You are on page 1of 11

See discussions, stats, and author profiles for this publication at: http://www.researchgate.

net/publication/232411770

Valuing local knowledge as a source of expert


data: Farmer engagement and the design of
decision support systems

ARTICLE in ENVIRONMENTAL MODELLING AND SOFTWARE · OCTOBER 2012


Impact Factor: 4.54 · DOI: 10.1016/j.envsoft.2011.09.013

CITATIONS DOWNLOADS VIEWS

15 98 136

6 AUTHORS, INCLUDING:

Michael Winter Chris J. Hodgson


University of Exeter Rothamsted Research
92 PUBLICATIONS 1,461 CITATIONS 16 PUBLICATIONS 222 CITATIONS

SEE PROFILE SEE PROFILE

Ann Louise Heathwaite Dave Robert Chadwick


Lancaster University Bangor University
190 PUBLICATIONS 3,999 CITATIONS 129 PUBLICATIONS 2,004 CITATIONS

SEE PROFILE SEE PROFILE

Available from: David M. Oliver


Retrieved on: 09 September 2015
Environmental Modelling & Software 36 (2012) 76e85

Contents lists available at SciVerse ScienceDirect

Environmental Modelling & Software


journal homepage: www.elsevier.com/locate/envsoft

Valuing local knowledge as a source of expert data: Farmer engagement and the
design of decision support systems
David M. Oliver a, Rob D. Fish b, *, Michael Winter b, Chris J. Hodgson c, A. Louise Heathwaite d,
Dave R. Chadwick c
a
Biological & Environmental Sciences, School of Natural Sciences, University of Stirling, Stirling FK9 4LA, UK
b
Centre for Rural Policy Research, Department of Politics, University of Exeter, Exeter, Devon EX4 6TL, UK
c
Rothamsted Research, North Wyke, Okehampton, Devon EX20 2SB, UK
d
Lancaster Environment Centre, Lancaster University, Lancaster LA1 4YQ, UK

a r t i c l e i n f o a b s t r a c t

Article history: Engagement with farmers and landowners is often undertaken by the research community to obtain
Received 9 June 2011 information relating to typical land, livestock and enterprise management and generally centres on
Received in revised form responses to questionnaire surveys. Farmers and land managers are constituted as expert observers of
1 September 2011
ground-level processes and provide diverse information on farming practices, enterprise economics and
Accepted 13 September 2011
underpinning attitudes towards risk. Research projects designed to inform policy and practice may rely
Available online 6 November 2011
on such data to understand better on-the-ground decisions that can impact on environmental quality
and the rural economy. Such approaches to eliciting local-level expert knowledge can generate large
Keywords:
Decision support
quantities of data from which to formulate rules relating to farm enterprise types. In turn, this can help to
Expert inform the structure of Decision Support Systems (DSS) and risk-based tools to determine farming
Farmer practices likely to impact on environmental quality. However, in this paper we advocate the need for
Local knowledge integrated farmer participation throughout the whole research process e from project inception through
Questionnaire survey to community qualitative validation and legitimation - and thus not just for the elicitation of ques-
Stakeholder participation tionnaire responses. With farm questionnaire surveys being adopted widely by the research community,
Uncertainty it is an opportune time to highlight a recent case study of the Taw catchment, Devon, UK. This serves as
Water quality
an example of co-construction of a DSS via a co-ordinated and integrated approach to expert elicitation
with a farmer questionnaire survey as a central methodology. The aim of the paper is to detail the core
aspects of an iterative cycle of participatory environmental management and DSS development for water
quality protection and consider the multiple benefits of co-ordinated programmes of engagement with
the farming community in this process.
Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction scientific knowledge. Alan Irwin has coined the term “Citizen
Science” to capture the lasting significance of this ‘lay’ expertise
Farmers and land managers are increasingly targeted by scien- (Irwin, 1995).
tists to help inform research and policy tools (e.g. Nettle et al., 2010; A range of research techniques are potentially available to help
Vignola et al., 2010; Winsten et al., 2010). Knowledge and experi- scientists elicit farmer and land manager knowledge as the basis for
ence grounded in everyday land management practices can provide designing integrated and sustainable approaches to land manage-
insight only partly accessible through scientific observation, and ment. In this paper, we examine the key dimensions of an iterative
can serve to clarify and extend the evidence base informing deci- cycle of participatory environmental management for the devel-
sions. In other words, farmers and land managers are custodians of opment of Decision Support Systems (DSS). In general terms, the
salient local knowledge. They bring new reasoning and insight to purpose of DSS is to translate wider policy concerns for sustainable
bear upon problems that might otherwise be missed by specialist agriculture and water resources under climate change into opera-
tional, flexible and adaptive ‘on the ground’ responses. They
therefore remain high on the environmental research agenda even
* Corresponding author. though few DSS are currently used to inform policy and to drive
E-mail address: Fish@exeter.ac.uk (R.D. Fish). policy analysis (van Delden et al., 2011). Opening up such

