We collected vegetation data from 189 plots (each 63 plots) based on the plot-based vegetation assessment protocol as mostly used in many studies in southwest Ethiopia (Senbeta and Denich, 2006; Schmitt et al., 2010; Hundera et al., 2015). Plots of 20 m by 20 m (400 m²) were laid out systematically where the first plot was randomly or arbitrarily selected and the next locations were spaced homogeneously throughout the survey. We selected coffee agroforest first and then coffee forest and natural forest subsequently along the transect line. The distance between the plots varied along the transect as a result of forest conditions.

Data were analyzed using the most commonly used metrics to estimate diversity such as richness, the Shannon–Wiener index, and the Simpson index. This is because richness is affected by sample size, the Shannon–Wiener index is affected by rare species, and the Simpson index is affected by common species, hence, the parallel use of these diversity measures is a general practice in ecological study (Yeom and Kim, 2011; Morris et al., 2014).

Woody species richness was computed for overall richness and included woody plants with a diameter greater than or equal to 10cm from recorded vegetation data in the coffee agroforest, coffee forest, and natural forest. We computed richness per plot for each forest type (coffee agroforest, coffee forest, and natural forest). All woody species recorded within 400 m² were converted into presence–absence data. Woody species richness is expressed as the number of species per forest type (Magurran, 2004; Magurran and McGill, 2011).

To test the difference of diversity for the three sample groups (coffee agroforest, coffee forest, and natural forest), data were tested for normality and homogeneity of variance before the analysis. Where these conditions were met, one-way analysis of variance (ANOVA) was used to compare diversity between the three forest types. When the assumptions were violated, the non-parametric Kruskal–Wallis H test was employed to compare the woody species richness among the three forest types. Data were organized in Microsoft Excel, and analyzed in SPSS version 25 and PAST software 3.24.

Diversity analysis was conducted for woody species with a diameter greater than or equal to 10 cm. The Shannon–Wiener index, Shannon evenness, and the Simpson diversity index were computed to compare the coffee agroforest, coffee forest, and natural forest (Magurran, 2004; Magurran and McGill, 2011). The Shannon–Wiener index (H’) was calculated as:

H ′ = − ∑ i = 1 s p i * ln p i , where pi is the proportion of individuals found in the i^th species and ln is the natural logarithm.

Shannon evenness (E’) was calculated as E ′ = H ln s where H is Shannon diversity and S is the number of species.

The Simpson diversity index (1-D) was calculated as 1 − D = ∑   p i 2 where pi is the proportion of individuals found in the i^th species. Data were organized in Microsoft Excel and imported to be analyzed in SPSS version 25 and PAST software 3.24.

The ecological importance of woody plants was studied through the relative importance of the species (Cottam and Curtis, 1956; Kacholi, 2014; Teketay et al., 2018; Asigbaase et al., 2019). It was computed based on the basal area, frequency, and density of woody plants (Cottam and Curtis, 1956; Asigbaase et al., 2019; Kunwar et al., 2020) with the equation I V I = D O + R D + R F , where DO is the relative dominance calculated as basal area per forest type, RD is the relative density calculated as the number of individual per ha, RF is the relative frequency calculated as the proportion of individuals per forest type. The Importance Value Index (IVI) was used as a proxy for a change in the ecological importance of the coffee agroforest, coffee forest, and natural forest during coffee management intensification. The higher the value, the greater the importance of woody species in the forest.

Ethnoecological data collection started with consulting the forest user group committee. It was guided by generating the required information rather than recruiting representative informants from the whole population. In this regard, purposive or convenience sampling was used to recruit the informants (Martin, 1995; Tongco, 2007; Longhurst, 2016; Kunwar et al., 2020). Potential participants were suggested by the forest user group committee. There was no payment for the participants except refreshment (coffee and tea). The interview and discussion were carried out in the informants’ residential area because here the interviewee would be most relaxed, as has been suggested by Dawson et al. (1993). The interview was held in the local language (Afaan Oromo and sometimes Amharic) and the researcher took notes in English or translated them into English soon after the discussion.

Resampling and the concepts of saturation and triangulation were used to reduce self-bias selection and respondent bias, respectively. Resampling refers to the selection of the correct informant each time. The study activities were divided on a case-by-case basis and participants were selected for each case. Data saturation refers the point where in-depth information is captured and there is no new further information obtained when interviewing a new respondent (Wray et al., 2007; Fusch and Ness, 2015). Data triangulation refers to collecting data from multiple sources (Wray et al., 2007; Fusch and Ness, 2015). Albuquerque et al. (2017) suggested a mix of methods to triangulate ethnoecological data. An effort was made to cross-check collected data through informal discussion among the informants and data were analyzed normatively.

