ArticlePDF Available

Navigating agricultural nonpoint source pollution governance: A social network analysis of best management practices in central Pennsylvania

PLOS
PLOS One
Authors:

Abstract and Figures

The Chesapeake Bay watershed is representative of governance challenges relating to agricultural nonpoint source pollution and, more generally, of sustainable resources governance in complex multi-actor settings. We assess information flows around Best Management Practices (BMPs) undertaken by dairy farmers in central Pennsylvania, a subregion of the watershed. We apply a mixed-method approach, combining Social Network Analysis, the analysis of BMP-messaging (i.e. information source, flow, and their influences), and qualitative content analysis of stakeholders’ interviews. Key strategic actors were identified through network centrality measures such as degree of node, betweenness centrality, and clustering coefficient. The perceived influence/credibility (by farmers) of BMP-messages and their source, allowed for the identification of strategic entry points for BMP-messages diffusion. Finally, the inductive coding process of stakeholders’ interviews revealed major hindrances and opportunities for BMPs adoption. We demonstrate how improved targeting of policy interventions for BMPs uptake may be achieved, by better distributing entry-points across stakeholders. Our results reveal governance gaps and opportunities, on which we draw to provide insights for better tailored policy interventions. We propose strategies to optimize the coverage of policy mixes and the dissemination of BMP-messages by building on network diversity and actors’ complementarities, and by targeting intervention towards specific BMPs and actors. We suggest that (i) conservation incentives could target supply chain actors as conservation intermediaries; (ii) compliance-control of manure management planning could be conducted by accredited private certifiers; (iii) policy should focus on incentivizing inter-farmers interaction (e.g. farmers’ mobility, training, knowledge-exchange, and engagement in multi-stakeholders collaboration) via financial or non-pecuniary compensation; (iv) collective incentives could help better coordinate conservation efforts at the landscape or (sub-)watershed scale; (v) all relevant stakeholders (including farmers) should be concerted and included in the discussion, proposition, co-design and decision process of policy, in order to take their respective interests and responsibilities into account.
This content is subject to copyright.
RESEARCH ARTICLE
Navigating agricultural nonpoint source
pollution governance: A social network
analysis of best management practices in
central Pennsylvania
Elsa L. DingkuhnID
1,2
*, Lilian O’Sullivan
2
, Rogier P. O. Schulte
1
, Caitlin A. Grady
3¤
1Farming System Ecology group, Wageningen University and Research, Wageningen, the Netherlands,
2Crops, Environment and Land Use Programme, Teagasc, Wexford, Ireland, 3Department of Civil and
Environmental Engineering, Rock Ethics Institute, Pennsylvania State University, University Park, PA, United
States of America
¤Current address: Department of Engineering Management and Systems Engineering, George Washington
University, Washington, DC, United States of America
*elsa.dingkuhn@wur.nl
Abstract
The Chesapeake Bay watershed is representative of governance challenges relating to agri-
cultural nonpoint source pollution and, more generally, of sustainable resources governance
in complex multi-actor settings. We assess information flows around Best Management
Practices (BMPs) undertaken by dairy farmers in central Pennsylvania, a subregion of the
watershed. We apply a mixed-method approach, combining Social Network Analysis, the
analysis of BMP-messaging (i.e. information source, flow, and their influences), and qualita-
tive content analysis of stakeholders’ interviews. Key strategic actors were identified through
network centrality measures such as degree of node, betweenness centrality, and clustering
coefficient. The perceived influence/credibility (by farmers) of BMP-messages and their
source, allowed for the identification of strategic entry points for BMP-messages diffusion.
Finally, the inductive coding process of stakeholders’ interviews revealed major hindrances
and opportunities for BMPs adoption. We demonstrate how improved targeting of policy
interventions for BMPs uptake may be achieved, by better distributing entry-points across
stakeholders. Our results reveal governance gaps and opportunities, on which we draw to
provide insights for better tailored policy interventions. We propose strategies to optimize
the coverage of policy mixes and the dissemination of BMP-messages by building on net-
work diversity and actors’ complementarities, and by targeting intervention towards specific
BMPs and actors. We suggest that (i) conservation incentives could target supply chain
actors as conservation intermediaries; (ii) compliance-control of manure management plan-
ning could be conducted by accredited private certifiers; (iii) policy should focus on incentiv-
izing inter-farmers interaction (e.g. farmers’ mobility, training, knowledge-exchange, and
engagement in multi-stakeholders collaboration) via financial or non-pecuniary compensa-
tion; (iv) collective incentives could help better coordinate conservation efforts at the land-
scape or (sub-)watershed scale; (v) all relevant stakeholders (including farmers) should be
PLOS ONE
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 1 / 30
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Dingkuhn EL, O’Sullivan L, Schulte RPO,
Grady CA (2024) Navigating agricultural nonpoint
source pollution governance: A social network
analysis of best management practices in central
Pennsylvania. PLoS ONE 19(5): e0303745. https://
doi.org/10.1371/journal.pone.0303745
Editor: Donato Morea, University of Cagliari:
Universita degli Studi Di Cagliari, ITALY
Received: March 16, 2023
Accepted: April 30, 2024
Published: May 23, 2024
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0303745
Copyright: ©2024 Dingkuhn et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All datafiles are
available from Zenodo repository under the DOI 10.
5281/zenodo.10043188, or through this URL:
https://doi.org/10.5281/zenodo.10043188.
concerted and included in the discussion, proposition, co-design and decision process of
policy, in order to take their respective interests and responsibilities into account.
I. Introduction
1.1. Background
Pennsylvania, with its strong historical and cultural connection to farming, benefits from a
large agricultural sector worth $81.5 billion of direct economic output and generating 301,900
direct jobs [1]. The sector has been facing the challenge of remaining competitive in a global
market while also reducing its ecological footprint for several years. Despite the State’s efforts
to reduce agricultural externalities, agriculture is still identified as a major cause of water pollu-
tion [2,3].
The Chesapeake Bay, the largest estuary in the United States (US), has faced widespread
water quality challenges, showcased by numerous classifications as “impaired” by the US Envi-
ronmental Protection Agency (EPA) [35]. In 2017, it was estimated that 42% of the nitrogen,
55% of the phosphorus, and 60% of the sediments entering the bay were from agricultural non-
point source pollution [6]. Consequently, the EPA has established a watershed-wide Total
Maximum Daily Load (TMDL), a monitoring and management plan focusing on restoring
clean water [4]. Despite positive results, a large part of the watershed remains on the EPA’s list
of polluted waters [3].
As societal pressure to reduce agriculture externalities has increased, concepts binding agri-
culture with environmental preservation and natural resources management have influenced
discussions around agricultural policy reform in recent years. Notions such as multifunctional-
ity of agriculture [7,8], provision of ecosystem services [9,10] and optimization of soil func-
tions [1113], attribute not only a primary production value but also ecological functioning
values to agriculture. This relatively new approach to valuing agriculture is being increasingly
institutionalized across the world [1417]. In the US, agri-environmental policies combatting
nonpoint source pollution are largely based on financial compensations of voluntary conserva-
tion practices, rather than on regulatory measures [14,18,19]. To date, these policy interven-
tions have yielded insufficient results as in 2022, 45,000 km of waterways and 28,000 lake ha
were still classified as “impaired” by the EPA [3]. Thus, the watershed’s states are under pres-
sure to comply with meeting the TMDL water quality requirements to reduce nonpoint source
pollution from nitrogen, phosphorus, and sediments, by 36%, 30%, and 29% respectively by
2025 [6].
This research focuses on the information exchange around conservation practices on small
and medium-scale dairy farms in central Pennsylvania. This sub-region of the Chesapeake Bay
watershed (Fig 1) is representative of governance challenges relating to agricultural nonpoint
source pollution and, more generally, of resources governance challenges in complex multi-
actor networks settings.
Central Pennsylvania is an important livestock raising region, with half of the farms in Cen-
tre county being cattle farms, one third of which are dairy [20]. Several studies put forth dairy
farmers potential contribution to meeting the local TMDL target from small scale dairy farm-
ers in the region [21,22]. In the following section, we present the theoretical framework
describing the guiding concepts of our research. We then present the mixed methods approach
used to collect and analyse both quantitative and qualitative data to understand information
exchange around management practices for small and medium scale dairy farms in central
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 2 / 30
Funding: This study was co-funded by The
Pennsylvania State University Institute for Energy
and the Environment (https://iee.psu.edu/home),
and by the LANDMARK (LAND Management:
Assessment, Research, Knowledge Base) project
(https://landmarkproject.eu). LANDMARK has
received funding from the European Union’s
Horizon 2020 research and innovation programme
under grant agreement No 635201. IRB approval
was obtained through Penn State study
#00011878 by author Caitlin Grady. The funders
had no role in study design, data collection and
analysis, decision to publish, or preparation of the
manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Pennsylvania. In the results section, we disclose the resulting actors’ networks and the distribu-
tion and strength of information flows, as well as insights on governance gaps and opportuni-
ties based on actors’ perspectives. Finally, we discuss strategies to optimize the coverage of
policy mixes, to reduce nonpoint source pollution from dairy farming in the study area.
1.2. Theoretical framework
Agricultural Best Management Practices (BMPs) are farm practices that generate public bene-
fit by targeting nonpoint source pollution reduction [23,24]. BMPs have been shown to effec-
tively reduce nutrients runoff from farms, thereby contributing to restoring water quality [22,
25,26].
Financial support, values and belief systems such as nonfinancial (stewardship) motivations
or (dis)trust in science and institution, farm characteristics (farm size, vulnerable land), opera-
tors characteristics (trainings and education level), economic factors (level of income, capital,
engagement in marketing arrangements), as well as access to quality information and connec-
tion to empowering networks, all influence BMPs adoption [23,2730]. Schall et al. (2018)
conclude that the amount of information a respondent has about BMPs is less relevant than
how this information comes to be interpreted and understood”. They highlight the necessity of
considering the cultural and political frames that shape the farmers understanding, representa-
tions and opinions in designing educational programs [27], which also applies to the design of
incentivization schemes and of governance mechanisms in general. This echoes Stuart et al.
(2014) and Ulrich-Schad et al.’s (2017) adoption studies, that suggest that the information itself
may be less decisive to a BMP adoption, than its origin (from where/whom the information
was emitted) and the pathway through which it reaches the farmer. In fact, several network
studies show that efficient (and effective) exchange of complex information depends on the
extent to which the intervention (in our case, the diffusion of the information) is adapted to
the topology (or configuration) of the network [3134]. They emphasize that information
flows need to be tailored to farmer types and farm systems specificities to enable innovation
uptake and successful information exchange.
