Use of cultivated foods and matrix habitat by Bale monkeys in forest fragments: Assessing local human attitudes and perceptions

Primates inhabiting human‐modified landscapes often exploit matrix habitat to supplement their diet with cultivated foods, at times resulting in economic losses and conflict with local people. Understanding human‐nonhuman primate interactions and the attitudes and perceptions of local people towards crop feeding species are crucial to designing effective species‐based management plans. Over a 12‐month period, we used scan sampling to study the consumption of cultivated foods and matrix use patterns by two habituated groups of Bale monkeys (Chlorocebus djamdjamensis), Ethiopian‐endemic bamboo specialists, in two forest fragments (Kokosa and Afursa) set amidst human settlements and farmland in the southern Ethiopian Highlands. Further, we conducted interviews with local people to document their attitudes and perceptions towards Bale monkeys at the two sites. We found that Bale monkeys at Kokosa, a more degraded habitat by most measures, consumed significantly more cultivated foods than their counterparts at Afursa. Moreover, Bale monkeys at Kokosa spent significantly more time in the matrix than in the forest habitat, while monkeys at Afursa spent significantly less time in the matrix than in the forest habitat. Not surprisingly, local people displayed a more negative attitude towards monkeys inhabiting Kokosa than those inhabiting Afursa. The differences in Bale monkey cultivated food consumption and matrix use patterns—as well as in local people's attitudes and perceptions towards Bale monkeys—between Kokosa and Afursa are probably associated with differences in habitat structure, degree of habitat alteration, and land‐use practices between the sites. We conclude that to ensure long‐term coexistence between Bale monkeys and local people in human‐modified landscapes, it is vital to incorporate nearby matrix habitats into management plans and to work closely with local communities to develop effective nonlethal crop protection strategies, thereby reducing the likelihood of negative interactions between Bale monkeys and humans.

In areas where local people depend primarily on subsistence farming, competition between crop feeding primates and the local community can be especially severe (Webber & Hill, 2014). In these areas, local people often view crop feeding species negatively and sometimes pursue retaliatory tactics like hunting these species, further jeopardizing primates already at risk of extinction due to habitat destruction (Chapman et al., 2016;Hill, 1997;McLennan, Hyeroba, Asiimwe, Reynolds, & Walls, 2012;Meijaard et al., 2011). Studying the relationship between humans and nonhuman primates in areas where they are at odds is essential for designing management plans that will ensure long-term coexistence between crop feeding species and local people (Chaves & Bicca-Marques, 2017;Hill & Webber, 2010;Lee & Priston, 2005;Spagnoletti, Cardoso, Fragaszy, & Izar, 2017).
However, the species is also known to persist in small and isolated forest fragments where bamboo populations have been degraded or nearly eradicated (Mekonnen et al., 2012). To persist in these fragments, the monkeys consume more diverse diets than in continuous forest (Mekonnen et al., 2012;. Our recent study found that habitat fragmentation and degradation significantly reduced food availability and habitat quality for Bale monkeys (Mekonnen et al., 2017;. Such changes in habitat structure and food availability, along with associated changes in diet  and terrestrial behavior (Mekonnen, Fashing, Sargis, et al., 2018), might be expected to affect the consumption of cultivated foods and associated matrix use patterns by Bale monkeys.
Like many other crop feeding African primates (Hill, 2018;McLennan et al., 2012), Bale monkeys face conflict with local people and are sometimes hunted in response to their crop feeding behavior (Mekonnen et al., 2012;. In fact, several Bale monkey populations are believed to have been extirpated in recent decades (Mekonnen et al., 2012). Despite these concerns, the interactions between humans and Bale monkeys have never been studied in detail. During a 12-month period, we assessed the crop-feeding behavior of Bale monkeys in two different forest fragments and explored how this behavior influenced the relationships between local humans and Bale monkeys at these sites. Specifically, we (a) evaluated the contribution of cultivated foods to the diets of Bale monkeys, (b) evaluated whether crop consumption by Bale monkeys is influenced by natural food availability, (c) examined the matrix use patterns of Bale monkeys, and (d) assessed the attitudes and perceptions of local people towards Bale monkeys. Because of differences in habitat structure, composition, and food availability between the two fragments in our study, Kokosa (or Patchy, the more degraded fragment) and Afursa (or Hilltop, the less degraded fragment) (Mekonnen et al., 2017;, we predicted that (a) Bale monkeys in Patchy would consume a greater percentage of cultivated foods than those in Hilltop, (b) the mean monthly consumption of cultivated foods by Bale monkeys in both groups would be inversely correlated with the overall availability of natural foods in the fragments, (c) Bale monkeys in Patchy would spend a greater proportion of their time in matrix habitat than those in Hilltop and (d) the attitudes and perceptions of local people towards Bale monkeys would be more negative near the fragment where the greatest crop consumption was occurring.

