Feeding ecology and diet of the southern geladas (Theropithecus gelada obscurus) in human‐modified landscape, Wollo, Ethiopia

Abstract Studying the dietary flexibility of primates that live in human‐modified environments is crucial for understanding their ecological adaptations as well as developing management and conservation plans. Southern gelada (Theropithecus gelada obscurus) is an endemic little‐known subspecies of gelada that inhabits human‐modified landscapes in the northern central highlands of Ethiopia. During an 18‐month period, we conducted this intensive study in an unprotected area of a human‐modified landscape at Kosheme in Wollo to investigate the feeding ecology of southern geladas and their dietary responses to seasonal variations. We quantified the monthly and seasonal diet data from a band of southern geladas using instantaneous scan sampling method at 15‐min intervals, and green grass phenology and availability using visual inspection from the randomly selected permanent plots. The overall average diet of southern geladas at Kosheme constituted grass blades 55.4%, grass undergrounds 13.2%, grass bulbs 5.6%, grass seeds 5.4%, herb leaves 4.0, fruits 7.3%, and cereal crops 5.6%. Grass blade consumption increased with increasing green grass availability, while underground food consumption increased with decreasing green grass availability, and vice versa. Southern geladas spent significantly more time feeding on the grass blades and herb leaves and significantly less time on bulbs during the wet season than the dry season. Underground grass items (rhizomes and corms) were not consumed during the wet season, but made up 22.3% of the dry season diet. Thus, although grass blades are staple diet items for geladas, underground diet items are important “fallback foods” at Kosheme. Our result shows insights into the dietary flexibility southern geladas adopt to cope with human‐modified landscapes of the north‐central Ethiopian Highlands. Thus, the study contributes to a better understanding of how changing environments shape primate ecology and evolution.

