The demography of a resource specialist in the tropics: Cecropia trees and the fitness of three-toed sloths

Abstract

Resource specialists persist in a narrow range of resources. Consequently, the abundance of key resources should drive vital rates, individual fitness, and population viability. While Neotropical forests feature both high levels of biodiversity and numbers of specialist species, no studies have directly evaluated how the variation of key resources affects the fitness of a tropical specialist. Here, we quantified the effect of key tree species density and forest cover on the fitness of three-toed sloths (Bradypus variegatus), an arboreal folivore strongly associated with Cecropia trees in Costa Rica, using a multi-year demographic, genetic, and space-use dataset. We found that the density of Cecropia trees was strongly and positively related to both adult survival and reproductive output. A matrix model parametrized with Cecropia–demography relationships suggested positive growth of sloth populations, even at low densities of Cecropia (0.7 trees ha⁻¹). Our study shows the first direct link between the density of a key resource to demographic consequences of a tropical specialist, underscoring the sensitivity of tropical specialists to the loss of a single key resource, but also point to targeted conservation measures to increase that resource. Finally, our study reveals that previously disturbed and regenerating environments can support viable populations of tropical specialists.

1. Background

Specialist species are those that exhibit a limited niche breadth, ultimately resulting from trade-offs between the ability to exploit a range of resources versus efficiency of using individual ones [1,2]. Consequently, strict specialists that rely upon only a few key resources typically exist within a narrow range of environmental conditions [3] and are more likely to occur in stable and homogeneous environments [1]. Variation in the availability of key resources and environmental perturbations can disproportionately affect the individual fitness of specialists [2]. Unsurprisingly, then, across taxonomic groups, specialists are more susceptible to environmental change [4] and suffer greater rates of decline globally relative to generalists [5].

Owing to the dependence of specialist species on critical resources, multiple investigations in temperate latitudes have evaluated how key resource variation affects the presence and distribution [6,7], density [8,9], dispersion [10], reproductive success [11], and mating systems [12] of specialist vertebrates. More importantly, some studies have evaluated how conservation actions can be designed to promote specialist species in the face of environmental change [13–15]. Despite the large number of specialist species in tropical areas compared with temperate ones [16], studies evaluating the variation of key resources and consequent effects on specialists are scarce; those few studies have focused on the broad patterns in distribution and abundance of specialist species related to resources [17,18]. To our knowledge, no one has directly evaluated how the variation of these resources across a landscape affects the vital rates of a tropical specialist species.

Mammalian arboreal folivores are a guild of resource specialists that have evolved a suite of specialized features [19,20] including (i) low and restricted range of body mass (1–14 kg) [21], (ii) strict dietary preferences [22], and pregastric organs to effectively digest chemically and structurally protected leaves [21], (iii) modes of locomotion that minimize transportation costs [23], and (iv) behaviours that minimize thermoregulatory costs [24]. These adaptations are thought to have evolved because arboreal folivores are highly constrained by their lifestyle of consuming an abundant, yet low-quality, food resource. Essentially, these species must be light enough to be supported in the canopy but sufficiently large to process plant matter, which is rich in fibre and low in digestible nutrients [21,22]. Possibly as a consequence of this strict specialization [24], mammalian arboreal folivores are rare with less than 5% of all mammal species occurring within this guild [19,24]. This group is also disproportionately threatened (figure 1) because of their dependence on forest habitats, specific food resources [25,26], and an inability to switch resources when habitats are degraded or lost [27].

Figure 1.

Percentage of mammal species threatened (i.e. listed as vulnerable, endangered, and critically endangered by the International Union for Conservation of Nature (IUCN) red list [2017]) mammalian arboreal folivores (white) compared to all other mammalian species (black), by order. Not included: order Carnivora given the low representation of mammalian arboreal folivores in this group, as well as the orders Dermoptera and Hyracoidea due to the lack of threatened species in these groups (see electronic supplementary material, methods for details).

