Introduction

Agroforestry, as a land management system that deliberately integrates trees, crops and/or animals into the same cropping system, has acquired ever more importance in the global quest for sustainable solutions for agriculture and food security (Atangana et al. 2014; Khadka et al. 2021). Trees are key components of the agroecosystem. Their multiple functions—improving soil fertility, regulating climate regulation and preserving biodiversity – enable them to significantly contribute to a farm’s ecosystem resilience and environmental sustainability (Dimobe et al. 2018; Santos et al. 2022). However, most technologies promoted in African countries are often based on exotic species (Suárez et al. 2012) and then proposed to farmers for implementation (Erakhrumen 2009). Such an approach often results in farmers’ expectations going unmet regarding ecosystem services that are important to them (He et al. 2015). In many African countries, for example, despite the implementation of various modern agroforestry technologies, the agroforestry landscape is still dominated by traditional agroforestry parks that promote local species, which results in local species only being valued by farmers who conserve them on their fields (Gram et al. 2018; Hernandez Marentes et al. 2022). Thus, the conservation and valorization of these local species often depend on the endogenous knowledge and preferences of farmers (Feintrenie et al. 2010). This knowledge is important because, according to Félix et al. (2018), in Africa, improving soil enhancement and can be derived from intensive management of existing local species in agricultural landscapes.

Given the important role of local species, there is a need to investigate farmers’ knowledge on this subject to understand how it can be integrated in agroforestry technologies to strengthen these interventions. Several studies have been conducted about plant conservation in agroforestry systems by farmers, to help incorporate local species and farmers' knowledge in agroforestry technologies that are effectively adapted to socio-cultural and ecological conditions, and to farmers' preferences for wider adoption and use of these technologies (Erakhrumen 2009; Anglaaere et al. 2011; Parihaar et al. 2014; He et al. 2015; Gram et al. 2018; Hernandez Marentes et al. 2022). These studies have found that in many contexts and environments, farmers generally prefer to preserve or cultivate species on their farms that have multiple socioeconomic purposes and about which they have a fair knowledge. Furthermore, considering the variability of environmental conditions, vegetation composition, social conditions and cultural factors in each country, investigations integrating farmers' knowledge in the choice of agroforestry species in each country are important to identify local agroforestry species adapted to their specific context (Pulido-Santacruz and Renjifo 2011; Silva-Galicia et al. 2023).

In Benin, a country in West Africa, where the agricultural landscape is characterized by several varieties and types of local species, there is a wide range of agroforestry studies that address very different aspects of agroforestry systems in Benin. They range from ethnobotanical and socioeconomic analyses of agroforestry systems in Benin (Assogbadjo et al. 2012; Koussihouèdé et al. 2020; Lokonon et al. 2021; N’Woueni & Gaoue, 2021; Odounharo et al. 2022), to the vulnerability of systems to human practices and climatic variability (Salako et al. 2019a, b; Kakpo et al. 2024), as well as assessing the contribution of these systems to the conservation of plant and animal biodiversity and ecosystem services (Ouinsavi and Sokpon 2008; Assogbadjo et al. 2010; Fifanou et al. 2011; Enagnon Lokonon et al. 2017; Dassou et al. 2020; Koussihouèdé et al. 2020). These studies provide a very valuable background for the development of agroforestry in Benin, as well as guiding future research in the field. However, an in-depth analysis of the existing literature reveals a major gap in the analysis of the role and implications of gender in the choice of this practice. Given the important role of gender in agriculture and even more in agroforestry in Africa (Assé and Lassoie 2011; Kiptot and Franzel 2012a), and in Benin, according to Benin's Department of Agricultural Statistics (DSA), research is needed to fully understand the interactions between agroforestry and gender dynamics in Benin for an equitable and sustainable promotion of this practice. Failure to understand, for example, the differences between men's and women's preferences and their social interactions can make it difficult to identify opportunities to improve agroforestry research and development and to promote equity (Kiptot 2015). To address this gap, this study uses a gendered lens to understand if and how farmers’ gender has a role in their species preferences.

Methodology

Study context

This study was conducted in three of the eight agro-ecological zones of Benin, a West African country located in the intertropical zone, between parallels 6°30' and 12°30' North and meridians 1°3°40' East. The nation is bordered to the north by the Republics of Niger and Burkina Faso, to the south by the Atlantic Ocean, to the east by the Federal Republic of Nigeria and to the west by Togo (see Fig. 1 below). It has 1,388,361 inhabitants for a surface area of 114,763km2 according to (World Bank 2023) statistics. Tropical ferruginous soils and ferralitic soils occupy around 80% of the country's surface area and almost 42% of the country is covered by forest area (Ahononga et al. 2021). Benin’s eight agro-ecological zones (AEZ) are a cluster of villages and communes defined by several specific variables, including soil, climate, land pattern and vegetation cover. Each AEZ has homogeneous climate, soil, vegetation and geomorphology and different challenges and opportunities related to land use (Govt. Benin and UNDP 2015).