Different primate species show variable responses to human disturbance environments (Bryson-Morrison et al., 2017). Understanding the feeding response of primates to landscape changes has received attention by primatologists only in the last two decades (Arroyo-Rodríguez & Fahrig, 2014). Long-term ecological studies on primate populations that live in human-modified landscapes are crucial to improve our knowledge on the capacity of primates to adapt to habitat disturbances (Chapman & Peres, 2001;Corlett, 2011;Hill, 2017;Mekonnen et al., 2017;Struhsaker, 2008). It is also a necessary precursor to primate conservation programs (Marsh, 2003;Mekonnen et al., 2017).
In addition, to understand the dietary adaptability of primates to human disturbances, not only the contents of the species' diet, but also the seasonal variation in the diet should be examined. Primates may consume low-quality "fallback foods" to deal with temporal variation in food availability by using different foraging strategies (Jarvey et al., 2018;Marshall et al., 2009;Marshall & Wrangham, 2007).
Thus, as a result of seasonal variation in food availability, primates may switch consumptions from one diet item to another to optimize nutrient intake (Jarvey et al., 2018).
Seasonal fluctuations in environmental variables influence food availability of animal's diet (Chouteau, 2006). This in turn influences animal's diet choices. Dietary shifts typically correspond with seasonal resource scarcity (Hanya, 2004;Yiming, 2006). Rainfall is a major determinant of plant productivity, where seasonal patterns correspond with resource availability. Variation in food availability is one of the main factors determining seasonal variation in the diet of primates (Fashing et al., 2014;Hanya, 2004;Jarvey et al., 2018;McConkey et al., 2003). For example, primates inhabiting tropical forests often ingest mature leaves and unripe fruits during lean-seasons when preferred foods are scarce (Chapman & Rothman, 2009;Marshall et al., 2009). During the rainy season, when ripe fruit was scarce, chimpanzees relied heavily on piths and leaves (Basabose, 2002). Colobine species such as Francois' langurs (Trachypithecus francoisi) feed on more low-quality, subsistence foods, such as petioles and stems, when high-quality foods, such as young leaves, were scarce (Zhou et al., 2006). Geladas (Theropithecus gelada) in the Simien Mountains National Park spent considerable time consuming underground food items in the dry season (Hunter, 2001;Jarvey et al., 2018).
Southern geladas (Theropithecus gelada obscurus) are an endemic subspecies of gelada that live in human-modified landscape across the northern central highlands of Ethiopia. Geladas exist across a wide variety of habitat types and altitudinal ranges where sleeping cliffs are available with variable levels of habitat degradations and alterations. They inhabit near human settlement areas where agricultural activities are intense. As the result of habitat losses throughout their geographical ranges geladas currently occupy ~10% of their original habitat (Gippoliti, 2010). Competition from domestic livestock has forced the geladas to remain on the less productive gorge slopes (Abu et al., 2018;Kifle et al., 2013). Since the habitats of geladas are occupied by humans and their livestock, the availability of grazing pastures are decreasing from time to time. In addition, geladas raid cereal crops, resulting in potential conflict with local farmers, and they are continually harassed during crop growing months (Kifle & Bekele, 2020a). Thus, because of such ongoing expansion of subsistence farming, human settlement, competition for grazing pasture, and conflict with local farmers, they are vulnerable to future decline and local extinction (Bergman & Beehner, 2013).
In addition to habitat disturbances and degradations by human agricultural and grazing practices, seasonal variation in food availability also influences the feeding ecology of geladas (Fashing et al., 2014;Jarvey et al., 2018). So far, the feeding ecology of gelada populations in human-disturbed habitat in unprotected areas has not been intensively investigated. In addition, most of the previous studies of gelada feeding ecology have been carried out in Afroalpine protected ecosystems (Dunbar, 1977;Fashing et al., 2014;Hunter, 2001;Iwamoto, 1979;Jarvey et al., 2018;Woldegeorgis & Bekele, 2015).
However, only a few brief (lasting a few weeks to a few months each) studies of gelada feeding ecology have been conducted in humanmodified landscapes of unprotected areas (Abu et al., 2018;Dunbar & Dunbar, 1974;Kifle et al., 2013). Thus, little is known about gelada feeding ecology living in human-dominated habitats. Therefore, we carried out this long-term intensive behavioral study on the feeding ecology of southern geladas in a human-degraded unprotected Afromontane habitat, Wollo, north-central Ethiopia. Understanding the feeding ecology of primates in human-modified habitats will contribute for understanding how changing environments shape primate ecology and evolution (Jarvey et al., 2018) and their capacity to coexist in the long term with their human neighbors (Hill, 2017). Thus, by studying the diet of southern geladas in human-modified habitats, we can understand their ecological and behavioral adaptations in disturbed environment. The objectives of this study were (a) to provide feeding ecology data on a band of southern geladas inhabiting humanmodified habitat; (b) to examine seasonal variation in the diet composition of the specified primate; and (c) to investigate how the level of green grass food availability affect the diet choice of geladas.

| Study site and habitat characteristics
The study was carried out at Kosheme near Mekanselam town,

| Study population
We selected a band of southern gelada populations for detailed behavioral study. Since this study was the first at Kosheme, we habituated the band to human observers for 3 months (from February to April 2015) by following the band from dawn to dusk. We confirmed habituation of the selected band when all fleeing and defensive behaviors disappeared, and travel and feeding activities took place in a relaxed manner as well as when the band tolerated us at a distance of 5-10 m. The group size of the band was 37 individuals on average.

| Feeding ecology
During the 18-month period (May 2015-March 2017), we collected activity budget data from individuals using instantaneous scan sampling (Altmann, 1974). After habituation period, we collected the first 12 months behavioral data every month and the last 6 months data on bimonthly basis. We collected the behavioral data for 5 days in each month at 15-min intervals for up to 5 min duration by following the band during the daylight hours (Mekonnen et al., 2017). During each scan, we collected activity data from the first 5 random individuals (adults or juveniles but not from infants) in order of occurrence from left to right that avoid possible biases toward eye-catching activities, recording the first activity they engaged in that lasted ≥3 s (Fashing et al., 2014). We began behavioral data collection, when the band left the sleeping cliff and climbed to its top in the early morning (typically at 07:00 hr). Then, we followed the band throughout the daylight hours until individuals returned back to their sleeping cliff (typically at 18:00 hr). Monthly sampling effort was evenly distributed throughout the study period. At each scan interval, when an individual gelada was observed feeding, we recorded the type of diet item and growth form (Di Fiore, 2004;Fashing et al., 2014;Jarvey et al., 2018;Mekonnen et al., 2018). We defined feeding behavior as picking, grazing, handling, foraging, chewing, excavating, or manipulating any potential food items. If an individual gelada was handling multiple diet items during a particular scan, we considered the most abundant food item in the hand to be the diet item. We categorized the food items as grasses (blades, rhizomes, seeds, corms, or bulbs), herbs (leaves, flowers, roots), shrubs (flowers, fruits, or buds), trees (fruits, piths, seeds, or gums), invertebrates (ants, termites, or alates), or others (Dunbar, 1977;Fashing et al., 2014;Iwamoto, 1979) as well as crops (grains, seedlings, or vegetative).

