Estimating the population size of the Sanje mangabey (Cercocebus sanjei) using acoustic distance sampling

The Sanje mangabey (Cercocebus sanjei) is endemic to the Udzungwa Mountains, Tanzania, and is classified as Endangered due to its putatively declining population size, habitat degradation and fragmentation. Previous population size estimates have ranged from 1,350 to 3,500 individuals, with the last direct survey being conducted 15 years before the present study. Previous estimates are now thought to have underestimated the population due to a limited knowledge of group and habitat size, nonsystematic approaches and the use of visual methods that are not suitable for surveying the Sanje mangabey with its semi‐terrestrial and elusive behaviors. We used an acoustic survey method with observers recording the distinctive “whoop‐gobble” vocalization produced by mangabeys and point transect distance sampling to model a detection function and estimate abundance. Twenty‐eight surveys were conducted throughout the two forests where Sanje mangabeys are found: Mwanihana forest in the Udzungwa Mountains National Park (n = 13), and the Uzungwa Scarp Nature Reserve (n = 15). Group density was found to be significantly lower in the relatively unprotected Uzungwa Scarp forest (0.15 groups/km2; 95% CI: 0.08–0.27) compared to the well‐protected Mwanihana forest (0.29 groups/km2; 95% CI: 0.19–0.43; p = .03). We estimate that there are 1,712 (95% CI: 1,141–2,567) individuals in Mwanihana and 1,455 (95% CI: 783–2,702) in the Uzungwa Scarp, resulting in a total population size of 3,167 (95% CI: 2,181–4,596) individuals. The difference in group density between sites is likely a result of the differing protection status and levels of enforcement between the forests, suggesting that protection of the Uzungwa Scarp should be increased to encourage recovery of the population, and reduce the threat of degradation and hunting. Our results contribute to the reassessment of the species' IUCN Red List status and informing management and conservation action planning.