Fig 1. The Chesapeake Bay watershed. This map was created by the authors using publicly available shapefiles from
the Homeland Infrastructure Foundation-Level Data (HIFLD) database.
https://doi.org/10.1371/journal.pone.0303745.g001
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 3 / 30
For example, targeting trusted brokers in a network, increasing the trust of farmers towards
central information senders, or reducing the number of information-sources (by bundling
information flows) to avoid information overload, are possible strategies to enhance the effec-
tiveness and efficiency of an information-network [3538]. This relates to the concept of “lati-
tude of acceptance” in the Social judgement Theory, in which messages acceptance depends
on the perception and evaluation of the message by the actor receiving it [39]. The receiver of
the message will judge the message by comparing it to his/her current standpoint, thereby
locating it in one of the three “zones of acceptance”: the latitude of acceptance, of rejection, or
of non-commitment [3941]. The latitude in which the message falls, and thus its acceptance,
is conditioned by a multi-layered judgement process relative to the current point of view of the
receiver. Layers of acceptance include the acceptance of the underlying problem, the accep-
tance of the proposed solution, and the relationship and perception of the sender with/by the
receiver (credibility and trust), among others [42,43].
This highlights the importance of message credibility, and thus of their diffusion through
strategic sources/senders in order to have greater chances to be accepted. Typically, farmers
who are wary of the government, may automatically reject messages from government agen-
cies [4446], regardless of the content of the message. Thus, BMP-message acceptance by
those farmers may meet less resistance if communicated via more credible senders, and thus
may increase chances of BMPs adoption [46].
In the context of BMPs information in central Pennsylvania, it is crucial to understand how
the actors network is structured, in order to determine how information diffusion can be
adapted to the network. Tailoring agri-environmental governance mechanisms to the diversity
of demands, motivations and attitudes that shape the farming profession calls for mixed devel-
opment models (where various production and marketing systems are combined), underlining
the need for pluralistic governance mechanisms [28,4749]. Stakeholders (other than farmers)
throughout the food supply chain, as well as the organizations composing the social network
around farmers, also have agency in addressing nutrient pollution issues [28,47,50,51].
Understanding information exchange across these stakeholders, as well as their positions in
the network, could unveil ways for more successful information exchange and for innovative
governance strategies. Comprehending the network structure and how information is valued
by farmers, can inform effective targeting of policy interventions, for example by identifying
tailored policy entry points.
We refer to BMPs-information fluxes between actors as persuasive “messages”, represent-
ing triggers or signals that can potentially lead to action undertaking or behavioural change
[37,42,43,52], in our case to adoption of BMPs. We conceptualize “messages” as interactions,
or information fluxes tied to a specific topic (here BMPs), that actors exchange and that could
elicit a response having implications for the said topic. The messages populating the gover-
nance context around an issue (here BMPs), inform and influence the decision-making pro-
cess of actors (here land-users) at a local scale (e.g. at farm-scale) [37,48]. Hence, here
messages designate information flows with consideration for their influence strength (per-
ceived credibility). In the context of our study, BMP-messages may influence the adoption (or
not) of those BMPs by farmers. These messages can consist of actual information (news),
knowledge, money (e.g. a subsidy, a grant), or any other interaction between actors, that relate
to BMPs.
Social Network Analysis (SNA) offers methodological approaches for assessing these mes-
sages and their flow, to inform policy design and intervention. It is a tool commonly used to
investigate social structures through the use of graphs theory, by determining key actors, their
roles, their interactions and influences [37,5355], and has recently gained interest from the
field of applied policy and agricultural development [37,5659].
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 4 / 30
1.3. Knowledge gap and research objective
In the case of Chesapeake Bay’s nonpoint source pollution issue, the network and flow of
BMPs-related information have not been studied. Increasingly, policy analysis highlights the
need for policy mixes where instruments are optimally combined to reach policy targets [60
63]. Relations, complementarities, and effects of agri-environmental policy instruments have
been widely assessed in literature [6,6264]. Less focus has been put on determining the actors
that they should involve related to their existing position in the governance context. In fact,
the targeting of entry-points based upon the network potential, the ultimate actors targeted by
the intervention, and the pathways through which those can be reached have, to date, received
limited investigation. Recent examples of such studies include the work of O’Sullivan et al.
(2022) and Valujeva et al. (2022), which sought to identify key (strategic) actors in farmers net-
works. However, both studies focus on information flows relating to the concepts of soil func-
tions, and do not address entry points and pathways for farm practices (such as BMPs)
diffusion [37,59]. This research contributes to filling this gap. Using SNA and the analysis of
BMPs messaging to detect governance opportunities, we explore potential entry points for
intervention to foster sustainable land management through improved BMPs communication.
The method is tested on information exchange around BMPs adoption by dairy farmers in
central Pennsylvania, with the following research question: How are farmers exchanging infor-
mation about BMPs and can this exchange be captured and modelled using social network
analysis to inform future policy interventions targeting?
Our hypotheses are that the information exchanged with farmers differs by actor and BMP
type and that farmers value information differently based on actor.
Owing to the context specific nature of SNA, identifying universal schemes and generaliz-
able policy guidelines is not appropriate, but rather, we want to demonstrate if/how different
points of intervention can be optimally combined in policy mixes within the context of our
study area to reach a wider range of farmers, and if strategies emerge to better integrate other
(intermediate) stakeholders in policy intervention.
II. Materials and methods
2.1. Mixed-methods and sampling procedure
The data for this research was collected following a mixed-methods approach combining sur-
veys and interviews, which were conducted between February and April 2019. The survey pro-
vided quantitative and qualitative data relating to BMPs information exchange (messages),
while the interviews provided complementary qualitative data about stakeholders’ perceptions
and contextual governance challenges and opportunities. The surveys and interviews were
conducted with the same group of participants. These were identified by applying a snowball
sampling method [65,66], where the answers of the respondents oriented us towards the next
actors to interview (Table 1). Some respondents were a part of the same organization; thus, the
number of respondents exceeds the number of organizations.
The analysis process encompassed three distinct phases:
1. SNA, which analysed information flow and key actors in the network;
2. BMP-messaging analysis, which characterized the frequency, strength, and the distribution
of BMP-related messages;
3. Thematic content analysis of qualitative data to assess governance challenges and
opportunities.
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 5 / 30
Phases 1 (SNA) and 2 (BMP-messaging analysis) draw on the survey results, while phase 3
(Thematic content analysis) draws on the qualitative interviews outcomes. All data were col-
lected in accordance with approval by the Penn State Institutional Review Board for study
#00011878.
2.1.1. Quantitative data. Survey. A preliminary review of scientific literature and infor-
mational discussions with experts guided the questionnaire formulation (S1 File), in particular
the determination of a set of 16 BMPs (S1 Table). For each selected BMP, quantitative and
qualitative data relating to message fluxes were collected, namely information about to whom
the actor sent and received messages relating to the listed BMPs, as well as the type and the
strength of the information (for messages type categorization refer to S2 File).
Social network analysis. The SNA was conducted on network graphs of BMP-message
fluxes, using Gephi software [67,68] for both visualization and calculation of network proper-
ties (Table 2). The graphs are composed of “nodes” (or vertices) representing the organizations
(actors), and of “edges” representing their ties. These ties, or connection between nodes, corre-
spond to the messages reported by the survey respondents, and are directed from the emitter
of the message (source) towards the receptor of the message (target). Duplicate messages were
merged into single edges.
Two network visualizations were constructed:
1. A socio-centric network graph, based on the full messaging dataset from all the survey
respondents and representing the larger governance structure (including actors who may
not interact directly with farmers);
2. A farmers-centric network graph, constructed with data from farmers-respondents exclu-
sively, and composed of the farmers themselves and of the actors from whom they directly
receive BMP-messages.
In order to protect the participant’s identities, the actors and organization names were
replaced by numerical pseudonyms in the network graphs. A complete list of the organizations
that populate the nodes is presented in S2 Table.
The socio-centric network. All organizations that were interviewed, and all those identified
by respondents as message senders, constituted a node of the socio-centric network (Table 3).
The participating farmers, and the farmers to whom other respondents referred to, were
Table 1. Number of organizations interviewed (per organization type).
Type of Organization Number of Organizations Number of Respondents
Farmers*Small-scale
1
6 9
Medium-scale
2
1 1
Large-scale
3
2 2
Local government (below state level) 4 5
Government (state or above state level) 4 4
Public institution of higher education 1 3
Non-profit organizations 4 5
For profit organizations 3 3
Total 25 32
1
Small scale: farm with <200 animals
2
Medium-scale: farm with >400 and <1,000 animals
3
Large-scale: Concentrated Animal Feeding Operation of >1,000 animals (CAFO)
*all farms are dairy operations, except of one small-scale farm (previous dairy, now crops).
https://doi.org/10.1371/journal.pone.0303745.t001
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 6 / 30
merged into a broader node labelled “Farmers”. The network boundaries were determined
when saturation was reached, based on two observations: (i) when pursuing the snowball sam-
pling added duplicate edges, and/or only added nodes that would have no 1st or 2nd degree
connection to farmers; (ii) when all the key actors (listed prior to the sampling process, based
on literature and informational discussions with experts and key-informants) appeared in the
network.
Table 2. Description of network property measures and node’s centrality measures in SNA [53,6974]. In the equations, vrepresents each vertex (node), erepresents
each edge. For Network diameter, a(s,t) denotes the number of edges in the shortest path from a node sto a node t. For degree centrality, iis the transmitter node of e,jis
the receiving node of e. For closeness centrality, d(v,t) is the geodesic distance between any node vand t(i.e. the sum of the edges on the shortest path). For betweenness
centrality, g
st
represents the number of geodesics (or shortest paths) connecting sto any other node tin the network, and g
st
(v) denotes the number of shortest paths from s
to tthat some node vlies on. For clustering coefficient, n
v
is the number of alters (neighbours) of v, and m
v
is the number of alter-to-alter edges in the neighbourhood of v.
Measure Equation Definition Explanation
Network diameter D¼maxstfaðs;tÞg The shortest path length in the network (the shortest
distance between the most distant nodes in the network).
Indicates how long it would take (or how many intermediary
nodes it would take) for information to circulate between the
two most distant nodes in the network.
Degree (in-degree,
out-degree)
In-degree The number of relations (edges) of the nodes. In directed
graphs, a distinction is made between in-degree (number of
incoming edges) and out-degree (number of outgoing
edges).
Effective measure to assess the importance of an actor in a
social network, but doesn’t take into consideration the global
structure of the network.
IDi =S|e
ij
|
k = 1 In-degree: frequency of message receiving.
Out-degree:
ODi =S|e
ij
| Out-degree: frequency of message sending.
k = 1
Closeness
centrality
The average length of all shortest paths from one node to all
other nodes in the network. The farness of a node is defined
as the sum of its distances from all other nodes, and its
closeness is defined as the reciprocal of the farness.
Broadcasters, measure of reachability: closeness to the entire
network, i.e. how easily a node can reach all the other nodes
in the network (ability to reach the entire network quickly, to
broadcast information).