| Monitoring temporal patterns of food availability
We monitored phenological data on the grass and herb greenness levels to evaluate patterns of temporal change in food availability at monthly intervals over the study period (Fashing et al., 2014;Hunter, 2001;Jarvey et al., 2018). We collected the phenological data from the randomly selected permanent plots (each 50 × 50 cm).
We constructed 22 plots within the home range of the study band to follow up the levels of greenness and desiccation of herbs and grasses using visual inspections. Depending upon their patterns of temporal greenness changes, we assigned a score for each plot from 0-3, where 0 = 0% (absence of green grass or/and herb within a particular plot), 1 = <33% (brown grass or/and herb with slight green of a particular plot), 2 = 31%-66% (light green grass or/and herb), and 3 = ≥67% (strictly green grass or/and herb) after detailed inspections. Monitoring of each plot was tightly correlated with behavioral observations of the study periods.

| Data analysis
We calculated the contribution of each food item to the total diet consumption using the proportion of the total number of feeding records spent on each diet type. We summed the daily food item consumptions within each month to construct monthly proportion of diet item composition. We then calculated mean seasonal and overall dietary composition by averaging the monthly percentage proportions. We used the following formula to calculate the availability of green grass and/or herb: monthly greenness level = (greenness score of a plot × n + …)/N, where n = number of plots that scored a particular greenness level; N = total number of plots examined per a month. For seasonal analysis, we included data from October to April as dry season and from May to September as wet season. Although May typically is known as a dry season in the area, we included both May 2015 and 2016 into the wet season as the result of unusually heavy rainfall during the study period.

| Statistical analysis
We conducted all statistical tests using SPSS version 20 software (IBM SPSS Inc., Chicago, USA). We used Shapiro-Wilk tests to test for normality and Levene tests to check homogeneity of variances (p > .05). We used a one-way ANOVA model to test for differences in the percentage consumption of each growth form and diet item between seasons. When the assumptions of parametric tests on the proportional food item did not meet, we performed arcsine square root transformations prior to statistical analysis to fulfill ANOVA model assumptions (Sokal & Rohlf, 1981). We tested the correlation between mean monthly green grass availability and the monthly percent feeding time of each diet category using a Pearson's rank correlation test. We set the level of statistical significance at p ≤ .05 for all analyses.

| Monthly diet
We recorded a total of 21,135 activity scans (10,471 recorded as "feeding" over 1,057 hr across 90 days. Feeding accounted for 49.5% of the band's overall activity time budget (Kifle & Afework in prep).
Southern geladas exhibited wide variability in their diet item consumptions across months (Table 1). Grass blades constituted the bulk for southern gelada's diet throughout the study months. The monthly proportion of consumption of grass blades ranged from 21.4% to 97.5%. The band also devoted its feeding time for grass rhizomes 0.0%-36.3%, grass seeds 0.0%-47.3%, and grass bulbs 0.0%-28.7% in monthly proportion of diet consumption. The band consumed considerable amount of food items from tree and bush parts, fruits ranging from 0.0% to 29.6% and flowers 0.0%-2.8% in the monthly proportion of consumption. In addition, southern geladas at Kosheme added a considerable amount of crop grains, seedlings, and its vegetative parts (0.0%-41.5%). Similarly, geladas ate invertebrates especially termite alates, each June when they appeared in mass shortly just on the onset of the rainy season.
Occasionally, the band also fed on other (limestone), possibly for mineral intake.

