1 Introduction

Agroecology is increasingly recognized as a transformative paradigm for building sustainable and resilient food systems in the face of climate change, land degradation, and rural inequality [1, 28]. As both a scientific discipline, a practice and a political movement, agroecology integrates ecological principles with local knowledge, farmer autonomy, and social justice, offering an alternative to industrial input-dependent models of agriculture [2, 20]. In many parts of the Global South, particularly Sub-Saharan Africa and Latin America, grassroots organizations, farmer movements and civil society coalitions have become central actors in driving transitions to agroecology [52,47, 51].

Agroecology actors often self-organize into what is known as “agroecology network organizations.” These are structured collaborations among coalitions, alliances, and other platforms comprising diverse stakeholders, such as farmer associations, NGOs, research institutions, and advocacy groups. Different agroecology network organizations have been established around the world to harness the opportunities offered in such networks. For instance, in the Philippines, Magsasaka at Siyentipiko para sa Pag-unlad ng Agrikultura (MASIPAG) promotes farmer-led seed systems through decentralized experimentation [23]. Colombia’s Red MAC combines local market development with territorial governance and gender inclusion [17]. In Latin America, movements in Ecuador, Bolivia, and Brazil have established agroecology as a bottom-up alternative development pathway [32, 52]. In Senegal, agroecological coalitions comprising farmers, NGOs and state actors are advancing climate-resilient territorial strategies [6, 7]. Likewise, in Uganda, several agroecology-focused networks have emerged, including the Participatory Ecological Land Use Management (PELUM) Uganda, East and Southern Africa Farmers’ Forum (ESAFF), the Advocacy Coalition for Sustainable Agriculture (ACSA), and the Food Rights Alliance (FRA) [9].

The networks provide avenues for fostering knowledge exchange, resource sharing, political advocacy and collaborative experimentation [16, 33]. They often engage in training, policy advocacy, farmer mobilization and participatory research to advance agroecological practices and food sovereignty [9, 36]. While the visibility and political engagement of these networks are increasingly well documented [9, 36], little is known about their internal organizational structures, specifically, the collaboration dynamics that shape their effectiveness and cohesion. The effectiveness of agroecology networks depends not only on shared values or mandates but also on their internal relational dynamics, how members interact, connect, and support one another [6, 7, 25].While agroecology networks play a central role in agroecology promotion, many such networks experience internal fragmentation, coordination failures, and asymmetric relationships between members [25].These challenges stem not only from ideological differences or competition over resources but also from weak communication structures, unequal resource flows and the absence of tools to assess "network health” [8, 31]. Network health describes the strength, balance, and inclusiveness of internal collaboration patterns [30]. Understanding how collaboration unfolds within and across such networks is crucial to grasping the internal forces that drive their broader impact. While some prior studies on agroecology networks in Uganda have explored discourse [35] or civil society-state interactions [9], there is a dearth of studies examining the structural relationships among the member organizations themselves.

Social Network Analysis (SNA) provides a valuable methodology for bridging this gap. By visualizing and quantifying connections among actors, it helps identify central hubs, peripheral members, and bridging organizations that facilitate (or obstruct) coordination and knowledge exchange [45]. It provides both a descriptive and diagnostic lens, revealing patterns of interaction that may be invisible in formal reporting or conventional surveys. This SNA method has been widely applied in sectors like public health, education, and agricultural innovation to inform strategy, governance, and institutional learning [10, 11, 15]. In the agroecology field, however, especially in African contexts, such methods remain underutilized.

SNA remains underutilized in agroecology, particularly among grassroots civil society networks in Africa [61]. This study offers an original contribution by applying a dual-tool SNA framework (Gephi and Kumu) to visualize and quantify collaboration patterns in an African agroecology network. This paper provides a novel diagnostic lens to assess and strengthen internal coordination, trust, and knowledge flows across agroecology networks. Hence, by applying SNA to an agroecology network in Africa, namely PELUM Uganda, this study examines the structural relationships among the member organizations.

Using PELUM Uganda as a case study, this study presents a replicable, participatory, mixed-methods approach to SNA for evaluating inter-organizational collaboration within agroecology networks [31, 57]. We apply two complementary tools: Gephi, which produces modularity-based visualizations to identify clusters and potential silos [22], and Kumu, which calculates key network centrality metrics, degree, betweenness, closeness, and eigenvector [43], to assess actor influence, brokerage, knowledge exchange, and collaborative learning. Together, these tools allow for both qualitative insights and quantitative measurement of how member organizations connect, support, and relate to each other. In doing so, this methodological paper was guided by the following research questions: What are the structural patterns of collaboration within the PELUM Uganda network? Which organizations serve as key connectors, brokers, or isolates within the network? How can visual and metric-based SNA tools be used to improve internal coordination, learning, and strategic alignment? By answering these questions, the study contributes a novel methodological framework for diagnosing internal functionality in agroecology networks. The paper provides a useful framework for assessing and strengthening partnerships, reducing silos, to scale agroecology through enhanced organisational connectivity.