CCvð Þ ¼ 1
Sdðv;tÞ
Betweenness
centrality
The extent to which a node lies between other nodes in the
network. The fraction of shortest paths that go through a
node divided by the total number of shortest paths between
all nodes.
Brokers or gatekeepers: interfaces between tightly-knit
groups, nodes with high a high betweenness centrality are
vital elements in the connection between different regions of
the network, and can control the flow of information
between communities.
CBðvÞ ¼ SgstðvÞ=gst
Clustering
coefficient
The fraction of the possible relation triangles that are
actually completed.
Node’s clustering coefficient: The extent to which a node’s
direct neighbours or alters (i.e. nodes to which it is
connected) are also likely to be neighbours (connected).
Indicates the level of cohesion between the neighbours of a
node.
A node’s (local) clustering coefficient is the fraction of its
possible relational triangles that are actually completed.
CClv¼mv=nvðnv1Þ
2
The (global) network clustering coefficient (computed for
the whole network), is the measure of all completed
relational triangles over all the possible relational triangles
in the network.
https://doi.org/10.1371/journal.pone.0303745.t002
Table 3. Actors categorization and related colour code.
Farmers farmers, agricultural land managers and/or landowners
Private for profit private agricultural enterprises including agricultural inputs suppliers, private consultants,
veterinarians, animal nutritionists, trade associations
Non-profit / NGO private non-profit and Non-Governmental Organizations including environmental and social
advocacy groups and farmers associations
Local government local government agencies with below state level outreach (e.g. central Pennsylvania, county,
township, or borough level)
Government
(state)
state or above-state government agencies with outreach at state, multi-state or federal level
Universities public institutions of higher education
Other multi-actor partnership (e.g. public-private partnership, consortiums), the media, civil
society, primary or secondary education)
https://doi.org/10.1371/journal.pone.0303745.t003
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 7 / 30
The farmer-centric network. The farmer-centric network includes the farmers-respondents
themselves (as individual nodes), as well as the actors they reported receiving information
from. The other actors were attributed the same numerical pseudonyms and colour code as in
the socio-centric network. An additional node Other farmers’, representing the rest of the
farmer community, was also added to the network.
The edges, or ties, of the farmers-centric network represent the messages received directly
by farmers, hence only those reported by the farmers-respondents were included.
In order to compare BMP-specific networks, the farmer-centric graph was constructed for
four different situations: (i) considering the messages for all BMPs, (ii) considering the mes-
sages relating to (forested and grass) riparian buffers exclusively, (iii) considering the messages
relating to no-till and cover-cropping exclusively, and (iv), considering the messages relating
to manure management exclusively (having a manure management plan, following a manure
management plan, and manure storage). These categories were based on grouping of the most
important BMPs into BMP-types: riparian buffers, soil conservation cropping practices, and
manure management.
BMP-messaging analysis. The BMP-messaging analysis first focused on the frequency and
distribution of all messages reported by the survey respondents (after merging of duplicates),
considering the BMP they were tied to, and the type of information they related to.
A second analysis was performed on the messages from the farmers-centric network (mes-
sages received and reported by farmers directly). Therefore, the comparative frequency and
distribution of the messages weights were analysed considering the BMP they related to, and
their source (actor who emitted the messages). Duplicate messages were merged, and their
weights averaged.
2.1.2. Qualitative data. Additional qualitative data was collected via semi-structured
interviews, guided by open-ended questions relating to governance challenges and opportuni-
ties of nonpoint source pollution management (S3 File).
The interviews notes and audio-recordings were transcribed into text format, on which
inductive content analysis was conducted through a systematic coding process [7578]. The
interviews coding process was split into three stages, using QDA Miner Lite software [79]:
decontextualization, recontextualization and categorization [75,77,80].
The textual data from interviews was first disassembled (decontextualization) by systemati-
cally coding repeated observations. Sections of text were coded if their content reflected ideas
or concepts mentioned in previous interviews (repetition), if they were explicitly indicated as
important, or/and if they echoed with information from literature or discussions with experts.
During the recontextualization process, the interviews were re-read and recoded in order to
refine the codes (adding, merging or deleting of initial codes). At the same time, coded text
sections were tagged:
with “H” (= hindrance) if the information was explicitly cited as a governance issue, a chal-
lenge or a hindrance, or appeared to be a factor that prevented BMPs adoption or nonpoint
source pollution reduction;
with “O” (= opportunity) if the information was explicitly cited as an opportunity or a possi-
ble solution, or appeared to be a factor that allowed for/favoured BMPs adoption or NPS pol-
lution reduction.
The data were then reassembled, by grouping the codes into concepts/themes. The fre-
quency of themes and codes, as well as the frequency and distribution of tagged codes were
then analysed, in order to determine the relative importance of the governance challenge and
opportunities from the stakeholder’s perspective.
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 8 / 30
III. Results
3.1. The socio-centric network
3.1.1. Graph properties. The socio-centric network is composed of 3919 edges (messages)
and 57 nodes (actors or organizations), over half of which are non-profits or state (or above
state) government agencies. Of the 9 nodes falling under the category “Other”, 6 correspond to
multi-actor partnerships (Table 4).
The graph diameter (5) indicates a rather small but highly connected network [70]: in order
for BMPs messages to be exchanged between the two most distant actors of the network, the
information would have to transit by four intermediary actors. The average degree of nodes
(68.8), indicates that a typical actor of the network exchanges (receives or sends out) as many
as 69 BMP-messages in average per year [58]. The fact that the network diameter is nearly
twice the value of the average path length, suggests a high communication efficiency across the
network. In addition, the average clustering coefficient (or average neighbourhood complete-
ness of nodes), reveals that nearly 1/4th of the possible relation triangles between three nodes
are completed, indicating that a fairly high level of cohesion is present in the network (Table 4)
[72].
3.1.2. Node properties. A few actors repeatedly scored the highest centrality measures.
The most active BMP-message senders are also the most important message receivers in the
network, but in different orders. For example, farmers (node 1) only rank at the 5
th
position as
message senders (Fig 2A.i and 2A.ii), while they are the most important BMP-message receiv-
ers (Fig 2B.i and 2B.ii). On the other hand, the top-3 message senders are the two most promi-
nent state (or above-state) government agencies (nodes 39 and 14) and an NGO (node 44),
while a private for profit organization is the third most important message receiver (node 48),
after farmers (node 1) and a government agency (node 14). This indicate that most messages
converge at farm level, and that farmers are (one of) the main information receivers.
The betweenness centrality measures (Fig 2C.i and 2C.ii) show that the active message
receivers and senders often also act as the shortest path between pair of nodes that are not con-
nected, thereby acting as bridges (brokers) between actors [53,73]. In fact, the two prominent
government agencies (nodes 14 and 39) are also in the top-5 message receivers and senders,
followed by the aforementioned NGO (node 44) and farmers (node 1). Public institutions of
higher education (node 7: Universities), show a nearly as high betweenness centrality measure
than the later, and are the fifth major broker of the network. These actors are vital elements of
Table 4. Graph properties of the socio-centric network calculated on Gephi.
Nr of edges 3919
Nr of nodes 57
Farmers 1
Non-profit 17
For profit 6
Local government agencies 8
Government (state) 15
Universities 1
Other 9
Network diameter 5
Average path length 2.3
Average degree 68.8
Average clustering coefficient 0.24
https://doi.org/10.1371/journal.pone.0303745.t004
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 9 / 30
Fig 2. Socio-centric networks. a = out-degree, b = in-degree, c = betweenness centrality, d = closeness centrality,
e = clustering coefficient; i = SNA graphs, ii = bar graphs showing the top eight centrality measure scores (Y-axis) of
nodes (X-axis); the color code reflects the actors type classification.
https://doi.org/10.1371/journal.pone.0303745.g002
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 10 / 30
the network, as they act as intermediaries to connect areas of the network that would otherwise
be disconnected. Universities (node 7) also record the highest closeness centrality measures
(Fig 2D.i and 2D.ii), suggesting that they are the most connected actors of the network, fol-
lowed by one of the prominent government agency (node 39) and farmers (node 1). The close-
ness centrality measure displays less heterogeneity across nodes of the network as the other
centrality measures.
New actors appear in the top-4 clustering coefficient measures results (Fig 2E.i and 2E.ii),
namely: a multi-actor partnership (node 15), two for profits relating to animal services (nodes
30 and 29), and the media (node 40), which suggests the existence of sub-networks or “small
worlds” within the larger network [70]. High clustering coefficients are an indicator of tightly
knitted neighbourhoods (cliques). These nodes with a high clustering coefficient may be worth
targeting to relay information to their neighbours while keeping the number of links to con-
nect the network to a minimum [37].
3.2. The farmer-centric network
3.2.1. Graph and node properties. Farmer-respondents reported having received mes-
sages from 21 actors over the 57 actors of the socio-centric networks (37%), resulting in a
farmers-centric network of 346 edges (messages) and 30 nodes (including the 9 farmers who
were interviewed) (Fig 3A), with an average degree of 11.5. Degree of nodes refers to in-degree
(received messages) for farmers nodes, and to out-degree (emitted messages) for the other
nodes. Farmers in-degree vary considerably, ranging from 16 (node 58) to 102 (node 64), illus-
trating the high variability in farmers exposure to BMPs information.
When considering all messages, without distinction between BMPs, the most active mes-
sage senders are two local government agencies (nodes 2 and 5, out-degrees = 65 and 60,
respectively), followed by private consultants (node 42, out-degree = 34), and the media (node
40, out-degree = 29) (Fig 3A).
However, this order changes when considering messages relating to specific BMPs, showing
that different (types of) actors communicate on different types of practices.
For example, the most active message senders for riparian buffers remain the local govern-
ment agencies, followed by an NGO (node 3) and a state government agency (node 12) (Fig
3B). On the other hand, no-till and cover-crops messages appear to be more strongly tied to
other farmers (node 1), the media (node 40), a local government agency (node 2) and agricul-
tural industries (node 19), in descending order (Fig 3C).
When looking at manure-management messages (manure management planning and
manure storage), the same local government agencies stand out (nodes 2 and 5), followed by
private consultants (node 42) and a state government agency (node 12). The larger farms
(nodes 64, 65 and 66) receive more messages relating to manure management than the small-
scale farms (nodes 58 to 63) (Fig 3D).
Our results show that the farmers reached by for-profit type of actors (light blue nodes) are
different than the farmers reached by non-profit type of actors (violet nodes) (Fig 3A). Fur-
thermore, only peer farmers (node 1) reach all types of farmers in the network (Fig 3A), but
these information fluxes relate to specific types of BMPs (rather agronomic and cropping prac-
tices such as cover-cropping and no-till), and exclude other BMPs such as riparian buffers (Fig
3A and 3B), while they are very limited for manure management related practices (Fig 3D).