| Relationship between green grass availability level and diet item consumption
Monthly green grass/herb availability level was significantly positively correlated with monthly grass blade (r = 0.647, p = .004) and herb aboveground (r = 0.526, p = .025) consumption. However, monthly green grass/herb availability (N = 18) was significantly negatively correlated with monthly underground grass (r = −0.825,
From the overall total 8,332 grass diet records, blades comprised 69.6%, rhizomes 14.1%, seeds 6.8%, corms 2.4%, and bulbs 7.1% of the overall diet. Similarly, from the overall total of 542 herb diet records, leaves comprised 76.8%, flowers 17.5%, and roots 5.7% of the overall diet. From the total of 839 tree feeding records, fruits accounted for 88.0%, seeds 4.3%, piths 3.9%, and gums 3.8%. At Kosheme, fruits from trees contributed 98.1%, and 1.9% from shrubs for southern gelada diet. The fruit of Ficus sycomrous accounted for 84.8% from a total 738 tree fruit records and 83.2% from overall 752 fruit records. In addition, from the total of 112 feeding records of shrubs, flowers accounted for 47.3%, buds 40.2%, and fruits 12.5%.
Southern geladas at Kosheme also raided cereal crops.

| Seasonal variation in the feeding ecology of southern geladas
Considering the overall food source/growth form categories, geladas consumed more grass parts during the wet season than the dry season; however, the difference was not significant ( Considering the food item consumption categories, the diet of southern geladas varied markedly across the two seasons ( Figure 3).
The band shifted its diet item consumptions from more grass blades during the wet season to a much greater dependence on