1364-8152/$ e see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.envsoft.2011.09.013
D.M. Oliver et al. / Environmental Modelling & Software 36 (2012) 76e85 77

frameworks to accommodate the skill and expertise of farmers is characteristics typical of farmed land in order to inform the output
a critical step in developing DSS that have utility to a range of of a risk indexing tool (essentially, a field and farm DSS; Oliver et al.,
catchment stakeholders working between catchment-based 2009a, 2010) to identify agricultural land most vulnerable for
research and real world management decisions. This includes the contributing towards diffuse microbial pollution of water. Key
farming community, extension officers and farm advisors and phases of farmer recruitment, relationship building, on-farm
policy makers (Macleod and Haygarth, 2010). monitoring and farmer qualitative validation were essential to
In this paper our focus is on a DSS that aims to help manage risks complement the core data obtained by questionnaire survey.
arising from agriculture for water quality and in particular, an The DSS resulting from this project demonstrates considerable
integrated research process that has a farmer questionnaire survey compatibility with prevailing approaches to environmental risk
as a central part of the DSS methodology. A parallel argument about management, most notably the source-pathway-receptor model
examining the use of questionnaire surveys to elicit specialist (e.g. Lytton et al., 2003). The case study of farmer participation for
scientific knowledge on land-water management has been dis- DSS co-construction therefore has relevance and applicability to
cussed elsewhere (Fish et al., 2009). Questionnaire surveys have a broad range of diffuse agricultural pollutants that can be
been a key and longstanding element of social science research into modelled using this generic conceptual framework. While the
issues of land management and conservation (e.g. Potter and questionnaire survey and farmer participation was conducted to
Gasson, 1988; Morris and Potter, 1995; Fish et al., 2003). draw on information relating to microbial transfers from land to
However, one of the emerging challenges of interdisciplinary water to inform a DSS, the Taw case study analysis below identifies
research is to relate this type of work to the specific needs of DSS general issues and complexities, in addition to the advantages and
and to embed such elicitation procedures within a more integrated limitations of the data, that can be of benefit to future participatory
framework that involves the participation of the farming commu- modelling studies.
nity from project inception through to legitimisation of the
research product (Barreteau et al., 2010). As a way of eliciting local- 3. Engaging farmers within DSS: underpinning rationales and
level knowledge, questionnaire surveys can generate large quan- approaches
tities of data from which to generate ‘model’ rules relating to farm
enterprise types and help to discern farming practices likely to In this section we explain how the process of designing and
impact on environmental quality. The proposition is that their value implementing a DSS relates to wider rationales for stakeholder
and impact can be magnified considerably through careful coor- engagement within environmental management, and the partic-
dination of the wider participatory process itself (e.g. Ticehurst ular function that questionnaire surveys and wider consultation
et al., 2011; Selfa et al., 2010; Oliver et al., 2009a). serve within this field. In principle, the idea of engagement with
We begin by providing the empirical context to this paper; an stakeholders - in this case the farming and land management
extensive questionnaire survey and programme of farmer partici- community - has tended to be based on three main arguments (see
pation undertaken by the authors in the Taw catchment in North especially Fiorino, 1990):
Devon, UK. The aim of the paper is to highlight an integrated
approach to participatory environmental management and DSS  Inclusiveness: that individuals have a legitimate right to influ-
development as exemplified using the Taw catchment project and ence processes that have a direct bearing on them. In this
its associated questionnaire survey as a case study. As a result, sense, engagement with farmers in DSS design should be seen
a sequential (and iterative) seven-step protocol is presented to as a pre-requisite of the research process because the system
promote farmer participation (through both informal dialogue and implies change;
formal survey-based procedures) as an integral part of DSS  Acceptability: that greater trust and legitimacy can be built into
evolution. what is being proposed by the environmental management
process. In other words engagement is a better way of
2. Case study context achieving particular ends. Farmers and land management
organisations and groups, will potentially be more accepting of
This paper bases its principal insights on a questionnaire survey the DSS if involved in the processes and reasoning that
of 77 farms, and a three year process of farmer engagement with 10 underpinned it;
of those farms in the Taw catchment, North Devon. The extended  Effectiveness: that outcomes and outputs are realised more
engagement component was limited to ten farms and built on the effectively. For example active engagement with farmers
assumption that greater resources of time could be committed provides a source of practical know-how that can help inform
towards nurturing a trusting working relationship with individual the DSS and lead to a better product.
research subjects (c.f. Jakeman and Letcher, 2003). The ten farms
were selected to represent a range of farm types and landscape/ In practical terms, it is useful to think of these three rationales in
geographical characteristics and were identified in advance of the terms of three main types of engagement (Chilvers, 2009), namely:
completion of all 77 farm questionnaires. The time and cost
constraints of typical 3-year research projects make it difficult to  Informing stakeholders: non-interactive engagement designed
undertake substantive engagement with a larger subset of farms. to disseminate information to constituencies of interested/
Thus, our study developed an integrated participatory process with affected people. Thus, engagement techniques are employed to
ten farmers and framed the experiences in the wider context of the make farmers and land managers aware of the existence and
77 farm systems. The participatory process is outlined in detail in utility of the DSS;
Section 4 but briefly it focused on a structured questionnaire con-  ‘Learning from’ stakeholders: interactive engagement designed
ducted face to face and designed to obtain information of on-farm primarily to take account of the views, interests or insights of
management practices likely to impact on the risk of microbial stakeholders. This type of engagement is therefore concerned
transfers to watercourses from livestock faecal material and with drawing out farmer and land manager insight to inform
manures deposited on farmed land. In general terms the ques- the DSS;
tionnaire survey was structured to obtain information relating to  ‘Collaborating with’ stakeholders: shared approaches to decision
land, livestock and manure management and general making among stakeholders based, for instance, on dialogue,
78 D.M. Oliver et al. / Environmental Modelling & Software 36 (2012) 76e85