| INTRODUCTION
Nonhuman primates are key to the successful functioning of their ecosystems; however, primates are currently facing an extinction crisis with approximately 75% of species declining and 60% threatened with extinction, with the largest threats including habitat loss to agriculture, logging and livestock farming, and hunting (Estrada et al., 2017). Research efforts into monitoring wild primate populations have proven crucial in conservation management as recording data on population abundance and distribution can provide insights into the response of a species to changes in habitat and [The copyright line for this article was changed on 20 January 2020 after original online publication] population trends over time (Campbell, Head, Junker, & Nekaris, 2016;Chapman et al., 2018;Jones, Hawes, Norton, & Hawkins, 2019;Lwanga, Struhsaker, Struhsaker, Butynski, & Mitani, 2011). By establishing an initial baseline and appropriate methodology, these data can be used to assess population trends in subsequent years and develop adaptive management plans that call for the implementation of improved methods to protect species (Lyons, Runge, Laskowski, & Kendall, 2008;Nichols & Williams, 2006).
The Sanje mangabey (Cercocebus sanjei) is endemic to the Udzungwa Mountains in south-central Tanzania (Ehardt, Butynski, & Struhsaker, 2008). Since its discovery in 1979 by Homewood and Rodgers (1981), it has been studied to elucidate its distribution and population size to determine its conservation status and required management (Ehardt et al., 2008;Ehardt, Jones, & Butynski, 2005;Rovero et al., 2006;Rovero, Mtui, Kitegile, & Nielsen, 2012), and an inferred declining population size has resulted in an IUCN Red List Endangered status (EN; McCabe, Rovero, Fernández, Butynski, & Struhsaker, 2019). The population is divided between two isolated forest blocks: Mwanihana forest in the well-protected Udzungwa Mountains National Park, and the Uzungwa Scarp Nature Reserve forest, which has a lower level of protection and regulations that are not strongly enforced (Ehardt et al., 2005). These forests are separated by 100 km of agricultural land and low elevation habitat unsuitable for mangabeys, preventing dispersal of individuals between forests, which could potentially limit the recovery of each population.
The current population size of the Sanje mangabey remains debated and with previous habitat loss and degradation and the current impact of hunting in the forests likely to impact the species, especially in the Uzungwa Scarp (Hegerl et al., 2017), current estimates and subsequent monitoring are essential to assess the conservation status and needs of the species. Previous population size estimates range from as little as 1,350 individuals to 3,500 (Dinesen, Lehmberg, Rahner, & Fjeldså, 2001;Rovero, Marshall, Jones, & Perkin, 2009, respectively; Table 1) with the last dedicated survey conducted by Ehardt et al. (2005Ehardt et al. ( ) between 1997Ehardt et al. ( and 2002. However, previous studies used methods that were not suitable for the elusive behavior of the Sanje mangabey, and the group size and habitat area calculations used to extrapolate group density were underestimations, resulting in an underestimated population size.
All previous surveys of the mangabeys have been nonsystematic or have used line transect methods to estimate population size.
These methods are now recognized to be inefficient for this species as unhabituated groups flee rapidly from humans and are difficult to detect in dense vegetation (Rovero & Struhsaker, 2007;Rovero et al., 2006Rovero et al., , 2012. This was supported by line transect observations when individuals were heard calling but were not seen by observers (Rovero et al., 2006).
This study aimed to conduct the first systematic survey of the Sanje mangabey population and provide the first inferential  surveys were suspended as the ability to detect calls decreased.
Each survey was conducted once and only early in the morning when the mangabeys are known to call at the highest frequency during the day (approximately 70% of calls before 1200 hr ;Ehardt et al., 2005). The surveys started when light levels were safe enough for observers to move through the forest such that survey times were variable. Observations were recorded from the time the observers arrived at the post (mean start time: 0642 hr ± 11.6 min) until 0900 hr; all surveys covered a core time of 0700 hr to 0900 hr.
The full survey time for each post was used so that the earliest calls (<0700 hr: 10.2% of calls) were not lost, which would have led to an underestimation of group density.
With each vocalization detected, observers recorded the time of the start of the call, a bearing from the post and estimated the distance to the origin of the call. Observers would estimate the number of groups heard whilst in the field, attributing individual vocalizations to an assumed group, to support later data analysis. All assistants had been a member of the Sanje Mangabey Project team before this study and therefore were well trained and reliable in identifying mangabey calls.
The method followed the assumptions of a point transect survey (Buckland et al., 2001). Individuals were detected with certainty at the posts and at the initial location of the call as observers were stationary which ensured groups would not be disturbed and therefore measured at their initial location. The assumption that measurements were exact was not met as distance to calls were estimates by the observers and the variation in terrain and loudness of calls may have affected the perceived distance by observers of each call. Groups could not be located by observers during the survey to validate distances as groups flee quickly if disturbed, making it difficult to locate groups at the original location and risked disturbing other groups. Before the study, observers underwent training whilst studying a habituated group to estimate distances and bearings of calls to minimize possible interobserver differences. The assumption that surveys were positioned at random was violated as listening posts were positioned on nearby ridges and vantage points which may have deviated from randomly assigned points.

| Estimating average group size
A mean average group size for the Sanje mangabey was calculated for each forest from focal follows of five groups (Mwanihana: n = 2; Uzungwa Scarp: n = 3) found opportunistically when in the field outside survey times, and from known average group sizes of an additional three habituated groups in Mwanihana.

| Estimating group density and population size
Vocalizations were plotted on a map in QGIS using the bearing and distance estimates recorded during the surveys. Call clusters were used to identify groups in a similar way to previous studies using indri vocalizations to identify distinct groups (Indri indri; Glessner & Britt, 2005;Pollock, 1986). Vocalizations that were within a 300 m distance of another call were assumed to be from the same group. If vocalizations were less than 30 min apart and more than 300 m apart, these were assumed to be separate groups (Figure 2b). If group definition was unclear (n = 13 out of 370 vocalizations) from the plotted vocalizations, notes from the field of assumed number of groups heard were used to attribute the individual calls to a group cluster.
F I G U R E 2 Diagrams of the acoustic distance sampling method used in this study: (a) The 3 × 1 array positioning of listening posts with observers (crosses) positioned 200 m apart, with the area of detection for each post (r = 1 km; shaded region), and (b) an illustration of an example of the call clustering method analysis and attribution of group identification to vocalizations. The time of the call is shown in brackets, dashed lines from posts show the posts that detected the call and the assumed group identification is shown by the color of lines. The two calls below the posts (red group) are assumed to be the same group as they are close in time and space; less than 30 min apart and less than 300 m apart. The call above the posts (blue group) is assumed to be a different group as it is over 300 m away and less than 30 min apart from the other calls Hermite polynomial adjustments, hazard-rate key with polynomial adjustments and uniform key with cosine adjustments, and the best model was selected using Akaike's information criteria (AIC).
The difference between the group density estimates for each forest was measured using a Student's t test, and the difference in group size between forests was measured using a Mann-Whitney U test. All summary statistics were calculated in R (R Core Team, 2018).

