Trade-offs between benefits and costs of forest proximity: farmers' practices and strategies regarding tree–crop integration and ecosystem disservices management

The impact of ecosystem disservices is among the issues that farmers have to consider in management of livelihoods and local landscapes. We investigated distinct practices developed within local communities in tree–crop integration and strategies to offset disservices. Forty-eight transects (24 at ≤1 km and 24 at ≥3 km from forest edges) were laid in the study sites. Woody and crop species were recorded from a total of 150 homegardens and farm fields along the 48 transects. In addition, farmers (n = 384) were interviewed using a semistructured questionnaire to assess their land-use practices and management strategies to counter ecosystem disservices. Data were analyzed using a linear mixed effects model of the statistical program R. A total of 72 woody and crop species belonging to 40 families were recorded. The mean number of woody species increased near to the forest. Wild mammals, such as olive baboons, bush pigs, warthogs, vervet monkeys, and porcupines were common crop raiders. Farmers used fences, guarding, noise, scare devices, and smoke to scare away crop-raiding animals. To protect beehives in the forest fragments, they have developed indigenous skills such as dusting ashes, spraying indigenous repellant suspensions, and destroying the nests of raiding ants. A biological control mechanism was also used by farmers where they cut part of the nest of Crematogaster sp. (locally called “Penie”) and glue it onto the trunk of trees with beehives. Crematogaster sp. safeguard the beehives from raiding ants as part of their efforts to protect their own nests. We recommend both ecological and socioeconomic studies in order to augment farmers’ strategies to balance disservices and corresponding management practices across the landscapes.