Tree sloths are arboreal folivores that lie at the extreme end of specialization [21,24]. Two phylogenetic groups of tree sloths, two- (Choloepus spp.) and three-toed (Bradypus spp.), occur in the Neotropics. While both are mid-sized foregut-fermenting arboreal folivores [20,21], two-toed sloths possess a more diverse diet of animal matter, fruits, and leaves, compared to three-toed sloths, which are strict folivores [28]. Moreover, individual three-toed sloths are specialists, roosting and consuming leaves from only a few tree species within the forest [28]. Several authors have proposed that three-toed sloths exhibit a strong relationship with Cecropia trees [29,30]—trees that exhibit a high ratio of protein content and a low fibre and chemical defences content [31,32]—and for many years, it was believed that three-toed sloths only consumed Cecropia leaves [33,34]. Indeed, the distribution of Bradypus spp. [35] is nested within the known distribution of Cecropia [36] (figure 2), supporting the idea of the tight association of three-toed sloths with Cecropia. Other studies, however, have found that three-toed sloths consume leaves from several plant species and even inhabit areas where Cecropia trees are relatively scarce [28,37]. In these regions lacking Cecropia trees, three-toed sloths can specialize on other species (e.g. Dipteryx panamensis [28]) or, in the case of the maned sloth (Bradypus torquatus), consume several trees from the families Sapotaceae, Apocynaceae, Moraceae, and Rosaceae [37]. Nevertheless, Cecropia trees appear to be a central resource for three-toed sloths across a broad geographical range, although its direct role on tree sloth demography has not been quantified.

Figure 2.

Distributional range of three-toed sloths (Bradypus spp.; grey polygon), and trees of the genus Cecropia (green hatched) across the Neotropics (see electronic supplementary material, methods for details).

At our study site in northeastern Costa Rica, three-toed sloths (Bradypus variegatus) select for tropical forests [38] and disperse to less suitable open habitats using narrow riparian forest [39]. Furthermore, three-toed sloths in this region disproportionately use only two species of trees: Cecropia obtusifolia and Coussapoa villosa [38]. Taking advantage of a multi-year demographic, genetic, and space-use dataset from three-toed sloths, we aimed to clarify the role of the resources on which this Neotropical specialist depends. Specifically, we evaluated how the abundance of C. obtusifolia and C. villosa affect the reproductive success, survival, and the overall fitness of three-toed sloths. We also tested if the amount of regenerating forest affected vital rates of three-toed sloths. We hypothesized that the abundance of key resources, especially C. obtusifolia, would drive vital rates, and ultimately, the fitness of three-toed sloths at our study site. Specifically, we predicted that individuals that possess higher densities of Cecropia within their core areas would exhibit higher reproductive output and survival than individuals with a lower amount of resources. More broadly, we explored the demographic sensitivity of tropical specialists to variation in their key resources and aimed to identify targeted conservation actions to promote these species persistence in human-modified landscapes.

2. Material and methods

(a). Study area

Fieldwork was conducted from March 2010 to December 2016 in an agro-ecosystem located in the Caribbean coastal plain of northeastern Costa Rica (10.32° N, −83.59° W). The climate is wet and warm, with two dry seasons: March to April and September to October [40]. The study area is a 4 km² agro-ecosystem consisting of a privately owned shade-grown cacao (Theobroma cacao) farm with sparse overstory of taller trees; late-stage regenerating tropical forests in patches and along narrow riparian forests (approx. 20 m); and cattle pastures containing living fence rows, isolated legacy trees, and planted non-native trees (see electronic supplementary material, figure S1). The cacao farm was bordered on two sides by monocultures of pineapple and banana grown without an overstory of trees—habitats not used by sloths [38].

(b). Captures and radio-tracking

We captured 54 three-toed sloths, 19 males, 21 females, and 14 juveniles in the shade-grown cacao plantation, surrounding cattle pastures and riparian forests. We determined the sex of each adult sloth from their external genitalia and classified individuals as juveniles or adult. A sloth was considered to be a juvenile when it was less than 3.7 kg [41], except in cases where exact age was determined in recently born individuals, where juveniles were considered to be less than 3 years old [42]. Individuals were captured by hand from trees, marked with uniquely coded passive integrated transponder (PIT) tags (Biomark, Boise, ID, USA) inserted subcutaneously between the shoulder blades, and fitted with VHF radio-collars (Mod-210, Telonics Inc., Mesa, AZ, USA). We obtained telemetry-locations for sloths five to six times per month.

