Trees as brokers in social networks: Cascades of rights and benefits from a Cultural Keystone Species

Indigenous trees play key roles in West African landscapes, such as the néré tree (Parkia biglobosa (Jacq.) R.Br. ex G.Don). We applied social–ecological network analysis to understand the social–ecological interactions around néré. We documented the benefits néré provides and the multiple social interactions it creates amongst a large range of actors. The flows of rights over the trees and benefits from them formed two hierarchical networks, or cascades, with different actors at the top. The two forms of power revealed by the two cascades of rights and benefits suggest possible powers and counter-powers across gender, ethnicity, and age. We documented how the tree catalyses social interactions across diverse groups to sustain vital social connections, and co-constitute places, culture, and relationships. We argue that a paradigm shift is urgently needed to leverage the remarkable untapped potential of indigenous trees and Cultural Keystone Species in current global restoration and climate change agendas. Supplementary Information The online version contains supplementary material available at 10.1007/s13280-022-01733-z.

This document provides details about the methods for building and analysing social-ecological networks to study social-ecological interactions around a tree species. The study followed six broad steps.
Step 1. Gathering data about the case study.
We conducted field work in the study site to collect qualitative and quantitative data about the social-ecological interactions around the tree. The aim was to achieve an inventory and a description of the interactions. The actors involved in Néré management, harvest, and transformation were first identified from observations and interviews and then confirmed through a household survey. The market related actors were identified through market studies.
The data used in this study came from various research projects conducted between 2013 and 2017 on land and tree tenure, value chains, and livelihoods related to Néré, in which two authors of this paper participated (Pehou et al., 2020). Our qualitative analysis was built on a synthesis of a rich dataset collected by four means.
First, ethnographic and participant observation and 36 qualitative interviews were used to gather data on the cultural, social and spiritual context and the access rights of different social groups. We observed the daily harvest activities in the fields, fallows, and woodlands. We documented land tenure and tree tenure regulations and restrictions. Through focus group discussions in 18 participatory workshops, we collected information on the actors involved in Néré activities, their relationships and exchanges, the importance of Néré for different social groups, the different types of fields and woodlands where it was harvested, the products used, the seasonality of use, and the threats to the species.
Secondly, we interviewed 180 women, randomly selected across ethnic groups: 62 Nouni, 81 Mossé, and 37 Fulani (these numbers are proportional to the population of each ethnic group across the selected villages). The semi-structured interviews dealt with the use of the Néré tree, the economic value of different Néré products, the economic and social exchanges around Néré, and the participation of household members in its management, harvest, and transformation activities.
Thirdly, we analysed the history of the sites to understand the evolution of access rights and changes in harvesting and use practices. For this, we listened to the life histories of six women aged 55 years or older and conducted semi-structured interviews with customary authorities, official authorities and technical staff from the Forest Service.
Finally, we studied the sumbala market to understand trade and use practices. For this, we surveyed 280 Néré traders and 133 consumers in 24 selling sites: traditional markets, food stores and shops around the study area and in the capital city of Ouagadougou.
Based on our knowledge of the field and the interactions described in the data, we defined the key human and non-human elements of the social-ecological system, which would become the nodes of the social-ecological network. We identified the ecological nodes (i.e., land, animals, trees, seeds on trees, and harvested seeds) and the social nodes (i.e., actors related to Néré). The list of social nodes included local inhabitants (described by their gender, marital status, age, and ethnicity, for example migrant Mossé or first spouses and other spouses of Nouni farmers) and external actors (described by their role, such as traders or consumers). The social nodes were types of actors (e.g. pastoralists) rather than individuals (e.g., Mr X, a pastoralist).
We also included a few meta-nodes (e.g., "any community member") to ease the coding of links. For example, as any community member may help people facing hardship, we create one interaction from "all community members" to "people in hardship". At the time of creating the mathematical network, the "all community members" node was replaced by several nodes (one for each type of community members) and the interaction was duplicated.
We created a spreadsheet with the node list and details (Table SI1).
We extracted information on social-ecological interactions to create a list of network links. For this, we browsed the collected data and recorded all mentioned interactions related to Néré. We excluded interactions that were mentioned only once to avoid overcomplicating the network with marginal or anecdotal links.
Step 4. Mapping the social-ecological system.
We drew the nodes on a whiteboard and mapped the interactions among them by drawing links on the board. The process was similar to a concept mapping or related approaches from systems thinking (Aubrecht et al., 2019;Davies, 2011). During this process, we defined a typology of links (Table SI2). The typology was not predefined but created from the data. Social to Social A helps B to conduct an action related to land or trees (e.g., participating in harvest) Payment Social to Social A pays B with cash or goods (goods that are not Néré products) (e.g., paying in cash to buy seeds) Ecology Ecological to Ecological An ecological element influences another (e.g., trees providing fodder to animals) Action Social to Ecological A conducts an action on an ecological element (e.g., harvesting or selling seeds) Contribution Ecological to Social An ecological element provides a contribution to the livelihoods or wellbeing of A (e.g., providing shade, income or nutrition) We built the map iteratively by adding new interactions at each iteration (see examples of initial and almost final maps in Fig. SI1 and SI2). At one point, it became clear that the drawing would be facilitated by separating the system into three levels, depending on where the interactions took place (land, tree, or product levels).
After each map was created, we checked it for inconsistencies, redundancies or omissions. Whereas the first map was drawn on a whiteboard, we drew the following ones with the VUE (Visual Understanding Environment) software (Tufts University, 2015). When the final map was agreed upon by the authors, we created a spreadsheet with the list of links included in the map. The links were described by the key attributes "Type", "Source Node", "Target Node", and "Level" (Table SI3).
Step 5. Creating the mathematical network.
We imported both spreadsheets (node list and interaction list) in R to build the mathematical network. We converted the meta-nodes into the corresponding nodes and duplicated the links from or to these meta-nodes. We transformed the information on interactions into an adjacency matrix, i.e. a square asymmetrical matrix with values of 0 and 1 and rows or columns representing nodes. A value of 1 in the matrix meant that there was a link from the row node to the column node.
In order to retain the information on the types of links (e.g., Rights, Work, or Payment), we built seven adjacency matrices, one per link type. We then created seven networks from the adjacency matrices with the "graph_from_adjacency_matrix" function in the R package igraph (Csardi, 2018). These seven networks were combined to create a multiplex network with seven layers (one by type of interactions). In this multiplex network (Fig. SI3), all nodes can exist in all layers and the links between different nodes occur within layers. The inter-layer links connect one node in one layer to the same node in another layer and are called "self-coupling" links.