| Ethics statement
This study did not capture or handle animals and was in adherence to individuals; n = 5), however, the difference was not significant (Table 2).
All detection function models fitted well with the data (ΔAIC < 2) and did not differ significantly in abundance estimations. The model using a uniform key with cosine adjustment was selected as the best fitting detection function model (ΔAIC = 0; goodness of fit: p = .46; Figure 3). Group density was estimated to be significantly higher in found. Therefore, the estimated total number of groups for the Sanje mangabey was 89.6 (95% CI: 60.8-131.9) groups and estimated population size a total of 3,167 (95% CI: 2,181-4,596) individuals (Table 3).

| DISCUSSION
The population size estimates in this study are in concordance with previous predictions by Rovero et al. (2009), but larger than previous surveys of the Sanje mangabey due to the larger average group size and habitat size sampled in the current study (Dinesen et al., 2001;Ehardt, 2001;Ehardt et al., 2005;  to 5,536 individuals (Table 3). This would suggest a possible decline; however, due to inaccuracies previously discussed of earlier population size estimates, it is not possible to definitively infer a temporal change.
Therefore, this study provides the first inferential estimate to allow future surveys to detect and estimate population trends. report a decline for several primate species (Rovero et al., 2012(Rovero et al., , 2015. In surveys conducted between 2002 and 2012, Rovero et al. (2015) found that populations of the arboreal Udzungwa red colobus  Ehardt et al. (2005) are reported as the original data presented in the study and as adjusted estimates (new values italicized) where the group densities from the original calculations have been used to calculate population size with the higher group size and habitat size estimates found and used in this study. disturbance in this time period through hunting and pole cutting, both likely to also impact the semi-terrestrial Sanje mangabey.
Group density in Mwanihana was significantly higher than that found in the Uzungwa Scarp, with the lower density found in the forest that presently and historically has had a considerably lower protection status and level of law enforcement. When using camera traps and occupancy modelling, which is likely an efficient method for the shy, semi-terrestrial mangabey, Hegerl et al. (2017) found Sanje mangabey occupancy in the Uzungwa Scarp was only a quarter of that found in Mwanihana. This difference, as with the difference in group density in this study, suggests that threats to other primates in the Uzungwa Scarp are likely also affecting the Sanje mangabey. Further, findings in this study reflect previous work examining group density for three arboreal primates in the Udzungwa Mountains: the Udzungwa red colobus, Angolan colobus and Sykes' monkey (Cercopithecus mitis monoides/moloneyi). Across Mwanihana, Uzungwa Scarp, and two other forests, group density of all three species was found to be lowest in the Uzungwa Scarp, which was attributed to the lack of active protection (Araldi, Barelli, Hodges, & Rovero, 2014). Lower densities have often been found for primates living in disturbed habitats compared to those in less disturbed regions due to factors such as reduced biomass, shelter, canopy cover and food availability (Phoonjampa et al., 2011). A study by Phoonjampa et al. (2011) of pileated gibbons (Hylobates pileatus) found group density was significantly associated with habitat disturbance, with higher densities found in forests that had been issued formal protection for longer than those that were more recently elevated.
While both the National Park and Nature Reserve were originally protected by Forest Reserve status, these regulations were weak and often poorly enforced. Mwanihana's protection was upgraded in 1992 when it was included within the Udzungwa Mountains National Park boundary; however, the Uzungwa Scarp was only upgraded to Nature Reserve protection in 2016, which strengthened regulations and management, but did not lead to active patrols or greater law enforcement on the ground. Human disturbance has increased in the Uzungwa Scarp since 2007 (Rovero, Mtui, Kitegile, Nielsen, & Jones, 2010) and the declining encounter rate for the mangabeys has previously been attributed to this escalation in encroachment (Rovero et al., 2012). A recent long-term study of the impact of protected areas in the Udzungwa Mountains found both species richness and encounter rates for the most commonly encountered medium to large-bodied mammals increased with level of protection status (Jones et al., 2019), which further supports the difference in density found in this study for the mangabey.