discussion, deliberation and collective problem framing. In this proposed here was trialled in the Taw catchment and novel insights
case farmers and land managers work directly with specialists from this case study are detailed within each of the seven steps. A
to create the DSS. precursor to the seven iterative steps will always be a clear defi-
nition of the research need and this was established (and evolved)
A framework for clarifying the role of participation in partici- in conjunction with the farming community. This is an important
patory research has been proposed elsewhere as the basis for first phase for ensuring that any questionnaire survey and pro-
promoting and facilitating quality control in participatory research gramme of engagement is structured in order to meet the objec-
(Barreteau et al., 2010). Taking these engagement rationales and tives of the science under investigation and to avoid a mismatch
types together, our general proposition is that questionnaire between the type of engagement desired (i.e. ‘inform’, ‘learn from’
surveys should primarily be seen as a mechanism for drawing out or ‘collaborate’) in the DSS formulation process and the type of
farmer knowledge as the basis for improving model effectiveness technique used. As we have explained, questionnaire surveys are an
but that the methodology needs to be couched within a much appropriate means of eliciting key information but should be
larger programme of engagement (e.g. a shift to ‘collaborating with framed within a more integrated approach to elicitation so that
stakeholders’) as outlined above. ‘users’ of models become active participants in fostering the
acceptability of DSS at project inception (Oliver et al., 2009b). They
are a ‘learning from’ approach designed to inform the evidence base
4. Co-construction of a DSS: The Taw catchment case study
behind DSS.
Fig. 1 outlines a protocol for effective farmer-scientist co-
construction of a DSS guided by seven iterative stages as outlined in 4.1. Establish relationships (Fig. 1 stage 1)
the following (4.1e4.7). The co-learning process is an iterative cycle
and Fig. 1 is not dissimilar to the approach of modelling with In the Taw case study the farming community were contacted
stakeholders highlighted in Voinov and Bousquet (2010). Indeed, using a four-pronged approach involving the use of: (i) leaflet
approaches to social learning are well documented as being central distribution (see Fig. 2); (ii) existing networks; (iii) word-of-mouth;
to the management of complex environmental issues (e. g.; Pahl- and (iv) cold-calling. The purpose was to make clear the importance
Wostl et al., 2011, 2010; Jakeman and Letcher, 2003). The cycle of farmer opinions and to dispel any suggestion that their

Fig. 1. Seven stage protocol for effective farmer-scientist co-construction of a DSS: key iterative steps needed in the development of DSS for farmers.
D.M. Oliver et al. / Environmental Modelling & Software 36 (2012) 76e85 79

Fig. 2. Leaflet distributed to farmers in the Taw catchment as a method of introducing the project ahead of making contact and engaging in person.

participation was simply a pro-forma exercise (c.f. Glicken, 2000). shot across the catchment and this was followed up with a tele-
The leaflet (Fig. 2) was designed to emphasise the point that the phone call to attempt to arrange an interview. The leaflet proved to
research team needed the assistance of the farming community to be an important component of the Taw study: it introduced the
take this DSS project forward. The leaflet was distributed via mail- project, expressed a call for help from the farming community,
80 D.M. Oliver et al. / Environmental Modelling & Software 36 (2012) 76e85