Ethnoecological data were collected through free listing and semi-structured interviews (Albuquerque et al., 2017; Furusawa et al., 2014; Dorji et al., 2019). Prior to free listing, the informants were briefed on the objective of the study. They were asked about the three types of forest identified for the study and all participants were in a position to distinguish coffee agroforest, coffee forest, and natural forest. Eight focus group discussions were undertaken with groups of forest users from four sites consisting of four to six individuals divided by age, either 18 to 35 years old (youth) or greater than 35 years old (old). During the interview process, the groups were asked about their perception of the benefits of the forest in their livelihoods. The questions asked were “What is/are the benefits of the forest in your surroundings?” and “Which forest type is more important to suggested forest benefits?”. The groups listed the general ecosystem services of the forest they have experienced in their surroundings and ranked the relative importance of each forest type out of 100. Participating informants had grade and junior high school education and they wrote down their answers on paper. The relative importance was estimated as a percentage. The researcher distributed the paper and played a facilitator role during the process.

Cited ecosystem services were grouped into provisioning, regulating, cultural, and supporting ecosystem services as per the Millennium Ecosystem Assessment (Millennium Ecosystem Assessment (MEA), 2005). Provisioning ecosystem services were aggregated into major categories and a semi-structured checklist was prepared for further individual interviews (Martin, 1995).

A checklist for semi-structured interviews was prepared based on the preliminary findings of the free listing. The checklist included but was not limited to questions such as “Do you collect forest product x (local name of the product)?” and “Where do you collect them?” A total of 136 forest users (107 men and 29 women) were interviewed. Furthermore, 15 focus group discussions (5 “old” groups, 5 “youth” groups, and 5 groups of women) were conducted to assess the relative importance of provisioning ecosystem services and forest types (coffee agroforest, coffee forest, and natural forest). The size of a group varied between 4 to 5 individuals. The duration of an interview and a focus group discussion differed case by case (an hour for focus group discussion and 30 mins to 40 mins for an interview).

The proportion of citations and ranking were used to organize and analyze the relative importance of provisioning ecosystem services and forest types (Martin, 1995). Indicators of forest products were used to associate forest products with the coffee agroforest, coffee forest, and natural forest (Gardener, 2014). The association was estimated based on Pearson residuals (Pearson residual = ( O b s e r v e d − E x p e c t e d ) / √ E x p e c t e d ). Gardener (2014) stated that a Pearson residual is normally distributed and a value of −2 was designated as significant.

The use value of woody plants was estimated based on the number of citations. Woody species recorded during the inventory were organized and listed for use value estimation.

Semi-structured interviews were conducted to assess the uses of woody plants. Forest users were asked, but not limited to, the following questions: “Local name of a plant (1^st, 2^nd, 3^rd,———– 64^th)”, “Do you know the species x (local name of the plant)?”, “What is/are the uses of the plant?” (The use of planted coffee in coffee agroforest was not recorded), and “Do you remove or maintain the plant in your coffee agroforest?”. A total of 96 forest users (85 men and 11 women) were interviewed. Previous studies by Gueze et al. (2014) and Soares et al. (2017) employed similar approaches to assess the uses of plants. The number of uses were calculated from use categories of woody species developed by Albuquerque and Oliveira (2007) and Albuquerque et al. (2009). The number of woody plant uses was expressed as the total number of citations of uses. The number of use citations helped us to order or rank the relative importance of woody plant species for specific uses. The number of uses was used to categorize woody plants into three categories, namely, generalist, specialist, and versatile, following Albuquerque et al. (2009). Woody plants were considered specialist with at most two uses, generalist with at least three to five uses, and versatile with more than five (Albuquerque et al., 2009). The number of woody species per use category was used to categorize woody plants into three categories: highly redundant (>75%), redundant (25% to 75%), or not redundant (<25%) (Albuquerque and Oliveira, 2007). The concept of redundancy was adopted from the utilitarian ecological redundancy concept (Albuquerque and Oliveira, 2007). The concept refers to species with similar uses to distinguish from woody plant species with specific uses (Albuquerque and Oliveira, 2007, Santoro et al., 2015). In forest resource use, the presence of redundant species guarantees the resilience of a given system (Albuquerque and Oliveira, 2007; Santoro et al., 2015).