| Study areas and fragments
We conducted our study on Bale monkey crop feeding and matrix use in and around two forest fragments (6°44′-06°45′N and 38°48′-38°51′E), Kokosa and Afursa, in the southern Ethiopian Highlands (Mekonnen et al., 2017). Our research on human-Bale monkey interactions was conducted in the villages adjacent to Kokosa and Afursa forest fragments.
Kokosa forest fragment (hereafter referred to as "Patchy" fragment) consists of trees, shrubs, lianas, and bamboo set amidst a larger matrix of cultivated land, shrubland, human settlement, and grazing land. In total, Patchy covers 162 ha at elevations ranging from 2,534 to 2,780 m asl. Most of the fragment is owned privately by local individuals, though a smaller portion is collectively owned by the local community (Mekonnen et al., 2017). Cutting of bamboo by local people is common in the fragment and remaining stands are dwindling, though the site was dominated by bamboo forest just three decades ago (Mekonnen et al., 2012). The bamboo in the fragment today is a mix of naturally occurring stands and stands planted more recently by local people in their yards.
Located only 9 km from Kokosa, Afursa forest fragment (hereafter referred to as "Hilltop" fragment) is centered on a hilltop and consists of a mix of trees, shrubs, lianas, and a Eucalyptus plantation with graminoid and forb cover underneath. Though abundant at Hilltop 30 years ago, bamboo has been nearly eradicated through intensive harvesting. Hilltop covers 34 ha at elevations ranging from 2,582 to 2,790 m asl and is surrounded by an anthropogenic matrix of cultivated lands, pastures, and human settlements. Currently, cutting trees and using the fragment for grazing are prohibited. However, the edge of the fragment, especially the ground cover underneath the Eucalyptus plantation, is used for grazing (Mekonnen et al., 2017).
The main source of income for most people living near Patchy and Hilltop is agriculture, consisting of both planted crops and livestock farming. The rainfall and temperature patterns are also similar between the two fragments due to their geographic proximity, an occurrence at similar elevations, and orientation in the same north-south and east-west directions . Annual rainfall averages 1676 mm and follows a bimodal pattern with a long wet season and a short dry season . Mean annual temperature averages 16.7°C . The only primates sympatric with Bale monkeys in each fragment are black-and-white colobus monkeys (Colobus guereza; Mekonnen et al., 2012), though other large mammals present include porcupines (Hystrix cristata), bushbuck (Tragelaphus scriptus meneliki), dik-diks (Madoqua saltiana), and mole rats (Tachyoryctes splendens).
There are several notable differences between the two fragments. First, Hilltop is surrounded by a busy road and denser human settlement than Patchy (A. Mekonnen, pers. obs.). Second, Hilltop is nearly five times smaller and contains fewer different habitat types, including much less cultivated land than Patchy (Mekonnen et al., 2017 (Mekonnen et al., 2017). On the other hand, though bamboo density in both fragments is far lower than in intact continuous bamboo forest, the density of bamboo is 35× higher at Patchy than at Hilltop (Mekonnen et al., 2017). Further details about the characteristics of the two fragments can be found in Table 1 of Mekonnen et al. (2017).
Further details about the study areas, study groups, and home range characteristics can be found in Table 1

