A demographic history of a population of howler monkeys (Alouatta palliata) living in a fragmented landscape in Mexico

Ethics statement

This study is based on observational data and there was no direct interaction with the study subjects. We were granted access to the study site by local communities, landowners, and the Los Tuxtlas Biosphere Reserve, part of the National Commission of Natural Protected Areas of Mexico (CONANP). All research adhered to the American Society of Primatologists Principles for the Ethical Treatment of Non-Human Primates and to the legal requirements of Mexico.

Study species

Five subspecies of mantled howler monkeys (Alouatta palliata) are currently recognised: A. p. mexicana, A. p. palliata and A. p. aequatorialis, A. p. coibensis, and A. p. trabeata. These subspecies are distributed from south-east Mexico to northwest Peru (Cortés-Ortiz, Rylands & Mittermeier, 2015).

Mantled howler monkeys are seasonally folivorous, with leaves contributing over 80% of food intake when fruit is scarce (Milton, 1980; Glander, 1981; Cristóbal-Azkarate & Arroyo-Rodríguez, 2007; Dunn, Cristóbal-Azkarate & Veà, 2010). This degree of folivory has been associated with their small home range size compared to other more frugivorous species (Milton & May, 1976) and primates living in small home ranges are considered to be more resistant to habitat fragmentation (Cowlishaw & Dunbar, 2000).

Gestation lasts six months in mantled howler monkeys (Glander, 1980) and weaning occurs at approximately 18–20 months of age (Carpenter, 1934; Clarke, 1990; Domingo-Balcells & Veà-Baró, 2009). Age of first reproduction for females is approximately 41–43 months and males reach maturity at approximately 48 months of age (Glander, 1980; Domingo-Balcells & Veà-Baró, 2009). This species is characterized by bisexual emigration of juveniles; males typically emigrate at around 22 months of age and females typically emigrate at around 33 months of age (Glander, 1992). However, it has been suggested that in Los Tuxtlas juveniles may occasionally emigrate as early as 14 months of age (Domingo-Balcells & Veà-Baró, 2009). Accordingly, the 11-year duration of our study covers an important proportion of a howler monkey’s life cycle. Recent evidence indicates that secondary dispersal (individuals emigrating from groups that they have previously immigrated into) also exists in mantled howler monkeys and that this can be driven by the sex ratio of groups (Clarke & Glander, 2010). Dispersal patterns can be disturbed by relatively low levels of fragmentation (Chiarello & De Melo, 2001), as howler monkeys are highly arboreal and spend almost all of their time in the upper canopy, very rarely coming to the ground (Mendel, 1976).

Study site

The Los Tuxtlas Biosphere Reserve represents the northernmost limit of tropical rainforest distribution in the Americas (Guevara-Sada, Laborde & Sánchez-Ríos, 2004). Our study site (18°39′21″–18°31′20″N and 95°9′14″–95°1′45″W; elevation 0–400 m a.s.l) covers approximately 7,500 hectares, and like many other regions throughout the tropics, it has suffered from extensive forest loss, transformation, and fragmentation, principally as a result of cattle farming (Fig. 1). This occurred predominantly between 1976–1986, and the great majority of the actual forest fragments were created during this time (Cristóbal-Azkarate, 2004). Nevertheless, compared to many other fragmented landscapes, it retains a relatively high level of connectivity, with live fences (i.e., several strands of barbed wire held up by a line of trees), riparian vegetation and isolated trees found between many fragments, and a mean distance to nearest fragment of 103 ± 172 m (Arroyo-Rodríguez, Mandujano & Benítez-Malvido, 2008). It also contains areas of continuous forest in close proximity (less than 500 m) to the fragments (Fig. 1).

The climate in Los Tuxtlas is warm and humid with a mean annual temperature of 25 °C and rainfall of 4,900 mm (Soto, 2004). There is a dry season between March and May and a wet season from June to February. During the wet season there is also a period of strong winds and a considerable reduction in temperature between October and February (Fig. 2A). Long-term records of phenological data in the region show that there are two distinct peaks in fruit production: a primary peak at the end of the dry season—beginning of the rainy season (April–June), and a shorter, less intense secondary peak in the wet season (August–October), while fruit production abruptly falls to very low levels between November and March (Fig. 2B). The howler monkeys in Los Tuxtlas respond to the reduction in temperature and fruit availability between November and March by increasing their consumption of leaves and their foraging effort (Dunn, Cristóbal-Azkarate & Veà, 2010), which, in turn, has been associated to higher levels of physiological stress (Dunn et al., 2013). Therefore, we refer to this period as the “period of energetic stress”.