A change in provisioning ecosystem services across the coffee agroforest, coffee forest and natural forest was assessed based on plant use value (Phillips and Gentry 1993; Castaneda and Stepp, 2007; Andrade-Cetto and Heinrich, 2011; Faruque et al., 2018). Use value was calculated as where u refers the number of uses mentioned by forest users and n refers the total number of forest users interviewed (Phillips and Gentry 1993; Faruque et al., 2018). The total use values of the coffee agroforest, coffee forest, and natural forest were calculated as the summation of the use value of all woody species recorded within each forest type (Andrade-Cetto and Heinrich, 2011; Ouedraogo et al., 2014). The Kruskal–Wallis H test was used to compare differences in the ecosystem services (benefits) between the natural forest, coffee forest, and coffee agroforest.

Relative frequency citations (RFC) were used as the consensus on woody species that were retained or removed from the coffee agroforest. Relative frequency citations were expressed as the number of times a particular species was mentioned as being retained divided by the total number of interviewees (Faruque et al., 2018). One way of understanding the effect of forest modification for coffee production is to relate ecologically important woody species and the uses of woody species (Gueze et al., 2014). Spearman’s rank correlation was conducted to investigate the relationship between the availability of woody species and plant uses (Sop et al., 2012; Gueze et al., 2014). Woody species availability across the coffee agroforest, coffee forest, and natural forest was based on phytosociological metrics (relative density, relative frequency, and dominance) (Albuquerque et al., 2009). Ethnoecological data were summarized descriptively (Jalilova and Vacik, 2012; Ahammad et al., 2019) using Microsoft Excel and imported to SPSS version 25 for the non-parametric analytical Spearman’s rank correlation test.

Woody plant diversity and availability are determinant factors in plant usage (Soares et al., 2017). The current status of forest biodiversity varies with intensity of land use (Chazdon et al., 2009; Phillips et al., 2017). Coffee production is one form of land use that modifies the natural forest. Coffee agroforests have great potential to conserve forest biodiversity (Ismail et al., 2014). Coffee management removes undergrowth or understory plants in coffee agroforests. Interest has grown in human-managed landscapes in forest biodiversity conservation. Woody species diversity indicates the status of forest biodiversity under human management systems. Species richness per plot (i.e., all woody plants recorded with abundance data) decreased from the natural forest to the coffee forest to the coffee agroforest.

In contrast to our expectations, woody species with a diameter greater than or equal to 10cm richness decreased from the coffee agroforest towards the natural forest. Higher numbers of woody species were found in the coffee forest and coffee agroforest than the natural forest. Silvicultural practices that encourage tree species in the coffee agroforest were maintaining desired trees and not clearing the seedlings of desired tree species. These contributed a greater number of pioneer species such as Albizia gummifera, Milletia ferruginea, and Cordia africana in the coffee agroforest. The higher species richness in the coffee agroforest compared to the natural forest was attributed to land use history and other factors related to the environment rather than silvicultural treatment. Studies have shown land use history affects woody plant species richness (Shumi et al., 2018; Kumsa et al., 2016; Arnell et al., 2019). As stated in literature four decades ago (McCann, 1997), Belete forest was under logging, implying the removal of timber tree species from the natural forest. Commercial logging was not carried out in the coffee agroforest. Moreover, coffee management practices involve slashing understory plants to create vacant space for planting coffee and avoid competing vegetation, as well as the thinning or stem reduction of canopy trees. The reduction of bigger trees (DBH >=10cm) is carried out to remove heavy shade on coffee plants. But the higher number of trees with a DBH greater than 10cm in the coffee agroforest implies the removal of understory plants for coffee intensification. The bigger trees (DBH>=10cm) are scattered and so there is no need to reduce the canopy trees. Decuyper et al. (2018) has stated that forest utilization in southwest Ethiopia has a greater effect on the undergrowth plants than on the canopy tree species. According to Decuyper et al. (2018), a coffee forest has canopy openness as when the undergrowth plants are removed the gaps created are sufficient for coffee and there is no need for further thinning of canopy trees. Existing bigger trees left in situ result in a coffee agroforest containing a greater number of woody plants. Natural forest modification to coffee agroforest has contributed to a reduction in commercial logging in coffee agroforests because commercial logging does not take place here.