The most prominent nodes of the socio centric network (nodes 39 and 14, 44 and 48), and
most state (or above state) government agencies, do not appear in the farmer centric network,
implying that they are more active in the macro governance structure than in direct interac-
tions with farmers. In contrast, the local government agencies that are most active in the socio-
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 11 / 30
Fig 3. The farmers centric network. Nodes’ color code relate to the actors’ type classification; a = all BMP-messages;
b = forested and grass riparian buffers messages; c = no-till and cover-crops messages; d = manure management plan
messages. Node sizes are relative to the node’s degree, and edge colors correspond to the source (i.e. the actor emitting
the message). Nodes 58, 59, 60, 61, 62, 63 = small-scale farms; node 64 = medium-scale farm node 65 and 66 = large-
scale farms (CAFOs).
https://doi.org/10.1371/journal.pone.0303745.g003
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 12 / 30
centric network (nodes 2 and 5), as well as universities (node 7) and the media (node 40), are
also important in the farmers-centric network. This indicates that these are active actors in
both, the macro governance structure as well as on the ground in closer interaction with farm-
ers. Hence, local government agents, universities and the media play important bi-directional
top-down and bottom-up brokerage roles. Private non-profit organizations with boards con-
stituted of a majority of farmers (“farmer led organizations”) are mainly active on the ground
(farmer centric network), but receive a lot of messages from the macro level too, therein acting
as top-down and horizontal information channels.
3.3. BMP-messaging analysis
3.3.1. Message characterization and distribution. In total 4344 BMP-messages were
reported by the survey respondents, after merging of duplicates. The BMPs for which the most
messages were reported (>350) are cover crops (395), forested riparian buffers (385), no-till or
conservation tillage (372), in descending order (Fig 4).
The BMPs for which the least messages were reported (<210) are reduced stocking density
(207), precision fertilization (184) and hedgerows plantation (127). Over half of the total infor-
mation flow related to knowledge exchange or knowledge dissemination; nearly one quarter
(22%) related to regulation or standards; 11% to funding; 10% to personalized technical assis-
tance to implement BMPs (assistance request or provision), and only 5% to networking and
actors linkage.
3.3.2. Message strength. In the farmers centric network, BMP messages are most numer-
ous for ‘Having a manure management plan’ (40 messages reported), ‘Forest riparian buffers’
(35) and ‘Manure storage’ (34); and are the least frequent for ‘Hedgerows plantation’ (0), ‘Pre-
cision fertilization’ (7), ‘Wetland and permanent grassland preservation’ (10), ‘Following a
nutrients management plan’ (13) and ‘Following a manure management plan’ (13) (Fig 5).
Hence, the relative importance of BMPs (in terms of message frequency) are different in the
farmers-centric network and in the socio-centric network, suggesting that the focus on specific
BMPs differs across governance scales. In fact, only forest riparian buffers messages are among
Fig 4. Messages distribution per BMP (X-axis) and per message kind, after merging of duplicates.
https://doi.org/10.1371/journal.pone.0303745.g004
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 13 / 30
the most prominent BMPs in both networks, while hedgerows plantation and precision fertili-
zation messages are the least important in both networks.
The BMPs recording the highest frequency of strong influence (rank 4) are, in descending
order: ‘Having a manure management plan’, ‘Manure storage’, ‘Grass riparian buffers’ and
‘Reduced stocking density’, followed by ‘Animal exclusion’ (Fig 5). Farmers assigned high
Fig 5. Messages distribution per BMP (includes only the messages received directly by farmers). X-axis: message strength
based on influence-rank assigned by farmers (1 = weak, 4 = strong); Y-axis: messages count; legend: message source (emitter of
the information); ‘For-profit (animal)’: animal specialist, *’: farmers-led organizations. Note: no messages were recorded for
BMP nr 8 ‘Hedgerows plantation’.
https://doi.org/10.1371/journal.pone.0303745.g005
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 14 / 30
influence strengths to these BMPs, while they do not seem to communicate much about them
among peers. In contrast, farmers seem to share more information about agronomic and crop-
ping practices such as no-till, cover cropping and strip-cropping; which are also strongly influ-
enced by farmers-led NGOs, i.e. private non-profit organizations, the boards of which are
composed primarily of farmers (Fig 5).
The sources of the messages (actors who emitted the information) recording the highest fre-
quency of strong influence (rank 4) are, in descending order: local government agencies
(nodes 5 and 2), a private non-profit organization (node 3), and private consultants (node 42)
(Fig 5).
While most BMPs are tied to many different actors, our results show that the influence of
some actors is limited to certain BMPs. State and above state government agencies for instance,
are only present when it comes to riparian buffers, manure management (planning and
enforcement) and animal exclusion. The mandatory aspect of manure management planning
may explain this strong link with state or federal authority, while the importance of riparian
buffers here illustrates the keen interest in this practice at a higher governance level. Local gov-
ernments are closer to farmers on the ground, and are influential agents for all BMPs, aside
from precision fertilization and precision feeding. In contrast, these BMPs seem to be more
tied to the profit sector, perhaps because of their reliance on inputs provision (fertilizers and
animal feed). Furthermore, specific profit agents are influential on specific BMPs. For instance,
not only are agricultural industries strong influencers on cover-cropping, but also on nutrients
management planning, precision fertilization and prescribed grazing. Additionally, private
consultants are influential on grassland management choices (grass riparian buffers, grazing
and permanent grassland management), as well as on manure management (reduced stocking
density, manure storage, having and following a manure management plan). Lastly, animal
specialists such as veterinarians and nutritionists, as one might predict, are influential agents
for discussing livestock density reduction precision feeding practices.
Although public universities (node 7) only reach a limited number of farmers (Fig 4), they
are influential on a wide range of BMPs, and their messages were systematically assigned high
influences by farmers (rank 3 or 4). Hence, not only do they play an important brokerage role
in the overall governance structure, but are also influential actors on the ground.
Messages originating from the media are better ranked when relating to technology and
technical equipment such as precision fertilization, precision feeding, manure storage. In con-
trast the media record rather low influence scores when relating to agronomic and cropping
practices (no till, cover-cropping, strip cropping, as well as for grassland management (graz-
ing) (Fig 5).
3.4. Thematic content analysis
The systematic coding of the interviews resulted in a set of 34 codes, that were grouped into 7
themes (Fig 6). A detailed list of the themes, codes, and their description is presented in S3
Table. The highest frequency codes for both hindrance and opportunity are showcased in
Table 5.
The first major theme, “Perception of policies and actors”, relates to stakeholders’ percep-
tions of the roles/responsibilities of other actors (government, for-profit, non-profit, peer-
farmers) or their relation to them. While government was frequently mentioned as (financial
and technical) facilitators of BMPs implementation, farmers distrust towards the government
was also stressed.
For instance, a farmer explained that connecting with local government agencies encour-
aged him to deploy BMPs on his farm, while (some) neighbouring farmers were much more
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 15 / 30
wary: “(. . .) what got the ball rolling (. . .),it started with the grant.Although,(we were) doing
some of this stuff ahead of time.But when (the local government agency) came in,and we formed
a relationship with them,it became pretty easy after that to take their recommendations and
transition with their suggestions.(. . .) And we saw the one programme work.So we tried another.
(. . .) So,it just kind of snowballed from there.(. . .) some of our neighbours (. . .) don’t want the
government around.They don’t trust them.They think if they’re coming it’s for something bad.
And they’re probably not the only ones that have that attitude.” Another respondent also clearly
stated his aversion: I have no interest in having any government money for inputs.I know what
I’m doing.In his opinion, the government is deliberately letting small-scale farms disappear:
the government is trying to force out all small farmers.Get rid of all small dairy farmers.(. . .)
Because they want to just control a few dairies.Quite big dairies.(. . .) It’s easier to monitor 2
dairies that have 10,000 cows than a hundred dairies that have 50 cows.”
This theme also relates to the participants’ (positive or negative) critical perception of (regu-
latory or voluntary) policy measures. While some farmers felt constrained by regulation, others
insisted on the need for more enforcement: “Better enforcement of compliance with (agricul-
tural) requirements is essential,and will drive improvement.Right now,in Pennsylvania,compli-
ance is cursory at best,no real zeal behind it.Without the threat of some consequences for
inaction,no change is the most attractive option”. This opinion was shared by several non-
farmer interviewees, while one mentioned conflict of interests as the main obstacle impeding
enforcement. Another participant suggested that this could be addressed by establishing a ded-
icated revenue-source for BMPs establishment and enforcement: “(. . .) if the Commonwealth
establishes a revenue source,or commits so much money each year to create a revenue source to
help pay for these practices,then I think you would see more of a political will to maybe start
enforcing some of these laws.(. . .) But without a revenue source to do that,no enforcement takes
place”. Representative of the contrasting opinions among stakeholders, the code “regulation
and enforcement” was cited as both an opportunity and a hindrance with similar frequencies.
On the other hand, “voluntary measures” were considered as an opportunity as frequently as
Fig 6. Relative importance of the emerging themes from the interviews contentanalysis. Each theme refers to a
cluster of codes, the frequency of which were computed to establish the relative importance of the theme (%).
https://doi.org/10.1371/journal.pone.0303745.g006
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 16 / 30
regulatory measures (frequency = 25), while they were rarely mentioned as a hindrance (fre-
quency = 9) (Table 5).
The second major theme, “Resources allocation and economic valuation”, relates to the
availability of financial, human and time resources, to the economic valuation of environmen-
tal/social services provided by farmers, and to the viability of farms and their (in)capacity to
self-finance BMPs. Farmers-interviewees emphasized that financial assistance was crucial in
their decision/ability to implement BMPs on their farms, while government agencies
highlighted that funds were, in general, not sufficient to cover farmers demands and to provide
appropriate extension services. The poor investment capacity of farmers and their struggle to
keep the farm viable was repeatedly mentioned as a main challenge, implying that farmers
would not prioritize BMPs as long as it does not increase profitability in the short/medium
term and help them keep their farm viable: right now,dairy farms are in an economic crisis.
So,some can’t even afford to feed their own families.(. . .) even when the milk price goes back up,
producers will spend years digging out of this whole.So,they are not going to use that extra
money to implement conservation practices”.
Table 5. Maximum frequencies of observations from interviews content analysis (Fq: frequency, Int: interview,
edu.: education).