The next section presents a literature review on SNA followed by an overview of the PELUM network. The subsequent section, “Step-by-Step Application of dual SNA Method”, details the research methods and a sample of findings through a five-step process to provide practical guidelines for applying a dual SNA approach. The paper concludes with a discussion of the findings and their implication for network-based approaches in agroecology.

2 Social Network Analysis (SNA) as a tool for network diagnosis

Social Network Analysis has gained prominence as a methodological and analytical approach to understanding the structure, dynamics, and influence of relationships among actors in various domains, including health, governance, agriculture, and sustainability transitions. Initially rooted in sociology and graph theory [21], SNA enables the visualization and quantification of actor relationships based on connectivity, influence and position in a network [5]. Key metrics, such as degree centrality, closeness centrality, betweenness centrality and eigenvector centrality, help diagnose how information and resources flow, identify where structural bottlenecks exist and determine which actors function as hubs, brokers, or isolates [5]. In their seminal work, Hoppe and Reinelt [31] applied social network analysis (SNA) to evaluate leadership networks, arguing that understanding both formal and informal linkages is essential for assessing effectiveness [31]. Similarly, Prasetyo [48] emphasized the importance of mapping informal leadership within community organizations [48]. These studies underscore SNA’s versatility, not only for measuring collaboration but also for revealing hidden dynamics that are often overlooked in traditional assessments. As argued by Helbling and Anderson [24], who applied SNA across five distinct network types (including innovation, work, and informal networks), such a layered approach highlights the complex architecture of collaboration [24]. These applications demonstrate that SNA functions not merely as a descriptive tool but also as a strategic resource for enhancing collaborative processes and enabling systems transformation [4].

SNA has been increasingly applied to diagnose networks functioning in diverse domains in different parts of the world [10, 11, 15, 25, 26]. For example, in the context of agricultural innovation, Cadena et al. [10] employed SNA to map knowledge flows among poultry trainers in California, revealing key boundary spanners and uncovering critical communication bottlenecks that constrained the diffusion of new practices. In a similar vein, Cunningham et al. [15] applied SNA to assess Indigenous health partnerships in Australia, demonstrating how network mapping can illuminate central actors, structural weaknesses and collaborative performance across health systems. Likewise, Christensen and O’Sullivan [11] employed SNA to investigate the evolution of food system collaborations in Canada, highlighting the importance of longitudinal analysis in capturing shifts in relational dynamics and the sustainability of grassroots alliances over time. Similarly, Hermans et al. [25] employed SNA to assess multi-stakeholder platforms, revealing weak inter-level ties and underrepresented voices, particularly from the private sector or grassroots actors. These insights resonate with the challenges faced by agroecology networks in sub-Saharan Africa, where fragmentation and funding dependencies weaken collective action [34, 42].

While the utility of SNA in assessing collaborative networks is widely recognized, several gaps persist. Many studies rely solely on a single tool (e.g., ORA, PARTNER, UCINET), thereby limiting the breadth of analytical possibilities [8]. Moreover, survey-based SNA studies often fail to incorporate stakeholder validation, leading to critiques of methodological extractivism [41]. There is also limited integration of spatial, directional and qualitative data, which are vital for understanding both formal and informal dimensions of collaboration. This study builds on and extends existing literature by using a dual-tool approach (Gephi and Kumu) that integrates both structural and relational dimensions of network health. Gephi’s modularity algorithms enable the identification of clusters and silos, while Kumu’s interactive dashboards facilitate the measurement of influence and information flow. More importantly, this research incorporates participatory validation, thereby aligning analytical insights with local knowledge of network members and enhancing the trustworthiness of the findings [41]. Furthermore, unlike earlier studies that emphasize leadership networks [31] or institutional scaling [11], this study focuses on strengthening operational collaboration among grassroots agroecology actors. It positions SNA not only as a diagnostic tool but also as a developmental methodology that can inform strategic decision-making, partner profiling, and capacity building within agroecology networks. The researchers used PELUM Uganda as a case study.