highlighted the central role of farmers in the project, informed that relationships, making regular visits and by maintaining commu-
we were independent of any government agency, and showed nication there is ample opportunity to undertake verification of
a photograph of the three researchers who would be undertaking questionnaire survey responses by direct observation as a cross-
the project and engaging with farmers and landowners. This comparison to other more formal data sources. So whether coin-
therefore served to introduce the team ahead of the farm visit. cidental or not, regular ‘checks’ for consistency between on-the-
Obtaining lay information from farmers can provide crucial data to ground operation and survey responses can be maintained and
the research community yet is often undervalued at project reveal interesting supplementary or confirmatory data.
inception stages. Appreciating the value of such information, and The successful recruitment of farmers for management surveys
anticipating farmer questions of ‘what’s in it for me’, the leaflet also and wider engagement can be enhanced through contacts at
informed of a free manure or slurry nutrient test to all those who respected environmental research institutes and the reputations of
participated in the survey (and access to the latest advice on how to organizations can therefore prove vital for research programmes
manage livestock manures). This not only provided the farmers that need the help of the farming community. Sterk et al. (2009)
with useful information free of charge but also facilitated discus- highlight examples of work undertaken in Ecuador whereby
sion around land and nutrient management issues and helped to extensive contact networks in environmental research institutes
conceptualise the DSS framework through agreed common needs facilitated positive research activities and enhanced farmer
(Quinn et al., 2010). engagement. Likewise, in the Taw catchment case study, the
The project team chose to adopt a face-to-face approach to grasslands research institute - North Wyke Research - provided
engagement. This approach allowed an ongoing rather than ‘one- a mechanism whereby respondents could relate to the presence
off’ relationship with the respondent to be established. While and reputation of the organization from past and extensive
repeat visits are a methodological ‘ideal’ in terms of resource and research undertaken in the local area. Furthermore, local farms
time commitments they can also help reduce the uncertainties were made aware that this institution is independent of govern-
associated with the acquisition of farm management data and ment policing agencies linked to land and water quality. These
unlock the depth of information central to a farmer’s day-to-day arguments can work in reverse too; if reputations are not held in
routine (on a limited number of the farm sample). For example, high regard, or seen as threatening, interviewers may be impeded
in the Taw catchment case study, a participating dairy farmer in their activities (Fairweather, 2008).
claimed in the original survey return that he had purchased enough However, long-term relationships cannot always be developed
land (for increased slurry spreading while avoiding overloading as part of a research project because of a mismatch in funding cycle
pasture with faecal material and associated contaminants) to allow duration. This issue is not confined to the UK; Nancarrow (2005)
for an increase in his dairy herd, and that when he spread slurry to highlights this as a major problem in Australia because of the
land he ensured he left a 10 m buffer between the watercourse and time required to develop a degree of trust and familiarity with the
the applied slurry. Further visits to the farm revealed new intents [farming] community and the race against the three year funding
and attitudinal positions. At a later date the same respondent cycle. Long-term research platforms or ‘catchment observatories’,
quipped once, ‘I get a thrill when I hear muck splash in the water’ and with farming communities forming an integral part of the research
suggested that ‘I always spread slurry onto the 3 or 4 most convenient network, offer attractive benefits to modelling studies since they
fields next to the slurry lagoon’. Both of these comments contradict can provide longitudinal data rooted in wider farmer insights and
the responses given within the formal questionnaire survey. It is ownership.
a debating point in this case regarding which information was more
true or accurate, but certainly our reasoning is that, as formalities 4.2. DSS conceptualisation (embedding expert dialogue into
are dropped and confidences built, deeper insights into system conceptualisation of DSS; Fig. 1 stage 2)
characteristics and behaviour are brought to bear upon the problem
at hand. This may relate to an initial lack of confidence on the McCown (2002) suggested that a critical problem with
farmers behalf in terms of how survey responses may be inter- ‘management science models’ is that managers practically never
preted, and may possibly have been linked to concern about the use them. Others have stressed the overarching need to ask ‘why
regulatory implications of supplying particular types of manage- are we building a model or DSS?’ (Nancarrow, 2005). If tools are to
ment information to ‘researchers’. In this sense, it proved especially be used directly by the farming community or in cooperation with
important for the project team to stress the funding mechanism of the regulatory community then there is an imperative for clear
the project (e.g. non-government, Research Council) and to clarify communication between science providers and science users to
the disassociation between the project and regulatory authorities. maximise utility of products to both parties. Active engagement in
In addition, face to face interviews in practical settings provided early stages is important to bridge rather than reinforce the
opportunities to ‘ground-truth’ claims made in a survey. For science-stakeholder divide and accommodate knowledge or
example a participating beef farmer in the Taw Study informed the understand the concerns of those who need to use emerging tools
project team of his clean and dirty water separation system within (Quinn et al., 2010; Barreteau et al., 2010). By embracing farmers as
the farm hard-standings that allowed clean water to be channelled experts and encapsulating their detailed local knowledge into DSS
to a clean water drain which in turn diverted the clean water to the the resulting products are more reflective of the farm system and
stream. When the interview team finished discussions with the perhaps more accessible or appealing to the farming community
farmer and left the farm office a farm worker was witnessed hosing because they are seen to support the decision making process rather
down a tractor that was covered in fresh farmyard manure directly than attempt to make the decision.
above the clean water drain that diverted to the stream. This latter Whatever the purpose of a DSS there are considerable benefits
example serves to demonstrate the value of on-farm survey to be gained by simply entering into a dialogue with end-users at
approaches because of the different levels (and mismatches) of early stages (preferably inception) of DSS conceptualisation and
information that can be obtained from first-hand experiences. then throughout the design process. This will ensure that their
Besides, such information is difficult to detect as a pollution signal knowledge and input is included within the design process but
in water quality because of the activity’s incidental nature. This ultimately it also facilitates social learning (van Delden et al., 2011).
example is of course coincidental with the timing of the research Through the formation of trusting relationships with the farming
team’s visit but the key point to make is that by establishing community such dialogue should take place organically and form
D.M. Oliver et al. / Environmental Modelling & Software 36 (2012) 76e85 81