Fig. 1
figure 1

Study areas

Full size image

This study, which focuses on the management of land and forest resources (agroforestry) in line with the Food and Agricultural Organization's recommendations for advanced applications of agroecological zones for study purposes, focuses on three agroecological zones (3, 4 and 5) (see Fig. 1 below).

These AEZs were selected for this study because they face particular challenges related to land and natural resource management (Govt. Benin and UNDP 2015). For example, AEZ 4 (Ouest-Atacora) and AEZ 5 (" cotton zone " of Central Benin) are among the four agro-ecological zones most exposed to the effects of climate change in Benin (Borde and Bokonon-Ganta 2015). In addition, all three AEZs selected for this study are characterized by heavy deforestation and are among the five zones currently targeted for the implementation of the land degradation neutrality program in Benin (MCDD 2019). They represent a climatic transition zone (between the sub-equatorial of the south and the Sudanian of the north) characterized by high production of both cash and food crops (Govt. Benin and UNDP 2015). Within these three zones, agriculture, which is the main activity of the local population, is mostly for subsistence purposes and done by traditional means (i.e. the hoe). As a result, the current production system is inefficient, requiring five to ten years of fallow after a period of cultivation (Govt. Benin and UNDP 2015). Furthermore, the choice of these zones considers the country's major sociolinguistic groups, a wide range of possible agricultural production systems and a wide range of tree species, thus encompassing all possible eventualities regarding farmers' preference for agroforestry species.

Data collection

Sampling

Ethnoagroforestry and ethnographic surveys were carried out in six municipalities in the three AEZs of the study (i.e. Kopargo, Djougou, N’dali, Pèrèrè, Tchaourou, and Ouèssè). These municipalities were selected based on their agricultural status and the size of the agricultural population. Specifically, 18 villages, three villages per municipality, were selected based on criteria such as accessibility and the practice of agroforestry by farmers. The survey sample size was determined using the Schwartz formula (1995), considering the size of the agricultural population using the following formula:

$${n= z}^{2}xp(1-p)/{m}^{2}$$

where:

N:

is the sample size;

Z:

is the confidence level according to the standard normal distribution with a confidence level of 95% (z = 1.96);

P:

is the estimated proportion of the population exhibiting the characteristic;

M:

is the tolerated margin of error (for instance, we aim to know the true proportion within 5% accuracy).

The total population size and the agricultural population size used for this determination were extracted from the document 'Village and City Neighborhood Notebook of the Borgou, Colline, and Donga Department' (INSAE 2016). In total, 364 farmers were interviewed with the assistance of the staff members of the Communal Cells of the Territorial Agency for Agricultural Development (ATDA).

In total, 144 women farmers, representing 39.56% of the total sample, and 220 men farmers, representing 60.44% of the total sample participated in this study. It should be noted that the survey was not addressed exclusively to household heads, at the risk of excluding the majority of women farmers (Mai et al. 2011). The survey therefore included all farmers aged at least 18, including household heads. In order to be considered in our surveys, these farmers should have at least one farm plot that is independent of those farmed by the household head. In general, agroforestry landscapes in the study area were dominated by agricultural plots interspersed with a few trees. For the agroforestry inventory, one plot per farmer with an area of at least 0.40 ha was considered (Perry et al. 2005). However, as women's dominated system plots are very small, exceptionally plots of less than 0.40 ha were considered for women. 364 agricultural plots were thus considered for the inventory of species, after having been georeferenced using a GPS. Five circular sub-plots (radius 12.5 m, 25 m apart) were established in plots measuring 1 ha or more, while in smaller plots (less than 1 ha), 2 rectangular sub-plots (Fig. 2) were established to inventory the species found (Bieng et al. 2022; Ndayambaje et al. 2013), as illustrated in Fig. 2 below. In the few cases where the initial plot form was irregular and difficult to identify, and its surface area was less than 1 ha, a third rectangular sub-plot was installed (Roshetko et al. 2002). In addition, the distance between plots and the size of plots on farmland of less than 1 ha depended on the actual initial size of the farm plot (Roshetko et al. 2002). The information collected on the species was: their name (vernacular in local or vulgar languages), and the number of trees of each species per plot.