3 Case study: PELUM Uganda

PELUM Uganda is part of a regional network operating in 14 African countries, including Uganda, Kenya, Tanzania, Botswana, Zimbabwe, among others. PELUM Uganda was founded in 1995 to promote sustainable land use, ecological farming and farmer empowerment. It consists of 74 members, including community-based organizations, individual, faith-based institutions and national NGOs working on food systems, natural resource governance and rural development (PELUM [46]). PELUM Uganda was selected as the case study for three reasons: (1) its wide geographic operations covering 122 of Uganda’s 146 districts, (2) its diverse and extensive membership; and (3) its central role in agroecology advocacy and practice in the region (PELUM [46]). In Uganda, the Ministry of Agriculture, Animal Industry and Fisheries recognize PELUM Uganda as a key driver of the national agroecology strategy [56], Biovision [3]. Alongside PELUM, the study recognizes the relevance of other regional agroecology networks such as Eastern and Southern Africa Farmers Forum (ESAFF), Advocacy Coalition for Sustainable Agriculture (ACSA), Alliance for Food Sovereignty in Africa (AFSA), and Food Rights Alliance (FRA), which often interact or overlap with PELUM Uganda members in various collaborative efforts across Uganda and the broader East African region [9].

The next section, “Step-by-Step Application of dual SNA Method”, details the research methods and a sample of findings through a five-step process to provide practical guidelines for applying a dual SNA approach. Step 1 covers data collection via interviews and focus groups. Step 2 involves organizing data for network tools (Gephi and Kumu). Step 3 applies visual and metric analysis to map collaboration patterns. Step 4 interprets findings to identify key actors and gaps. Step 5 validates results through stakeholder engagement.

4 Step-by-step application of dual SNA method

Step 1: data collection

Data was collected through Focus Group Discussions (FGDs), Semi-Structured Group Interviews (SGIs), and In-Depth Interviews (IDIs). During data collection, NGO staff of PELUM member organizations were asked to indicate the organisations with which they currently collaborate, what they do together, and how these collaborations could be strengthened. The interviews and discussions were audio-recorded with permission from the respondents, transcribed verbatim and supplemented by secondary sources such as organizational reports and websites. Direct quotations were used to ground the findings in real-world experiences [40, 50]. Organizations were anonymized to uphold confidentiality [55]. Participants' anonymity was maintained by assigning codes to represent names of organizations [55]. The code MOS(X) represents the member organizations where “X” is the number that identifies a particular organisation. For example, MOS1 and MOS25. The partners of the organisations are presented by POS (X), for example POS1 and POS2. The researchers recorded the names and codes of each organization and used this background knowledge of who each organization is to guide the data analysis. The researcher maintained only codes in the report for ethical reasons.

Step 2: data preparation

To explore the collaboration dynamics and structural roles within PELUM Uganda’s organizational ecosystem, we applied a mixed-methods network analysis using Gephi and Kumu, two complementary tools widely used in social network research for visual and metric-based analysis [22, 43].The data from the field work was input in two templates, Kumu's data template for visualisation in Kumu and a Gephi data template visualisation in Gephi. The Kumu's Stakeholder Template is a structured MS Excel spreadsheet comprising two primary sheets: elements and connections. The Elements sheet (Table 1) lists individual stakeholders and organizations, categorized under default types such as “Person” and “Organization,” along with relevant descriptive attributes like roles or affiliations. The connections sheet (Table 2) defined the relationships between these stakeholders by specifying a source ("From"), a target ("To"), and the type of connection (e.g., "Affiliated with" or "Collaborates with"). This structured format ensured compatibility with Kumu’s import system, enabling a clear and organized visualization of stakeholder dynamics. The prepared dataset was then imported into Kumu (https://kumu.io/), where visual and structural adjustments were applied to refine the stakeholder map for analysis [13, 38, 44]. Connections between organizations were revealed through interviews and document reviews. For example, one member of staff revealed that “MOS2 has worked with MOS6 in farmer training programs and was introduced to organic farming through […] partnerships." (FGD, MOS2). During an FGD with MOS6, a participant remarked: “MOS6 mobilized beneficiaries to participate in training at MOS2” (FGD, MOS6). Following this revelation, connections were identified between MOS2 and MOS6, as well as between MOS6 and MOS2.