(c). Estimating resource abundance

To estimate the density of C. obtusifolia within areas used by sloths, we first calculated core home range areas for 19 male and 21 female three-toed sloths over 51 months (range = 30–400 locations per individual) using 50% fixed-kernel density estimates [43]. We then mapped all C. obtusifolia trees with a diameter at breast height (DBH) > 20 cm, using visual surveys over each sloth core area. Trees with smaller DBH were assumed to be unimportant because they are not frequently used by three-toed sloths [44]. To test if other tree species influence the vital rates of three-toed sloths, we also estimated the density of C. villosa following the same methods. Finally, we tested whether the abundance of regenerating forest habitat is associated with the vital rate of three-toed sloths by calculating the abundance of the forest as the proportion of this habitat in the core area of each individual sloth. We defined habitats using 2013 RapidEye satellite imagery (IntraSearch Inc.) with 5 m resolution and five-band multispectral imagery to visually digitize polygons representing these habitat types (ArcGIS10.1; ESRI, Redlands, CA, USA) and estimated the proportion of forest within each sloth core area [40].

(d). Estimating reproductive output

Male reproductive output was assessed based on genetic paternity assignments because males do not care for offspring. We used 13 previously developed microsatellite markers to genotype all captured three-toed sloths at our study site [45]. We extracted sloth DNA, conducted PCR reactions, and genotyped individuals according to previously developed procedures [41,46]. We included all genotyped adult males and juveniles with weights between 3 and 3.7 kg as potential fathers to account for the possibility that some juveniles could have been reproductively active. Of the 36 males we genotyped, 19 had sufficient data to calculate their core areas, while none of the remaining 17 males exhibited reproductive success. We used standard likelihood approaches implemented in the program Cervus v3.0.7 [47] to assign paternity for 33 juveniles that were sampled with what we assumed were their biological mothers (juvenile was held by or located immediately adjacent to a female); maternity was confirmed by checking that the mother–offspring dyads shared at least 1 allele at all loci. We assumed a genotyping error rate of 0.01 and set the proportion of sampled candidate fathers at 85% based on the fact that almost all males encountered in our study area had previously been captured and sampled [46]. We used logarithm of odds ratio (LOD) scores, the difference in the likelihood of the two most likely candidate fathers, to assign paternity and to assess confidence in the assignment [41]. We used both a relaxed confidence level of 80% and a strict confidence level of 95% to assess paternity, and in cases where confidence was only 80%, we only assigned paternity if there was spatial congruence between the mother and putative father.

We estimated the reproductive output of adult females using visual observations of radio- and colour-marked of 21 individuals [48]. We considered an adult female to have successfully reproduced if she was observed with a recently born juvenile on a single occasion from January to June, the breeding season activity for female three-toed sloths at our site [48]. Similarly, we considered an adult female to not have reproduced if it was observed without a juvenile on at least five occasions from January to June [48]. We used separate multiple linear regression analyses to test whether the reproductive output of males and females was related to the density of C. obtusifolia, C. villosa, and proportion of forest in the core area of sloths.

(e). Estimating survival

We estimated monthly survival of radio-tagged adult and juvenile sloths using known-fate models [49] in the program MARK [50]. For adults, we tested whether survival was constant (model 1) or varied by month (model 2), sex (model 3), density of C. obtusifolia (model 4), density of C. villosa (model 5), and per cent of regenerating forest in the core area (model 6). Then, we explored if there were additive effects among these variables (models 7–9) (table 2). We placed juvenile sloths into one of two stage classes based on their ages, juveniles 0–1 years old and juveniles 1–2 years old. We excluded juveniles 2–3 years old because they were frequently observed dispersing and no longer inhabited their natal home range, precluding the estimation of a core area. We first evaluated whether survival was constant (model 1), time varying (model 2), or differed by stage (model 3). Then, we maintained stage because model 3 had the greatest support, and we explored the effect of C. obtusifolia, C. villosa density, and per cent of forest in the core area (models 4–6). Finally, we tested if there were additive effects among these variables (models 7–9). Models were ranked based on Akaike Information Criterion (AICc) [51]. We used the analysis of deviance (ANODEV) in the program MARK [50] to determine the proportion of the total variation in survival explained by the top-ranked model [52]. To calculate the survival of juveniles 2–3 years old we tested whether survival was constant (model 1) or varied by month (model 2) (see electronic supplementary material, table S1). We also tested if individuals that survived or died differed in the density of C. obtusifolia in their core area using a t-test for unequal variances.