Fig. 2
figure 2

A example of a 1 ha agricultural plot (200 m x 50 m) composed of 5 circular plots of 12.5 m radius, separated by 25 m. B example of a 0.25 ha agricultural plot composed of 2 circular plots of 20 m x 15 m separated by 40 m

Full size image

Survey and inventory procedure

The questionnaire used for this study was written in French and administered to farmers in their respective local languages. Ethics approval was obtained from the local research committee of ethic at University of Parakou and informed consent was sought from each farmer by first having initial contact at which time a member of the research discussed with them the study’s objectives, and asked if they would be interested to participate. A respondent's consent to participate in the survey was expressed by agreeing to reveal his identity, his contacts and arranging an appointment with research team member. Following this initial step, only those farmers who agreed to take part in the study were then invited to lead a member of the team onto their farm plots where the interview was then conducted. During the interview, the tree species present on the plot were inventoried together with the farmer with the researcher asking the farmers specific questions included in the pre-developed questionnaire about each species found on the individual plots. The ethno-agroforestry survey was based on a mixed questionnaire with open and closed questions. During the survey, information was collected on: the socio-demographic factors of each farmer (i.e. gender, age, origin, farm size, access to land, social background, level of education, number of years of farming activity, main crops); reasons for choice of species in relation to their knowledge of the ecosystem services rendered by these species in agroforestry systems and the cropping benefits they derive from them, as well as their knowledge of how these species are propagated. During the interviews, farmers were asked to rank, on a scale of 1 to 5, the tree species present on their plots in relation to selected key ecosystem services, which relate to the provisioning and supporting services listed in the questionnaire using the classification of ecosystem services of trees proposed by Millennium Ecosystem Assessment (2005) and revised by Haines-Young and Potschin-Young (2018). This classification consisted of the farmer scoring the species present on their plot according to whether the species was better than other species present.

Data analyses

A "Yes" (1) or "No" (0) distribution analysis was applied to identify farmers' knowledge about ecosystem services provided by species and their propagation methods. Ecosystem services were grouped into three major categories following the classification of (MEA 2005): provisioning services, support services and cultural services. Then, a comparative analysis of the average amount of knowledge per species between the two distinct clusters was carried out using a moustache box, incorporating a student’s t analysis to assess the statistical significance of the differences observed between the groups. Based on the socio-demographic and agro-ecological data collected (Table 1), a hierarchical classification was performed using the FactoMineR and Factoextra packages. This was used to group farmers into homogeneous groups. To describe these groups, the difference between the class values and the overall values was measured. These statistics were converted into a criterion known as a value test (V-test) to make a selection on the variables, and thus designate the most characteristic variables (Lebart et al. 1984; Husson et al. 2010). The most characteristic parameters of a class were those for which the associated value test was greater than 3 in absolute value. The abundance of local species identified during the inventory operations were assessed for each specific group of farmers. The median score and associated rank based on the scores attributed to each species on a scale of 1 to 5 during farmers species, revealed the 5 priority species for farmers in each group. Similarly, information relating to farmers' assessment of the importance of different species and their associated ecosystem services on a rating scale of 1 to 5 was used to create ecological networks using the NetworkX package with python software version 3.9. These networks are based on undirected models, which means that they do not account for the direction of interactions between species and ecosystem services. Furthermore, network nodes are split into two categories: species and ecosystem services. For each species assessed, a node has been created in the network, identified as "species". Similarly, for each ecosystem service assessed, a node has been added, identified as "service". In this way, the network represents a combination of species and ecosystem services. The edges (connections) between the nodes were added according to the scores assigned by the farmers. This methodological approach thus offers a graphical representation of the interactions between species and ecosystem services and enables us to understand farmers' knowledge of the ecosystem services of the species on their agricultural plots, the services most important for them and why. Additional analyses were carried out using Rstudio 9.5 software.

Table 1 Variables and their meanings
Full size table

Results

Discriminating characteristics of farmers

Socio-demographic variables, land access methods, gender, level of education, status of household head (Defined as the family's main decision-maker), marital status, second activity (Defined as having an activity other than agriculture), sociolinguistic affiliation, and origin (Defined as being native to the area or having migrated there), were the most discriminating of farmers (p < 0.05; Table 2). The same applies to quantitative variables such as agroforestry plot size and farming experience (p < 0.05; Table 3). As for the agroecological variables, it was affiliation to agroecological zones (ZAE) that was discriminatory for farmers (p < 0.05; Table 3). Two homogeneous groups representative of the farmer population were identified, reflecting the existence of two types of agroforestry farming systems present in the study area.