| Behavioral observation
We collected data on the consumption of cultivated foods and matrix use patterns from July 2013 to June 2014 via instantaneous scan sampling at 15 min intervals for up to 5 min duration on up to 5 visible individuals excluding infants, from 0630 to 1830 (Mekonnen et al., 2017;. During scan sampling, when a monkey was observed feeding, we recorded the type of food item and the species to which it belonged as well as whether it was from a natural or cultivated food source. In instances where Bale monkeys consumed bamboo, we distinguished between wild bamboo (natural) and bamboo obtained from the backyards of farmers (cultivated).
To evaluate temporal changes in the availability of potential natural food resources, we carried out monthly phenological assessments over an annual cycle for eight plant food species selected based on an earlier 8-month study of the species' diet in continuous forest (Mekonnen, Bekele, Fashing, et al., 2010). We monitored 10-15 individuals of each plant species on a monthly basis, including trees (≥10 cm DBH), bamboo (Arundinaria alpina) and shrubs. For each plant, we recorded the relative abundance score (ranging from 0 [item absent] to 8 [fully laden with the item]) for each of its potential food items (young leaves, mature leaves, flowers, ripe fruits, and shoots) via visual inspection as well as using binoculars where necessary (Mekonnen et al., 2017;. We analyzed phenological data from eight food species that cumulatively accounted for 50.9% and 44.5% of the annual diets of the study groups at Patchy and Hilltop, respectively (Mekonnen et al., 2017;. The somewhat low contribution of monitored plants to the diets of the study groups resulted from these groups consuming much less bamboo as well as a greater variety of food species-including insects, graminoids and forbs, which are difficult to monitor (Fashing, Nguyen, Venkataraman, & Kerby, 2014)-than our original study group in continuous forest, which ate primarily bamboo plus several other plant species . We calculated the monthly food availability index (FAI) for each plant part by multiplying the mean phenology scores of species i with the mean basal area of species i and density of the corresponding species i per ha (Mekonnen et al., 2017;. At the same 15 min intervals when scan sampling for diet was conducted, we also recorded the GPS location (Garmin GPSMap 62 s with a precision of ±3 m) representing the "center of mass" (Fashing, 2001) of most of the group members and the corresponding habitat occupied: "forest" or "matrix" (Mekonnen et al., 2017). Matrix habitat was defined as a habitat that predominantly served as a human use area either for grazing land, human settlement, tree plantation or cultivated land. Forest habitat was defined as a habitat dominated by natural forest habitat with little human use including habitats other than grazing land, human settlement, tree plantation, or cultivated land (Mekonnen et al., 2017).

| Ethnoprimatological methods
We conducted semistructured interviews with local people in the communities near our two study sites. We interviewed 40 people near Hilltop and 65 people near Patchy (Table S1). In each community, we randomly sampled households and carried out an interview with one adult per household (Gavin & Anderson, 2007;Gollin, McMillen, & Wilcox, 2004). We completed the interviews in the local languages of the interviewees (Sidamigna or Oromifa). showed each interviewee a photo of a Bale monkey and a sympatric black and white colobus monkey which are phenotypically very different, minimizing the risk of potential confusion. Local people also differentiate between them by their local names, a distinction that exists in both local languages. Next, we assessed the attitudes and perceptions of local people towards Bale monkeys, including whether they (a) knew Bale monkeys are endemic to Ethiopia, (b) were aware that Bale monkeys are legally protected, (c) considered it necessary to conserve Bale monkeys, (d) had ever reported crop feeding by Bale monkeys to government authorities, and (e) had one of the following attitudes toward Bale monkeys: "like," "dislike" or "indifferent." Finally, the participants were asked: (a) the types of crops they grow on their farms, (b) to list, based on their experience, the top three crop feeding wildlife species in decreasing order of crop consumption, (c) whether Bale monkeys consume cultivated species and, if so, (d) which species, (e) the protection measures employed by local people to prevent crop loss by Bale monkeys and their perceived effectiveness ("highly effective," "less effective," or "not effective"), and (f) their reactions ("agree," "disagree," or "indifferent") to seven possible protection measures to mitigate adverse human-Bale monkey interactions, and if they had any additional measures to recommend.
The mean time taken to complete an interview was 20.5 min ± standard deviation (SD) 8.83 (range 8-33). The mean age of interviewees was 37.6 years ± SD 12.23 (range 20-75). Most of the participants were males (84.8%; Table S1). After checking that responses by males and females followed similar patterns, we combined the data for both sexes in our analyses. Most participants were farmers that engage in both agriculture and livestock farming (Table S1). The two sites, however, differ in both ethnicity and religion: Afursa is inhabited mostly by Christian Sidama people and Kokosa is inhabited mostly by Muslim Oromo people (Table S1).