Codes occurrence
Code Fq. count Int. concerned % Int.
funding 109 21 81%
government 107 20 77%
edu./knowledge and outreach/exposure 92 22 85%
regulations and enforcement 52 14 54%
innovation, adaptability/flexibility 48 18 69%
viability 47 15 58%
diversity and (mis)calibration 46 19 73%
responsibility distribution 36 18 69%
Tags occurrence
Code Fq. count
Hindrance government 31
funding 28
cost-share 24
regulations and enforcement 21
viability 19
priorities 17
human and time resources 17
diversity and (mis)calibration 16
selectivity 14
Opportunity funding 69
government 66
edu./knowledge and outreach/
exposure
62
innovation, adaptability/flexibility 34
collaboration 33
regulations and enforcement 25
voluntary measures 25
tradeoffs or co-benefits 24
human and time resources 24
https://doi.org/10.1371/journal.pone.0303745.t005
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 17 / 30
Interviewees recognized that farmers’ role was undervalued compared to the results/out-
puts generated. Better economic valuation of environmental and social services provided by
farmers was cited as a possible solution to address this challenge: “if society was ready to pay
the price associated with producing the food,given the costs to maintain the environment,the
way that the other 90% of the population that’s not involved in agriculture want to see the envi-
ronment maintained,then they should probably pay 20,30% more for their food”;“Chesapeake
Bay (. . .) generates,what,like $6.8 billion in revenues as a result of clean water.(. . .) all these
benefits,job creation,recreation,(. . .) value will increase,the cleaner the bay gets”. One farmer
expressed that we should “measure success a little differently”, for example by rewarding soil
restoration.
The third theme, “Social responsibility and recognition”, relates to power and responsibilities
distribution, to social recognition, and to farmers (political) representation and inclusion in dis-
cussions and decisions. Sense of (un)fairness and unequitable distribution of pressure stood out
as a sensitive subject in many interviews. For instance, one farmer said “there’s always things we
can be doing to make improvements,and do a better job of what we’re doing.But I don’t think
we’re the sole reason there’s a problem.(. . .) the general public doesn’t recognize that.(. . .) if
you’re going to put us under the thumb,everybody else should be held to the same standard,and I
feel like this is something we’ve been working on for a solid 20 years,where townships,a lot of
these other communities are not held to the same standard we are”. Relating to farmers’ inclusion
and consultation, this same farmer stated: “it’s important we need regulations,or may. . . (but)
farmers need to be at the table.(. . .) one of the biggest frustrations is when you get a bunch of gov-
ernment and university people together,that all don’t do this every day.And on paper,it looks like
a really great logical solution,but the application in reality is very different.(. . .) Farmers will be
more open to it,if it’s done in a good way,and presented in a good way”. Likewise, another farmer
put forth the need to consult farmers and make use of their knowledge: “their (= the farmers’)
voices aren’t heard.(. . .) the people writing the regulations (. . .) haven’t lived there and haven’t
experienced it,(. . .) they may be book smart but they are not necessarily
street smart. (. . .) they write the laws (. . .) using general math, and don’t necessarily use
common sense.”
Finally, the theme “Knowledge, exposure, connectedness” refers to education and knowl-
edge exchange, to connectedness with other networks/actors, to collaboration and coordina-
tion between actors, and to farmers geographic (im)mobility (isolation or propensity to
travel). The code “education/knowledge and outreach/exposure” recorded the third highest
opportunity-tag frequency, and “collaboration” (i.e. multi-actor and multi-scale partnerships,
coordination between actors) the fifth highest (Table 5). For instance, one of the participants
highlighted the need for stronger inter-municipal and watershed-scale collaboration: “more
cooperation and collaboration between municipalities (is needed),because what oftentimes you
see is an upstream problem causing (. . .) downstream water quality issues.(. . .) one thing that
could possibly help is having multi-municipal,or having some kind of watershed impact be a
part of funding requirements.(. . .) And getting input and support from other municipalities to
see how this project is going to benefit others.
Interestingly, “government” and “funding” are the most redundant codes (overall, but also
in terms of both “opportunity-tags” and “hindrance-tags” occurrence), but were reported over
twice more frequently as opportunities than as hindrances.
IV. Discussion
We have shown that different actors communicate to farmers on different types of BMPs, that
farmers exposure and networks vary considerably and that farmers may value information
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 18 / 30
differently based on its source. We also showed that intermediary actors play important bro-
kerage roles according to their position in the macro-governance network. Drawing on the
governance gaps and opportunities that we identified, in this discussion we articulate potential
strategies that may provide opportunities for governance and management of non-point
source pollution as well as several limitations of this study.
4.1. Targeting interventions based on network topology
The coverage of policy mixes may be improved by targeting a specific variety of messages sources,
representative of the high variability in farmers’ networks and exposure [46,59]. Our SNA results
suggest that government agencies are connected with all types of farmers (Fig 3), yet the results of
the qualitative data analysis reveal that farmers’ avoidance of government-related actors is com-
mon. We may have failed to capture the “mistrustful” farmers networks, as they may not be repre-
sented in our respondents’ sample. Hence, although BMPs dissemination through government
and universities can be effective, it may not be enough. In fact, studies show that farmers partici-
pation in research and government-led programs is limited to certain types of farmers, and that
certain farmers are unlikely to ever participate to such programs [46,81,82].
Our SNA also found that some farmers are very connected to the for profit sector, which is
increasingly viewed as a potential entry-point for BMP intervention [83,84]. In fact, BMPs val-
uation through markets could present interesting opportunities given that government pro-
grams have several limitations including a lack of funding, mandate, or farmers reluctance to
participate. This would also allow to target intervention and incentivization towards other
actors of the supply chain and alleviate the concentration of pressure and information on
farmers. This aligns with a recent European-level study [37], where targeting “higher” entry
points for messages diffusion was identified as a strategy to reduce message overload on farm-
ers. Furthermore, targeting intervention towards other (higher) actors in the network may
improve messaging coherence and alignment through “bundling” of messages, i.e. grouping
messages to reduce the number of information-sources [37], thereby reducing the “noise” per-
ceived by the end-target actors (here farmers).
Contextualised to BMPs adoption in Pennsylvania, adaptations such as supply-chain per-
spectives (through sustainable and ethical sourcing for example) are promising avenues that
have raised the interest of shareholders and investors and thus, are being increasingly adopted
by agri-food businesses [85,86]. Such initiatives are emerging in the region, such as the Turkey
Hill Clean Water Partnership in Lancaster County, a collaborative effort between a dairy pro-
cessor, milk producers cooperatives, and non-profit organizations, to incentivize conservation
planning on farmland [87]. Scaling up such initiatives could have a massive impact, given the
importance of Pennsylvania’s food processing and manufacturing sub-sector, accounting for
60% of the state’s agricultural economic output [88].
Several studies have shown that agricultural cooperatives can play an active role in encour-
aging sustainable farm practices [8992], while government support programs can support
these cooperatives through capacity building [93]. However, despite the density of BMP-mes-
sages flow in the network, none of the survey respondents has reported receiving or emitting
BMP-messages to/from processors or dairy cooperatives, suggesting that these are, at this time,
poorly involved in conservation efforts in central Pennsylvania. This could be addressed
through incentivization schemes encouraging cooperatives and processors to adopt sustain-
ability standards and supply-chain governance approaches. Certain private non-profit organi-
zations, or state level government agencies could facilitate this process [94,95].
Furthermore, private farm consultants may constitute possible entry points for manure
management policies, revealing a potential role for private companies in compliance control.
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 19 / 30
For example, in the same way that private nutrients and manure management planners are
certified by the state [51], compliance control of the actual application of these plans by farm-
ers could be conducted by accredited private certifiers. This may be a more popular option
among farmers, than compliance-control being conducted by local government agencies. In
fact, farmers increasingly rely on private consulting firms [96,97]. Interviewees reported that
this may be due to distrust towards government and to reduced public resources (human and
financial) for enforcement and extension [98100]. Private farm advisors can also be conserva-
tion intermediaries [98], by providing technical assistance to farmers for BMPs implementa-
tion. This was considered a ‘risky opportunity’ by some interviewees, because of possible
competition between public conservation interests, and the advisors’ private economic inter-
est, as also acknowledged in previous research [51,98,101103]. Nonetheless, the role of the
for profit sector in fostering BMPs adoption, and the extent of its outreach, should not be
underestimated and should be considered when designing policy mixes. As participation in
BMPs by certain farmers may be undermined by unalignment of their (relational) values with
government-led programs [46,104] public-private partnerships may offer opportunities for
more effective policy and extension outreach [46,83]. Our results show that most of the for
profit organizations are in close interaction with farmers, but are less connected to the overall
governance structure. They also reveal that some actors, such as farmers organizations, trade
associations, local governments and universities can be bridges between the government and
some of these “small worlds”.
The SNA results suggest that farmers are the actors who are most likely to reach a wide
range of other peer-farmers, which is congruent with the findings of Valujeva et al. (2022) and
O’Sullivan et al. (2022) who identified peer-to-peer farmers interactions as a major communi-
cation channel in European countries. Hence farmers exposure should be strongly supported
to foster inter-farmers interaction. Therefore, programmes focusing not simply on BMP adop-
tion, but on farmers’ coordination, mobility, training, and knowledge-exchange, could be
developed. In fact, numerous studies showcase how programmes’ impact and outreach were
multiplied by inter-farmers interaction [31,32,105108], and how collective action contrib-
utes to enhancing farmers participation and coordination [109113]. Different approaches
have been identified by Valente (2012) to frame interventions based on network topology. The
most common one being the targeting of individuals such as opinion leaders (local champi-
ons), bridging individuals (who may be more amenable to change and diffuse change than
leaders who have interest in maintaining the status quo), and actors at the periphery of the net-
work (to avoid their exclusion, but also because they may be sources of innovation as they are
likely to be connected to other communities or networks). Another approach is through segmen-
tation (identifying groups of people to change at the same time), which may be useful to reach
communities with established norms and processes that will only change if the whole group
changes. Finally, induction interventions, such as word-of-mouth situations or media marketing
campaigns, stimulate interpersonal communication persuading others to adopt a new behaviour.
These interventions do not necessarily use network data, but depend on the network for their
effects. Thus, induction is an effective strategy to stimulate peer-to-peer exchanges which are likely
to create cascades in information diffusion and behavioural change [114].
4.2. Inclusiveness, valuation, and concerted responsibilities distribution
Although farmers are voluntarily responsive to (expected forthcoming) regulation and to
intense policy scrutiny regarding BMPs adoption [16], government intervention alone cannot
solve complex issues like cross-boundary nonpoint source pollution. In fact, interviewees
(including farmers) recognized the need for regulation and compliance control, but also
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 20 / 30
stressed the lack of economic valuation and social recognition for voluntary adoption of
BMPs. Especially, small-scale farmers reported oppressive economic, political, and social pres-
sure, with limited support or recognition. Inequitable distribution of pressure and responsibil-
ities across sectors was also repeatedly mentioned.