Table 2.

Rankings of known-fate models estimating survival (ϕ) for adult three-toed sloths in northeastern Costa Rica. AICc: Akaike information criterion corrected for small sample sizes, K: number of parameters, w: AICc weight. Models are ranked based on AICc. The order in which the models were computed is shown in bold numbers.

model	K	AICc	Δ AICc	w
4 S (Cecropia)	2	67.974	0	0.363
7 S (Cecropia) + (Coussapoa)	3	69.426	1.452	0.176
8 S (Cecropia) + (forest)	3	69.895	1.920	0.139
1 S (.)	1	69.917	1.943	0.137
3 S (Sex)	2	71.546	3.572	0.061
6 S (forest)	2	71.78	3.806	0.054
5 S (Coussapoa)	2	71.949	3.974	0.049
9 S (forest) + (Coussapoa)	3	73.777	5.802	0.019
2 S (t)	49	146.257	78.283	0

(f). Estimating factors influencing fitness

We used finite rate of population change (λ) as an estimate of fitness because it implicitly incorporates age-specific survival and fecundity [53,54] and can be interpreted as the mean fitness across individuals within a year [54,55]. To calculate λ, we constructed a female stage-based demographic matrix model using vital rate estimates described above. We used a four-stage-class birth-pulse model [48] matrix model: juveniles 0–1 years old, juveniles 1–2 years old, juveniles 2–3 years old, and adults (greater than 3 years), where only adults were able to produce offspring, considering the age of first breeding as 3 years [42] (table 1). We assumed a birth-pulse model because most juveniles are produced from February to April [48]. We estimated λ from the dominant right eigenvalue of the deterministic matrix model, using the best models of survival (ϕ) and fecundity (m), the fecundity was divided by two assuming a ratio 1 : 1 sex. We substituted the best models of survival (ϕ) and fecundity (m) for the parameters in the demographic matrix (table 1). To simulate the effect of the variation of resource abundance on λ, we varied the abundance of resources in the models that explain the vital rates (table 1), within the ranges exhibited in Neotropical areas.

Table 1.

Matrix model used to estimate annual population growth rates for three-toed sloths in northeastern Costa Rica. m: reproductive output; ϕ_0–1: annual juvenile 0- to 1-year-old survival rate; ϕ_1–2: annual juvenile 1- to 2-year-old survival rate; ϕ_2–3: annual juvenile 2- to 3-year-old survival rate.

0	0	m + ϕ2–3	m + ϕ2–3
ϕ0–1	0	0	0
0	ϕ1–2	0	0
0	0	ϕ2–3	ϕ2–3

3. Results

(a). Resource abundance

The density of C. obtusifolia was 0.96 trees ha⁻¹ in our study area, but this varied by habitat, with C. obtusifolia reaching its highest density in forests (2.41 trees ha⁻¹), followed by cacao (0.69 trees ha⁻¹) and pastures (0.31 trees ha⁻¹). Of the 40 adult sloths, 36 had C. obtusifolia trees in their core area (electronic supplementary material, figure S1). Mean density of C. obtusifolia in male core areas was 4.82 trees ha⁻¹ (n = 19, range = 0–25.59 trees ha⁻¹) and 5.88 trees ha⁻¹ for females (n = 21, range = 0–21.04 trees ha⁻¹). C. villosa was relatively scarce (21 trees), exhibiting a site-level density of 0.11 trees ha⁻¹. C. villosa was primarily found in the forests (18 of 21 trees, 0.54 trees ha⁻¹) and was very low in the other two habitats (pastures = 0.09 trees ha⁻¹, cacao = 0.01 trees ha⁻¹). Mean density of C. villosa in the core areas of males was 0.69 trees ha⁻¹ (n = 19, range = 0.00–13.12 trees ha⁻¹), and 1.68 trees ha⁻¹ in female core areas (n = 21, range = 0.00–13.82 trees ha⁻¹). Core areas of males contained a mean of 49% of forest (n = 19, range = 0–100%), and females exhibited a mean of 45% of forest (n = 21, range = 0–99%).

(b). Reproductive output

We assigned paternity to 29 juveniles with 95% confidence and the remaining 4 juveniles with 80% confidence. All mothers and genetically assigned fathers shared at least one allele at all loci with their putative offspring. Our parentage assignments were further corroborated by the fact that all genetically assigned mothers were sampled holding their offspring and genetically assigned fathers occurred in close proximity to the assumed mother, having spatial contact at some point during the breeding season.