Table 1 Elements in Kumu data template
Full size table
Table 2 Connections in Kumu data template
Full size table

On the other hand, for the Gephi data template (https://gephi.org/), the data was also structured into an MS Excel template with two sheets: a Nodes and an Edges file. Gephi is an open-source platform licensed under the GNU—General Public License (GPL3), designed for the interactive visualization and analysis of networks and complex systems, including dynamic and hierarchical graphs. It supports the visualization of social networks and offers tools for computing key SNA metrics, as well as partitioning, ranking, and filtering of nodes and edges [27]. Building on the Element and Connection data identified from the Kumu data template, the Nodes file, as sampled in Table 3, contained the identifiers for each stakeholder which define nodes in the network while the Edges file contains the connections between the stakeholders based on the Connection File in the Kumu data template. Unlike the connection file in Kumu, the edges file of Gephi (Table 4) captured the connections between these stakeholders by specifying a “Source” and “Target” for each relationship. For example, MOS6 was connected to multiple nodes, including MOS2, PMO3, POS22, and POS23, illustrating its central role in the network. After cleaning the data for consistency, such as ensuring correct ID formatting and removing duplicates, the Kumu data template and the Gephi data template were imported into the respective software, Kumu and Gephi.

Table 3 Simplified excerpt of the Gephi nodes dataset showing unique identifiers (Id) and corresponding labels (Label) for selected Member Organizations (MOS20, MOS12, MOS6) within the PELUM Uganda agroecology network)
Full size table
Table 4 Simplified excerpt of the Gephi edges dataset showing relationships
Full size table

Table 3 presents a simplified excerpt of the “nodes” dataset used in Gephi for social network analysis. Each row corresponds to a unique actor (node) in the network. The Id column provides a unique identifier for the node, which must be consistent across both the Nodes and Edges tables to ensure proper linkage in Gephi. The Id is a mandatory field in Gephi datasets, serving as a unique key for each node. It ensures consistency when linking nodes to their corresponding edges in the Edges table. The Id cannot be duplicated. The label column gives the readable name of the node, which is displayed in the network visualization for easier interpretation. In this example, the nodes correspond to specific Member Organizations (MOs) within the network under study, coded as MOS20, MOS12, and MOS6.

The label is an optional but recommended field, providing a human-readable name for each node. While the Id is primarily for internal reference, the Label is what appears on the network graph, making the visualization interpretable for researchers, practitioners, or policymakers.

Table 4 illustrates a simplified excerpt of the ‘edges’ dataset used in Gephi for social network analysis. Each row represents a connection (edge) between two nodes (actors) in the network. The Source column indicates the origin node, while the target column identifies the destination node of the relationship. The node id and node label columns list the identifiers and labels of the connected nodes, ensuring consistency with the Nodes table. In this example, edges capture relationships between Member Organizations (MOS) and Partner Member Organizations (PMO) within the agroecology network, such as links between MOS20 and PMO1, MOS12 and PMO2, and MOS6 and MOS2. In the Gephi edges dataset, the Source represents the originating node of the relationship, while the Target refers to the node to which the relationship is directed. The Node Id column contains the unique identifiers of the connected nodes, ensuring consistency with the Nodes table, whereas the Node Label provides the readable names of these nodes, which are displayed in the network graph for easier interpretation.

Step 3: Network visualization and structural mapping

The first visual mapping of the relationships among the stakeholders was done using Gephi (Fig. 1), which allowed for an intuitive mapping of the network’s topology [19, 60]. Nodes represented individual organizations, including participating PELUM member organizations (MOS), partners (POS) and referenced but non-participating members (PMO). Gephi visualization (Fig. 1) identified hubs and peripheral nodes, with clusters indicating subgroups and silos. The network visualisation conducted with Gephi revealed the collaboration pathways of PELUM Uganda member organizations. The analysis illustrates the structure, connectivity, and relationships within the network by identifying key nodes (i.e., significant organizations), sub-network clusters, and potential areas for strengthening connections. In the network diagram (Fig. 1), each node represents an organization labelled (for example, as MOS1, POS4, or PMO6). The network structure indicates that some organizations, such as MOS6 and MOS17, have a significantly higher number of connections and function as key players. In contrast, organizations such as MOS4, MOS13, MOS11, MOS5, and MOS12 exhibited fewer connections, indicating a more peripheral role.

Fig. 1
figure 1

Network diagram of PELUM members from Gephi software. MOS are PELUM member organizations who participated in the study, POS  are PELUM partners, PMO  are PELUM member organizations who didn’t participate in the survey; the number after the organization acronym indicated the number of the organization.

Full size image

Step 4: Quantitative network metrics

To complement the visual insights, we conducted a metrics-based analysis using Kumu, focusing on four core measures of centrality (Table 5) commonly used in organizational network studies [18,5]. The integration of Gephi’s visualization with Kumu’s metric analysis allowed for both a macro-level understanding of network cohesion and micro-level insights into specific organizational roles. This dual approach facilitated the identification of central actors, bridging nodes, and under-connected organizations whose integration could enhance network efficiency and resilience [14, 49].

