−0.47; Table 3). These species are often preferred by hunters throughout the Neotropics (Robinson & Redford, 1991). Preference by hunters is often determined by prey body size and taste of game meat, but may be strongly influenced by cultural factors (Escamilla et al., 2000), whereas actual offtakes are more closely related to the reproductive productivity of game species and local depletion of game stocks (Bodmer, 1995; Jerozolimski & Peres, 2003). In the southern portion of the Yucatan Peninsula, paca, agouti, brocket deer, great curassow, collared peccary and armadillo comprised the top-ranking hunted species accounting for the bulk of local offtakes (Escamilla et al., 2000; Weber, 2000). In contrast, coati, plain chachalaca, paca, agouti and collared peccary are the most frequently hunted species in central Quintana Roo (Jorgenson, 1993). Nonetheless, nominal payments at the end of Jorgenson's study may have influenced the delivery of a large number of carcasses of coatis and chachalacas (Escamilla et al., 2000), which are generally considered a nuisance and are frequently killed to protect crops.

Most harvest-sensitive game species, as inferred from between-site differences in ER, were those both favoured by hunters and associated with low intrinsic rates of natural increase (r_max) [Great curassow, r_max=0.28; brocket deer, r_max=0.40 (Robinson & Redford, 1986; Pereira, Daily & Roughgarden, 2004)] or low survival rates (ocellated turkey) (Gonzalez et al., 1998). However, negative abundance responses that were likely related to hunting intensity were also detected for small-bodied species with a high rate of natural increase (Agouti, r_max=1.1). Despite its high reproductive potential, the behaviour of this species can help to explain its sensitivity to hunting as agouties are reluctant to leave their home ranges, and so can be run to ground by dogs and killed with machetes (Reid, 1997).

On the other hand, it is difficult to establish whether low ER were a direct consequence of harvest. Animals can adapt behaviourally, increasing nocturnal activity where they are heavily hunted (e.g. collared peccaries, coatis and white-tailed deer, Reid, 1997) or selecting habitats less frequently visited by humans (Reyna-Hurtado & Tanner, 2005). For the species examined here, it remains to be seen whether selection of habitat refugia is suboptimal, thereby affecting fecundity. In sum, species sensitivity to hunting may be affected by the complex relationships between species reproductive potential, habitat-carrying capacity and species behavioural responses.

It is not possible to entirely disentangle effects of single human disturbance factors, as they are often correlated (Lopes & Ferrari, 2000; Naughton-Treves et al., 2003). In this study, hunting pressure and forest cover were largely independent, although hunting was linked to a greater extent of cultivated land. Synergistic interactions between forest fragmentation and hunting augment large vertebrate extinction risk in tropical forests (Peres, 2001), yet the swidden successional mosaic has been identified as an important macrohabitat for several herbivore species, providing supplemental resources in the form of crops and browse (Medellín & Equihua, 1998; Naughton-Treves et al., 2003). However, because many farmers practiced garden hunting to both control levels of crop raiding and obtain additional protein sources (Smith, 2005), agricultural areas may function as ecological traps affecting the overall population dynamics of species that are tolerant of second growth (Battin, 2004).

Despite several reserve administration problems (Smardon & Faust, 2006), Sian Ka'an undoubtedly plays an extremely important role in wildlife conservation in the Yucatan Peninsula. Threatened species like the tapir and white-lipped peccary were only detected in this biosphere reserve, and for many species the highest ER were observed in old-growth semi-evergreen medium-statured forest within the reserve. However, this forest type is underrepresented in the reserve (≈25%) (Challenger, 1998), and our results show that areas dominated by low-statured forest and open areas supported lower abundances of old-growth and second-tolerant species (Fig. 3). In particular, the Sian Ka'an Reserve alone may not be sufficiently large to protect highly arboreal species, such as howler and spider monkeys, which is a major concern given their threatened status, susceptibility to human disturbance (Daily et al., 2003; Urquiza-Haas et al., 2009) and their ecological role in forest regeneration (Julliot, 1997).

In comparison with other neotropical forest sites, Jorgenson (1993, 2000) showed that densities of some mammal and bird species in Quintana Roo were extremely low. A long history of human occupation, agricultural activities and wildlife use has been held responsible for the low levels of animal abundance in this region (Robinson & Bennett, 2004). However, differences in wildlife population density can hardly be attributed to single factors, as many environmental (e.g. soil fertility, climate) and anthropogenic variables may be interacting (Emmons, 1984; Lopes & Ferrari, 2000). A wider comparison with numerous neotropical forest sites where large vertebrates have been censused to date using line-transect techniques reveals that ER for most species in the Yucatan Peninsula are in fact within the range reported for other hunted and non-hunted Mesoamerican (Fig. 4) and South American forest sites (Bodmer, Eisenberg & Redford, 1997; Lopes & Ferrari, 2000; Gómez, Wallace & Veitch, 2001; Aquino & Calle, 2003; Peres & Palacios, 2007). ER of game species (except Odocoileus sp.) in our survey sites compare more closely to sites that report high or severe hunting pressure elsewhere (Fig. 4). Further, these comparative data show that (1) agouti, brocket deer, white-lipped and collared peccaries were considerably more abundant in sites with low hunting pressure; (2) highly vulnerable group-living species, such as white-lipped peccary and spider monkey, are often driven to local extinction in both mainland and island sites that have been persistently overhunted. On the other hand, there are striking differences among sites for some species; agouti ER in this study were very low compared most sites, in particular those in Barro Colorado and the Gigante Peninsula in Panama, and several continuous forest sites in Brazilian Amazonia (Fig. 4, Dalecky et al., 2002; Peres, Barlow & Haugaasen, 2003; Barlow & Peres, 2006). Ateline primate ER also showed remarkable differences in relation to most Panamanian and South American sites surveyed (Fig. 4, Wallace, Painter & Taber, 1998; Dalecky et al., 2002; Barlow & Peres, 2006). Gonzalez-Kirchner (1998) showed that howler monkey densities in Muchukux, Quintana Roo, were comparable or even higher than those reported for Alouatta pigra elsewhere. Nonetheless, howlers in this study were rarely encountered, even in continuous old-growth forest sites. High ER were only obtained for spider monkeys in Yodzonot Laguna, where key food trees such as Brosimum and Manilkara are abundant (Table 2). This is consistent with the regional-scale determinants of patch occupancy for these arboreal frugivores across 147 sites surveyed throughout the Yucatan Peninsula (Urquiza-Haas et al., 2009).