We did not detect significant relationships between the reproductive output of either males or females and the density of C. villosa or the proportion of regenerating forested habitat in the core area (all ps > 0.1). We found a positive and strong linear relationship between the density of C. obtusifolia and reproductive output for both males (F_1,13 = 6.25, p < 0.01; r² = 0.89; figure 3a), and females (F_1,17 = 7.31, p < 0.01; r² = 0.41, figure 3b). Given that male sloths present high reproductive skew with one male exhibiting high reproductive output (42% of offspring for which paternity was assigned), we removed this influential point, yet still detected a positive linear relationship between the density of C. obtusifolia and the reproductive output for males (F_1,17 = 18.61, p < 0.01; r² = 0.62) (electronic supplementary material, figure S2).

Figure 3.

Reproductive output of (a) male and (b) female three-toed sloths (Bradypus variegatus) as a function of Cecropia obtusifolia density in an agro-ecosystem in northeastern Costa Rica.

(c). Survival

Adult survival of sloths was high as only 5 of the 40 radio-marked individuals died during our study. Notably, individuals that died had lower densities of C. obtusifolia in their core areas compared to those that did not (1.99 trees ha⁻¹ versus 5.86 trees ha⁻¹; t₃₃ = −3.15, p < 0.01). For adult sloth survival, we found the most support for the model that includes the density of C. obtusifolia (table 2 and figure 4 inset graph), with a monthly estimate of true survival of 0.998 (s.e. = 0.001), and annual estimate of true survival of 0.984 (s.e. = 0.012). This model explained 40% of the variation in survival based on ANODEV, treating model {ϕ(.)} as the constant model and model {ϕ(t)} as the global model. The only other competitive model (ΔAIC < 2.0) for adult survival was the constant model (w = 0.207).

Figure 4.

Influence of Cecropia obtusifolia density on the population growth rate (λ) of three-toed sloths (Bradypus variegatus). Density of C. obtusifolia (coloured dots) was based on previous estimates of C. obtusifolia in different habitat types (see electronic supplementary material for details). Filled colours represent the range of expected population growth in different habitat types. Inset graph illustrates four-stage life-history diagrams including the relationship between C. obtusifolia density and adult survival and reproductive output (m = reproductive output; A = Adults; ϕ_0–1 = annual juvenile 0 to 1-year-old survival rate; ϕ_1–2 = annual juveniles 1- to 2-year-old survival rate; ϕ_2–3 = juveniles 2- to 3-year-old survival rate; ϕ_A = annual adult survival rate) used to derive the estimated population growth rate.

We did not find support for any single model predicting the influence of density of C. obtusifolia, C. villosa, and proportion of forest in the core areas on monthly juvenile survival (table 3). However, given the recurrence of the juvenile stage in all top models, we used the model that incorporated stage in our fitness analysis. Based on our ANODEV, this model explained 33% of the variation of the juvenile survival. The model with the only stage as a variable generated a monthly estimate of true survival of 0.968 during the 0–1 years old stage (s.e. = 0.014), and an annual estimate of true survival of 0.678 (s.e. = 0.168). No mortality was observed for 1–2-year-old juveniles, the monthly and annual estimate of true survival were 1.000. The best model for the survival of juveniles 2–3 years old exhibited a monthly estimate of true survival of 0.973 (s.e. = 0.019) and the annual estimate was 0.723 (s.e. = 0.228) (table 1).

Table 3.

Rankings of known-fate models estimating survival (ϕ) probabilities for juvenile three-toed sloths in northeastern Costa Rica. AICc: Akaike information criterion corrected for small sample sizes; K: number of parameters; w: AICc weight; st: stages (dependent, independent). Models are ranked based on AICc. The order in which the models were computed is shown in bold numbers.

model	K	AICc	Δ AICc	w
5 S (stage) + (Coussapoa)	2	46.892	0	0.239
7 S (stage) + (Coussapoa) + (forest)	3	47.329	0.438	0.192
6 S (stage) + (forest)	2	47.935	1.043	0.142
4 S (stage) + (Cecropia)	2	48.285	1.393	0.119
3 S (stage)	2	48.296	1.404	0.118
8 S (stage) + (Coussapoa) + (Cecropia)	3	48.886	1.994	0.088
2 S (.)	1	49.973	3.081	0.051
9 S (stage) + (forest) + (Cecropia)	3	49.988	3.097	0.051
1 S (t)	50	153.664	106.773	0