the second iterative step in our seven-stage sequential protocol. In respondents; (iv) questionnaire survey conduct; and (v) question-
the Taw catchment case study there were ample occurrences naire survey analysis. It is not the intention of this study to offer
whereby qualitative information was provided through informal a roadmap of questionnaire survey design as there is extensive
conversations during routine water quality monitoring that took information available detailing such procedures (e.g. Oppenheim,
place on farms. These points were expanded upon during the more 2000). From the experiences of the Taw catchment study
formal questionnaire survey and allowed for open and purposeful whereby the land, manure and livestock management practise of
discussion of the DSS requirements linked to microbial pollution of 77 farmers were recorded across a 1200 km2 area, there were clear
watercourses. For example, the importance of ‘within field hot- advantages in using particular approaches over others. Face-to-face
spots’ of pollution sources was raised by many of the farmers meetings allowed for a first-hand visualisation of the farm
during early discussions of the practicality of on-farm monitoring management regimes providing critical secondary data to
strategies. This in turn contributed to the development of both complement the dialogue between farmers and researchers. Closed
a field- and farm-scale DSS to allow farmers to identify not only the questions were used to obtain key numerical data linked to herd
overarching drivers of risk within their farm systems but also finer size and tanker volumes, but more open-ended discussion was
resolution field-by-field vulnerability to microbial loss from land to promoted around the use of visual maps annotated by the farmers
water. This latter point was critical in that the resulting DSS and through open-ended questions towards the latter stages of the
provided farmers with a method to prioritise land for a shift in survey. The face to face approach provided a context in which
management or introduction of a mitigation option. different information elicitation techniques could be deployed in
conjunction with the structured survey. For example, in our study
4.3. Farmer survey & expert elicitation (Fig. 1 stage 3) we employed map-based questions as a way of collecting and
visualizing information about the farm landscape (Winter et al.,
There are a number of key criteria to evaluate before under- 2011). Fig. 3 shows an example of data recorded from farmers
taking a questionnaire survey or elicitation procedure with using these map-based approaches to elicit knowledge of drainage,
respondents. Careful consideration must be given to: (i) question- land use, soil properties, grazing regimes and manure applications
naire survey design; (ii) sample population; (iii) recruiting associated with their property. In principle, map orientated

Fig. 3. Map of dairy farm annotated by farmer to show grazing patterns, manure applications, drainage patterns and no-spread zones.
82 D.M. Oliver et al. / Environmental Modelling & Software 36 (2012) 76e85

discussions allow the linkage of management approaches and Otherwise such errors may compromise the conceptualisation and
expert knowledge with spatial coverage of the farm area. This execution of a DSS reliant on such input data. It is not difficult to
interactive graphical component in the research process provided envisage a researcher being unaware of the limitations of survey
a detailed spatial representation of the farmscape adding subtlety returns particularly if s/he was not responsible for the original
and depth to the results of the central questionnaire survey survey design, and simply used the template to ‘extract’ informa-
methodology. Our experience was that this mapping exercise also tion from a respondent. This point strengthens the argument made
served to make the farmer feel more comfortable with the in Section 4.1 with regard to familiarisation and an evolving rela-
discussion rather than being “interrogated” over management tionship between researcher and farmer in order to reap most
approaches via a rigid questionnaire. In the Taw study farmers reward. Uncertainties, limitations and sources of error can easily
populated the map using coloured pencils and pens, though it is become disguised, forgotten and embedded within DSS frame-
possible to envisage more technologically advanced approaches, works without necessary caveats being associated with the source
such as participatory GIS (Macleod et al., 2007). The questionnaire data. The risk then is that erroneous data and anecdotal evidence
surveys were trialled within the research team before being become inherent to the functioning of a DSS (Davern et al., 2008).
employed on-farm. A hard-copy survey was preferred over an Sources of uncertainty can include, for example, aspects of
electronic version on a laptop simply because of the awkwardness missing data, potential for contradiction and a lack of confidence in
in loading the computer operating system and the delay in being the reliability of some responses. Certainly, large structured ques-
able to get started in conversation because of powering-up issues. tionnaires will often comprise multiple sections meaning that