Figure 4.  Patterns of abundance of 11 mammal species across non-hunted to lightly hunted (sites 1–11), moderately hunted (12–15) and heavily hunted (16–27) Mesoamerican forest sites surveyed to date using line-transect census methods. Abundance is expressed in terms of encounter rates of largely solitary (individuals/km walked) or group-living species (groups/km walked). Sites are ordered left to right in terms of their hunting pressure (HP) category. Grey bars indicate the eight forest sites surveyed in this study. Site numbers correspond to: (1) Barro Colorado Island (Glanz, 1991; Wright et al., 1999; Wright et al., 2000); (2) Peña Blanca; (3) Gigante (Glanz, 1991); (4) Rio Macho (Wright et al., 2000); (5) Sian Ka'an Uayamil, (6) Sian Ka'an (this study); (7) Muchukux (Gonzalez-Kirchner, 1998); (8) Selva Lacandona (Cuarón, 2001); (9) Najil Tucha, (10) Muchukux (Gonzalez-Kirchner, 1999); (11) Calakmul Biosphere Reserve (Weber, 2000); (12) Yodzonot Laguna; (13) Zapotal; (14) Las Palmas (this study); (15) Nuevo Becal (Weber, 2000); (16) Paraiso, (17) Limite, (18) Rio Mandinga, (19) Plantation Road (Wright et al., 2000); (20) X-Conha (this study); (21) Sendero Las Cruces (Wright et al., 2000); (22) Tierra Negra (this study); (23) Carretera C25 (Wright et al., 2000); (24) Pipeline Road (Glanz, 1991); (25) X-Hazil (Jorgenson, 1993); (26) Tezoco Nuevo (this study); (27) El Refugio and La Mancolona Villages (Weber, 2000).

There has been an increasing interest in the conservation effectiveness of partially protected areas based on community-based conservation schemes (e.g. Noss & Cuellar, 2001). Although the conservation role of communally managed ejidos in the region has been widely discussed (Bray et al., 2004; Dalle et al., 2006), evaluation of success has mainly focused on the persistence of forest cover, while wildlife has been largely ignored. High hunting pressure, such as documented in this study for X-Conha, Tezoco Nuevo and Tierra Negra, may hinder effective wildlife conservation in the region. While more ecological and social research are needed to evaluate the full potential of communal landholdings for both wildlife and forestland conservation, results presented here caution against focusing conservation actions solely on halting tropical deforestation, especially given the key role of wildlife in ecosystem function and forest regeneration (Stoner et al., 2007).

Sustainable wildlife use and conservation of large tracts of forest will undoubtedly be one of the greatest challenges faced by local communities, researchers and organizations working towards sustainability in the region, in particular given currently the high development pressure and burgeoning human population growth driven mainly by the mass tourism and real estate industry. Human population has grown in the state of Quintana Roo from 88 150 in 1970 to 1 135 309 in 2005, and the GDP of the construction sector rose from about US$384 million pesos in 1993 to more than US$3 billion in 2004 (Manuel-Navarrete, Pelling & Redclift, 2009; for a detailed discussion of conservation challenges in the region, see Urquiza-Haas et al., 2009). Furthermore, traditional values are changing, unleashing uncontrolled overharvesting (Anderson, 2001), which will profoundly affect populations of many species, even if forest cover is maintained. On the other hand, ER for both game and non-game species were relatively high at Yodzonot Laguna (Fig. 3), which leaves room for optimism. This supports the notion that areas set aside and managed by traditional Mayan communities – with or without the aid of a local conservation agency – can accrue substantial conservation dividends [Anderson, 2001; Ramos-Fernández et al., 2005; but see García-Frapolli et al. (2009) on emergent difficulties when ejidos are decreed as official protected areas]. Currently, over 100 000 hectares have been identified as voluntary communal reserves in the central portion of Quintana Roo state (Elizondo & López-Merlín, 2009). However, it is critical that conservation-oriented landholding units do not become isolated from the surrounding landscape. To ensure the continued persistence of most biodiversity in remaining Mesoamerican forests, it is critical that private, communal and public sector reserves are integrated into a coordinated management approach across the entire landscape.

Acknowledgements

This work was supported by a grant from the Wildlife Conservation Society and a conacyt (Consejo Nacional de Ciencia y Tecnología) fellowship to T.U.-H. We are grateful to Alejandro Tuz, Arsenio Xool and Cosme Caamal for assistance in the field. T.U.-H is indebted to all Mayan communities from the study area for their help and hospitality and wants to thank Juan Carlos Faller, Cecilia Elizondo, David López and her family for their continuing support. We are grateful to the editor and two anonymous referees for helpful comments.