| D ISCUSS I ON
Only a few studies have been carried out on the feeding ecology of geladas in unprotected areas of human-disturbed habitats, and these studies have been conducted for short duration that lasted a few weeks or few months (Abu et al., 2018;Dunbar & Dunbar, 1974;Iwamoto & Dunbar, 1983;Kifle et al., 2013). Thus, the feeding ecology and diet composition of geladas inhabiting unprotected landscapes remains largely unstudied. Therefore, this long-term intensive study was conducted to describe the feeding ecology and diets of southern geladas inhabiting a disturbed environment and to quantify seasonal differences in the diet along north-central highlands of Ethiopia.
At Kosheme, a human-disturbed unprotected afromontane habitat in north-central Ethiopia, we found that the overall diet of southern geladas consisted of 79.6% grass parts, 5.2% herb parts, 8.0% tree parts, and 1.1% shrub parts. In addition, the diet of southern geladas comprised 5.6% cereal crops and 0.5% invertebrates. We also found that the diet of southern geladas at Kosheme was diverse and showed wide variability across months and seasons. Grass blades and herb leave consumption correlated positively, and underground food items consumption correlated negatively, with monthly level of green grass/herb availability. Our result offers insights into the dietary strategies southern geladas adopt to cope with humanmodified landscapes of the north-central Ethiopian Highlands.
In general, previous studies showed that geladas primarily consume greater proportions of grass blades (Dunbar, 1977;Fashing et al., 2014;Hunter, 2001;Iwamoto, 1993;Jarvey et al., 2018;Woldegeorgis & Bekele, 2015). In line with this, the overall diet for geladas at Kosheme consisted of 55.4% grass blades followed by underground grass items (rhizomes, corms, and bulbs: 18.8%) and fruits (7.3%). However, the proportions of gelada diets are highly variable across study sites (  menu. In a similar finding, geladas at Kosheme also ranged through crop fields and fed on considerable amounts of grain (5.6%) at the sowing time, and fallen grain from the previous harvest. However, Fashing et al. (2014) and other studies in different sites showed that geladas did not include any grain from cereal crops.
These differences in gelada diet across study sites may be attributed to several factors. The length of studies (most previous studies were shorter in duration, ranging from 3 to 9 months, missing the full annual variation of ecological conditions of the year) and variations in the rainfall pattern, elevation, and human disturbance may influence dietary differences across studies. Study duration or seasonally biased sampling may account for differences on food item consumptions and compositions among monkey populations (Xiang et al., 2012). Studies also showed that human disturbance and land use severely affect primate diets (Fashing et al., 2014;Hunter, 2001;Irwin, 2008;Riley, 2007;Singh et al., 2001). Tesfaye et al. (2021)   . In addition, the availability of a particular diet item at a particular site is likely to be a primary contributor to the variability in the diets of geladas across study sites. For example, elevation and habitat composition in terms of vegetation types influence the diet of colobine monkeys (Bennett & Davies, 1994) and gray snub-nosed monkeys (Xiang et al., 2012).
In Ethiopia, habitats at relatively lower elevations (Afromontane ecosystem e.g., Kosheme) tend to be more floristically diverse and typically contain a wider range of food types available to geladas when compared to habitats at higher elevations (Afroalpine ecosystem e.g., Gich, Sankaber and Guassa).
Studies revealed that primates in disturbed forest habitats consumed less diverse diets than intact habitats (Riley, 2007;Tesfaye et al., 2013). However, geladas are unique among primates in that they permanently inhabit the grassland habitats of the Ethiopian  (Fashing et al., 2014), Gich (Woldegeorgis & Bekele, 2015), and Sankaber (Hunter, 2001;Jarvey et al., 2018). On the other hand, as the preferred food items decline in human disturbance habitats, primates feed more widely on less preferred foods species, which are easier to obtain in the area (Chaves et al., 2012;Dela, 2007;Dunn et al., 2012;Grassi, 2006).
Geladas, like other primates, are characterized by seasonal changes in diet throughout the year (Fashing et al., 2014;Jarvey et al., 2018). Our study also showed that southern geladas at Kosheme exhibited wide variability in diet item consumptions across months and seasons likely to reflecting temporal variations in particular food availability during those periods. Although geladas are predominant graminivore, they show dietary variability associated with seasonal declines in grass leaf availability (Dunbar, 1977;Hunter, 2001;Iwamoto, 1979;Jarvey et al., 2018). For example, studies on gelada populations at Guassa (Fashing et al., 2014) and in the Sankaber region of the Simien Mountains National Park (Jarvey et al., 2018), the time geladas spent consuming grass blades decreased when green grass blade availability was low and the time spent consuming underground foods increased. In line with this, we found that the diet of southern geladas at Kosheme was highly inclined on grass blades consumptions, which make up 71.7% of the consumed diet items during the wet season compared to 43.4% during the dry season when green grass blade availability became low.
At the beginning of the wet season, freshly growing grass blade consumptions by southern geladas increased sharply when grasses recovered rapidly from dryness due to the presence of rainfall.
Such higher consumption of grass blades might be due to their easy availabilities and higher nutritional qualities during the wet season compared to the dry season. Temporal variation in food availability is one of main factors determining seasonal variation in the diet of primates (Hanya, 2004;McConkey et al., 2003;Poulsen et al., 2001).
Grass blades consumption also varied among wet season months (maximum 97.5% and minimum 33.8%) at Kosheme and maximum 93.2% and minimum 80.2% at Sankaber (Jarvey et al., 2018). Such monthly variation in grass blade consumptions might be due to young grass blade availability at the beginning of wet season months, and maturity of grass blades and switch on grass seed consumption at the end of the wet season months.
Seasonal dietary shifts are a common feature of primates (Guo et al., 2007;Li, 2006;Xiang et al., 2007). Southern geladas at 8 Dunbar & Dunbar, 1974. 57.7% of the dry season months for southern geladas at Kosheme, Wollo, Ethiopia. Similar studies also showed that underground diet items comprised a larger part of the dry season diet than the wet season for geladas (e.g., Jarvey et al., 2018). Primate diets are severely affected by seasonal variation in preferred food availabilities (Irwin, 2008;Jarvey et al., 2018;Riley, 2007). Thus, when the availability of the preferred green grass blades decreased during the dry season, southern geladas at Kosheme increased underground food item consumption. Study suggested that underground foods are fallback foods for geladas (Fashing et al., 2014;Jarvey et al., 2018), and the consumption of underground food does not appear to be influenced by availability, but rather by the lack of preferred diet items (Altmann, 2009;Marshall & Wrangham, 2007). When green grass availability remains higher during the dry season, underground food items constitute less of the dry season diet (Jarvey et al., 2018). Iwamoto and Dunbar (1983) and Jarvey et al. (2018) at Sankaber, Zewdu et al. (2013) at Wonchit Valley, andFashing et al. (2014) at Guassa noted that the consumption of underground food items increased during the dry season and the consumption of such underground items exhibited significant negative correlations with rainfall (Fashing et al., 2014;Jarvey et al., 2018). The present study also suggests that the scarcity of green grass blades might be the reason for shifting the consumption of underground food items during the dry season when the rainfall reached minimum level. During the dry season, when the green grasses dry out, geladas can shift their foraging profile to digging for more subterranean food items (Hunter, 2001). Such subterranean plant parts like tubers provide an alternative source of energy in the form of carbohydrates (Byrne et al., 1993). Likewise, addition of underground diet items like bulbs and rhizomes by geladas for their diet menu may be good sources of carbohydrates, proteins, and other nutrients during the dry season.
Exploration of underground diet items represents an adaptation that allowed all the Theropithecus to tap subterranean storage food sources that are unavailable to the gelada's main competitors, namely wild ungulates (Dunbar & Bose, 1991). Similarly, in human-dominated landscapes, livestock and pack animals are the main competitors of geladas on the aboveground grass items at Kosheme (Kifle, personal observation). Geladas rely more heavily on underground foods in habitats more heavily influenced by humans (Fashing et al., 2014;Jarvey et al., 2018).
Higher consumption of fruits from trees and shrubs is another interesting finding of this study to understanding the dietary flexibility of southern geladas in human-disturbed habitats. During the dry season, the contribution of fruits for geladas diet at Kosheme was high. For example, they consumed fruits up to 29% in some months. This finding is incongruent with studies that argued geladas to be obligate graminivores (Dunbar, 1977;Dunbar & Dunbar, 1974;Iwamoto, 1993;Iwamoto & Dunbar, 1983;Jarvey et al., 2018). The main sources of fruit for geladas diet at Kosheme were Ficus sycomorus trees. Fruits from Rosa abyssinica are also good sources of diet item at Sankaber (Hunter, 2001). The fruit of Ficus species contains excess amount of sugar (Byrne et al., 1993) and are known to be keystone species that help sustain frugivores (Byrne et al., 1993).

