Land use change has converted natural contiguous forests into patches of different size, threating the maintenance of global biodiversity on an unprecedented scale and rate (Newbold et al., 2016; Taubert et al., 2018). This situation is of particular importance in the tropics, where deforestation rates are the highest worldwide and increasing (Hansen et al., 2013). Understanding how these changes affect species in human-modified landscapes is needed to improve conservation strategies (Arroyo-Rodríguez et al., 2020). However, research in fragmented forests focus on only a few patch attributes, such as patch size (Fahrig, 2003), overlooking the effect of other important variables, such as patch quality. Also, there is a paucity of research assessing the effect of the regional context (e.g., land use history, regional deforestation) and interspecific variance in responses to land use changes (Pardini et al., 2010; Villard and Metzger, 2014). This can lead to incomplete or inaccurate conclusions, and the adoption of inappropriate conservation measures.

The research bias towards patch size effects is related to the strong influence of island biogeography theory and metapopulation theory in ecology and conservation biology (Fahrig, 2013; Haila, 2002). However, an increased number of studies have documented that biodiversity patterns in fragmented landscapes are not related to patch size everywhere (Laurance, 2008; Onderdonk and Chapman, 2000; Watling et al., 2020). Although patch size is positively related to population abundance and species richness in some regions (Mohandass et al., 2017), it is a poor predictor of species presence, abundance, and richness in other regions (Mellink et al., 2017; Onderdonk and Chapman, 2000; Pardini et al., 2010). For instance, Onderdonk and Chapman (2000) did not find a relationship between patch area and the presence of primates in Uganda. Mellink et al. (2017) also found that bird communities depend on the presence of trees and not on patch size in a semi-arid region of central Mexico. Thus, other variables, such as patch quality, can play important roles for biodiversity maintenance in fragmented landscapes.

Patch quality for arboreal mammals is often indexed using variables such as tree density and size, canopy height, vegetation type, and disturbance level (Arroyo-Rodríguez and Mandujano, 2009). Tree size, for instance, is a good indicator of food availability (Grant et al., 1992; Chapman et al., 1992) and has been proven to be an important predictor of forest species distribution in fragmented landscapes, as has patch tree species richness (Korboulewsky et al., 2016). Yet, which ecological variable bests predicts success in fragmented landscapes varies among species and their dietary flexibility (Dunlop et al., 2017; Hannibal et al., 2020; Marsh and Chapman, 2013). Therefore, our understanding of this topic is far from complete.

Primates are involved in numerous ecological processes, such as herbivory, seed dispersal, predation, and pest control, representing an important component of forest ecosystems (Andresen et al., 2018; Chapman et al., 2013; Estrada et al., 2017). Yet, as most primate species are forest dwellers (Galán-Acedo et al., 2019), they are particularly vulnerable to deforestation (Marsh and Chapman, 2013). In fact, ∼60% of the world's primate species are threatened with extinction (Estrada et al., 2017). The loss of primate species will have significant implications as they are of critical cultural and ecological importance (Estrada et al., 2017).

Here we assessed the effect of patch size and quality on the encounter rate (ER) and immature-to-female ratio (IFR, a proxy of reproductive output) of three endangered and forest-specialist primates: Mexican mantled howler monkey (Alouatta palliata mexicana), black howler monkey (Alouatta pigra), and Geoffroy's spider monkey (Ateles geoffroyi). We sampled 36 forest patches from three regions with different forest cover (50%, 17% and 5% forest cover). As patch colonization increases with patch size, and larger patches usually have more food resources (Arroyo-Rodríguez and Mandujano, 2006a) and lower anthropogenic pressures (Chapman et al., 2006; Peres, 2001), we predicted that both ER and IFR would increase with patch size in most regions. Yet, following the fragmentation threshold hypothesis (sensuPardini et al., 2010), patch size effects are predicted to be weaker (or absent) in the 5% forest cover region, as the lack of landscape connectivity in this region limit inter-patch animal movements, making populations more dependent to the local conditions in the patches (also see Villard and Metzger, 2014). Thus, patch quality effects are predicted to be stronger in this region. In particular, ER and IFR of both primate species are predicted to be positively related to total tree basal area (a proxy of tree size and forest maturity positively related to food availability, Arroyo-Rodríguez and Mandujano, 2006b; Chapman et al., 1992) and basal area and species richness of top food species. Finally, although howler monkeys can consume many early successional (pioneer) tree species in disturbed habitats (Arroyo-Rodríguez and Dias, 2010), basal area of pioneer species is indicative of forest disturbance. Thus, we expect that basal area of pioneer trees will be negatively related to ER and IFR of both species.