Table 2 Link between the cluster variable and the categorical variables (chi-square test)
Full size table
Table 3 Link between the cluster variable and the quantitative variables
Full size table

Description of identified farming agroforestry systems (cluster)

The first system, referred to here as the men's dominated agroforestry farming system, is principally composed of farmers from agroecological zone 3 (40%), followed by farmers from agroecological zones 4 and 5 (30%) in smaller proportions (Fig. 2). This type of agroforestry farming system is mostly composed of male (96%) household heads (Defined as the family's main decision-maker) (97%), who principally cultivate legacy land (98%) (Table 4). In these men’s-dominated agroforestry farming system, the level of education can reach secondary school level (90%). Most members of this group belong either to the ethnic group of Bariba (84%) or Yom (73%). The group is largely composed of natives (74%) who are engaged in crafts and livestock farming (95%) as secondary activities (Defined as having an activity other than agriculture), with 78% who engage in agriculture as their main income-generating activity, as shown in above table. In this predominantly male group, farmers have larger agroforestry areas and more agricultural experience (v.test = 3.42, mean = 2.88 ± 1.39, p-value < 0.001) (Table 5). The second group refers to the second type of agroforestry farming system found in our study area, and we will refer to it here as the women-dominated agroforestry farming system. Farmers in this group are mainly women, cultivating land they have received primarily as gifts (81%) or loans (64%). This group of farmers is characterized by a low level of schooling (44%), by contrast with farmers in the first system, while most are widows from the Fon sociolinguistic group (98%). In this group, most of whom are migrants or non-natives (61%), trade is the dominant secondary activity (defined as having an activity other than agriculture. This group is more present in agroecological zone 5 (41%), but also in lower proportions in agroecological zones 4 (38%) and 3 (20%) as shown in Fig. 3, with limited access to land, explaining the restricted size of agroforestry plots in this type of agroforestry farming system, unlike farmers in the first system (v.test = -3.42, mean = 2.34 ± 1.17, p-value < 0.001) (Table 5).

Table 4 Description of each cluster by qualitative variables
Full size table
Table 5 Description of each cluster by quantitative variables
Full size table
Fig. 3
figure 3

Distribution of each farming agroforestry systems in the 3 agro-ecological zones (AEZ). 1 (men’s dominated agroforestry system); 2 (Women’s dominated agroforestry system); ZAE (agroecological zone)

Full size image

Local knowledge about local agroforestry species

The results of the analysis of farmers' knowledge of the ecosystem services provided by species reveal that farmers have at least some knowledge of the ecosystem services provided by species, whatever the species. However, this knowledge diversity varied from one group of farmers to another. Indeed, in the category of provisioning ecosystem services as shown in Fig. 4 below, trees are strongly recognized by farmers for their role in providing medicinal products (91.62%), products for sale (87.64%), edible products (78.83%), fodder (66.61%), firewood (52.27%), and wood in general (42.47%). In the category of supporting ecosystem services, shade provided by trees (90.62%) and soil fertilizer benefits (5.39%) are also recognized, although to a lesser extent than provisioning services. Ritual utility (4.54%) and the mystical power of specific species (1.70%) are the essential services recognized by farmers in the category of cultural services provided by species.

Fig. 4
figure 4

Farmers' knowledge about tree ecosystem services classification. Note: PMP: Provide medicinal products, PSP: Provide sale products, PEP: Produces edible products, FP: Fodder provide, Fire WP: Fire wood provide, FWP: Furniture wood provide, BWP: Building wood provide, NPPP: Natural phytosanitary product provides; SP: Shade provide, SF: Soil fertilizer, RP: Rituals plants

Full size image

Moreover, in the male-dominated agroforestry system of agroecological zone 3, the number of services or knowledge varies between 1 and 9 per species, with a total average of 5.66 ecosystem service knowledge (Fig. 5). Additionally, the median (Fig. 5) suggests that more than half of farmers in this system hold an average of more than 6 pieces of knowledge per species. In contrast, in the women-dominated agroforestry system, the number of pieces of knowledge varies between 2 and 8 in total, with an overall average of 5 pieces of knowledge per species (Fig. 5). This disparity suggests that farmers in the male-dominated agroforestry system have a higher average knowledge of tree services in their fields by species than farmers in the female-dominated agroforestry system.

Fig. 5
figure 5

Comparative analysis of ecosystem service knowledge between the male-dominated agroforestry system (cluster 1) and the women-dominated agroforestry system (cluster 2)

Full size image

Furthermore, analysis of the knowledge of farmers' capacity to produce these species reveals that most of the species found on farmers' plots are natural regeneration. Furthermore, farmers in the male-dominated agroforestry system have more knowledge (Fig. 6) about the breeding of such species as Mangifera indica (Mango tree), Elaeis guineensis(Palm oil tree), Khaya senegalensis(Senegal mahogany), than farmers in the female-dominated agroforestry system, who declared that they were only able to produce the species Mangifera indica or mango tree (Fig. 6, Table 6). These results suggest a gendered dynamic of knowledge.