The acoustic survey method used in this study addressed previous issues from line transect surveys as it did not rely on visual observations and did not disturb the mangabeys that are shy and quick to move away.
Therefore, the estimates from this method are likely to be a more accurate representation of the current population size and future surveys of this species should include this approach. Anecdotal observations from the long-term study of the habituated group suggest that it is rare for the groups to not vocalize in the morning (G. McCabe pers. obs.); however, the method in this study could be adapted to bolster estimates by surveying the same location over multiple days to increase detection likelihood. Extrapolating average group density to the full extent of the forest assumed that groups were evenly distributed which may be unlikely given the wide elevation gradient and habitat heterogeneity of both forests. The survey posts were positioned at random and were successful in achieving a mostly full coverage of the forest extent, however, future studies should aim to cover the full extent of each forest and aim to determine whether a difference in group density is found in different habitat types, accounting for possible uneven distribution of groups across forests when estimating population size. Responses to food abundance, quality of forest, habitat structure and proximity to recent human disturbance have been found to influence group density in other studies of primates (Agetsuma, Koda, Tsujino, & Agetsuma-Yanagihara, 2015 (Mbora, Wieczkowski, & Munene, 2009). This was suggested to be attributable to increased parasite prevalence and/or increased competition for food in degraded forest resulting in lower fecundity and increased fitness costs, which may be also applicable in the Sanje mangabey subpopulation in the Uzungwa Scarp with further study.
The Sanje mangabey has shown behavioral and dietary flexibility in its ability to adapt to the use of both primary and secondary forest (Ehardt et al., 2005;McCabe et al., 2013), which suggests continued and improved protection of the forests to continue the recovery of currently unsuitable degraded habitat to usable secondary forest may encourage an increase in group density. This has been seen in conservation projects aimed at the San Martin titi monkey (Plecturocebus oenanthe), for example, where regeneration of forest by increased protection and active reforestation increased group density (Allgas et al., 2017). Similarly, increased tree density due to active forest protection led to increased group density for the graycheeked mangabey (Lophocebus albigena) in the Kibale Forest Reserve, Uganda (Olupot, Chapman, Brown, & Waser, 1994).
This study has provided the first inferential estimate of the Sanje mangabey population size which was essential due to previous estimates being considered inaccurate and the last direct survey being conducted over 15 years before this study (Ehardt et al., 2005). It is key to the survival and protection of species to monitor any changes in the population and the responses to changes in their environment, by natural disaster or anthropogenic disturbances. Populations can be slow to respond to such changes; therefore, long-term and regular monitoring PADDOCK ET AL. | 7 of 9 can provide an insight into population trends. Recently, Newmark and McNeally (2018) described the predicted "sizable" extinction debt due to the fragmentation of forests within the Eastern Arc Mountains, including forests of the Udzungwa Mountains, and the threat to the survival of species within these biodiversity hotspots. Considering this for the Sanje mangabey, we recommend continuing regular population surveys with the acoustic method described here, adapted following recommendations, to regularly monitor the population and to use the results from this study as the baseline population size estimates. The isolation of the two forests preventing migration of individuals and recovery of a population, and the lower group density found in the Uzungwa Scarp, underlines the need for increased protection and active enforcement in this region.
Continued active protection of the National Park is required for maintaining the population and potentially aiding an increased group density as highly degraded habitats recover. Active protection of the Uzungwa Scarp needs to be established to prevent the continued impact of hunting and habitat degradation and declining trend in primate populations in the region.

ACKNOWLEDGMENTS
Grateful acknowledgments go to Tanzania National Parks (TANAPA),