(d). Factors influencing fitness

Given the estimated relationships of female reproductive output (figure 3) and the adult survival rate with C. obtusifolia (figure 4), we varied the density of C. obtusifolia within the range of values based on previous estimates of C. obtusifolia in Neotropical areas (0 to 20 trees ha⁻¹, in increments of 1 tree ha⁻¹). The predicted population growth rate for the three-toed sloth was 0.977 when the density of C. obtusifolia is 0; λ increased when the density of C. obtusifolia increased and λ reach 1.0 when the density of this tree reached 0.7 tree ha⁻¹ (figure 4). Based on the density of C. obtusifolia in our locality and other Neotropical areas, shade-grown cacao and regenerating forests would support positive population growth for three-toed sloths, whereas forests and pastures would not.

4. Discussion

Our results provide evidence that vital rates of three-toed sloths are related to a key resource within a Neotropical ecosystem. Specifically, we found that the reproductive output and the adult survival for three-toed sloths were strongly associated with the density of C. obtusifolia within the individual's core area. Field ecologists have long speculated that the leaves of Cecropia trees are a critical dietary resource for a number of three-toed sloth species [29,30,33]; our study is the first, however, to directly link the relative abundance of this resource to demographic consequences among three-toed sloths.

Cecropia trees possess a number of characteristics that make them a high-quality food resource for arboreal folivores, in general, as their leaves possess high concentrations of nitrogen, low fibre and secondary metabolite content [31,32], making it a digestible and high-quality food item [56]. This nutritional composition appears to be the consequence of multiple factors: (i) the roots of several species of Cecropia present symbiosis with mycorrhizal fungi with which they increase nitrogen fixation [57], (ii) like other pioneer species, Cecropia exhibits a strategy of rapid growth for structural and photosynthesis tissues, rather than chemical defences [32], (iii) many Cecropia trees possess a symbiosis with ants of the Azteca genus [58,59] that defend the Cecropia from competing plant species as well as herbivores [60], in exchange for dietary resources in the form of Müllerian bodies and refuge in hollow internodes [61], which reduces chemical defences and increases nitrogen concentration within Cecropia leaves. In addition to its nutritional composition, Cecropia is one of the most abundant species in early successional and human-altered Neotropical ecosystems [62]. Cecropia trees also continuously produce leaves throughout the year [63], which benefits arboreal folivores, including three-toed sloths, that preferentially consume new leaves in which fibre content and chemical defences are particularly low [30,37]. Finally, the relative height and structure of leafless branches allow access to the sun, which is important given the relative heterothermy observed in three-toed sloths [24] and their regular sun-basking habits [28]. Owing to its abundance, nutritional composition, and continuous production of leaves and architecture, Cecropia appears to be an ideal resource for three-toed sloths.

Despite the important role of C. obtusifolia in the vital rates of adult three-toed sloths, and the tendency for juvenile mammals to select high-quality dietary resources [64] and accumulate somatic stores to reduce the amount of time in a vulnerable size [65], the abundance of C. obtusifolia was not related with juvenile survival of three-toed sloths. Montgomery & Sunquist [28] reported that pregnant three-toed sloths selected for Cecropia trees but upon parturition, switched to other tree species. They suggest that females moved to rare tree species with higher nutritional value; however, the other trees species used by juveniles in our locality (e.g. Tabebuia rosea) do not possess higher nutrition compared with C. obtusifolia [32]. Given that juvenile mammals are faced with a trade-off between predation risk and starvation [65], it is possible that juvenile three-toed sloths selected trees with greater cover to avoid predation. Although Cecropia leaves have high nutritional value, it is possible that the open structure of the tree could increase the detection of three-toed sloths by predators. Similar strategies have been found in other species of arboreal folivores, where juveniles avoid places with high levels of predation despite the presence of high-quality resources and choose microhabitats with more cover [65]. It is also possible that Cecropia trees are not only beneficial as forage but also in increasing forest for mating opportunities. Three-toed sloths vocalize during their mating season to attract mates [29], and the height and openness of Cecropia trees likely increase the range of communication [66]. The strong positive relationship of the density of C. obtusifolia and the reproductive success of the three-toed sloth supports the idea that one of the roles of this tree relates to the mating system of tree sloths even if not in the growth and development of juvenile sloths.