Figure S3. The multiplex network. Each layer corresponds to a type of interaction. The same node can exist in several layers. A grey vertical link connects a node in one layer with the same node in another layer (self-coupling link)
The multiplex network was created as igraph object with an attribute "Level" for links. We plotted the full network using a force-directed layout algorithm (Fruchterman and Reingold, 1991), in which two nodes are close to each other if they are linked (which is visually depicted as a group of connected nodes that are in the same region of the plot) (see flattened version of the multiplex network in SI4).
With the multiplex network, we described the types of observed interactions between pairs of nodes and their number of occurrences in the network. These interactions were either reciprocal (two links of opposite directions connect two nodes) or not. We also calculated the degree of each node (number of links to or from a node). Figure S4. Flattened representation of the multiplex network (the area of each node is proportional to its degree and the node label shows the node degree d).
We focused on two sub-networks, both built using the full set of nodes and a subset of selected links. The first one, called 'network of rights', showed how people exchanged rights among themselves and acted on ecological elements (selected links were "Right" and "Action"). The second one, called 'network of benefits', showed how ecological elements interacted and produced contributions to people (e.g. supplying seeds) and how people interacted to exchange these contributions (e.g. giving seeds) (selected links were "Ecology", "Contribution", and "Transfer"). We drew these two subnetworks and analysed their hierarchical nature. A hierarchical network is one in which paths (sequences of links) are not reciprocated, for example, if node A influences node B, node B cannot influence node A, directly or indirectly through other nodes. Hierarchical networks show the pyramidal structure of a group, in which dominant actors can be identified.
We calculated a simple index to identify the degree of hierarchy of the full network and the two subnetworks (the Krackhardt hierarchy score, which has a value of 1 for a fully hierarchical network and 0 for a non-hierarchical one) (Krackhardt, 1994). The Krackhardt hierarchy measure was calculated with the hierarchy function in the sna package for (Butts, 2016).
In the hierarchical sub-networks, we identified the dominant actors, i.e. the nodes that connect to a large number of other nodes following the directed links of a network. In the sub-network of rights (respectively benefits), this would be the nodes that can transfer rights (benefits) to many other nodes, directly or indirectly. The two sub-networks were plotted using a layout algorithm for hierarchical graph drawing (Sugiyama et al., 1981), in which the most dominant actors are at closest to the top (Fig. SI5). The dominance was measured as the length of subcomponents with igraph, and more specifically with the function "subcomponent(myGraph , j, mode="out")", which determines all nodes reachable from a give node j via a directed path through the network myGraph.