In the 50% forest cover region, spider monkeys were recorded in 6 of 12 patches and black howler monkeys in 9 of 12 patches (Table 1). The encounter rate (ER) of spider monkeys averaged 0.8ind/km (range among patches=0.1–2.24ind/km) and mean immature-to-female ration (IFR) averaged 0.6 (range=0.3–1.21). Mean ER of black howlers was 1.03 (range=0.28–1.75ind/km) and mean IFR was 1.17 (range=0.67–2). In the 17% forest cover region, we only recorded mantled howler monkeys, which were present in 5 of 12 patches, with a mean ER of 0.55ind/km (range=0.03–1.05ind/km) and a IFR averaging 0.44 (range=0.38–0.52). Finally, in the 5% forest cover region, we only recorded mantled howlers in 9 of 12 patches and they averaged 1.2ind/km (range=0.62–1.92ind/km) and 0.69 immatures per female (range=0.5–1).

The predictor variables explained >48% of deviance of all response variables (Fig. 2). Its predictive power was higher for spider monkeys in the 50% forest cover region (Fig. 2a) and for mantled howlers in the 17% forest cover region (Fig. 2c). In the 50% forest cover region, ER of spider monkeys (∑wi=0.9) was mainly and positively related to patch size, although it was also associated with the richness of top food species and total basal area (Fig. 2a; Tables 2 and 3). Patch size was also the most important predictor of mantled howler populations, positively affecting ER (∑wi=0.96) in the 17% forest cover region (Fig. 2c; Tables 2 and 3).

Fig. 2.

Predictor variables included in 95% set of models (bars) selected for the two response variables (encounter rate and immature-to-female ratio) of three primate species in three rainforest regions with different forest cover. The importance of each variable is shown by the sum of Akaike weights (∑wi). We assessed the impact of patch size and four vegetation attributes: richness of top food species (R top food), basal area of pioneer species (BA pioneer), basal area of top food species (BA top food), and total tree basal area (BA total). The percentage of deviance explained by each complete model is indicated in each panel as a measure of goodness-of-fit of each complete model. Positive (striped grey bars) or negative (black bars) responses to each predictor are indicated. Filled grey bars represent the cases in which the unconditional variance was higher than the model-averaged parameter estimate, indicating that such parameter can include zero (i.e., a null effect; see parameter estimates and unconditional variances in Table 2).

(0.31MB).

Patch quality was important for black howler monkeys in the 50% forest cover region (Fig. 2b) and for mantled howlers in the 5% forest cover region (Fig. 2d). In the first case, ER of black howler monkeys was mainly positively related to basal area of pioneer species (∑wi=0.38), but total basal area (∑wi=0.19) and basal area of top food species (∑wi=0.19) were also positively related to ER (Fig. 2c; Tables 2 and 3). IFR of black howlers was also positively related to basal area of top food species in this region (∑wi=0.32; Fig. 2b); yet, this relationship is weak, as the most plausible model included only the intercept (i.e., null model; Table 3). Regarding mantled howlers in the 5% forest cover region, the basal area of pioneer species was the best predictor of ER (∑wi=0.65) and IFR (∑wi=0.55), negatively affecting both responses (Fig. 2d; Table 3). Surprisingly, ER of mantled howlers was also negatively related to top food species richness (∑wi=0.56) in this region.