INTRODUCTION
Forest patches and trees in mosaic agricultural landscapes provide many ecosystem services for smallholder farmers (Ricketts 2004, Millenium Ecosystem Assessment (MA) 2005, van Damme and Kindt 2012). Alongside of these provisions, ecosystem functions produce some disservices (defined here as "functions of an ecosystem resulting in unwanted and economically harmful effects on human well-being") (Escobedo et al. 2011, von Döhren and Haase 2015, Lyytimäki 2015. The concept of ecosystem disservices has been used in different studies as negatively impacting nuisances, such as pests (Lemessa et al. 2013, Ango et al. 2014) and food insecurity (Zhang et al. 2007), biological hazards, such as diseases (Chevalier et al. 2016), human and livestock predations (Vailshery et al. 2013), allergies caused by pollen, volatile organic compounds, and other pollutants (Escobedo et al. 2011), and geophysical hazards, such as floods (Dunn 2010), noise pollution (Basner et al. 2014), and heat stress and storms (Xu et al. 2016). The concept has also been used in studies of crop raiding by wild mammals and the effect of ants on honeybees negatively impacting the livelihoods of smallholder farmers, particularly those in close proximity to forests (Lemessa et al. 2013). Forest-dwelling wild mammals, such as baboons, monkeys, bush pigs, porcupines, chimpanzees, and elephants, have been identified as significant crop raiders causing substantial damage to cereals, root crops, and fruits in different regions of Africa (Seifu and Beyene 2014). Such crop loss poses a serious problem, resulting in food shortages for many farmer households living near to forests (Hill 2000, Fungo 2011). Furthermore, crop raiders caused substantial indirect costs, such as labor for guarding and negative effects on children's schooling (Kleijn et al. 2006, Ango et al. 2014, Byg et al. 2017).
Ants attack beehives and negatively affect the livelihoods of smallholder farmers. They are also the most important enemies of honeybees, causing serious problems for the honey production sector (Hansen and Klotz 2005, Petit and Burel 2006, Stamps and Groothuis 2010. Studies have shown that forests are preferred nesting and foraging sites for ants (Andersen et al. 2006, Philpott et al. 2006, Schultheiss and Nooten 2013 because forests have structurally complex habitats with abundant litter, higher amounts of coarse, dead woody debris, sufficient humidity, cold climate, high soil organic matter, and fewer disturbances (Chala et al. 2012, Desalegn 2015.
In response to these ecosystem disservices, farmers have invented various indigenous management strategies against crop raiders and ants that differ spatially in relation to damage types and intensities (Lemessa et al. 2013, Ango et al. 2014, von Döhren and Haase 2015. However, these management actions carried out to mitigate the impact of crop raiders may lead to habitat modifications, e.g., tree clearance, which in turn impact on other ecosystem processes and result in loss of biodiversity (Zahawi and Augspurger 2006, Zhang et al. 2007, van Damme and Kindt 2012, Swanepoel et al. 2017. Despite the influence of ecosystem disservices at close forest proximity on their agricultural production systems and livelihoods, farmers integrate trees into their croplands (Zahawi andAugspurger 2006, Ango et al. 2014, Mohammed and Asfaw https://www.ecologyandsociety.org/vol25/iss4/art36/ 2015) and they retain forests and woodlands. Tree-crop integration (defined here as "managing some multipurpose natural/planted tree species at different settings of their land plots together with introduced crops") is among the basic activities of farmers and needs thorough analysis to determine how it contributes to farmers' livelihoods in terms of domestic consumption and as a source of income (Harvey 2000, Tolera et al. 2008. Furthermore, farmers allow multipurpose trees to grow inside their croplands and manage them with their crops (Posada et al. 2000, Harvey et al. 2011). These trees, which grow from the soil seedbank of croplands, if located at required sites within the cultivated field, are protected by farmers and managed according to their traditional agroforestry practices (Perfecto et al. 1996, Schroth et al. 2004, Harvey et al. 2006, Ruelle 2014. Farmers account for yield loss in their decision for growing crops near to and far from forest sites. But they continue to grow crops that are regularly attacked by raiders because they either lack alternative crops less prone to attack or have developed different protection mechanisms (Lemessa et al. 2013, Mohammed andAsfaw 2015).
Various studies have shown that farmers managed trees in and around homesteads and on agricultural lands to obtain various benefits (Harvey 2000, Zahawi and Augspurger 2006, Tolera et al. 2008, such as live fences (Chaco 2003, Francesconi 2006, Pulido-Santacruz and Renjifo 2011, grazing lands (Posada et al. 2000, Harvey et al. 2006, 2011, and shade trees for coffee production (Perfecto et al. 1996, Schroth et al. 2004, Ango et al. 2014. However, these studies have not explicitly addressed the influence of forest proximity and farmers' knowledge and practices to balance ecosystem disservices and conservation. They ignore the aspects of the influence of forest proximity on farmers' crop production system and management strategies to overcome crop loss due to wildlife raiders that play key roles in balancing between ecosystem disservices and forest conservation. Our approach tried to address this question and provides alternative insights into farmers' knowledge and skills for tree-crop management and ecosystem disservices trade-offs at various forest proximities. Based on the reconnaissance survey of our study sites and earlier studies elsewhere (Hill 2000, Kleijn et al. 2006, van Damme and Kindt 2012, von Döhren and Haase 2015, Sinare et al. 2016, Byg et al. 2017, Swanepoel et al. 2017, we assumed that the major crop raiders in the area are vervet monkeys (Cercopithecus aethiops), warthogs (Phacochoerus africanus), bush pigs (Potamochoerus larvatus) and porcupines (Hystrix cristata). Several studies have shown that greater crop damage was found near to rather than far from the forest (e.g., Fungo 2011, Linkie et al. 2007, Lemessa et al. 2013, Ango et al. 2014. We also expected greater crop damage in close proximity to rather than far from forests in our study landscape. Although crop damage by wildlife substantially affects the livelihoods of a large number of smallholder farmers, their indigenous analytical management practice is still relatively poorly understood, and little is known about farmers' attitudes regarding the trade-off between ecosystem disservices and conservation. Thus, we hypothesize that: (1) There are significant differences in tree-crop integration and crop raiding for farmers living near to and far from the forests. As a result, we expect that the management strategies of local communities increase as distance to the forest decreases. (2) Farmers have indigenous strategies to avoid beehive raiding by ants. We tested these hypotheses by investigating tree-crop composition in 150 homegardens, farmlands, and farm boundaries (referred as geometric borders of cropped areas) at different distances from the forest edges in a pair-wise design. In addition, we conducted a semistructured questionnaire to investigate farmers' knowledge and practices in managing crop raiders and ant-mediated damages on beehives.