As stated by several farmers-participants, the inclusion and active participation of all rele-
vant stakeholders in the discussion, proposition, co-design and decision process of policy, is
necessary in order to take their respective interests and responsibilities into account. Such par-
ticipatory processes increase the chances to reach consensus among stakeholders, and to
develop realistic, equitable, and better calibrated policy mixes. Evidence of collaborative prac-
tice and multi-stakeholder partnerships in addressing complex environmental issues is plenti-
ful [115118]. In the case of Chesapeake Bay watershed, partnerships including representatives
from land grant universities, trade associations and businesses, agricultural and environmental
government agencies, were established to provide expertise and leadership on policy and pro-
grams development. However, this requires available time and human resources from the par-
ticipating organizations. For example, farmers lack time for off-farm activities, making their
inclusion difficult [46]. This could be addressed by valuing or compensating their participation
in such initiatives. Administrators (of farmers organizations or cooperatives for example)
could play a role in facilitating and incentivizing (farmers) engagement in multi-stakeholders
partnership [102].
As inter-farmers interactions were shown to be an effective information channel, incenti-
vising farmers cooperation and collaboration through collective engagement may be an inter-
esting avenue. Recent agricultural policy reforms in Europe have institutionalised such
collective schemes [111,112,119,120], which are increasingly recognised as an effective way
to coordinate conservation effort at landscape level [103,109,110,112,113]. In France for
instance, agri-environmental schemes can be contracted by collective organizations such as
pastoral groups or municipalities [120], while collective incentives (e.g. where farmers are only
paid once a certain area coverage is collectively reached) are being implemented [121]. Since
2016, the Netherlands went a step further in institutionalising collective action by choosing to
implement agri-environmental schemes through collective applications exclusively (mostly
through Environmental Cooperatives) [112,119,122]. These novel approaches have potential
to rapidly diffuse BMPs, as they may be fostered by peer-to-peer interactions [111]. Such col-
lective schemes can be more effectively diffused by using network characteristics to identify
strategic adopters (farmers) and diffusers of the scheme (farmers and non-farmers). This may
be achieved by targeting actors who appear as bridging individuals or opinion leaders in the
network, or by identifying communities of practice which are more likely to change at the
same time (segmentation) [114].
V. Limitations and future research needs
Throughout this discussion, we have presented potential future policy avenues for increased
BMPs implementation based upon both our analyses and existing literature. However, the
sample size of this research (number of survey-respondents) is too small to be considered rep-
resentative of the larger region (Pennsylvania or Chesapeake Bay watershed), or of dairy farm-
ers in general. The method could be replicated with much larger and more diverse samples
(for example, by including other types of production and farming systems), in order to deter-
mine statistically significant typologies and relations between actor types and messages charac-
teristics. In addition, more qualitative research is needed to understand the common
characteristics of farmers who share similar networks. In fact, as we demonstrated that differ-
ent farmers are linked to different actors, future research should elucidate the criteria of
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 21 / 30
differentiation. This would allow to investigate and compare the networks of farmer types,
which may yield more precise insights for intervention targeting. Similarly, for profit indus-
tries could be better characterized, with a differentiation between agricultural inputs suppliers
and agricultural outlets. This would allow to determine the role of these different types of
industries in the sustainability transition, and the frameworks through which they can be
included. Other geographic areas should be targeted, in order to identify commonalities and
trends in relational patterns.
We acknowledge that information needs may vary per farmer and per BMP, which might
also explain patterns in messages distributions and information flows. This factor of differenti-
ation has not been taken into account in our analysis, and needs to be further investigated in
future studies. Moreover, data collection should not be limited to BMP-messages, and should
be extended to “farm practices messages”, as some actors may strongly influence farmers prac-
tices, but may not communicate about BMPs (e.g. dairy cooperatives). These actors could also
constitute potential policy targets or entry points, that could have a powerful influence in
changing farmers practices.
We have defined strategies to enhance BMP-messages acceptance by farmers, by targeting
trusted intermediaries (i.e. message senders who are considered more credible by the farmers).
However, it is important to note that, although this may lead a wider range of farmers to con-
sider adopting certain BMPs, it may not lead to actual practice uptake. In fact, in the words of
Social Judgement Theory, we have focused on the acceptance “of the credibility and trustwor-
thiness of the intervening actor” [42,43,52]. Other factors and layers of acceptance may hin-
der BMPs adoption despite the message being received by a trusted source, such as the
acceptance of the problem definition, or of the perceived consequences or risks associated
with the intervention [42,43,52].
Farmers preferences, needs, and motivations to engage in conservation action vary greatly,
requiring “menus” of conservation measures, where a variety of options are proposed to farm-
ers [123,124]. Follow-up research could focus on determining types of farmers based on their
trusted sources of information, and on studying relational patterns between types of measures,
message source (i.e. the actors through which they are convened to farmers), message accep-
tance, and actual practice uptake. This would be especially useful to assess if “harder to reach
farmers” (e.g. who are unwilling to work with government and research) can be reached via
“non-traditional” conservation intermediaries such as supply chain actors.
In addition, more research is needed to understand the subtleties of cascade effects of different
policy interventions. As discussed, tailored governance entry points may change BMPs adoption
and information exchange by dairy farmers in central Pennsylvania. Further research could both
refine policy mixes and provide more content about adoption itself, instead of information
exchange specifically. There remains significant uncertainty, not only about governance instru-
ments, but also about the ability for BMPs to positively impact large watershed change [125].
Finally, persistent agri-environmental governance challenges remain rooted in the valuation and
accounting system that frame the entire governance structure. Addressing those would require
the redesign of standards and valuation frameworks, inclusive of the true-costs of environmental
degradation and the true-return of restoration and conservation. Hence, methods and standards
to assess, quantify and monitor the degree of liability of actors in environmental deterioration,
and their credits/share in natural resources preservation need to be further investigated.
VI. Conclusion
Water quality governance is a complex and challenging task, complexity which becomes even
more apparent through our study. In the context of Chesapeake Bay’s nonpoint source
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 22 / 30
pollution issue, this study has assessed for the first time how information about specific BMPs
is exchanged with dairy farmers in a sub-region of the bay, unveiling how the actors network
is structured. Therewith we demonstrate how using network data allows to identify (BMP-)
message senders whose messages are more likely to be accepted by certain farmers.
Our findings suggest that a variety of tailored policy measures might be optimally combined
through specific actor entry points to build better information exchange across stakeholders.
For instance, our study confirms the effectiveness of inter-farmer interaction in disseminating
farm practices such as BMPs. In addition, certain actors of the for profit sector who are in
close interaction with farmers could be viewed as potential conservation intermediaries. This
is especially relevant to reach farmers who are unwilling to work with the government, and
would allow to alleviate the concentration of pressure and information on farmers.
We have demonstrated that we can increase policy mixes coverage by tailoring governance
entry points to specific BMPs and actors, and by building on variability in farmers’ networks
and exposure. We recommend that policy makers consider the variety of entry points and pos-
sible intermediaries that can be involved in encouraging BMPs uptake when designing policy
mixes. Thereby the policy mix for BMP diffusion could be broadened, through measures that
complement the “menu” of existing incentives, potentially allowing to reach more (and differ-
ent types of) farmers. In particular, we suggest that:
i. Focus could be put on designing conservation incentives that target supply chain actors as
conservation intermediaries, such as cooperatives, processors, or private consultants,
instead of directly targeting farmers;
ii. Compliance-control of manure management planning could be conducted by accredited
private certifiers;
iii. Policy should focus on encouraging inter-farmers interaction; thus, farmers’ mobility,
training, knowledge-exchange, and engagement in collaborative multi-stakeholders initia-
tives should be encouraged via financial or non-pecuniary compensation;
iv. Collective incentives could be considered to better coordinate conservation effort at the
landscape or (sub-)watershed scale;
v. All relevant stakeholders (including farmers) should be concerted and included in the dis-
cussion, proposition, co-design and decision process of policy, in order to take their respec-
tive interests and responsibilities into account.
Further research is needed to better understand if and how information needs differ per
BMP, to characterize farmer-type networks, and to determine with which policy instruments
and financing schemes these entry points can be optimally combined.
Supporting information
S1 File. Survey questionnaires. I. questionnaire for farmers; II. Questionnaire for organiza-
tions.
(DOCX)
S2 File. Categorization of message types.
(DOCX)
S3 File. Guiding questions for semi-structured interviews.
(DOCX)
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 23 / 30
S1 Table. List and description of BMPs.
(DOCX)
S2 Table. List and description of the organizations from the socio-centric network.
(DOCX)
S3 Table. Codes and themes from interviews content analysis.
(DOCX)
Acknowledgments
The authors would like to thank all the stakeholders who participated in this research by vol-
untarily accepting to share their experience and opinions with us, as well as all the persons
who helped facilitating the data collection process.
Author Contributions
Conceptualization: Elsa L. Dingkuhn, Lilian O’Sullivan, Rogier P. O. Schulte, Caitlin A.
Grady.
Data curation: Elsa L. Dingkuhn.
Formal analysis: Elsa L. Dingkuhn.
Funding acquisition: Rogier P. O. Schulte, Caitlin A. Grady.
Investigation: Elsa L. Dingkuhn.
Methodology: Elsa L. Dingkuhn, Lilian O’Sullivan, Rogier P. O. Schulte, Caitlin A. Grady.
Project administration: Elsa L. Dingkuhn, Rogier P. O. Schulte, Caitlin A. Grady.
Resources: Rogier P. O. Schulte, Caitlin A. Grady.
Supervision: Lilian O’Sullivan, Rogier P. O. Schulte, Caitlin A. Grady.
Validation: Elsa L. Dingkuhn, Lilian O’Sullivan, Rogier P. O. Schulte, Caitlin A. Grady.
Visualization: Elsa L. Dingkuhn, Rogier P. O. Schulte, Caitlin A. Grady.
Writing original draft: Elsa L. Dingkuhn.
Writing review & editing: Elsa L. Dingkuhn, Lilian O’Sullivan, Rogier P. O. Schulte, Caitlin
A. Grady.
References
1. Team PA. The Economic Impact of Agriculture in Pennsylvania: 2021 Update. 2020. Available: https://
www.agriculture.pa.gov/Documents/2021%20Pennsylvania%20Agriculture%20Economic%20Impact
%20Report.pdf.
2. Berg E, Kim H-J, Rumpler J. Troubled Waters Industrial Pollution Still Threatens American Water-
ways. PennEnvrionment Research and Policy Center—Frontier Group. 2018.
3. PDEP. 2022 Pennsylvania Integrated Water Quality Report. 2022. Available: https://storymaps.arcgis.
com/stories/b9746eec807f48d99decd3a583eede12.
4. EPA. Chesapeake Bay TMDL Executive Summary. 2010.
5. PDEP. 2016 Final Pennsylvania Integrated Water Quality Monitoring and Assessment Report—Clean
Water Act. Pennsylvania Department of Environmental Protection. 2016.
6. Shortle J. Policy Reforms Needed for Better Water Quality and Lower Pollution Control Costs.
CHOICES. 2017; 32: 1–7.