| CON CLUS ION
Anthropogenic habitat disturbances can dramatically affect the quality, availability, and distribution of food resources and the addition of anthropogenic food sources into the diets with both positive and negative effects on the survival of primates (Higham et al., 2009;Hoffman & O' Riain, 2010;Mekonnen et al., 2018;Tesfaye et al., 2021). This study contributes to our understanding of the ecological and dietary flexibility of primates in human-modified environments, as well as conservation implication of such flexibility. In addition, determining how primate diets, such as that of geladas, shift in response to human disturbances is important to understanding how changing environments shape primate ecology and evolution (Hill, 2017;Jarvey et al., 2018).
The results of this study shows that geladas inhabit human-disturbed environment consumed more diverse diet items. The consumptions of such diversified diet items are the major dietary strategies of geladas to survive in human-modified landscapes. Similar study on the Bale monkeys (Chlorocebus djamdjamensis) in southern Ethiopia showed that those groups that inhabit fragmented landscapes exploited far more plant species to broadening their diet than conspecific in continuous forest . Southern geladas can respond to anthropogenic habitat disturbances by incorporating more alternative food items such as cereal crops, fruits, bulbs, roots, rhizomes, corms, piths, buds, and seeds. Such dietary flexibility in the consumptions of these food items is key factors for the survival strategies of geladas in human-modified environments.
Our results also indicate that southern geladas at Kosheme have relied on underground diet items during periods of food scarcity.

ACK N OWLED G M ENTS
This work was supported by the Rufford Small Grant Foundation, the Primate Conservation Incorporated, the IDEA WILD, the Bahir Dar University, and the Addis Ababa University. We thank the Amhara National Regional State Environment, Forest, Wildlife Protection, Development Authority Amhara National Regional State for permission to carry out this research. We additionally thank all of the local residents for their welcome and positive cooperation during the study period in the field. Lastly, we also acknowledge the two anonymous reviewers for their critical and helpful comments that greatly improved this manuscript.

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data of this study are available from the corresponding author upon reasonable request.