Table 5 Metric-based analysis
Full size table

Adapted from [5, 18, 45]

The Kumu network visualisation (Fig. 2) highlighted central hubs, influencers and bridges. Edges represented by the arrows in the diagram (Fig. 2) show the ties or relationships between nodes (organizations). Directed edges denoted reported collaborations, with arrow direction indicating the flow of support (e.g., knowledge, funding, or resources). The variation in the number of edges, from highly connected hubs to isolated nodes, illustrates differences in interaction intensity. Clusters of interconnected nodes (organizations) indicate that some entities work closely together while others remain less integrated, potentially operating in silos.

Fig. 2
figure 2

Network diagram from the Kumu software. Circles represent PELUM member organizations that participated in the study, while triangles indicate PELUM partners identified by the participating organizations. Diamonds represent PELUM members who were not part of the study but were mentioned as partners by the participating organisations. Directed lines indicate connections between organisations, and arrows point to the organisation receiving support.

Full size image

By examining the direction of arrows in the network diagram, the authors were able to understand how resources such as knowledge, funds, or information flowed within the network. For example, Fig. 2 shows that organizations MOS6 and MOS7 are central hubs on the network since many arrows are pointing towards them. This suggests they receive support from many organizations. Arrows pointing away means that the organization is providing support to others. For pairs of nodes with arrows pointing in both directions, it means reciprocal relationships. This indicates a mutual or reciprocal relationship. For example, MOS67 and MOS64 have arrows pointing at each other, suggesting a two-way interaction.

Further use of the Kumu metric revealed the structural dynamics of PELUM member organizations using key network metrics such as degree closeness, betweenness, and eigenvector centrality (Fig. 3). Based on the Degree centrality, MOS6 and MOS4 are the main hubs within the PELUM network, highlighting their role in facilitating frequent direct connections. However, these organizations are not necessarily well connected to the broader system. Size metrics (Fig. 3) reinforced this observation, with MOS4 and MOS6 maintaining intense direct engagements, while organizations with lower degrees of centrality and size values occupied peripheral positions with limited interaction. Measuring the closeness centrality, which reflects the efficiency of information dissemination across the network, ranked MOS5 highest, followed by MOS4 and MOS17. These organizations are well positioned to access and spread information across different network areas. While the betweenness centrality, measuring the extent to which organizations act as a bridge between others, revealed MOS4 as the most critical intermediary, followed by MOS6 and MOS7. These organisations play a vital role in controlling the flow of information and facilitating collaboration pathways. Finally, the Eigenvector centrality identified MOS2 and MOS7 as the most influential organizations due to their strong connections with other well-connected organizations. Other key players, such as MOS6, MOS22 and MOS4, also exhibit strong influence, positioning them as strategic leaders within the PELUM network.

Fig. 3
figure 3

PELUM Uganda Network Metrics. Degree centrality: Organizations with the most direct connections, such as MOS6 and MOS4, play a key role in frequent interactions but may not always connect across the broader network. Information flow and closeness centrality: MOS5 leads in information dissemination, followed by MOS4 and MOS17. Bridging organizations and betweenness centrality: MOS4, MOS6, and MOS7 act as critical intermediaries, controlling information flow and influencing collaboration pathways. Influential nodes and eigenvector centrality: MOS2 and MOS7 emerge as the most strategically influential organizations, potentially driving decision-making and advocacy within the PELUM network. Optimizing network reach: high-reach organizations like MOS17 and MOS4 can be leveraged for better information dissemination.

Full size image

Collaboration pathway analysis using Kumu revealed interconnected clusters, with some organisations working closely together while others remained less integrated, operating in silos. This pattern aligns with previous findings by, which showed that hierarchical networks limit inclusivity and resource sharing. According to Lee et al. (2020), organisations often operate in silos due to limited trust, scarce resources, and centralised communication structures. Organizations with lower centrality and size values tend to be on the periphery and require better strategic integration into network activities. Organizations with fewer connections may be less integrated into the PELUM network, reducing their potential contributions, a concern echoed in other sustainability networks [5]. The presence of distinct sub-network clusters suggests that some groups are more interconnected than others, often based on regional or thematic collaborations [26]. Researchers observed during fieldwork that shared interests and proximity drive the formation of clusters within PELUM. Additionally, specific project teams or donor-aligned collaborations may also contribute to these clusters [39]. According to a PELUM Secretariat staff member (IDI, PMO2), existing regional clusters facilitate networking among member organizations. However, interview findings indicate a lack of awareness about these clusters among members, who continue requesting the formation of clusters.