Three-toed sloths are generally considered to be forest specialists [28,37,48]. However, the proportion of regenerating forest in the core area of individuals was not associated with any vital rate. Research is increasingly finding that three-toed sloths are capable of using and persisting in modified habitats [38,39,67], especially shade-grown agricultural systems due to the canopy cover, connectivity, and food resources that they possess [67,68]. Nevertheless, late-stage regenerating forests are critical in specific stages of the life history of sloths; Garcés-Restrepo et al. [39] found that during the natal dispersal, juvenile three-toed sloths were strongly tied to riparian forests, which was critical for their dispersal and overall population viability. It is possible that the important role of C. obtusifolia and the high densities that this tree species can reach in shade cacao agro-ecosystems [69] attenuates the conversion of forest to agriculture for three-toed sloths.

While Cecropia is a relatively rare tree in heavily disturbed environments such as monocultures or cleared pastures [70], it is among the first to colonize clearings and form a canopy in Neotropical ecosystems [42] and reaches its greatest densities in regenerating forests [71]. Indeed, C. obtusifolia is relatively rare in pristine environments [62], because its presence depends upon the formation of tree gaps which are spatially unpredictable [72]. Often, resource specialists are associated with stable and constant environments [1]. It is somewhat surprising, then, that the three-toed sloths exhibit a positive fitness in early regenerating and disturbed habitats compared to pristine or constant environments [73]. We speculate that the disturbance and subsequent regeneration of forests have created an ideal habitat for early secondary species, including C. obtusifolia, which in turn benefits the three-toed sloths. Generally, disturbed environments are considered as hostile habitats for the stability of populations of specialist species [1]; however, our results show that the capacity of heterogeneous and regenerating environments for the maintenance of viable populations of specialist species may have been underestimated. The persistence of specialist species in a disturbed landscape would depend on how the disturbance and consequent forest regeneration specifically affects the distribution and abundance of the key resources on which specialists depend.

While the dependence of specialists on a narrow set of resources increases their vulnerability to environmental change [2,5], this dependence can also present opportunities for their conservation, given that targeted conservation on a single resource can bring about rapid recovery. For example, the Red-Cockaded woodpecker (Picoides borealis), a specialized cavity nester of North American forests, experienced significant declines due to forest loss [13]; the installation of artificial cavities for nesting led to the recovery of their populations [15]. Similarly, populations of specialized predators of the European rabbit (Oryctolagus cuniculus), the Imperial eagle (Aquila adalberti) and the Iberian lynx (Lynx pardinus) [14,74] collapsed with the local decline of rabbit populations. The subsequent reintroduction of rabbits to the Iberian Peninsula has been central to recovery for both predators [74]. Although three-toed sloths are not globally threatened, they are of conservation concern in Costa Rica [75] and many local extirpations have been reported in Central America [76]. In these areas, conservation actions including the planting of Cecropia trees in agro-ecosystems and the conservation of riparian forests are prescriptive approaches managers can use to increase population viability.

Ethics

Capturing, handling, and tracking were conducted as stipulated and authorized by Institutional Animal Care and Use Committee protocol A01424 of the University of Wisconsin-Madison and adhered to the guidelines for the use of mammals in research set forth by the American Society of Mammalogists [77]. Access was granted by the landowner and our project was approved by the Ministerio de Ambiente, Energía y Telecomunicaciones, Sistema Nacional de Áreas de Conservación, Costa Rica (179-2012-SINAC).

Data accessibility

All radio-telemetry and survival data are available via the Dryad Digital Repository.

Authors' contributions

All the authors conceived the idea for the manuscript and contributed to the design of the analysis, data analysis, and the writing of the manuscript. M.G.R. mapped the resources. All authors gave final approval for publication.

Competing interests

We declare we have no competing interests.

Funding

M.G.R. is supported by Colombia's COLCIENCIAS program. Research was supported by the grants from the National Science Foundation (grant DEB-1257535), Disney Foundation Conservation, University of Wisconsin-Madison and the American Society of Mammalogists.