Study Area
The study was conducted in Gimbo district (Fig. 1), Kafa Zone, southwestern Ethiopia, which is located about 461 km southwest of Addis Ababa. It is situated between 07° 23'N to 07° 49' N latitude and 36° 00'E to 36° 47'E longitude with altitudinal range of 500 to 3300 m a.s.l. The study area is characterized by a variety of landforms resulting in a highly diverse climate, soil, and vegetation. Topographically, the area is characterized by a complex system of highlands, steep valleys, and large flatlands, which drop to the lowlands in the South. The moist afromontane forests of the area have diverse animal and plant assemblages (Friis et al. 2010). The black and white colobus monkey (Colobus guereza) is the main feature in the forests of the study area. Syzygium guineense, Olea welwitschii, Millettia ferruginea, Pouteria adolfi-friedericii, and Schefflera abyssinica are the dominant trees species (NABU 2017). The forests provide provisional ecosystem services, i.e., common spices such as Aframomum corrorima and Piper capense grow naturally across the landscape and are harvested by the local communities.

Sampling Design
Forests, farmlands, settlements, and other landscape features of our study area were identified using Google Earth (Ango et al. 2014) followed by a reconnaissance survey made from 2-6 September 2018. Subsequently, eight sites were identified; four sites (Baka, Beymo, Tula, and Zingaji) are far from (≥3 km) forest edges (Lemessa et al. 2013), and the other four sites (Kayikela, Shorori, Ufudo, and Wakaraba) are near to (≤1 km) the forests. Accordingly, we laid a total of 48 transects (six in each selected study site). The length of each transect line is from 2-3 km with 1 km width on each side of the transect. Subsequently, we selected 150 homegardens (defined here as a homestead area for growing and cultivating plants, although its composition varies considerably in size and crop diversity among farms), farmlands (referred as farm fields for plants cultivated and grown on a large scale), farm boundaries, and riparian areas, (75 at near to and 75 far from forest edges) along all 48 transects to collect data on woody and crop species diversity. All landscapes and homegardens were visited with the permission of local administrative bodies and homegarden owners.

Woody and Crop Species Data Collection
All woody and crop species were recorded by their vernacular and scientific names and identified in the field by using the Flora of Ethiopia and Eritrea (Hedberg and Edwards 1989, Edwards et al. 1995, Hedberg et al. 2004. Plant specimens were pressed and dried, and field identifications were confirmed at the National Herbarium of Ethiopia, Addis Ababa University. Authenticated specimens' replicates were deposited both in the National Herbarium of Ethiopia and in the Botany laboratory of Arba Minch University with their specific identification/reference number. In addition, a semistructured interview questionnaire was conducted focusing on tree and crop management.

Data Collection Regarding Farmers' Knowledge of Management of Ecosystem Disservices and Tree-Crop Integration
We conducted a semistructured questionnaire that was translated to the local language (Kafi nono) to explore data on farmers' practices of tree-crop integration, the most susceptible crop types, type of raiders attacking crops regularly, beehive raiding by ants, and management strategies of the communities to minimize crop loss. Accordingly, we sampled a total of 384 farmers (193 near to and 191 far from the forests) by following Kothari (2004). Key informant interviews and focus group discussions (FGD) were used to validate the collected data. The key informants (n = 10) were identified through snowballing and are mostly elder farmers. All of them have been recognized to be knowledgeable with regard to specific issues of land use and forest cover changes, tree-crop integration, types of wild animals affecting their livelihoods by raiding crops, and indigenous management strategies. Moreover, a total of seven FGD were conducted with 32 informants (women (n = 12) and men (n = 20)) in separate and mixed groups of seven to ten participants. Each group was composed of model farmers, farmers' representatives, development agents, district agricultural officers, and NABU experts, which were sampled systematically. The major issues discussed were the strategies about how farmers manage trees and crops and types of ecosystem disservices and management practices. We carried out repeated field walks with farmland owners to verify and collect additional data. The data were collected from September to November 2019.