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 24 / 30
7. Renting H, Rossing WAH, Groot JCJ, van der Ploeg JD, Laurent C, Perraud D, et al. Exploring multi-
functional agriculture. A review of conceptual approaches and prospects for an integrative transitional
framework. J Environ Manage. 2009; 90: S112–S123. https://doi.org/10.1016/j.jenvman.2008.11.014
PMID: 19121889
8. Moon W. Conceptualising multifunctional agriculture from a global perspective: Implications for gov-
erning agricultural trade in the post-Doha Round era. Land use policy. 2015; 49: 252–263. https://doi.
org/10.1016/j.landusepol.2015.07.026
9. de Groot RS, Alkemade R, Braat L, Hein L, Willemen L. Challenges in integrating the concept of eco-
system services and values in landscape planning, management and decision making. Ecological
Complexity. 2010; 7: 260–272. https://doi.org/10.1016/j.ecocom.2009.10.006
10. Simoncini R, Ring I, Sandstro
¨m C, Albert C, Kasymov U, Arlettaz R. Constraints and opportunities for
mainstreaming biodiversity and ecosystem services in the EU’s Common Agricultural Policy: Insights
from the IPBES assessment for Europe and Central Asia. Land use policy. 2019;88. https://doi.org/10.
1016/j.landusepol.2019.104099
11. Schulte RPO, Creamer RE, Donnellan T, Farrelly N, Fealy R, O’Donoghue C, et al. Functional land
management: A framework for managing soil-based ecosystem services for the sustainable intensifi-
cation of agriculture. Environ Sci Policy. 2014; 38: 45–58. https://doi.org/10.1016/j.envsci.2013.10.
002
12. Coyle C, Creamer RE, Schulte RPO, O’Sullivan L, Jordan P. A Functional Land Management concep-
tual framework under soil drainage and land use scenarios. Environ Sci Policy. 2016; 56: 39–48.
https://doi.org/10.1016/j.envsci.2015.10.012
13. Schulte RPOO’Sullivan L, Vrebos D, Bampa F, Jones A, Staes J. Demands on land: Mapping compet-
ing societal expectations for the functionality of agricultural soils in Europe. Environ Sci Policy. 2019;
100: 113–125. https://doi.org/10.1016/j.envsci.2019.06.011
14. Baylis K, Peplow S, Rausser G, Simon L. Agri-environmental policies in the EU and United States: A
comparison. Ecological Economics. 2008; 65: 753–764. https://doi.org/10.1016/j.ecolecon.2007.07.
034
15. Amdur L, Bertke E, Freese J, Marggraf R. Agri-Environmental Policy Measures in Israel: The Potential
of Using Market-Oriented Instruments. Environ Manage. 2011; 47: 859–875. https://doi.org/10.1007/
s00267-011-9655-y PMID: 21431445
16. Savage JA, Ribaudo MO. Impact of environmental policies on the adoption of manure management
practices in the Chesapeake Bay watershed. J Environ Manage. 2013; 129: 143–148. https://doi.org/
10.1016/j.jenvman.2013.06.039 PMID: 23916836
17. Meyer C, Matzdorf B, Mu¨ller K, Schleyer C. Cross Compliance as payment for public goods? Under-
standing EU and US agricultural policies. Ecological Economics. 2014; 107: 185–194. https://doi.org/
10.1016/j.ecolecon.2014.08.010
18. Shortle JS, Ribaudo M, Horan RD, Blandford D. Reforming Agricultural Nonpoint Pollution Policy in an
Increasingly Budget-Constrained Environment. Environ Sci Technol. 2012; 16: 1316–1325.
19. Drevno A. Policy tools for agricultural nonpoint source water pollution control in the U.S. and E.U. Man-
agement of Environmental Quality: An International Journal. Emerald Group Publishing Ltd.; 2016. pp.
106–123. https://doi.org/10.1108/MEQ-12-2014-0177
20. USDA. 2021–2022 Agricultural Statistics Annual Bulletin Pennsylvania. Harrisburg; 2022. Available:
www.agriculture.pa.gov.
21. Toor GS, Sims JT, Dou Z. Reducing phosphorus in dairy diets improves farm nutrient balances and
decreases the risk of nonpoint pollution of surface and ground waters. Agric Ecosyst Environ. 2005;
105: 401–411. https://doi.org/10.1016/j.agee.2004.06.003
22. Amin MGM, Karsten HD, Veith TL, Beegle DB, Kleinman PJ. Conservation dairy farming impact on
water quality in a karst watershed in northeastern US. Agric Syst. 2018; 165: 187–196. https://doi.org/
10.1016/j.agsy.2018.06.010
23. Baumgart-getz A, Prokopy Stalker L, Floress K. Why farmers adopt best management practice in the
United States: A meta-analysis of the adoption literature. J Environ Manage. 2011; 96: 17–25. https://
doi.org/10.1016/j.jenvman.2011.10.006 PMID: 22208394
24. Liu T, Bruins RJF, Herberling MT. Factors Influencing Farmers’ Adoption of Best Management Prac-
tices: A Review and Synthesis. Sustainability. 2018; 10: 1–26. https://doi.org/10.3390/su10020432
PMID: 29682334
25. Rittenburg RA, Squires AL, Boll J, Brooks ES, Easton ZM, Rebecca A, et al. Agricultural BMP effec-
tiveness and dominant hydrological flow paths: Concepts and a review. J Am Water Resour Assoc.
2015; 51: 305–329. https://doi.org/10.1111/1752-1688.12293
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 25 / 30
26. Liu Y, Wang R, Guo T, Engel BA, Flanagan DC, Lee JG, et al. Evaluating efficiencies and cost-effec-
tiveness of best management practices in improving agricultural water quality using integrated SWAT
and cost evaluation tool. J Hydrol (Amst). 2019; 577: 1–16. https://doi.org/10.1016/j.jhydrol.2019.
123965
27. Schall D, Lansing D, Leisnham P, Shirmohammadi A, Montas H, Hutson T. Understanding stake-
holder perspectives on agricultural best management practices and environmental change in the
Chesapeake Bay: A Q methodology study. J Rural Stud. 2018; 60: 21–31. https://doi.org/10.1016/j.
jrurstud.2018.03.003
28. Greiner R, Patterson L, Miller O. Motivations, risk perceptions and adoption of conservation practices
by farmers. Agric Syst. 2009; 99: 86–104. https://doi.org/10.1016/j.agsy.2008.10.003
29. Petrzelka P, Korsching PF, Malia JE. Farmers’ attitudes and behavior toward sustainable agriculture.
Journal of Environmental Education. 1996; 28: 38–44. https://doi.org/10.1080/00958964.1996.
9942814
30. Prokopy LS, Floress K, Arbuckle JG, Church SP, Eanes FR, Gao Y, et al. Adoption of agricultural con-
servation practices in the United States: Evidence from 35 years of quantitative literature. J Soil Water
Conserv. 2019; 74: 520–534. https://doi.org/10.2489/jswc.74.5.520
31. Isaac ME. Agricultural information exchange and organizational ties: The effect of network topologyon
managing agrodiversity. Agric Syst. 2012; 109: 9–15. https://doi.org/10.1016/j.agsy.2012.01.011
32. Aguilar-Gallegos N, Muñoz-Rodrı
´guez M, Santoyo-Corte
´s H, Aguilar-A
´vila J, Klerkx L. Information net-
works that generate economic value: A study on clusters of adopters of new or improved technologies
and practices among oil palm growers in Mexico. Agric Syst. 2015; 135: 122–132. https://doi.org/10.
1016/j.agsy.2015.01.003
33. Sanya LN, Sseguya H, Kyazze FB, Baguma Y, Kibwika P. Actor diversity and interactions in the devel-
opment of banana hybrid varieties in Uganda: implications for technology uptake. Journal of Agricul-
tural Education and Extension. 2018; 24: 153–167. https://doi.org/10.1080/1389224X.2017.1401549
34. Cofre
´-Bravo G, Klerkx L, Engler A. Combinations of bonding, bridging, and linking social capital for
farm innovation: How farmers configure different support networks. J Rural Stud. 2019; 69: 53–64.
https://doi.org/10.1016/j.jrurstud.2019.04.004
35. Szreter S, Woolcock M. Health by association? Social capital, social theory, and the political economy
of public health. Int J Epidemiol. 2004; 33: 650–667. https://doi.org/10.1093/ije/dyh013 PMID:
15282219
36. Klerkx L, Proctor A. Beyond fragmentation and disconnect: Networks for knowledge exchange in the
English land management advisory system. Land use policy. 2013; 30: 13–24. https://doi.org/10.1016/
j.landusepol.2012.02.003
37. O’Sullivan L, Leeuwis C, de Vries L, Wall DP, Heidkroß T, Madena K, et al. Trust Versus Content in
Multi-functional Land Management: Assessing Soil Function Messaging in Agricultural Networks.
Environ Manage. 2022; 69: 1167–1185. https://doi.org/10.1007/s00267-022-01647-2 PMID:
35451645
38. Klerkx L, Hall A, Leeuwis C. Strengthening agricultural innovation capacity: Are innovation brokers the
answer? International Journal of Agricultural Resources, Governance and Ecology. 2009; 8: 409–438.
https://doi.org/10.1504/IJARGE.2009.032643
39. Sherif M, Hovland CI. Social judgment: Assimilation and contrast effects in communication and attitude
change. Yale studies in attitude and communication. 1961;4.
40. Sherif CW, Sherif M, Nebergall RE. Attitude and attitude change: The social judgment-involvement
approach. Philadelphia, Saunders; 1965.
41. Mallard J. Engaging students in social judgment theory. Communication Teacher. 2010; 24: 197–202.
https://doi.org/10.1080/17404622.2010.512869
42. Leeuwis C. Communication for rural innovation: rethinking agricultural extension, with contributions
from Anne van den Ban. Blackwell Science; 2004.
43. van den Berg LM, Dingkuhn EL, Meehan N, O’Sullivan L. Investigating bottlenecks hampering the
adoption of water quality-enhancing practices for sustainable land management in Ireland. J Environ
Manage. 2023; 345: 118741. https://doi.org/10.1016/j.jenvman.2023.118741 PMID: 37604104
44. Gronewold KL, Burnett A, Meister M. Farmers’ Cynicism Toward Nature and Distrust of the Govern-
ment: Where Does that Leave Conservation Buffer Programs? Applied Environmental Education &
Communication. 2012; 11: 18–24. https://doi.org/10.1080/1533015X.2012.728061.