| Ethical statement
We obtained informed voluntary consent from all participants before their inclusion in the interview survey. Permission to carry out this project was granted by the Ethiopian Wildlife Conservation Authority. This project also adhered to the legal requirements of Ethiopia and complied with the American Society of Primatologists Principles for the Ethical Treatment of Primates.

| Data analysis
We tested all data for normality and homogeneity of variances using Shapiro-Wilk and Levene tests, respectively (p > .05). To normalize the data, we carried out logit transformations of proportion data and log transformation of food availability data before conducting statistical analysis (Warton & Hui, 2011).
We assessed cultivated food consumption for each group by determining the proportion of feeding scans accounted for by cultivated food species. We compared the consumption of cultivated food species between groups using an independent Student's t-test.
We used linear regressions to evaluate whether monthly values for overall food availability were a good predictor of the monthly proportion of cultivated food species consumption.
To assess the matrix use patterns, we classified the home range (95% Kernel Density Estimation, KDE) of Bale monkeys (Mekonnen et al., 2017) into matrix and forest using the corresponding GPS locations collected during group follows in ArcGIS 10.3 (ESRI, 2016).
We then calculated the areas of both matrix and forest habitats in ArcGIS. To determine whether Bale monkeys avoid matrix habitat or not, we calculated matrix use ratios by dividing the frequency of observed use (% of range scans in matrix vs. forest) by the frequency of expected use (% of home range area accounted for by matrix vs. forest) (Krebs, 1999;Manly, McDonald, Thomas, McDonald, & Erickson, 2002). We calculated the observed use of matrix and forest by dividing the total number of GPS points recorded in each category with the total number of GPS points recorded in the home range. To calculate the expected use of matrix and forest habitats, we multiplied the percentage representation of matrix and forest areas with the total number of range points recorded within the home range. A selection ratio close to 1 indicates no selectivity for that habitat, <1 indicates a habitat is avoided and >1 indicates a habitat is selected (Krebs, 1999). We also used χ 2 goodness-of-fit tests to compare the observed versus the expected frequencies of matrix and forest utilized by each group in their home range.
Finally, we used descriptive statistics to examine differences between the human communities near Patchy and Hilltop in demographic and socioeconomic variables as well as their attitudes and perceptions towards Bale monkeys. We conducted all statistical tests using R version 3.5.2 (R Development Core Team, 2018) with significance set at p ≤ .05.

| Consumption of cultivated food species by Bale monkeys
Bale monkeys in forest fragments cumulatively exploited five cultivated food species including bamboo planted on farms near the fragments (Table 1; Figure 1). Patchy group consumed all five of these species whereas Hilltop group consumed only two of them (Table 1). The percentage of the overall annual diet accounted for by cultivated species was also much greater in Patchy group (10.3%) than in Hilltop group (0.2%). Further, the percentage of the overall diet accounted for by cultivated food species each month was significantly higher for Patchy group than for Hilltop group (Student's t-test: p < .001; n = 12 months).
Monthly percentage consumption of cultivated foods varied widely in Patchy group (1.7-19.8%) but never exceeded 1.1% in

| Matrix effect
The available matrix habitat was much greater for Patchy group, accounting for 73.5% of its home range, than for Hilltop group, whose range included only 26.9% matrix habitat (Table 2; Figure S2). Bale monkeys in Hilltop group also used the matrix habitat less frequently than expected by chance based on its availability in their home range (χ 2 = 63.79; df = 1; p < .001; Table 2). In contrast, Bale monkeys in Patchy group used matrix habitat more frequently than expected by chance based on its availability in their home range (χ 2 = 24.45; df = 1; p < .001; Table 2).
T A B L E 1 The proportion of annual feeding records (Patchy, n = 5239 records; Hilltop, n = 3950 records) devoted to each cultivated food item by Bale monkey study groups at Patchy (Kokosa) and Hilltop (Afursa) fragments