Statistical Analyses
The variation in tree-crop integration among homegardens was analyzed by both the mean number of species per homegarden and the cumulative number of species for all transects near to and far from forest edges using a linear mixed effects model (the lmepackage) of the statistical program R (R Development Core Team 2010, Ango et al. 2014). As categorical variables, we used distance from the forest edges (near to and far from), the interaction between main factors, and altitude as fixed factors and transect pair (each pair consisting of one transect near to and one far from forests) as random factor (Lemessa et al. 2013). In addition, we tested if the species composition differed for all transects/ homegardens near to and far from forest edges with a multiresponse permutation procedure using Bray-Curtis dissimilarity index in the vegan-package of R (Oksanen et al. 2011). Crop raiding as a function of distance from the forest and ant-mediated impacts were analyzed with independent samples t-test. Similarly, we analyzed farmers' responses at near to and far from the forests with Pearson's Chi-square test. One-way ANOVA and Kruskal-Wallis tests were computed to see the variations at P < 0.005 and P < 0.001 with farmers' responses to crops and beehive raiding at different distances from the forests.

Ethical Issues
The study was approved by the Arba Minch University, Ethiopia. Written permissions for the study were obtained from Bureaus for Kafa Zone Administration, Gimbo District Administration, Environment Protection and Forestry, Agricultural and Natural Resources Development through a brief explanation of objective of the study. Verbal consent was obtained from the selected households prior to interviews.

Farmers' Strategies in Tree-Crop Integration
A total of 72 species of both woody plants and crops belonging to 40 families were identified (Appendix I). The mean number of species per farm plot was 31.04 (range, 1.3-83.7) near to and 27.3 (1.3-89.4) far from forest edges, and the difference was statistically significant at P < 0.001. Ensete ventricosum was the most frequent food crop grown across the landscape. Roots crops (e.g., Solanum tuberosum)) were mostly cultivated far from forests. Farmers noted that root crops were the most susceptible to crop raiders, and therefore, there is a tendency to cultivate less of these crops near to forests.
We found significant variations in mean numbers of tree species and food crops in the homegardens, inside farmlands, along the farm boundaries, and in riparian areas (defined here as a buffer zone not appropriate for crop cultivation) along the transects near to and far from forest edges (Fig. 2). The difference was statistically significant at P < 0.005 (Table 1). More perennial crops were recorded in the homegardens and farmlands far from forest edges. Perennial fruits, such as Citrus sinensis, Persea americana, and Mangifera indica, were less commonly cultivated in close proximity to the forests. In addition, more annual crops were found in the homegardens far from the forest edges (Fig. 2). Trees in the coffee agroecosystem of southwestern Ethiopia were categorized by farmers as either plantations (referred to, in this case, as cultivated woody plants in contrast to naturally grown trees) or naturally grown in the homegardens, farmlands, along farmland boundaries, and in riparian areas. The distribution pattern of trees varied significantly in these areas. We found a greater number of plantations in the homegardens and along farmland boundaries in study sites, and the differences were significant near to and far from the forests (Fig. 2). More plantations were recorded in the homegardens and along the farmland boundaries in Baka, Beymo, Tula, and Zingaji, whereas fewer plantations were found in the farmlands and riparian areas. Furthermore, naturally growing trees were abundant near to the forests (P < 0.001; Fig. 3A, B, C). Planted or naturally growing trees exhibited less variation near to and far from forests in riparian areas ( Fig. 3D; P < 0.07).
In all study sites, greater tree-crop integration was recorded in the homegardens and farmlands, but they were generally fewer in the riparian areas (Table 1). Results of the present study indicate that farmers used a number of criteria to integrate trees with crops. Among these, access to seedlings, farm size (Table 2), fast growth, multiple-use values (Table 3), and drought resistance were described as principal. In FGD and interviews, farmers pointed out that tree-crop integration trees such as Millettia ferruginea and Albizia gummifera were the most preferred as shade for Coffea arabica. The reasons they mentioned were that the leaves of these trees were thin, small, and light, having no impact on coffee trees when they fell off; these trees were chosen because they allow a sufficient amount of light to reach the crop. Farmers also integrated trees in farmlands with Eragrostis tef, Vicia faba, and Zea mays. As well, trees such as Cordia africana, Schefflera https://www.ecologyandsociety.org/vol25/iss4/art36/   Table 2. abyssinica, Ficus sycomorus, and fruit plants, e.g., Citrus sinensis, Persea americana, and Mangifera indica are integrated with Sorghum bicolor and Pisum sativum. From this study, we found that the knowledge of integrating multipurpose trees with crops has been acquired from parents, communities, or lifelong experiences (Table 4). Despite some discrepancies, the patterns of treeU-crop integration were generally similar in study sites near to and far from forest edges, and the difference was not statistically significant ( Fig. 4; P > 0.05).