Study groups

We carried out the first census of our study site in 2000. Of the 55 forest fragments that are found in our study site, we found 21 to be inhabited with at least one howler monkey and recorded a population of 316 individuals living in 43 groups (Cristóbal-Azkarate et al., 2005). We began studying four of these groups intensively in 2000. Over the following 10 years, we studied six more groups as part of a programme of interdisciplinary research, for a total of 10 groups, which provided the data for our analyses (e.g., Cristóbal-Azkarate et al., 2006; Cristóbal-Azkarate et al., 2007; Dunn, Cristóbal-Azkarate & Veà, 2010; Dunn et al., 2013). Despite the wide-ranging nature of the research, we gathered basic demographic data, such as the number of individuals, age-sex composition, births, deaths and migrations, over the 11-year period.

Data collection

We present demographic data from 10 groups of howler monkeys, representing 454 monthly group visits, which were carried out between 2000 and 2011 (Table 1). Given that the data has been pooled across several different studies, there is some discontinuity, with certain groups being studied for longer and/or more frequently than others (mean ± SD = 45.1 ± 29.7 monthly visits per group; Table 1). The study groups inhabited eight different forest fragments, which varied in size, shape and connectivity (Fig. 1).

Table 1:

Sampling chart of monthly visits to 10 groups of mantled howler monkeys in Los Tuxtlas, Mexico, between 2000 and 2011.

Months with at least one visit to a group are shown in dark grey. Detailed demographic data are also provided for all events for which we were able to determine the date with a maximum error of one month.

Group	Year	Month
		January	February	March	April	May	June	July	August	September	October	November	December
MT(1)	2000
	2001
	2002			3F↓
	2003					1SF↓					1*	2M↓1J ↑	1I
	2004											2*1S1J↑	2*1M
	2005		1I 1*				1M↓	1*	2M4F1J4I↑			1F↑
	2007
	2008	1*					1*
	2009					2M↑				3F2J1I↑2M2F1I↑
	2010		2*		4M↑			1J↑			1M1J↑	1M↑
	2011	2M↑	1M↑
MT(2)	2000
	2007
	2008
	2009		1*			1M↑2F↑	1*			3M1F↑2F1SM1I↑		2F↑ 1*
	2010	1M↑		2*1F1I	1*1SF↑	1SF↑	1I		1M1F↑				3*
	2011
RC3	2000
	2002
	2003
	2004								1*	1J		1*
	2005		1I
	2006			1SF↑				1*	1SM↑			1*
	2007				1I
	2008		1*								1F↑	1*
	2009									1*	1F↑	1SM↑	1F↑
	2010	1J↑	1I 2M1J↑						1M↑		2*
	2011
2AB	2001
	2002
	2003
	2004	1M 1*										1*	1J↑
	2005						1F↑	1J↑	1*
	2007
	2008
	2009						2M1F1S↑		1*1I↑
	2010								1SM↑		1*	1SM↑
	2011	1J↑
RH	2001
	2002
	2003									1*	1*		1I
	2004		1J↑	1*							1*	1I	1*
	2005			1*								1*1J↑
	2006							1*	1F↑	1I
	2007
	2008
JIC	2001
	2002
	2004
	2005
	2007
	2008											1*
	2009
RC5	2007
	2008
	2009	1F1S1I↑	1*										1F↑
	2010										1*	1I	1*
	2011
RC4(1)	2000
	2001
	2004												2M1F↑
	2005	1F↑				1F↑1I
RC4(2)	2004
	2005
RC2	2000
	2001
	2004												2F1I↑ 1M
	2005	1*	1I							1F1SF↑		2*

DOI: 10.7717/peerj.3547/table-1

Notes:

*

Birth

↓

immigration

↑

emigration

death

M

Adult male

F

Adult female

S

Sub-adult (SM, Subadult male; SF, Subadult female)

J

Juvenile]

I

Infant

We identified group members by the distinguishing colour patterns on their feet and tails, which are characteristic of this subspecies. We created an identity sheet for each individual as a reference in the field, drawing and making notes on the distinctive features. Each time we recorded a new individual in a group, we assigned it an age and sex using the classification system developed by Domingo-Balcells & Veà-Baró (2009), which allows an age range to be estimated on the basis of morphological and behavioural characteristics.

Demographic events

Throughout the study, we registered all demographic events in the groups, including: emigration, immigration, birth and death. However, given the low probability of observing these events directly, some of the events were also assumed to have occurred on the basis of changes in group composition and supporting evidence.

Statistical analyses

For the calculation of the seasonality of demographic events, IBI and birth rates, we considered only those events that were registered during periods in which the study groups were observed continuously and that could be assigned to a date with a maximum error of one month. In order to control for the effect that our slightly unbalanced sampling effort could have on the seasonality data, we weighted the original data by dividing the frequency of events per month by the number of different visits to the same group within a month (mean ± SD average visits per month = 37.6 ± 3.1, range = 35–42; Table 1). We used these weighted values to calculate the percentage of demographic events in each month.