| Human-Bale monkey interaction management practices and strategies
Local people reported using several strategies to try to reduce the consumption of cultivated foods by Bale monkeys at both study sites (  , only those at Kokosa (or Patchy) fragment intensively exploited cultivated foods and relied much more heavily on matrix habitats than those at Afursa (or Hilltop) fragment.
Correspondingly, local human attitudes and perceptions toward Bale monkeys were much more negative near Patchy than near Hilltop.
Here, we discuss the patterns revealed by our study in greater detail and place them within the context of the literature on primate crop feeding behavior and human-nonhuman primate interactions.
Crop feeding behavior is a common response by some primates to the reduction of their preferred wild food resources, often resulting from human disturbances. For instance, populations of several primates have been found to spend more time feeding on crops when wild food availability is low ( still quite scarce, especially when compared with its typical abundance in continuous forest (Mekonnen et al., 2017). As a result, we contend that Bale monkeys in Patchy are consuming a much greater (53-fold difference) proportion of cultivated foods than those at Hilltop, at least partly because of the lower overall food availability and habitat quality in Patchy.
Several studies have found that it is not reduced natural food availability but rather the spatiotemporal availability of crops that predicts crop feeding by primates (Cancelliere et al., 2018;Hill, 2017 [Gabriel, 2013;Kelley, 2013], Colobus vellerosus [Wong & Sicotte, 2007] and Cebus capucinus [McKinney, 2011]). Furthermore, highly fragmented and irregular shaped home ranges often contain more edge than forest interior habitat, thus forcing species to use matrix habitats more frequently (Fahrig, 2003). Indeed, in our study, the group that used matrix habitat most (Patchy) occupied a more irregularly shaped home range with lots of edge habitat. Hilltop after dark when we were not observing them. We also acknowledge that people's individual experiences, societal values, and cultural norms can contribute to their perception of crop feeding by primates and other animals (Hill, 2018).
Local people near both our study sites reported that guarding was the most commonly used and effective strategy to protect their farmland (Table 4a). Though guarding can be an effective crop protection strategy, it requires human power and this task is primarily carried out by children in many countries (including Ethiopia), coming at the expense of their time at school (e.g., Mackenzie, Sengupta, & Kaoser, 2015;A. Mekonnen, pers. obs.). In general, our finding that most people at both study sites are recommending and implementing nonlethal crop protection strategies is a good sign for Bale monkeys in these forest fragments.
However, some individuals near Patchy stated that lethal or potentially lethal strategies such as culling problematic animals (11%), throwing stones (20%), and patrolling with dogs (59%) are "highly effective." Because Bale monkeys are protected by Ethiopian law as a Threatened species (Butynski, Gippoliti, Kingdon, & De Jong, 2008;, all of these strategies are illegal. If all other possible nonlethal protection methods do not succeed, translocation from fragments where Bale monkeys are engaging in problematic levels of crop feeding to less disturbed habitats was supported by some locals, especially near Patchy fragment. However, in practice, translocation is expensive and it would be a tremendous challenge to find remaining suitable but unoccupied habitats for Bale monkeys to move to (Mekonnen et al., 2012). Similar constraints have limited the utility of translocation for conserving other crop feeding primates as well (Osborn & Hill, 2005).
We suggest that to ensure the long-term survival of Bale monkeys in forest fragments, it will be necessary to incorporate nearby matrix habitats into Bale monkey management plans, and to work closely with local communities to help improve their livelihoods, reduce crop losses to Bale monkeys and other crop feeding species, and reduce logging of natural habitats (Garriga, Marco, Casas-Díaz, Amarasekaran, & Humle, 2018). Any restoration projects in degraded forest fragments must focus on mitigating adverse human-Bale monkey interactions, planting appropriate indigenous food plant species, protecting natural habitats, minimizing edge effects, increasing fragment sizes, and enhancing forest fragment connectivity (Anderson et al., 2007a;Estrada, Raboy, & Oliveira, 2012;Hill, 2017).
We hope that this study documenting the variability in Bale monkey crop feeding behavior in and around forest fragments, as well as the variation in human attitudes and perceptions towards Bale monkeys near these fragments, will contribute to future such conservation initiatives in the fragmented forests of southern Ethiopia.