It is important to reiterate a number of critical assumptions that answers to questions in earlier sections may be related to responses
are inherent to the conduct of questionnaire surveys. For example, given in later aspects of the survey return. Technology can play an
a questionnaire survey tends to proceed from the assumption that important role for improving consistency in the logic of answers:
interviewees will: electronic surveys can be programmed to prevent contradictions
being made through the embedding of logic rules into the frame-
 understand what an interviewer is asking of them (i.e. question work that restrict text entry into response boxes if certain answers
comprehension) were given previously. Innovation in web-based surveys (e.g.
 speak when they are spoken to (i.e. willingness to answer) Fleming and Bowden, 2009) can use visual stimuli to alert the
 say what they mean when they speak (i.e. capacity to respondent to missing data or contradictory responses. However,
articulate) technology (e.g. computers) can also represent an awkward 3rd
party in questionnaire situations so the Taw catchment survey
Equally the questionnaire survey format also assumes that work was better conducted using a hard-copy paper survey and
interviewers will: digitised at a later date (though this can allow for potential errors in
data transfer and so double data entry would therefore be good
 understand what the interviewee is saying to them; (i.e. practice). In the case of paper-based surveys it is important to
answer comprehension) understand how best to deal with contradictions raised in ques-
 respond appropriately to points of clarification (i.e. capacity to tionnaires. There is a realisation that we need to develop meth-
articulate) odologies that can incorporate rather than deny contradictions, but
 conduct the survey in circumstances where completion is contradictions raise problems of coherent statistical analysis. This
guaranteed. would suggest an urgency to self-validate questionnaires and
devise strategies to cross-compare data sources (c.f. Voinov and
On this basis it is argued that a survey is amenable to ‘stand- Gaddis, 2008) from a wider programme of farmer engagement
ardised’ knowledge leading to consistent, comparable results from (e.g. steps 1e3).
which we can begin to generalise. However, it is not surprising to
also identify that with such a range of assumptions the credibility of 4.5. Qualitative validation (Fig. 1 stage 5)
the source of data can often be put in jeopardy. It is highly plausible
that erroneous responses will arise (through misunderstanding or We advocate a drive for increased use of farmers as experts not
through misdirection) and that the reliability of data may be just for informing model input and output, but also for qualitative
dubious through identification of contradictions in survey returns. validation of agricultural DSS prediction (thereby moving toward
In some cases answers may well appear to lack credibility because a more integrated and collaborative approach). This critical step in
responses appear to be improbable. In the context of all of the the iterative cycle of Fig. 1 is fundamental as it provides a ground-
assumptions above, questionnaire surveys generally take place in level ‘sanity check’ from those who have everyday expertise in their
a less than perfect way. All of these problems can be mitigated to land. In addition, continued involvement of farmers in the DSS
some extent in advance, such as: trialling questions; training validation should in turn help to promote subsequent uptake of the
interviewers; anticipating where reluctance to engage, and DSS by the farmer community. Validation is therefore not only
potential confusions, may exist; and being explicit about practical about comparing observed and predicted outcomes from DSS but
questionnaire survey needs, such as length of time needed. also about evaluating DSS output against expert (farmer in this
However, neither structured nor semi-structured interviews can by case) opinion (Le Gal et al., 2010) and allows for refinement of the
themselves provide a complete insight into farm operations but by science by local stakeholders (Raymond et al., 2010). Oliver et al.
embedding this methodology within an iterative programme as (2010) proposed farmer engagement for such qualitative valida-
depicted in Fig. 1 we can begin to cross-interview with a diverse tion of a risk indexing tool which currently forms a first approxi-
range of alternative information. mation for the protection of watercourses from microbial pollution
brought about by farming activity. Qualitative validation removes
4.4. Expert driven DSS output (Fig. 1 stage 4) a somewhat biased view that DSS will always result in sensible
output when in reality the peculiarity of heterogeneities in the
The research community must give careful consideration to the landscape may result in differing trends. This concept is outlined by
range of potential errors and misinterpretations of results that may Landry et al. (1996) who suggest that any model that is not legiti-
arise via questionnaire survey return and wider qualitative data. mate in the eyes of those for whom the DSS (or model) has been
D.M. Oliver et al. / Environmental Modelling & Software 36 (2012) 76e85 83