The current study findings show that many woody species are maintained in coffee agroforests. Study findings from Mexico by Valencia et al. (2014) showed lower species richness in a coffee agroforest at the plot level and comparable species richness at the landscape level compared to an adjacent natural forest.

We found that the Shannon diversity index of the natural forest, coffee forest and coffee agroforest was 3.33, 3.42, and 3.07, respectively. The Shannon diversity index usually ranges between 1.5 and 3.5, and rarely surpasses 4.5 (Bibi and Ali, 2013; Travlos et al., 2018). The Shannon diversity index of the coffee agroforest (i.e., 3.07) was found to be high (Magurran, 2004; Arzamani et al., 2018). The Shannon diversity index result showed that the coffee forest had the highest woody species diversity. The finding supports the intermediate disturbance hypothesis in that species diversity is highest at an intermediate disturbance level (Bongers et al., 2009). Similarly, the Simpson diversity index of the coffee agroforest, coffee forest, and natural forest was found to be 0.92, 0.96, and 0.95 respectively. Likewise, the Simpson diversity index result showed the highest diversity in the coffee forest. The Simpson diversity index ranges between 0 and 1. The Simpson diversity index value of 0 shows similarity within a community and a value of 1 shows diversity (Bibi and Ali, 2013; Travlos et al., 2018; Atsbha et al., 2019). The present study finding shows high diversity across the three forest management regimes.

The forest is a source of livelihood for local people (Ouedraogo et al., 2014). Our study highlighted forest users’ perspectives of ecosystem services provided by modified forests in general and the coffee agroforest, coffee forest, and natural forest in particular. Forest ecosystem services can be expressed as provisioning, regulating, supporting, and cultural benefits of the forest (Millennium Ecosystem Assessment (MEA), 2005). Forest users reported these four major categories of ecosystem services. However, our findings showed that forest users value the provisioning ecosystem services of the forest more than the other ecosystem services, showing the local relative importance of the coffee agroforest to forest users. In total, 17 out of the 26 freely listed forest ecosystem services were related to provisioning ecosystem services. Forest users interact with the forest mainly for coffee and fuelwood collection and to a lesser extent for other forest products. Comparing the three forest types with regard to the most important forest products, forest users unequivocally value the coffee agroforest the highest. This is because the coffee agroforest is a source of managed coffee. A previous study in southwest Ethiopia by Tadesse et al. (2014) showed that coffee is valued highly for its high cash value. Studies from other areas have also shown that forests are most used for provisioning ecosystem services. For example, a study from India showed that a traditional agroforest was a source of provisioning ecosystem services such as fruit, timber, fuelwood, fodder, and medicinal plants (Dhanya et al., 2014). Another study from southeastern Burkina Faso by Ouedraogo et al. (2014) showed that provisioning services were the most cited ecosystem services.

The relative importance of provisioning ecosystem services and the forest types showed the value of those services and their sources to forest users. Most provisioning ecosystem services were extracted for subsistence use from the coffee agroforest. An interesting finding of the study is that there is a difference in potential and actual ecosystem services of the forest. Forest users gave greater priority to the economic benefits of the forest than to the ecological and social benefits of the forest. A study by Ango et al. (2014) showed that coffee and honey were the most important cash-generating ecosystem service for most forest users in southwest Ethiopia.

Three types of plant uses were identified: specialist, generalist, and versatile (Albuquerque et al., 2009). Woody plants with at most two uses were grouped as specialist and those with three to five uses were grouped as generalist. Woody plants with more than five uses were grouped as versatile species (Albuquerque et al., 2009). Only a few woody species were grouped as specialist species. For example, two uses were reported for Brucea antidysenterica and Dracaena steudneri. Alangium chinense, Albizia gummifera, and others were among the generalist species. Allophylus abyssinicus, Apodytes dimidiata, Olea welwitschii, Prunus africana, Syzygium guineense, and others had versatile uses (Albuquerque et al., 2009). More than 90% of the woody plants were used for fuelwood. Furthermore, more than 80% and 50% of woody plants were used for construction and medicinal value, respectively.

Woody plant diversity and availability are determinant factors in plant usage (Soares et al., 2017). Availability and plant use across the natural forest, coffee forest, and coffee agroforest were studied through phytosociology (relative density, relative frequency, and dominance) and use value. Woody plants were categorized into three categories, namely, highly redundant, redundant, and less redundant, based on specific uses with arbitrary cut-off points (Albuquerque and Oliveira, 2007) greater than 75%, between 25% and 75%, and less than 25%, respectively. These showed the benefit lost as a result of woody plant removal during coffee intensification. For instance, the benefit that derived from a specific sources (i.e., woody plant species) might be lost along with tree removal. High redundancy showed that specific uses could be obtained from more than 75% of the available woody plant species. Similarly, redundant and less redundant showed that species uses could be obtained from between 25% to 75%, and less than 25% of available woody plant species respectively.