Fig. 6
figure 6

Farmers' ability to produce the locales agroforestry species in the male-dominated system (cluster 1 and the female-dominated system (cluster 2)

Full size image
Table 6 Species able to be produced by farmers
Full size table

Local knowledge and local agroforestry species abundance management

In the men's dominated agroforestry farming system, the rank-frequency graph (Fig. 7) reveals that the most abundant species is Vitellaria paradoxa,( Shea tree) followed by Parkia biglobosa(African locust bean), Mangifera indica(Mango tree), Elaeis guineensis(palm oil tree) and Pterocarpus erinaceus(African kino tree). This distribution suggests that Vitellaria paradoxa is the main dominant species in this system, while the other species are less abundant but still present. In contrast, in the women's-dominated agroforestry system, although the rank-frequency graph shows that Vitellaria paradoxa also remains the most abundant species, it is followed by Mangifera indica, Parkia biglobosa, Vitex doniana(African black plum) and Elaeis guineensis. The first three species are the same as in the human system, but with slightly different levels of abundance. However, the last two species differ between the men's and women’s dominated-systems, with Pterocarpus erinaceus in the men's and Vitex doniana in the women's-dominated system. These results suggest that some species are common to both agroforestry farming systems but differ in abundance. For example, species such as Vitellaria paradoxa, Mangifera indica, Parkia biglobosa and Elaeis guineensis are present in both systems, but in different proportions. The results of the rank-frequency graphs show similarities and differences in species composition and abundance between the men's and women's-dominated agroforestry systems. Based on results from the farmers' species knowledge analysis, we also found that species abundance on plots is not only associated with their knowledge of species multiplication, but may also be dependent on their knowledge of ecosystem services provided by these species. Even if some of the species for which farmers had declared they were able to breed were among the most abundant (Mangifera indica and Elaeis guineensis), farmers were unable to breed the first most abundant species (Vitellaria paradoxa and Parkia biglobosa) because of lack of knowledge.

Fig. 7
figure 7

Rank-frequency curves for the top five most abundant local species in (A) men’s- dominated and (B) women’s- dominated agroforestry system

Full size image

Local knowledge and species prioritization according to priority ecosystem services

The median score allows us to rank each species in order of priority according to the two-agroforestry system identified (Table 7). Vitellaria paradoxa, Elaeis guineensis, Pterocarpus erinaceus, Prosopis africana(African mesquite) and Afzelia Africana(African mahogany) emerged as the top five species nominated by farmers in the men's dominated system as priority species. Meanwhile, Vitellaria paradoxa, Elaeis guineensis, Mangifera indica, Parkia biglobosa and Vitex doniana are the top 5 priority species identified by farmers in the women's dominated group. Considering the overall ranking of species, we note that 4 of the 5 priority species are identical to those identified by men, with differences in species ranking but different from those identified by women. This supports the necessity to include socio-demographic characteristics such as gender and agronomic factors to select species for agroforestry technology development. The result also suggests that priority species in the female-dominated agroforestry system were the most abundant in this system. However, in the male-dominated system, only two of the most abundant species (Vitellaria paradoxa and Elaeis guineensis) were among the priority species in this system. This shows that beyond species availability, several other factors explain species prioritization by farmers on their agricultural plots.

Table 7 Median score according to the importance of species in the farmers' environment
Full size table

Furthermore, the species ecosystem services provided, those motivating their prioritization by farmers in both two agroforestry farming systems, essentially include species ability for food, commercial products, medicinal products, fodder and shade provide. But the ability of these species to provide these services is different from each individual species (green lines, Fig. 8, Table 8). However, individual analysis of each farmer group suggests that the most important ecosystem services justifying the prioritization of the five species among the male-dominated agroforestry farming systems are, the species' ability to provide food, tradable produce, medicinal products, fodder, wood for construction and furniture, and shade (green lines, Fig. 9, Table 9). Meanwhile, the agroforestry farming system dominated by women focuses on the capacity of species to provide edible, commercial, medicinal products and shade products as priority ecosystem services (green lines, Fig. 10, Table 10). This prioritization differs from men's priorities, who, in addition to all these services, conserve some species because they provide wood for construction and furniture-making. This species and their ecosystem services Prioritization by the two groups of farmers confirms their individual preferences for species characteristics and priority species and supports the necessity of considering each farmers' specific preferences for species in the development of agroforestry technologies. Overall, the species prioritized by both groups of farmers are multi-purpose species, with the potential to provide several benefits and services specific to each category of farmer. This shows that for each group of farmers, the most important factor in the choice of species is their knowledge about tangible services and benefits that these species can provide.