We used ANOVAs to analyse the differences in annual birth rate and IBI among groups, and reported eta squared (η²) as a measure of effect size (which is analogous to R² in regression analyses). Values of η² vary from 0 to 1 and values of 0.02, 0.13, and 0.26, and can be, as a rule of thumb, considered as small, medium and large effects, respectively (Cohen, 1973). We also used a Student’s T test to test the hypothesis that the death of a suckling offspring, ≤14 months of age (Domingo-Balcells & Veà-Baró, 2009), shortens the IBI by comparing the mean IBI of females with surviving and non-surviving offspring, and reported Cohen (1977) as a measure of effect size. For Cohen’s d effect sizes of 0.2, 0.5, and 0.8, can be thought of as small, medium and large, respectively (Cohen, 1977).

To test for differences in the frequency of demographic events between the season of energetic stress (November–March, see above) and the rest of the year, as well as to test for statistically significant differences between peaks in demographic events at certain times of year compared to the rest of the year, we conducted Chi-squared (X²) goodness of fit tests, with expected values being proportionally calculated according to the number of months used in the analysis. We calculated effect sizes for Chi-square tests using Cramer’s phi coefficient (φ), whereby 0.1, 0.3, and 0.5 can be interpreted as small, medium and large effects (Cramer, 1999). Furthermore, in order to account for the underlying continuity of the time variable, we also used circular statistics to test for seasonality of demographic events (Batschelet, 1981). This approach has several advantages over those traditionally used by primatologists to test for seasonality (Janson & Verdolin, 2005; Gogarten et al., 2014). The mean vector length (r) obtained from circular statistics is well suited as an index of seasonality, as it provides a measure of how evenly events are distributed throughout the year. When events are spread evenly across months (not seasonal), r is close to zero and when events are highly clustered at the same time of year (highly seasonal), r is close to one. We tested the statistical significance of the r statistic using the Rayleigh test (Batschelet, 1981), which compares the data with the null hypothesis that demographic events have a random distribution across months. As we used monthly data for demographic events, rather than specific dates, we also used a correction factor (c = 1.0115) when calculating the r statistic (Batschelet, 1981). To test for bimodal distribution in the data, we also calculated r by doubling the angle calculated for each demographic event (Batschelet, 1981; Janson & Verdolin, 2005; Gogarten et al., 2014).

We carried out analyses in R 2.13.1 (R Core Development Team, 2008), testing for normality in the data and considering p < 0.05 as significant.

Births

We registered 75 births and at least two births were observed in all 10 of our groups (Tables 1 and 2). Of these, we were able to determine the date of birth to within one month on 49 occasions.

The mean birth rate per group was 0.42 ± 0.32 births per female per year (N = 39 births; Table 3). There were groups with no births in some years, while other groups had a birth rate as high as 1 in some years (indicating that all females of reproductive age gave birth in that year). Although mean birth rate ranged from 0.18 ± 0.24 to 0.56 ± 0.40 births per female per year across groups (Table 3), we found no statistically significant differences in mean birth rate among groups (F_9,29 = 0.57, p = 0.81, η² = 0.15).

The mean IBI was 21.6 ± 13.3 months (N = 18; Table 3). Although mean IBI varied from 11.0 ± 4.2 months to 39.5 ± 24.7 months across groups (Table 3), we found no statistically significant difference among groups (F_4,13 = 1.68, p = 0.22, η² = 0.34). We observed 12 IBIs in which the offspring from the first birth had survived until weaning (mean = 26.1 ± 14.1 months) and six IBIs in which the offspring from the first birth had died before weaning (mean = 12.7 ± 4.5 months), and found a significant reduction in IBI when the first offspring had died before weaning (mean difference = 13.42, 95% CI [3.89–22.95], t = 3.01, df = 16, p = 0.009, d = 1.28).

Although we observed births throughout the whole year, births were clearly seasonal. Seventy-four per cent of births occurred between October and March, with a main peak in November and a smaller second peak in February (Fig. 3A & Table 1). Accordingly, the bimodal r statistic was highly significant (unimodal r = 0.20, p = 0.14; bimodal r = 0.43, p < 0.001). As the gestation time of howler monkeys is 6 months, and weaning occurs at approximately 18–20 months, these data indicate that conceptions leading to births and weaning peaked in May, after the period of energetic stress. In fact, the number of conceptions leading to births was not equally distributed, with fewer conceptions than expected during the period of energetic stress (10 conceptions between November and March) and more than expected during the rest of the year (39 conceptions between April and October) (X² = 9.38, df = 1, p < 0.005, φ = 0.19).