45. Rust NA, Stankovics P, Jarvis RM, Morris-Trainor Z, de Vries JR, Ingram J, et al. Have farmers had
enough of experts? Environ Manage. 2022; 69: 31–44. https://doi.org/10.1007/s00267-021-01546-y
PMID: 34633488
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 26 / 30
46. Hurley P, Lyon J, Hall J, Little R, Tsouvalis J, White V, et al. Co-designing the environmental land man-
agement scheme in England: The why, who and how of engaging ‘harder to reach’ stakeholders. Peo-
ple and Nature. 2022; 4: 744–757. https://doi.org/10.1002/pan3.10313
47. Shortle J, Horan RD. Nutrient Pollution: A Wicked Challenge for Economic Instruments. Water Eco-
nomics and Policy. 2017; 3: 1–39. https://doi.org/10.1142/S2382624X16500338
48. O’Sullivan L, Wall D, Creamer R, Bampa F, Schulte RPO. Functional Land Management: Bridging the
Think-Do-Gap using a multi-stakeholder science policy interface. Ambio. 2018; 47: 216–230. https://
doi.org/10.1007/s13280-017-0983-x PMID: 29178059
49. Dingkuhn EL, Wezel A, Bianchi FJJA, Groot JCJ, Wagner A, Yap HT, et al. A multi-method approach
for the integrative assessment of soil functions: Application on a coastal mountainous site of the Philip-
pines. J Environ Manage. 2020; 264: 1–14. https://doi.org/10.1016/j.jenvman.2020.110461 PMID:
32250894
50. Forney J, Rosin C, Campbell H. Agri-environmental governance as an assemblage: multiplicity,
power, and transformation. Earthscan Food and Agriculture. 2018.
51. Perez MR. Regulating Farmer Nutrient Management: A Three-State Case Study on the Delmarva
Peninsula. J Environ Qual. 2015; 44: 402–414. https://doi.org/10.2134/jeq2014.07.0304 PMID:
26023959
52. Sherif M, Hovland CI. Social judgment: Assimilation and contrast effects in communication and attitude
change. Yale studies in attitude and communication. 1961;4.
53. Butts CT. Social network analysis: A methodological introduction. Asian J Soc Psychol. 2008; 11: 13–
41. https://doi.org/10.1111/j.1467-839X.2007.00241.x
54. Grady CA, He X, Peeta S. Integrating social network analysis with analytic network process for interna-
tional development project selection. Expert Syst Appl. 2015; 42: 5128–5138. https://doi.org/10.1016/
j.eswa.2015.02.039
55. Balfour B, Alter TR. Mapping community innovation: Using social network analysis to map the interac-
tional field, identify facilitators, and foster community development. Community Development. 2016;
5330: 1–18. https://doi.org/10.1080/15575330.2016.1153493
56. Bandiera O, Rasul I. Social Networks and technology adoption in Northern Mozambique. The Eco-
nomic Journal. 2006; 116: 869–902.
57. Geerling JMFGW Smits EMAJM. Analyzing Collaborative Governance Through Social Network Analy-
sis: A Case Study of River Management Along the Waal River in The Netherlands. Environ Manage.
2016; 57: 355–367. https://doi.org/10.1007/s00267-015-0606-x PMID: 26324060
58. Rudnick J, Niles M, Lubell M, Cramer L. A comparative analysis of governance and leadership in agri-
cultural development policy networks. World Dev. 2019; 117: 112–126. https://doi.org/10.1016/j.
worlddev.2018.12.015
59. Valujeva K, Freed EK, Nipers A, Jauhiainen J, Schulte RPO. Pathways for governance opportunities:
Social network analysis to create targeted and effective policies for agricultural and environmental
development. J Environ Manage. 2023;325. https://doi.org/10.1016/j.jenvman.2022.116563 PMID:
36308958
60. Schader C, Lampkin N, Muller A, Stolze M. The role of multi-target policy instruments in agri-environ-
mental policy mixes. J Environ Manage. 2014; 145: 180–190. https://doi.org/10.1016/j.jenvman.2014.
06.016 PMID: 25038517
61. Howlett M, Rayner J. Design Principles for Policy Mixes: Cohesion and Coherence in New Gover-
nance. Res Policy. 2007; 26: 1–18. https://doi.org/10.1016/S1449-4035(07)70118-2
62. Trencher G, Van Der Heijden J. Instrument interactions and relationships in policy mixes: Achieving
complementarity in building energy efficiency policies in New York, Sydney and Tokyo. Energy Res
Soc Sci. 2019; 54: 34–45. https://doi.org/10.1016/j.erss.2019.02.023
63. Parson EA, Kravitz EL. Market Instruments for the Sustainability Transition. Annu Rev Environ
Resour. 2013; 38: 415–440. https://doi.org/10.1146/annurev-environ-061311-111640
64. Claassen R, Hansen L, Peters M, Breneman V, Cattaneo A, Feather P, et al. Agri-environmental Pol-
icy at the Crossroads: Guideposts on a Changing Landscape. Agricultural Economic Report. 2001; 1–
73. https://doi.org/10.1080/01635581.2014.888898
65. Goodman LA. Snowball sampling. The Annals of Mathematical Statistics. 2019; 32: 148–170.
66. Heckathorn DD, Cameron CJ. Network Sampling: From Snowball and Multiplicity to Respondent-
Driven Sampling. Annu Rev Sociol. 2017; 43: 101–119. https://doi.org/10.1146/annurev-soc-060116-
053556 Copyright.
67. Chu K-H, Wipfli H, Valente TW. Using Visualizations to Explore Network Dynamics. Journal of Social
Structure. 2014;14.
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 27 / 30
68. Bastian M, Heymann S, Jacomy M. Gephi: An open source software for exploring and manipulating
networks. International AAAI Conference on Weblogs and Social Media. Denmark: Alborg Univeristy;
2009. pp. 361–362.
69. Brandes U. A faster algorithm for betweenness centrality. Journal of Mathematical Sociology. 2001;
25: 163–177. https://doi.org/10.1080/0022250X.2001.9990249
70. Iijima R, Kamada Y. Social distance and network structures. Theoretical Economics. 2017; 12: 655–
689. https://doi.org/10.3982/te1873
71. Martı
´nez-Lo
´pez B, Perez AM, Sa
´nchez-Vizcaı
´no JM. Social network analysis. Review of general con-
cepts and use in preventive veterinary medicine. Transbound Emerg Dis. 2009; 56: 109–120. https://
doi.org/10.1111/j.1865-1682.2009.01073.x PMID: 19341388
72. Sarama
¨ki J, Kivela
¨M, Onnela JP, Kaski K, Kerte
´sz J. Generalizations of the clustering coefficient to
weighted complex networks. Phys Rev E Stat Nonlin Soft Matter Phys. 2007; 75: 2–5. https://doi.org/
10.1103/PhysRevE.75.027105 PMID: 17358454
73. Tabassum S, Pereira FSF, Fernandes S, Gama J. Social network analysis: An overview. Wiley Inter-
discip Rev Data Min Knowl Discov. 2018; 8: 1–21. https://doi.org/10.1002/widm.1256
74. O’Sullivan L, Leeuwis C, de Vries L, Heidkroß T, Madena K, Wall D, et al. Harnessing the network
potential, bundling and multifunctionality of soil functions for sustainable land management (submitted
for publication). 2021.
75. Erlingsson C, Brysiewicz P. Orientation among multiple truths: An introduction to qualitative research.
African Journal of Emergency Medicine. 2013; 3: 92–99. https://doi.org/10.1016/j.afjem.2012.04.005
76. Vaismoradi M, Turunen H, Bondas T. Content analysis and thematic analysis: Implications for con-
ducting a qualitative descriptive study. Nurs Health Sci. 2013; 15: 398–405. https://doi.org/10.1111/
nhs.12048 PMID: 23480423
77. Bengtsson M. How to plan and perform a qualitative study using content analysis. NursingPlus Open.
2016; 2: 8–14. https://doi.org/10.1016/j.npls.2016.01.001
78. Braun V, Clarke V, Hayfield N, Terry G. Thematic Analysis. In: Liamputtong P, editor. Handbook of
Research Methods in Health Social Sciences. Singapore: Springer; 2019. pp. 843–860.
79. Rahman J, Noronha A, Thiele I, Rahman S. Leigh Map: A Novel Computational Diagnostic Resource
for Mitochondrial Disease. Ann Neurol. 2016; 81: 9–16. https://doi.org/10.1002/ana.24835 PMID:
27977873
80. Castleberry A, Nolen A. Thematic analysis of qualitative research data: Is it as easy as it sounds? Curr
Pharm Teach Learn. 2018; 10: 807–815. https://doi.org/10.1016/j.cptl.2018.03.019 PMID: 30025784
81. McGurk E, Hynes S, Thorne F. Participation in agri-environmental schemes: A contingent valuation
study of farmers in Ireland. J Environ Manage. 2020; 262: 110243. https://doi.org/10.1016/j.jenvman.
2020.110243 PMID: 32250775
82. Weigel C, Paul LA, Ferraro PJ, Messer KD. Challenges in Recruiting U.S. Farmers for Policy-Relevant
Economic Field Experiments. Appl Econ Perspect Policy. 2021; 43: 556–572. https://doi.org/10.1002/
aepp.13066
83. Krell RK, Fisher ML, Steffey KL. A Proposal for Public and Private Partnership in Extension.J Integr
Pest Manag. 2016; 7: 1–10. https://doi.org/10.1093/jipm/pmw001 PMID: 26949567
84. Eanes FR, Singh AS, Bulla BR, Ranjan P, Fales M, Wickerham B, et al. Crop advisers as conservation
intermediaries: Perceptions and policy implications for relying on nontraditional partners to increase U.
S. farmers’ adoption of soil and water conservation practices. Land use policy. 2019; 81: 360–370.
https://doi.org/10.1016/j.landusepol.2018.10.054
85. Pei E, Guo CQ, Van Luu B. Environmental, social and governance transparency and firm value. Bus
Strategy Environ. 2018; 27: 987–1004. https://doi.org/10.1002/bse.2047
86. Tisenkopfs T, Kilis E, Grivins M, Adamsone A. Whose ethics and for whom? Dealing with ethical dis-
putes in agri-food governance. Agric Human Values. 2019; 36: 353–364. https://doi.org/10.1007/
s10460-019-09921-6
87. Turkey Hill Dairy. 2023. Available: https://www.turkeyhill.com/press/turkey-hill-dairy-press/news-
releases/2020-07-14-THCWP.
88. Team PA. Pennsylvania Agriculture: A Look at the Economic Impacts and Future Trends. Team Penn-
sylvania; 2018.
89. Sereiroth B, Ryuichi Y. The contributions of agricultural cooperatives to promote sustainable farming
in prey kabbas district, cambodia. Journal of the International Society for Southeast Asian Agricultural
Sciences. 2021; 27: 77–86.
90. Candemir A, Duvaleix S, Latruffe L. Agricultural Cooperatives and Farm Sustainability—a Literature
Review. J Econ Surv. 2021; 35: 1118–1144. https://doi.org/10.1111/joes.12417
PLOS ONE
A social network analysis of agricultural best management practices in central Pennsylvania
PLOS ONE | https://doi.org/10.1371/journal.pone.0303745 May 23, 2024 28 / 30