Fig. 8
figure 8

Ecological network showing main ecosystem services of the five most important species in all agroforestry systems

Full size image
Table 8 Average score matrix used for the ecological network of all farmers
Full size table
Fig. 9
figure 9

Ecological network showing main ecosystem services of the five most important species in the men’s dominated agroforestry system

Full size image
Table 9 Average score matrix used for the ecological network according to men’s dominated system
Full size table
Fig. 10
figure 10

Ecological network showing the main ecosystem services of the five most important species in the women’s dominated agroforestry system

Full size image
Table 10 Average score matrix used for the ecological network of women dominated s system
Full size table

Discussion

This study identified two distinct agroforestry farming systems (male-dominated and female-dominated) based on various socio-cultural, economic and demographic factors. The major factor factors that discriminate farmers, is gender (man and woman). Moreover, education level, sociolinguistic affiliation, additional activities, land access methods, land surface area, marital status, agricultural experience, household head status, origin and geographical location are also discriminate. Such factors are important in discriminating between farmers in the study sample. Identification of these two agroforestry systems based on the earlier mentioned factors, combined with the significant differences observed between the two systems, suggests that these factors have significant influences on farmers' knowledge of local agroforestry species and choice on their farm (Kiptot 2015). This supports findings by Rozaki et al. (2021) who demonstrated the influence of various socio-demographic factors, including gender, in the categorization of agroforestry farmers in Indonesia. The authors also found a difference in agroforestry practices between the two groups, including species selection. Similarly, this result supports the importance of socio-economic, cultural, environmental and agricultural factors on knowledge and choice of local agroforestry species demonstrated by Assé and Lassoie (2011); Assogbadjo et al. (2012); Feyssa (2012); Kiptot et al. (2014) Elias (2015); Dimobe et al. (2018). Moreover, this result opposes the findings of Moore et al. (2014), for whom gender is not a discriminating factor for agroforestry farmers.

Farmers' knowledge about local agroforestry species

The results show that farmers have a high knowledge about ecosystem services provided by local species on their plots, but a low knowledge about species propagation methods. Farmers' strong knowledge about ecosystem services provided by species is correlated with the fact that their livelihood in rural areas depends on the management of natural resources present in their immediate environment. Therefore, over time, farmers develop a profound knowledge about services they might or not extract from these resources, including trees (Lokonon et al. 2017; Hernandez Marentes et al. 2022). A variation in ecosystem service knowledge was identified between the two agroforestry systems (male- or female-dominated) in this study and demonstrated a gendered variation in ecosystem service knowledge. This difference in knowledge between men and women about ecosystem services is probably a result of the gendered division of work between men and women in African societies (Assé and Lassoie 2011). Thus, each farmer group's knowledge about ecosystem services provided by species may directly reflect the specific needs of each farmer group that require these species to satisfy (Dimobe et al. 2018). Moreover, farmers reported being able to replicate very few of the species that provided them these services. Some even had no knowledge about species' breeding organs. This is explained by increased mono-promotion of exotic species, despite the recognized importance of local species. In addition, limited diffusion of existing scientific information on species propagation could explain this lack of knowledge among farmers. However, as with ecosystem service knowledge, there is a significant difference between farmers' knowledge about species propagation. Farmers in male-dominated systems claimed to be able to produce their own planting material for 3 species present on their plot, whereas farmers in female-dominated systems could only produce material for one species. This gender disparity in knowledge could perhaps be explained by women's unequal access to extension training on plant breeding. Another important factor that could explain this disparity is cultural norms that make women believe it is inappropriate to engage fully with men, thus barring them access to training sessions on how to produce these species (Duffy et al. 2021). Taken together, the findings on farmer knowledge showing a gendered distinction in knowledge support the knowledge dynamics hypothesis, speculating that various individual socio-cultural and demographic traits such as gender, age and level of literacy or formal education are all correlated with an individual's knowledge level about plants (Gaoue et al. 2017).