Ecosystem Disservices and Management Practices
The results of this study indicate that crop raiding by large wild mammals was greater near to forest edges and was less prevalent far from the forests (Fig. 5A, Table 5). Olive baboons, bush pigs,  Table 2. warthogs, vervet monkeys, and porcupines were often listed by farmers as crop raiders, especially in the homegardens and farmlands near to forests; their effects are less common far from forests (Table 6).
There were differences on the effects of crop-raiding animals near to and far from forests. In FGD and interviews, most farmers (71.9%) near to the forest perceived the damage caused by bush pigs, baboons, and vervet monkeys as severe (a loss of more than a tenth of total yield). Some farmers (24.5%) living far from forest edges have reported similar crop damage, mainly by bush pigs and vervet monkeys. This difference is statistically significant (P < 0.001; Table 6).
When farmers were asked how to minimize the risk of crop damage by wild mammals, they listed several indigenous strategies such as guarding, smoking, shouting from different directions, chasing away the animals directly, or using dogs, fencing, and scare devices (e.g., scarecrows) in homegardens and at the edge of farmlands (Table 7). As well, children and women participate in chasing away the raiders from both homegarden and field crops Table 3. Results of preference ranking of trees based on multiple use-values exercised by key informants (KI) and focus group discussants (FGD). Scores in the table indicate ranks given to multipurpose trees based on their use-values (highest number (10) was given for the tree species which informants thought most-preferred in its use-values, and the lowest number (1) was given for its least-preference). The highest total scores were the most-preferred tree species by the farmers. Letters A-J represent KI and FGD.

Fig. 5. Farmers' responses regarding (A) crop raiding and (B)
ant-mediated impacts at specific forest proximities: a high degree of raiding in homegardens near to forest edges (Kayikela, Shorori, Ufudo, and Wakaraba) and relatively little raiding far from the forest edges (Baka, Beymo, Tula, and Zingaji). The statistical interpretations of the patterns are found in the text. Error bars are standard error. The sample size is found in Table 3.
during the day. Men usually patrol homegardens by shouting and using torches to protect their crops from bush pigs and porcupines. However, farmers explained that guarding their crops during the night had health effects such as cold and lack of sleep.
Farmers have developed different management strategies, such as dusting ash under trees with beehives, plastering the trees with plastic materials, finding and killing the queen ant, or destroying the ant nests to reduce ant attacks on beehives (Table 7). They also spray trees with suspensions of garlic and Eucalyptus leaves as a repellent. Another innovative strategy is the use of a biological control mechanism where they cut part of the nests of Crematogaster sp. and glue it to trees where beehives are located. These species attack beehive-raiding ants.
Beehive raiding by ants was greater in the forest and near to the forest than far from forests (Fig. 5B, Tables 5 and 6; P < 0.001). Moreover, independent samples t-test results showed that there was significant difference in the responses of farmers to the impacts of ants near to the forest (M = 2.27, SD = 0.624) and far from the forest (M = 2.34, SD = 0.657; t = 0.866, df =382, P = 0.002). The magnitude of the differences in the means was very small (η² = 0.00196). Despite, the great effect of ants on beehives, most farmers still believe that the benefits obtained from the forests outweigh the costs from disservices due to ants. They invented different strategies (Table 7) to mitigate ant attacks on beehives that are significantly different near to and far from forests (P < 0.001).