Some of the woody species were highly encouraged in coffee agroforest; as a result many woody species were commonly maintained in the coffee agroforest. There was no coffee management practice such as weeding and cutting that would discourage these plant species from the system. Their seedlings were encouraged to grow by removing competing grasses around them. Woody species such as Milletia ferruginea and Albizia gummifera were highly encouraged for coffee shade whereas Cordia africana and Aningeria adolfi-friederici were some of the highly encouraged woody species for timber. Cordia africa is widely used in the area for making furniture. Woody species that are mainly discouraged in the coffee agroforest, such as Bersama abyssinica, Brucea antidysenterica, Justicia schimperiana and Maesa lanceolata, had medicinal values. Lianas, which are almost absent from the coffee agroforest, can be used for fuelwood, construction material (building material for traditional houses, fencing, and traditional beehive making), bee forage, and as income through generating cash. Astropanax abyssinica is known as the honey tree for its popularity as bee forage. Coffee shade and multiple uses of woody species did not justify the reason for encouraging some trees and discourage others in the coffee agroforest. For instance, eight uses were mentioned for Clausena anisata and Calpurnia aurea where highly encouraged coffee shade trees Albizia gummifera and Diospyros abyssinica had five and four uses, respectively.

Coffee forest biodiversity has been receiving increasing attention for conservation. Some woody species are removed and others are maintained in coffee agroforests in southwest Ethiopia. Our study findings showed that woody species are encouraged in coffee agroforests not only for shade but also for other uses. Albizia gummifera and Milletia ferruginea are encouraged mainly for shade whereas Cordia africana and Aningeria adolfifriederici were encouraged for timber. Diospyros abyssinica was cited most for construction materials and Polyscia fulva was cited for traditional beehive making. Astropanax abyssinica was cited for bee forage. Forest users interact with the forest for plant use (Maroyi, 2012). The present study findings showed that although 33 different uses of plants were identified, only three to five were utilized most of the time. Fuelwood and construction materials were the main uses of woody plants. The potential uses of woody species do not indicate the actual use of woody plants in most cases (Ahammad et al., 2019).

Woody plant use citations show that forest users have the knowledge but forest modification is a matter of immediate benefit priority. The literature has shown that location, locally available resources, and plant knowledge increases the use and conservation of forest biodiversity (Pieroni and Soukand, 2018). Plant uses are also one form of the forest biodiversity conservation model (Albuquerque et al., 2009). The specialist, generalist, and versatile uses of woody plants indicate the importance of woody plants and their conservation value. Twelve different uses were reported for Apodytes dimidiata, whereas, the relative frequency of citation (RFC) was less than woody species with a fewer number of uses such as Cordia africana, Milletia ferruginea, Aningeria adolfi-friederici, and Albizia gummifera.

We showed that woody species availability and uses differ across the natural forest, coffee forest, and coffee agroforest. Woody plants can be categorized into high redundant, redundant and less redundant categories based on number of specific uses per species. This shows the diversification of plant use (Albuquerque et al., 2009) and has implications for woody plant conservation. The use of highly redundant species reduces pressure on woody species whereas use of less redundant species increases pressure on woody plants (Albuquerque and Oliveira, 2007). Forest users reported that they use available dried wood and branches of woody plants for fuelwood instead of specific woody species for fuelwood. The study findings also showed there was a change in most use values during forest modification to coffee production. Nevertheless, coffee agroforest increased the shade and timber use value of woody species. The well-known timber tree Cordia africana is more abundant in the coffee agroforest than in the coffee forest and natural forest. Similarly, the shade value of woody plants was apparent in the coffee agroforest.

A previous study by Gueze et al. (2014) in the Bolivian Amazon on the relationship between the Importance Value Index and useful value tree species showed a positive relationship between the Importance Value Index and overall use value. Kunwar et al. (2020) reported a weak relationship between plant use value and phytosociological indicators in Nepal. Our findings showed that there are moderate positive correlations between the Importance Value Index and overall woody plant use value for the natural forest and coffee forest and a strong positive correlation for the coffee agroforest.