This study assesses the effect of patch size and quality on the encounter rate (ER) and immature-to-female ratio (IFR) of three endangered Mexican primates in three rainforest regions with different deforestation level. Our findings indicate that the effects of patch size and quality are regional and species specific. Patch size and quality (i.e., species richness of top food species) was the best predictor of spider monkey populations in the 50% forest cover region, positively affecting the encounter rate (ER). Patch size was also the best predictor of mantled howler monkeys, positively affecting ER in the 17% forest cover region. However, ER and IFR of black howler monkeys in the 50% forest cover region and mantled howlers in the 5% forest cover region were mainly related to vegetation attributes, thus indicating that patch quality can be the main driver of primates in some regions.

As expected, patch size is an important driver of primate populations in two of three regions. The ER of two species increased with patch size in the 50% forest cover (spider monkeys) and 17% forest cover (mantled howler monkeys) regions. This is not surprising, as patch area is viewed to be generally positively related to food availability and patch connectivity (Hanski, 1998), positively influencing patch occupancy, population abundance, and species richness (Harcourt and Doherty, 2005; Mohandass et al., 2017). Species persistence in large patches can also be enhanced by the fact that hunting pressure and diseases are usually lower in these patches (Arroyo-Rodríguez and Dias, 2010). However, consistent with previous studies (Mellink et al., 2017; Onderdonk and Chapman, 2000), our results indicate that patch size is not always the best predictor of primate populations in fragmented forests.

Consistent with our hypothesis, patch size was a poor predictor of mantled howler monkeys in the 5% forest cover region, possibly because the remaining forest patches were small, scarce, and highly isolated between each other. Together, these conditions can increase the exposition of patches to hunting and predation (Chapman et al., 2006; Peres, 2001). Also, the anthropogenic matrix in this region is dominated by open areas (i.e., cattle pastures), which can significantly limit the interchange of individuals between forest patches. Thus, following the fragmentation threshold hypothesis, such limited availability of habitat and limited flow of individuals between patches (i.e., low landscape connectivity) can decrease the abundance of individuals in both small and large patches, especially the abundance of forest dependent mammals (Pardini et al., 2010; Villard and Metzger, 2014). Therefore, it is not surprising that the encounter rate of mantled howlers in the 5% forest cover region does not depend on patch size.

Patch size was also a poor predictor of black howler monkeys in the 50% forest cover region. This finding also follows the predictions of the fragmentation threshold hypothesis, which predicts weak associations between patch size and abundance in high forested regions because the remaining patches in these regions are usually very large, and very close to each other (Pardini et al., 2010). Under this context, there can be enough flow of individuals to maintain populations in small and large patches, which can potentially weaken the association between population structure and patch size. However, patch size was an important predictor of spider monkeys in the 50% forest cover region. This could be related to the notably larger home range size requirements of spider monkeys, which is 20 times larger than that of black howlers. This large spatial requirement can strengthen the dependency of this and potentially other mammal species with similar home range requirements (e.g., tapir: 564ha home range, peccary: 4097ha home range; Jones et al., 2009) on patch size, even in high forested regions.

Our findings also indicate that patch quality plays a key role in maintaining primate populations in fragmented forests. In particular, vegetation structure drove the ER of spider monkeys and the ER and IFR of black howler monkeys in the 50% forest cover region. Also, vegetation structure affected mantled howler monkey responses in the 5% forest cover region. This is consistent with previous studies showing the importance of vegetation structure for the presence and abundance of different species, including primates (Arroyo-Rodríguez et al., 2007; Bohenek et al., 2017; Gardiner et al., 2018). In particular, we found a positive association between the ER of spider monkeys and species richness of top food species. This is probably related to the need of a nutritional balance in the macronutrient composition of their diets. Spider monkeys require a protein and non-protein (e.g., lipids) balance to allow their energetic intake (Felton et al., 2008). Thus, patches with higher richness of top food species will offer a better nutritional balance, increasing the survival and maintenance of spider monkeys.