Farmers' Strategies in Tree-Crop Integration
Most of the studies concerning forest ecosystems focus on the benefits and inventories of services (Cerdán et al. 2012, Lemessa et al. 2013, Ruelle 2014, Mohammed and Asfaw 2015. Harvey (2000), Zahawi and Augspurger (2006), and Tolera et al. (2008) have reported various benefits of forests such as live fences (Chaco 2003, Francesconi 2006, Pulido-Santacruz and Renjifo 2011, grazing lands (Posada et al. 2000, Harvey et al. 2006, 2011, and shade trees for coffee production (Perfecto et al. 1996, Schroth et al. 2004, Ango et al. 2014. However, these studies have not explicitly addressed the influence of forest proximity and management practices on the associated disservices. They ignore the aspects of the influence of forest proximity on agricultural production systems and management strategies. Our interdisciplinary investigation provides alternative insights about farmers' knowledge and management practices regarding tradeoffs/disservices at varying forest proximities. Below, we first discuss the pattern of farmers' knowledge and practices in treecrop integration, and subsequently, crops and honeybee raiders and local management strategies. Our findings show that (1) statistically significant variations of tree species and food crops were observed in the homegardens, inside farmlands, along the farmland boundaries, and in riparian areas along the transects near to and far from forest edges; (2) more tree species were recorded in homegardens, along boundaries of farmlands, and in riparian areas compared with inside farmlands (Fig. 2); (3) more plantations were found far from the forest edges whereas there are more naturally grown tree species near forests (Fig. 3A, B, C); 4) farmers' integration of trees   with crops both near to and far from the forest edges did not exhibit statistically significant patterns (Fig. 4); and (5) Cerdán et al. 2012, Ruelle 2014) have shown that Croton macrostachyus, Olea europaea, Cuspidata sp., and Cordia africana were commonly integrated with food crops in farm landscapes in northern Ethiopia and other areas of the country (Mohammed and Asfaw 2015). Whereas the integration of fruit plants into the homegardens of smallholder farmers is a new source of income, Aregawi et al. (2018) have reported that about 35% of the household income is generated by fruits and vegetables cultivated in homegardens in Tigray, Ethiopia.
A number of criteria are used by the farmers to integrate trees with crops. These are farm size, fast growth, compatibility, multiple-use values, drought resistance, and access to seedlings. Among these, farm size was indicated as one of the determinant factors. We observed that some farmers often had several plots and more land, allowing them to plant more trees in addition to food crops than those with smaller-sized farms. These farmers use their entire land to grow food crops even though they were interested in growing a greater number of trees as well. Similar results were reported from Brazil and Kenya (Cerdán et al. 2012, van Damme andKindt 2012). It was found that farmers prefer to plant more trees in areas that are not productive, including very steep slopes, along farm boundaries, and beside their homes and in riparian areas. The same trend was reported in southern and northwest Ethiopia (Ruelle 2014, Mohammed andAsfaw 2015).