Organizations such as MOS4, MOS6, and MOS7 are central hubs capable of bridging members (Fig. 3) because they do best at information sharing and have high visibility within the PELUM network. Since the PELUM network involves organizations working on agroecology and climate adaptation, these central organizations are vital for leadership and knowledge-sharing platforms, like other findings in agricultural innovation networks [33]. This highlights the importance of leveraging central organizations to improve communication and collaboration with periphery organisations. Effective partnership and collaboration require multi-level coordination among the MOS, between MOS and farmer organizations, with the PELUM secretariat and across external networks such as POS2, POS22, and POS24.

Step 5: Validation

Participant validation is a strategy to enhance the trustworthiness and ethical integrity of qualitative research by presenting data or preliminary analyses to study participants for assessment. This approach can generate new data and serve as an opportunity to verify the rigour of data collection and analysis [41]. To confirm the data obtained, a meeting was convened with study participants after data collection. On March 19, 2025, a validation meeting was held via Zoom to discuss the findings. Participants engaged in rich discussions and called for PELUM Uganda to profile member organizations, develop a partnership strategy, encourage joint fundraising, and proactively address collaboration barriers within the network. Regular debriefing sessions were conducted on Zoom with all co-authors. On Mar 29, the draft findings were presented to the Directorate of Graduate Studies, Research and Enterprises of Uganda Martyrs University for further scrutiny and comments.

On 25 April 2024, the researcher presented the study findings to the entire PELUM Uganda network member organization during their Country Working Group Meeting, an annual general assembly held at Hotel Africana in Kampala. This followed an earlier online validation meeting that was limited to PELUM Member Organizations who directly participated in the study. The physical gathering provided an opportunity to share not only the key findings and recommendations but also the methodological steps taken to collect and analyze the data. The response from participants was affirming and constructive. One of the members emphasized the credibility of the research, noting that the findings resonated with what is happening on the ground. Participants asked whether PELUM could consider mechanisms to recognize and reward members that demonstrate strong collaboration and effective partnerships. The Chairperson of the PELUM Uganda Board welcomed the study as a much-needed external evaluation that revealed the ‘health’ of the network and encouraged members to be more transparent and communicative. Lastly, some members expressed concern that the PELUM Secretariat is overstretched, likening it to “a hen with many chicks failing to take care of all of them.” This metaphor effectively captured the sentiment that the Secretariat may be unable to adequately support all its 74 members under the current structure, and it highlighted the need for improved internal communication and strategic focus. The presentation at the Country Working Group Meeting not only validated the research's relevance but also generated actionable feedback that informed the refinement and communication of the findings.

5 Discussion

This study set out to develop and test a replicable, participatory methodological approach for diagnosing and strengthening collaboration among agroecology network organizations, using the PELUM Uganda network as a case study. By combining Gephi’s modularity-based visualization with Kumu’s centrality metrics, the study addresses a key methodological gap in agroecology systems thinking: the absence of robust tools for analyzing internal network dynamics.

The findings demonstrate that this dual-tool SNA approach offers a nuanced understanding of how agroecology organizations interact, highlighting patterns of connectivity, influence, and fragmentation that often remain hidden in conventional assessments. The central actors; MOS6, MOS7, and MOS17 not only held structural prominence on the network maps but also actively sustained collaboration through concrete practices. For instance, MOS6 frequently partnered with fellow organizations to co-host farmer training programs, thereby creating shared learning spaces and reinforcing trust across clusters. MOS7 cultivated reciprocal partnerships with both core and peripheral members, enabling the circulation of resources and knowledge in two directions rather than through one-way flows. Similarly, MOS17 played a bridging role by convening knowledge dissemination events that brought together disconnected members, ensuring that innovations and experiences reached a wider audience. These practices: joint trainings, reciprocal partnerships, and inclusive knowledge-sharing platforms proved critical for ensuring connectivity, creating influence, and overcoming silos. They illustrate that operational collaboration, embedded in everyday activities, is what transforms structural centrality into functional influence, thereby strengthening the overall health and resilience of the PELUM network [37, 58].