Emigration

We recorded emigrations in all but one of our study groups, with 62 individuals emigrating from groups (Tables 1 and 2). Thirty-four of these were adults (13 adult males and 21 adult females), 7 were sub-adults (2 males, 2 females, and 3 of unknown sex), 13 were juveniles and 8 were infants.

We recorded 24 natal emigrations (52.8% of emigrations of known origin). One by an adult male, 6 by adult females, 3 by subadults, 11 by juveniles and 7 by infants. Of the 7 infants, 4 left the group soon after the birth of a sibling and 8 left the group soon after one or more individuals had immigrated into their group.

We recorded 22 secondary emigrations (47.8% of emigrations of known origin). Nineteen of these were adults (11 males and 8 females), 1 was a male sub-adult, 1 was a juvenile and 1 was an infant that transferred to the group together with her mother, and then left with her after one month of permanence in the group. The mean time that the individuals spent in a group prior to secondary transfer was 18.4 ± 21.1 months (range = 1–90 months).

On 16 occasions, we were not able to determine whether the emigrating individuals were born in the group they emigrated from or whether they had previously immigrated into the group. Accordingly these emigrations were of unknown origin and were not classified as natal or secondary.

Emigrations occurred throughout the year, but there were clear differences in emigration patterns among the age-sex classes (Fig. 3B, Tables 1 and 2). Male emigration peaked in August (33.1% of cases) and January–February (43.9%), when more emigrations occurred than expected (X² = 8.57, df = 1, p < 0.005, φ = 0.66; and X² = 8.14, df = 1, p < 0.005, φ = 0.63, respectively), female emigration showed a very clear peak in August–September (49.6% of cases), when more emigrations occurred than expected (X² = 14.49,  df = 1, p < 0.001, φ = 0.69), and subadult and juvenile emigration showed a peak in November–December (38.9% of cases), when more emigrations happened than expected (X² = 8.87, df = 1, p < 0.005, φ = 0.48). However, the r statistic for seasonality was non-significant in all cases, only trending towards significance in the females (unimodal r = 0.34, p = 0.08).

Immigration

We recorded immigrations in all the forest fragments that we studied and in all but one of our study groups (this group was only followed for one year). We recorded 57 individuals immigrating into new groups (Tables 1 and 2); 46 were adults (22 males and 24 females), 6 were sub-adults (3 males and 3 females), 3 were juveniles and 2 were infants. Of these, we were able to determine the date of immigration to within one month in 41 cases (Table 1).

Immigration occurred throughout the year, but there were clear differences in immigration patterns among the age-sex classes (Fig. 3C, Tables 1 and 2). Adult male immigrations peaked in April–May (40.5% of cases), when more immigrations occurred than expected (X² = 5.88, df = 1, p < 0.05, φ = 0.39) and again in September–December (52.2%), though this was not significantly more than expected by chance. Female immigration showed a clear peak from September–December (79.3% of cases), when more immigrations occurred than expected (X² = 12.89, df = 1, p < 0.001, φ = 0.92), and subadults and juveniles showed a peak in immigration between August–September (65.3% of cases), when more immigrations occurred than expected (X² = 12.10, df = 1, p < 0.001, φ = 1.51). Despite these peaks in immigration, the r statistic for seasonality was non-significant in all cases.

Deaths

We registered 18 deaths, and at least one death was registered in eight of the 10 groups (Table 2). Thirteen of these individuals were infants: eight were younger than four months of age, three between four and eight months, and two were 10 months of age. Of these, we only observed one death directly, when, a one-month old infant died shortly after we found it lying by its dead mother which was seemingly killed by another howler monkey (see below). On another occasion we assumed that an infant had died shortly after its mother had died and it was observed falling in a tree. We assumed one juvenile to have died having shown signs of physical weakness and struggling to keep up with the group. The remaining four deaths were all adults. We recovered the body of one female, which had several serious bite marks. Post-mortem examination by a veterinarian found the cause of death to be lung perforation, consistent, in terms of bite shape, breadth and depth, with an attack by another howler monkey (M Escorcia-Quintana, pers. comm., 2008). One adult male probably died after we observed it with severe open wounds resulting from an attack by two immigrating males. Another adult male showed signs of paralysis and lethargy before his assumed death. A further adult male showed signs of old age, lack of appetite and was unable to keep up with the group.

We registered deaths in most months of the year, but there was a clear peak between November and March when 75.3% of deaths occurred (Fig. 3D, Table 1). There were more deaths than expected during the period of energetic stress (N = 14), and fewer than expected in the rest of the year (N = 4) (X² = 9.66, df = 1, p < 0.005, φ = 0.54). However, the r statistic for seasonality was not significant (unimodal r = 0.08, p = 0.89; bimodal r = 0.25, p = 0.30).

Disappeared

We were unable to interpret the history of 27 individuals from the data, which we recorded as disappeared (Table 2).