Crop raiding
Crop raiding by wild mammals negatively influences the livelihoods and food security of smallholder farmers (Escobedo et al. 2011, Fungo 2011, von Döhren and Haase 2015, Lyytimäki 2015, Xu et al. 2016. Farmers in our study have reported that olive baboons, bush pigs, warthogs, vervet monkeys, and porcupines were the major crop raiders posing the most serious threat to their livelihoods, particularly near forests rather than on sites far from forests (P < 0.001) (Fig. 5, Table 4). This result corroborates several studies from many landscapes across Uganda, Cameroon, and Ethiopia (Hill 2000, Weladji and Tchamba 2003, Seifu and Beyene 2014. In particular, farmers living adjacent to the forest edges estimated more than a quarter of their crop yield was lost. Furthermore, more than 40% of food production was being lost globally to pests, pathogens, and weeds; pest impacts occur for the most part at close proximity to the forest (Kleijn et al. 2006). Similarly, Zhang et al. (2007) reported that agricultural pests cause significant economic losses to smallholder farmers worldwide.
In our study, the majority of the informants (82.3%) agreed that the extent of crop raider impacts decreases as distance from the forest increases because the forests provide protection for these animals. Farmers near to forests claim more yield loss due to cropraiding animals than those living far from forests (Weladji and Tchamba 2003, Fungo 2011, Lemessa et al. 2013, Seifu and Beyene 2014, von Döhren and Haase 2015. This loss is beyond their management strategies (mainly guarding), and it was common to hear such stories in villages along forest edges (van Damme andKindt 2012, Ango et al. 2014). Diurnal crop raiders, such as olive baboons, common monkeys, and warthogs, are less common far from forests than near to forests because the latter provides easy protection for them when chased away. Similar findings were reported by other studies (Hill 2000, Sinare et al. 2016. Furthermore, the incidences of crop damage from warthogs are greater near to forested areas in southern Ethiopia (Byg et al. 2017). On the other hand, bush pigs and porcupines travel longer distances from forests for crop raiding than olive baboons, and their nocturnal behavior makes management strategies by farmers more difficult. Naughton-Treves (1997) reported similar findings, i.e., that protecting crops from bush pigs even at long distances from the forest edge was found to be difficult due to their nocturnal behavior.
Farmers have reported that the problem of crop raiding by wild animals was severe on early and late-cultivated crops, which corroborates previous studies (Kleijn et al. 2006, Seifu andBeyene 2014). The reason is that crop raiding is not spread over many fields, and thus the damage in any particular field is maximized (Naughton-Treves 1997 Lemessa et al. 2013). Swanepoel et al. (2017) have reported that the lack of integration in cultivation time and types of crops as well as lack of cooperative approaches will heighten crop damage by raiders. Moreover, there was a significant spatial variation, with crops adjoining forest edges more susceptible than those away from forest edges. Similar results were reported by several other studies (e.g., Hill 2000, von Döhren and Haase 2015, Sinare et al. 2016) from Uganda, Burkina Faso, and France. Crop raiding not only caused loss of food crops but also caused substantial indirect costs, such as additional labor for guarding and negative effects on children's schooling (Ango et al. 2014, Shackleton et al. 2016. Although farmers were challenged by crop raiders, they have developed their own management strategies, including fences, guarding, noise, scare devices, and smoke, to minimize such ecosystem disservices (Naughton-Treves 1997, Weladji and Tchamba 2003, Escobedo et al. 2011, Fungo 2011, Seifu and Beyene 2014, Lyytimäki 2015, Xu et al. 2016. Although farmers in the study area have used different management strategies, they reported that olive baboons and vervet monkeys are intelligent enough to adapt to human behavior easily and became less afraid of children and scare devices, making protection of crops less effective and more challenging (see also Lemessa et al. 2013, Ango et al. 2014. In line with this, Shackleton et al. (2016) underlined that farmers should adopt appropriate indigenous management practices to reduce crop loss by raiders.