Compared to traditional tools such as stakeholder matrices or perception-based evaluations, this approach provides greater analytical depth and practical relevance. It equips researchers and practitioners with evidence-based insights to improve network cohesion, learning, and strategic coordination. Key actors specified above (MOS6, MOS7, and MOS17) emerged as structurally central within the network. They played significant roles in brokering knowledge, facilitating resource flows, and linking disconnected clusters. Their influence extended beyond formal mandates, rooted instead in trust-based relationships and histories of collaboration, as confirmed during validation sessions. These insights support Helbling and Anderson’s [24] view that centrality, particularly betweenness, is a useful proxy for identifying influential actors shaping collaboration flows through both formal and informal means.

The study also builds on Cadena et al. [10] and Cunningham et al. [15], reinforcing the importance of mapping relational structures to inform capacity-building and coordination strategies. Unlike Hoppe and Reinelt [31], who focus on leadership development, our approach emphasizes operational collaboration among grassroots actors. This shift in focus reveals how everyday interactions, not just strategic positions, shape the effectiveness of agroecology networks. While highly central actors are vital, the presence of weakly integrated, peripheral organizations such as MOS11, MOS13, and MOS5 exposes risks of fragmentation. These findings echo concerns raised by Hermans et al. [25] and Cunningham et al. [15], who warn that without intentional strategies to foster inclusion, networks may replicate hierarchical or siloed dynamics. In this case, Gephi’s visualization revealed distinct sub-clusters—often shaped by region or donor alignment, indicating uneven integration. Participant feedback called for increased awareness of existing clusters and more deliberate efforts to bridge them, reinforcing the utility of SNA for network health diagnostics.

The combined use of Gephi and Kumu offers a robust, multi-dimensional approach to evaluating collaboration. As Christensen and O’Sullivan [11] and Cramer et al. [12] have shown, SNA enables organizations to assess both structure and function [12]. In our case, centrality metrics provided actionable insights: MOS5's high closeness centrality suggests strong potential for disseminating information, while MOS2 and MOS4’s high eigenvector scores indicate their influence through well-connected ties. These findings can guide strategies such as mentorship pairings, peer-learning platforms, and cross-cluster collaborations. Furthermore, the study highlights asymmetries in directional collaboration—some organizations consistently provide support while others mostly receive. This raises questions about equity, capacity, and sustainability within the network. As Cadena et al. [10] suggest, identifying and integrating under-connected actors is crucial for strengthening collective performance.

This research also contributes methodologically by advancing multi-layered SNA approaches. Building on Christensen and Helbling’s (2016) calls for richer network typologies, our integration of modularity analysis and four centrality dimensions offers a more comprehensive understanding of collaboration. While Cadena focused on poultry trainers and Cunningham on healthcare partnerships, this study uniquely applies SNA within a grassroots agroecology context, capturing the relational dynamics of a civil society-led network. Moreover, unlike Bowen et al. [8] or Hoppe and Reinelt [31], who emphasize strategic impact or leadership evaluation, this study centers on functional diagnostics—mapping the everyday realities of coordination, knowledge exchange, and support flows. Moreover, by combining technical analysis with participatory validation, the study reflects agroecology’s values of co-creation, inclusivity, and grounded learning [53].

Unlike previous SNA applications focused on leadership or policy networks [62], this study provides a methodological innovation by combining Gephi’s modularity mapping with Kumu’s centrality analysis to examine inter-organizational collaboration in a grassroots agroecology network. This dual-tool approach captures both formal and informal relationships and is further strengthened by participatory validation sessions that surfaced trust-based dynamics and hidden barriers. This layered methodology is uniquely suited to agroecological contexts where relational, horizontal collaboration is essential but often poorly documented. This paper fills a methodological gap by offering a replicable framework for diagnosing network health and identifying leverage points for improved coordination and equity [11, 40].

The findings are not only diagnostic but developmental. They offer a foundation for practical interventions, such as member profiling, partnership strategies, and incentive structures to foster collaboration. The insights into informal dynamics, such as trust and historical ties, emphasize that strengthening agroecology networks requires attention to both structural and relational dimensions. Finally, in comparing methodologies, our study distinguishes itself through its integration of qualitative coding, visual mapping, and participatory feedback from 194 actors across the PELUM Uganda network. Unlike survey-based SNA studies such as MILES [24] or those by Cadena et al. [10], our approach blends stakeholder narratives with structural analysis to enhance contextual validity and applicability.

6 Conclusion

This study offers a significant methodological contribution to the analysis of collaborative agroecology networks by integrating Gephi’s modularity-based visualization with Kumu’s centrality metrics. This dual-tool approach enables a multi-dimensional diagnosis of both formal and informal inter-organizational relationships, moving beyond conventional, single-tool or solely structural analyses. By visualizing modular clusters and quantifying network centrality roles, including degree, betweenness, closeness and eigenvector metrics, this methodology captures the internal dynamics of agroecology networks with greater granularity and interpretive power.