Local knowledge and species management

Overall, results show differences in species abundance between the male-dominated and female-dominated systems. These differences could be influenced by various factors such as specific farming practices, plot management, access to land, knowledge and preferences of male and female farmers. However, this study identified that farmers' management of species abundance on their plots was more closely linked to their knowledge of the ecosystem services provided by these species, and varied between the two systems. Thus, differences in species composition and abundance were found between the two agroforestry systems identified in this study. The five most abundant species in the female-dominated agroforestry systems were edible species, whereas in the male-dominated system they were a combination of woody and edible species. This difference may reflect men's and women's preference for species and ecosystem services (Wagner et al. 2019; Hernandez Marentes et al. 2022). Furthermore, ritual restrictions, or taboos that restrict women's ability to select species and numbers of trees to keep or plant on their farms, may justify this difference in composition and abundance between the two systems (Kiptot and Franzel 2012b). Moreover, despite belonging within the same agricultural development pole, the fact that they belong to different agroecological zones may explain the difference in species composition and abundance between the two groups of farmers. Indeed, agroecological zones are homogeneous areas characterized by specific vegetation types, cropping systems and agro-pedological parameters (Gbemavo et al. 2014). As male-dominated agroforestry farming systems are much more predominant among farmers in agroecological zone 3, and female-dominated agroforestry farming systems are more predominant among farmers in agroecological zone 5, the species present on their plots in each group identified in this study can reflect the characteristics of the vegetation that characterizes the AEZ they belong to. Indeed, the difference in abundance observed for some species such as Elaeis guineensis and Parkia biglobosa (more abundant in male-dominated systems than in female-dominated ones) may be more a result of the agro-pedological conditions specific to each ZAE. This result therefore confirms the availability hypothesis, which suggests that the use and abundance of species is significantly affected by their availability in the immediate environment of local populations (Gaoue et al. 2017). Finally, species values determined by farmers' knowledge about the species in each agroforestry system may explain the difference in species abundance and composition observed on agricultural plots. According to Valencia et al. (2015), farmers based on their knowledge and the associated values to specific species attempt to maximize the services provided by these species by modifying the richness, abundance and composition of tree species combined with their agricultural production. They privilege trees with useful functional traits and eliminate species they consider unimportant or whose functional traits are prejudicial to the crop. This supports the dynamic knowledge hypothesis, which states that the difference in knowledge about a species influenced by socio-demographic factors can determine its use and therefore its presence. But differences in knowledge of the silvicultural practices required for species management or planting may be responsible for variations in diversity and abundance in our two agroforestry farming systems.

Despite being in the same agricultural development pole, belonging to different agroecological zones may explain the differences in species diversity and abundance between the two groups of farmers. Indeed, agroecological zones are homogeneous areas characterized by vegetation types, cropping systems, and specific agro-pedological parameters (Gbemavo et al. 2014). As male-dominated agroforestry farming systems are concentrated in agro-ecological zone 3 and female-dominated systems in agro-ecological zone 5, the species observed on farmers' plots in each group identified in this study can reflect the vegetation characteristics of the ZAE to which they are attached. Indeed, the difference in abundance observed for some species such as Elaeis guineensis and Parkia biglobosa (more abundant in male-dominated systems than in female-dominated ones) may be more a result of the agro-pedological conditions specific to each ZAE. This result therefore confirms the availability hypothesis, which suggests that the use and abundance of species is significantly affected by their availability in the immediate environment of local populations (Gaoue et al. 2017). Finally, species values determined by farmers' knowledge about the species in each agroforestry system may explain the difference in species abundance and diversity observed on agricultural plots. According to Valencia et al. (2015), farmers based on their knowledge and the associated values to specific species attempt to maximize the services provided by these species by modifying the richness, abundance and composition of tree species combined with their agricultural production. They privilege trees with useful functional traits and eliminate species they consider unimportant or whose functional traits are prejudicial to the crop. This supports the dynamic knowledge hypothesis, which states that the difference in knowledge about a species influenced by socio-demographic factors can determine its use and therefore its presence. But differences in knowledge of the silvicultural practices required for species management or planting may be responsible for variations in diversity and abundance in our two agroforestry farming systems.

Local knowledge and species prioritization

Species prioritization showed that most farmers, whatever gender, prioritized multi-purpose species which could provide several specific priority services responding to their daily livelihood requirements. Several studies had reported how farmers attach importance to native species' versatility for their agroforestry uses. In Mali, Faye et al. (2010); in Togo Padakale et al. (2015) and in Tanzania, Wagner et al. (2019); had reported that most native agroforestry species were identified as priorities by farmers because of their multipurpose functions providing several products and services, including many sources of income. This supports the versatility hypothesis, in support of the optimal foraging theory, which states that for the general use of plant species or their integration into technologies, local populations give more importance to versatile species that produce more benefits per unit of foraging time, and that as the abundance of higher-value plants increases, lower-value plants will no longer be used (Gaoue et al. 2017).