developed will not have much opportunity for acceptance and of the DSS. This effectively allowed the farming community to seek
wide-spread uptake by that particular community. Legitimacy in clarification as to why certain information may not have been acted
DSS is therefore highly desirable because it reinforces the value of upon, consider the impact of their knowledge that had been
that particular tool and also receives a seal of approval from end- incorporated into the DSS and also provided an opportunity for
users. farmers to cross-examine other spatial risk maps derived for
A recent study combining farmer and scientist expertise in differing farm systems and question particular influences on the
discussions of sustainable agriculture drew attention to initial DSS output in an informal environment. It is clear that research
conflicts in opinion over the value of scientific knowledge versus processes cannot involve all necessary decision makers in the co-
farmer knowledge. However, conceptions of scientific knowledge construction of environmental management tools. Community
being ‘superior’ and ‘more valid’ than farmer’s local experience qualitative validation therefore provided an additional mechanism
were abandoned (Eshuis and Stuiver, 2005). Farmer legitimation of to capture feedback from those who were not involved directly in
a DSS is important in terms of providing a sense of ownership of the the research but who would use the results and outputs of this
process but does of course mean that the resulting DSS must be research process. This phase allowed for the expression of agree-
transparent and straightforward in operation (Voinov and ments and disagreements with the proposed DSS to facilitate
Bousquet, 2010; Voinov and Brown Gaddis, 2008). Obviously collective learning and DSS refinement (Daniell et al., 2010). The
quantitative validation to complement qualitative validation is community qualitative validation phase provided a novel and
needed, though as noted by Oliver et al. (2010) on-farm verification worthwhile extension to the project and led to clarifications of
of field and farm-scale tools is often difficult to achieve due to terminology used in guidance documents. Other methods for
inherent uncertainties associated with pollutant sources and extending the involvement to additional excluded decision-makers
pathways within farm systems. Indeed, Merot and Bergez (2010) have been documented in Hare et al. (2003). The management of
suggest that it is very difficult to validate decision support tools participatory validation needs to be undertaken with as much care
because farmers tend not to record their day-to-day activities, and and attention as attributed to the environmental monitoring and
certainly not on a field-by-field basis. modelling procedures if the research tool is to benefit from useful
The Taw case study project produced a briefing document for information from the correct set of actors and for a common
each of the 10 farmers who participated in the environmental agreement to be reached among all involved. This latter point is
monitoring phase of the project. This document was designed to unrealistic of all stakeholders involved in contentious environ-
highlight the key research findings linked to each farm and provide mental issues but for the purposes of the Taw case study the DSS
a spatial risk map for the farmscape. This provided a further was considered legitimate by the farming community involved.
opportunity for feedback and comment as the documents were
delivered in person and immediate comments noted. All farmers 4.7. Graphic user interface (Fig. 1 stage 7)
were encouraged to get in touch should further queries or ques-
tions arise from the DSS output that had been provided to them. The apathy of farmers and landowners towards DSS may be
Key feedback from the farming community tended to focus on their linked to their (lack of) packaging for end-user uptake. A clear and
desire to be able to generate the DSS output themselves (or via well constructed Graphical User Interface (GUI) that allows end-
a farm advisor) and therefore comments were made regarding the users without specific modelling skills or knowledge of the
need to refine these DSS into ‘user-friendly’ operating systems (see modelling system to take advantage of the DSS is likely to yield
Section 4.7). Suggestions of the provision of ‘look-up-tables’ to higher interest and usage among the farming community than
accompany the DSS were made because some farmers commented a complex DSS that is essentially inaccessible to user groups. This
that if a farm advisor was not in attendance then some terminology has been effectively demonstrated for data management interfaces
of risk factors would require clarification. The nature of these in the drinking water treatment sector (Worm et al., 2010) who
particular DSS (see Oliver et al., 2009a, 2010) were regarded as noted that the acceptance and appreciation of a DSS is increasingly
having strong visual impact and suggestions were made regarding determined by the ‘look and feel’ of the system. The Taw study was
the coupling of these tools to DSS that consider other pollutants not able to develop a GUI in the project timeframe but additional
also. Finally, one farmer queried the use of arbitrary risk scores and funds are being targeted specifically for this purpose. This is seen as
suggested that accompanying textual information may be appro- a critical next stage to ensure that the DSS is actually used by the
priate for linking to each score so that the significance of the results user-community rather than shelved as an inaccessible product of
were made more clear in the absence of a farm advisor. academic research deemed ‘unfit for purpose’. It is proposed that an
enhanced user interface will allow extension officers to use the DSS
4.6. Community qualitative validation (Fig. 1 stage 6) to demonstrate potential implications of land management
changes for a range of pollutants and scenarios.
Briefly, as with all models and DSS, their transferability to
different geographical areas is essential for wider, national scale, 5. Conclusions
uptake. The discussion of Section 4.5 must be applied beyond
individual farm gates and extended to communities of farmers and Valuing farmers’ knowledge reflects a move towards more
landowners to determine the geographical constraints of DSS. integrated approaches to land and water management: with
Qualitative validation from researcher, advisor and farmer points of natural scientists, sociologists, economists, farmers and regulators
view following simulations of prospective scenarios in the agri- all contributing knowledge that can assist in resolving
cultural environment all provide valuable data with which to agriculturally-derived issues (Hewett et al., 2010) Farmer knowl-
constrain the DSS capability (Vayssieres et al., 2009). In the case of edge is a vital form of expert data and can be acquired via a suite of
the Taw study, the project team hosted an open-day at North Wyke formal and informal stages throughout the development of
Research and invited the ten core farmers to attend along with a working relationship between researchers and the farming
other catchment stakeholders. This was an opportunity for the community. The analysis presented in this paper complements
farming community to view outputs of the project via poster previous reviews of DSS co-construction but provides a UK case
presentations and provide feedback and comment on the data thus study analysis of DSS development through a seven-step iterative
far which helped constrain features important to the development process.
84 D.M. Oliver et al. / Environmental Modelling & Software 36 (2012) 76e85