A further strength of this study lies in its commitment to participatory validation. Engaging network members in reviewing and reflecting on the SNA findings not only enhanced the ethical robustness of the research but also improved the contextual accuracy and uptake of its insights. This participatory dimension aligns with agroecology’s ethos of inclusivity, co-creation, and grounded learning and stands in contrast to more extractive or technocratic approaches to network evaluation. Validation sessions revealed high levels of resonance between the mapped relationships and stakeholders lived experiences, offering a foundation for practical reforms such as improved profiling of member organizations, incentive frameworks for collaboration, and regionally grounded clustering strategies.

Empirically, the study has shed light on the relational architecture of PELUM Uganda, one of the most expansive agroecology networks in the region. The analysis identified key central actors (e.g., MOS6, MOS7, MOS17), peripheral organizations at risk of disconnection, and directional collaboration flows that suggest asymmetries in support, influence, and access to information. The presence of tightly clustered subgroups alongside isolated nodes reflects both functional sub-networks and structural silos, which can simultaneously support innovation and undermine cohesion if not addressed strategically. These insights mirror concerns in the literature about fragmentation and the overstretching of hub organizations in multi-stakeholder platforms [15, 25].

Despite its contributions, the study has several limitations. The simplified categorization of actors—using labels such as MOS, POS, and PMO, may flatten nuanced roles, especially in a fluid and evolving organizational environment. Furthermore, the reliance on participant-reported ties introduces subjectivity into the mapping process, as collaboration may be interpreted differently by respondents depending on institutional culture, funding relationships, or personal history. This challenge resonates with findings from Helbling and Anderson [24], who noted that network perceptions often diverge among actors, complicating efforts to establish a shared understanding of collaboration. Additionally, the cross-sectional nature of the analysis captures only a snapshot of the network at a given point in time. Relationships in agroecology networks are dynamic, often shifting due to funding cycles, leadership changes, or external policy interventions. As recommended by Cunningham et al. [15], future studies should adopt longitudinal social network analysis (SNA) approaches to track the evolution of collaboration patterns. Such longitudinal data would allow network managers to identify trends, respond to disruptions, and measure the impact of interventions more effectively. Similarly, integrating spatial data could uncover how regional proximity, ecological context, or geographic access influence the formation and sustainability of collaborative ties, an extension supported by Cadena et al.’s [10] findings on regional disparities in training effectiveness.

SNA may obscure dynamics of social ties, particularly in complex and evolving contexts like agroecology. Static representations of the SNA may fail to capture the dynamic evolution of networks, while the focus on structure can overlook the functional and qualitative aspects of knowledge exchange and power relations [54]. In this study, after generating network maps and metrics, researchers convened validation sessions with PELUM Uganda members. This step ensured that informal ties, trust-based relationships, and hidden barriers were surfaced and cross-checked against the visual outputs. The study combined quantitative centrality measures with qualitative data from interviews, focus groups, and document analysis [29]. This approach helped contextualize structural positions (e.g., hubs, brokers, isolates) with the functional realities of collaboration, such as knowledge-sharing practices, asymmetries in support flows, and historical trust relations.

Notwithstanding these limitations, the methodological approach developed here holds a substantial promise for both research and practice. It is readily adaptable to other agroecology coalitions, donor networks, civil society alliances, or policy advocacy platforms that seek to strengthen their internal coordination and strategic planning. As Hoppe and Reinelt [31] argue, SNA visuals are particularly valuable for stimulating dialogue, building consensus, and identifying leverage points for improvement. This study reinforces that insight by demonstrating how SNA can catalyze self-assessment and action within complex, diverse, and mission-driven networks like PELUM Uganda.

In summary, this study bridges a critical methodological and practical gap in agroecology systems thinking. It offers a structured, replicable, and context-sensitive framework for evaluating network “health,” not merely in terms of membership size or institutional presence, but in how organizations connect, communicate, and collaborate in pursuit of common goals. Drawing on the work of Helbling and Anderson [24], it affirms that strong internal structure and strategic connectivity are foundational to collective effectiveness. As agroecology continues to gain traction as a pathway to resilient, just, and sustainable food systems, the ability to diagnose and enhance internal collaboration becomes ever more urgent. Tools like Gephi and Kumu, coupled with participatory methodologies, can support agroecology networks in building trust, deepening synergy, and strategically aligning their efforts to scale transformation across local, national, and regional contexts.