Ant-Mediated Ecosystem Disservices
Farmers in our study area have reported that ants raid their beehives, affecting honey production. Similar problems have been reported from the northern part of Ethiopia and elsewhere (Hansen and Klotz 2005, Andersen et al. 2006, Petit and Burel 2006, Desalegn 2015. There are multiple ecosystem disservices originating in forests, such as pests, predators, parasites, honey badgers, birds, spiders, snakes, lizards, and wax moths. But ants pose direct constraints on the productivity of beehives because they attack honeybees to steal the honey. Previous studies have also reported that ants were the most important enemies of honeybees, and they are considered to be a serious problem for the honey production sector (e.g., Stamps andGroothuis 2010, Schultheiss andNooten 2013).
Farmers have reported that ants enter the beehives and feed on the honey, brood, and pollen, and destroy the bee wax, causing Ecology and Society 25(4): 36 https://www.ecologyandsociety.org/vol25/iss4/art36/ the bees to abscond; they even feed on young bees and can destroy the entire bee colony. Our results corroborate previous studies (Hansen andKlotz 2005, Andersen et al. 2006). Desalegn (2015) found that ants have attacked 40% of honeybee colonies, in addition to eating honey and pollen.
Beekeeping is a tradition of the Kafa community of the study area, and 90.3% of farmers use predominantly traditional methods of apiculture, i.e., they place beehives inside forests permanently (Chala et al. 2012). We recorded that more beehives inside/near to forests were attacked by ants than those far from forests (Andersen et al. 2006). This is due to the presence of structurally complex habitats with abundant litter, having higher amounts of coarse, dead woody debris, sufficient humidity, cold climate, high soil organic matter, and often less disturbed shaded areas in forests (Petit andBurel 2006, Desalegn 2015). These complex habitats provide nesting sites for the ants. Abiotic factors, such as soil properties, daily temperature variations, humidity, and wind, and biotic drivers such as the structure of plant communities, determine local ant assemblage (Petit and Burel 2006, Stamps and Groothuis 2010, Schultheiss and Nooten 2013. Furthermore, the availability of food resources determines the severity of ant foraging behavior (Andersen et al. 2006, Stamps andGroothuis 2010). In our study area, the farmers practice coffee agroforestry in forests, where there are abundant nesting sites for ants, thus explaining the severity of the attacks on beehives by ants. Philpott et al. (2006) reported that ants prefer coffee and cacao forests due to physiological and ecological factors.
Farmers have reported that there is a temporal variation in attacks on beehives by ants. The most severe attacks occur from the beginning of October to December, during the drought season. During hot days, ants forage individually (Stamps andGroothuis 2010, Schultheiss andNooten 2013), carrying out severe attacks on beehives (Desalegn 2015) and they also exhibit dynamic foraging behavior, with worsening effects on beehives (Andersen et al. 2006). Furthermore, our results corroborate previous studies where ants were found to exhibit predatory behavior due to levels of coarse woody debris, food availability, litter depth, and soil temperature (Hansen andKlotz 2005, Petit andBurel 2006).
Farmers use different management strategies, such as dusting ashes, spraying indigenous repellant suspensions, and destroying the nests of raiding ants to minimize ant attacks on beehives. A biological control mechanism was also used by farmers, where they cut part of a nest of Crematogaster sp. (locally called "Penie") and glue it to the trunk of trees with beehives. Crematogaster sp. safeguard the beehives from the raiding ants as part of their efforts to protect their own nests (Desalegn 2015). The observed higher prevalence of beehive attacks near to the forests is due to abundant nesting sites for the ants. As distance from the forest increases, nesting sites of ants decrease, resulting in fewer attacks on beehives. Therefore, more transplantation of Crematogaster nests were observed near to than far from forests to reduce ant attacks on beehives (Chala et al. 2012).

CONCLUSIONS
Farmers have acquired the knowledge of integrating multipurpose trees with crops from past generations, their communities, or their own lifelong experiences. They integrate trees with crops selectively and strategically based on the multiple-use values. Although ecosystem services predominate, forests do produce a number of disservices. Even with spatial variation, ecosystem disservices could influence local livelihoods and the crop production system of farmers, particularly at close proximity to the forests. Crop raiding and ant-mediated impacts on beehives were found to be the major ecosystem disservices in our study area. However, farmers continue to live there because they have developed their own strategies to mitigate these ecosystem disservices. We recommend both ecological and socioeconomic studies in order to augment farmers' strategies to trade-offs between ecosystem disservices, traditional coping mechanisms, and biodiversity conservation.

Declaration of Competing Interest
The authors declare no conflicts of personal, commercial, political, academic, or financial interest. There was no disagreement in authorship. Authors made substantial contributions to the intellectual content from conceptualizing and designing the study to submission for publication. Authors also confirm that this work is entirely original, has not been published elsewhere, nor is it currently under consideration for publication elsewhere, and referenced words and/or ideas of others have been appropriately cited or quoted.
Responses to this article can be read online at: https://www.ecologyandsociety.org/issues/responses. php/12100