Interestingly, the basal area of pioneer tree species has contrasting effects on howler monkeys in different regions. As predicted, this variable was negatively related to ER and IFR of mantled howler monkeys in the 5% forest cover region. Nevertheless, contrary to our expectations, it was positively related to the ER of black howler monkeys in the 50% forest cover region. This variable usually increases in more disturbed forests, so it can be considered a proxy of forest disturbance (Bongers et al., 2009). Thus, in the 5% forest cover region, where inter-patch individual movements may be strongly limited, a high basal area of pioneer tree species likely indicates that the patch is very disturbed and that resource availability can be limited because there are fewer large trees from preferred food species (Arroyo-Rodríguez and Mandujano, 2006a). However, in landscapes with 50% forest cover, animals can easily move between fragments. Thus, even though the resident patch is disturbed (i.e., with high basal area of pioneer trees), the monkeys can make use of resources from other patches. Furthermore, pioneer trees can also represent valuable resources for howler monkeys in fragments (Arroyo-Rodríguez and Dias, 2010; Milton et al., 2019), as the fast-growing leaves of these trees typically have higher protein and lower fiber and phenolic levels than those of old-growth shade-tolerant tree species (Coley, 1987). Therefore, the positive response of howler monkeys to basal area of pioneer trees in landscapes with 50% forest cover can be, at least partially, explained by the food resources that these trees provide to primates.

Another surprising result is the mantled howler monkeys’ negative association between ER and IFR and top food species richness in the 5% forest cover region. This is counterintuitive and we cannot find a reasonable explanation. It may be the result of uncontrolled confounding factors, such as human pressures or unknown historical events (e.g., disease). Such factors are difficult to identify in regions deforested long ago (∼70 years of human disturbance).

Also puzzlingly is the fact that this region has the highest number of inhabited patches (together with the 50% forest cover region) and the second highest regional abundance. This finding suggests that monkeys can survive on the few available resources, supporting claims of very high behavioral flexibility of this species (reviewed by Arroyo-Rodríguez and Dias, 2010). However, this may also reflect an extinction debt: in long-lived species such as howler monkeys, there can be a time lag in species extinctions (Cowlishaw, 1999; Lira et al., 2012; Metzger et al., 2009). Therefore, collecting long-term data, over a relatively large spatial scale is needed to better understand and predict population trends in this highly deforested region.

We conclude that both patch size and quality must be considered in conservation plans. Prioritizing the preservation of large patches is likely the most effective in low deforested regions (∼50% forest cover), especially for terrestrial mammal species with large home ranges. However, for species with dietary flexibility (e.g., Eulemur sp., Sato et al., 2016; Chlorocebus djamdjamensis, Mekonnen et al., 2018), conserving landscapes with multiple patches and preventing further vegetation disturbance may be the most appropriate strategy in high forested regions. In regions with intermediate level of deforestation (e.g., ∼17% forest cover), patch size may be more for primates and other mammal species (Garmendia et al., 2013; Magioli et al., 2015; Norris et al., 2010); thus, conservation strategies in these regions should rely on the preservation of large patches. Finally, in strongly deforested regions (e.g., ∼5% forest cover), where populations are isolated and dependent on local vegetation conditions, conservation and management plans should focus on forest restoration. However, given the potential extinction debts in the study landscapes (Cowlishaw, 1999; Metzger et al., 2009), additional long-term monitoring studies are needed to better understand population trends, and thus improve conservation plans in these landscapes. Such a conservation planing can also be improved by better understanding the role of other arboreal land covers (e.g., secondary forests, living fences, isolated standing trees; see Arroyo-Rodríguez et al., 2020), not only old-growth forests, in biodiversity conservation.

None.