Furthermore, the results of the present study also reveal a difference between the two agroforestry farming systems for ecosystem services and species priority. This difference in species prioritization is probably a result of the different social roles and responsibilities of men and women. Men generally play a central role as the main providers within their families, which means engaging in high-value economic activities, such as wood production, in order to generate income to support their households; this can explain their priority for wood-producing species. While women are more likely to prioritize trees for livelihood essentials such as firewood, improving soil fertility, producing food and commercial products according to their traditional role of household head feeding and their own economic empowerment. This difference in species prioritization, moreover, may reflect a difference in species knowledge between men and women. According to Brandt et al. (2012), the difference in species knowledge between men and women determines the importance attached to species. Thus, the difference in prioritization of species by farmers in this study potentially correlates with farmers' knowledge of these species including they ecosystems services providing. Overall, in agroforestry farming systems dominated by both men and women, the ecosystem services provided by trees to support agricultural production by reducing soil erosion and improving soil fertility were not the main reasons behind farmers' prioritization of tree species. This may reflect farmers' lack of knowledge about the ability of these species to improve soil fertility. In addition, the availability of other forms of fertilizer on the market and farmers' investment in these new fertilizers may also explain this fact. Most priority ecosystem services use by farmers to designate their respective priority species were much more related to their livelihood support. The implication is that, above all, farmers actively pursue improved livelihoods through the acceptance of new technologies that can add value to their agricultural production (Feintrenie et al. 2010). Whereas this research indicates a prioritization of agroforestry species and their ecosystem services according to gender, Moore et al. (2014) report that farmers' species preference was more related to social class and agricultural experience and not gender. Similarly, Gram et al. (2018); Wagner et al. (2019) respectively in Uganda and Tanzania had obtained that there is no significant difference between men and women species and ecosystem services choice. This research demonstrates the importance of the integration of farmer-specific characteristics and gender inclusion in agroforestry species selection to optimize their adoption by farmers.

Conclusion

Species choice is essential for the success of sustainable agroforestry. Farmer involvement, consideration of local forest resources and farmer-specific factors are key to resilience and successful adoption of agroforestry technologies. But the top-down model often used to develop these technologies can overlook local resources present in farmers' immediate environment, their knowledge and attachment to these resources, and the gender specificity of agroforestry. This study contributes to addressing this gap, which is often cited as an obstacle to the expansion of agroforestry technologies, especially in Africa, where the choice of trees in agricultural landscapes is largely influenced by socio-cultural factors. As a result, it adopts a participatory approach, based on the resources present in the farmers' agricultural environment, their knowledge of these resources, their preferences for agroforestry species and the specific factors that characterize them, to propose canopy species that satisfy the needs of individual farmer categories and provide the greatest benefits to women and men. The results of this study provide a distribution of local species on agricultural plots, as well as a prioritization of these species and their ecosystem services according to gender and agroecological factors. Similarly, within the various agroforestry systems, farmers, despite of gender-dominance, prioritize versatile agroforestry. Contrary to expectations justifying the presence of trees in agricultural plots, the primary reasons for farmers' species prioritization do not align with the supportive functions for agricultural production. Consequently, their understanding of the roles of these species within their plots remains limited. Based on our findings, this study suggests that agroforestry development initiatives should consider local small-scale variations in social, economic, and ecological factors for agroforestry species choice. Moreover, it is recommended that local versatile species be further integrated into agroforestry technologies, with consideration of the specific needs and preferences of both men and women and their propagation methods should more widely be promoted. In an agroforestry context, species selection should aim to maximize the provision of ecosystem services while minimizing potential drawbacks, adapting to local agroecological conditions, and addressing the farmers' preferences and constraints, thus ensuring ecological, economic, and socio-cultural sustainability. In addition, agroforestry policies are required to improve farmers' knowledge of the agronomic benefits of these species, through in-depth education of farmers on the ecosystem services that these species can provide to support agriculture. This will enable the conservation of a wide variety of species in these environments. Similarly, these policies must ensure equality and equity in the promotion and extension of agroforestry practices, including training sessions on species propagation techniques, by taking specific measures that enable each male and female farmer to participate fully and reap the full benefits. For example, using specific women's groups for extension and learning sessions, ensuring that these sessions take place at times that are convenient for women in terms of their social responsibility, could improve women's knowledge of propagation methods for their priority species. Alternatively, agroforestry studies are necessary to test the performance of the priority species identified in this study under a wider range of soil and topographical conditions. Field tests of these species on agricultural plots, with the participation of farmers, should enable us to learn directly from them about the challenges and opportunities associated with planting these local species on their land. In order to provide practical recommendations that will support policies for the real inclusion of these species in agroforestry technologies, studies will be conducted to determine the ideal density and spacing for the growth and development of these species in a specific type of agroforestry system with a specific crop.