From the perspective of natural science one of the strongest Davern, M.J., Mantena, R., Stohr, E.A., 2008. Diagnosing decision quality. Decision
Support Systems 45, 123e139.
benefits of this process was the ability to improve DSS accuracy
Eshuis, J., Stuiver, M., 2005. Learning in context through conflict and alignment:
through the establishment of a trusting relationship with the farmers and scientists in search of sustainable agriculture. Agriculture and
farming community. This was demonstrated best with the subset of Human Values 22, 137e148.
the ten farms and their involvement in the DSS evolution. Without Fairweather, J.R., 2008. The number of farms and farmers in New Zealand: a plea for
researcher access to farm samples and the inclusion of farmer data. New Zea-
the building of trust there are clearly opportunities for errors and land Journal of Agricultural Research 51, 485e488.
uncertainties to be propagated through the elicitation procedure Fish, R., Seymour, S., Watkins, C., 2003. Conserving English landscapes: land
and into DSS output. As outlined in Section 4.7, it is hoped that managers and agri-environmental policy. Environment and Planning A 35,
19e41.
further uptake and use of these DSS can be promoted once proto- Fish, R., Winter, M., Oliver, D.M., Chadwick, D., Selfa, T., Heathwaite, A.L.,
typical versions are refined to allow for greater accessibility by the Hodgson, C.J., 2009. Unruly pathogens: eliciting values for environmental risk in
farming community. Farmer feedback (as outlined in Section 4.5) the context of heterogeneous expert knowledge. Environmental Science and
Policy 12, 281e296.
has been encouraging and suggests that such approaches have, in Fiorino, D.J., 1990. Citizen participation and environmental risk - a survey of insti-
part, been welcomed as qualitative risk assessment tools. tutional mechanisms. Science Technology & Human Values 15, 226e243.
The cyclical nature of the process is also important to empha- Fleming, C.M., Bowden, M., 2009. Web-based surveys as an alternative to traditional
mail methods. Journal of Environmental Management 90, 284e292.
sise. Although additional iterations are likely to incur further costs Glicken, J., 2000. Getting stakeholder participation ‘right’: a discussion of partici-
and take place beyond the life-cycle of the initial funding phase it is patory processes and potential pitfalls. Environmental Science & Policy 3,
important that DSS’s evolve according to emerging research needs 305e310.
Hare, M., Letcher, R.A., Jakeman, A.J., 2003. Participatory modelling in natural
and evidence. This also reinforces a point made by van Delden et al.
resource management: a comparison of four case studies. Integrated Assess-
(2011): that iterations allow for adaptations to the prototype DSS ment 4, 62e72.
once they have been developed and a better understanding of their Hewett, C.J.M., Doyle, A., Quinn, P.F., 2010. Towards a hydroinformatics framework
possibilities and limitations can then be obtained. to aid decision-making for catchment management. Journal of Hydro-
informatics 12, 119e139.
Such adaptations depend on cultivating the long-term buy-in of Irwin, A., 1995. Citizen Science: A Study of People, Expertise and Sustainable
experts in the farming community. Farmer relationships are the Development. Routledge, London.
‘social infrastructure’ behind DSS evolution and these take time to Jakeman, A.J., Letcher, R.A., 2003. Integrated assessment and modelling: features,
principles and examples for catchment management. Environmental Modelling
nurture. They should therefore be maintained where possible, such & Software 18, 491e501.
as via longer-term catchment initiatives. Thus rather than finish Landry, M., Banville, C., Oral, M., 1996. Model legitimisation in operational research.
abruptly, engagement processes should ideally inform the European Journal of Operational Research 92, 443e457.
Le Gal, P.Y., Merot, A., Moulin, C.H., Navarrete, M., Wery, J., 2010. A modelling
continuing refinement of the DSS. framework to support farmers in designing agricultural production systems.
There is little doubt that in recent years we have seen a shift in Environmental Modelling and Software 25, 258e268.
farmer-scientist relationships from one of knowledge transfer to Lytton, L., Howe, S., Sage, R., Greenaway, P., 2003. Groundwater abstraction pollu-
tion risk assessment. Water Science & Technology 47, 1e7.
one of knowledge exchange, learning and two-way communication Macleod, C.J.A., Scholefield, D., Haygarth, P.M., 2007. Integration for sustainable
of information and advice. This is a positive development for catchment management. Science of the Total Environment 373, 591e602.
interdisciplinary researchers with ambitions to build credible DSS Macleod, C.J.A., Haygarth, P.M., 2010. Integrating water and agricultural manage-
ment under climate change. Science of the Total Environment 408, 5619e5622.
and needs to be embraced. Our conclusion is that programmes of
McCown, R., 2002. Changing systems for supporting farmers’ decisions: problems,
participatory modelling need to provide sufficient opportunity paradigms, and prospects. Agricultural Systems 74, 179e220.
throughout and beyond the research process to deal with and Merot, A., Bergez, J.E., 2010. IRRIGATE: a dynamic integrated model combining
accommodate local expert knowledge within new and emerging a knowledge-based model and mechanistic biophysical models for border
irrigation management. Environmental Modelling and Software 25, 421e432.
DSS. Morris, C., Potter, C., 1995. Recruiting the new conservationists: the adoption of UK
agrienvironmental schemes. Journal of Rural Studies 11, 51e63.
Nancarrow, B.E., 2005. When the Modeller Meets the Social Scientist or Vice-versa.
Acknowledgements www.mssanz.org/modsim05/papers/nancarrow.pdf (accessed 04.05.11).
Nettle, R., Paine, M., Penry, J., 2010. Aligning farm decision making and genetic
The authors would like to acknowledge the 77 participating Taw information systems to improve animal production: methodology and findings
from the Australian dairy industry. Animal Production Science 50, 429e434.
Catchment Farmers for their valuable and insightful contributions Oliver, D.M., Fish, R.D., Hodgson, C.J., Heathwaite, A.L., Chadwick, D.R., Winter, M.,
to the wider findings of our research project RES-224-25-0086 2009a. A cross-disciplinary toolkit to assess the risk of faecal indicator loss from
which was funded by the Rural Economy and Land Use (RELU) grassland farm systems to surface waters. Agriculture, Ecosystems and Envi-
ronment 129, 401e412.
programme who receive financial support from the Biotechnology
Oliver, D.M., Heathwaite, A.L., Fish, R.D., Chadwick, D.R., Hodgson, C.J., Winter, M.,
and Biological Sciences Research Council (BBSRC), the Natural Butler, A.J., 2009b. Scale appropriate modelling of diffuse microbial pollution
Environment Research Council (NERC), the Economic and Social from agriculture. Progress in Physical Geography 33, 358e377.
Oliver, D.M., Page, T., Hodgson, C.J., Heathwaite, A.L., Chadwick, D.R., Fish, R.D.,
Research Council (ESRC), the Department for Environment, Food
Winter, M., 2010. Development and testing of a risk indexing framework to
and Rural Affairs (Defra) and the Scottish Government. Additional determine field scale critical source areas of faecal bacteria on grassland.
thanks are extended to Donald Barr for design of the Microsoft Environmental Modelling and Software 25, 503e512.
Excel questionnaire template and data extractor. Finally, we thank Oppenheim, A.N., 2000. Questionnaire Design. Continuum, London.
Pahl-Wostl, C., Jeffrey, P., Issendahl, N., Brugnach, M., 2011. Maturing the new water
the constructive comments of the three reviewers and the editorial management paradigm: progressing from aspiration to practice. Water
board. Resource Management 25, 837e856.
Pahl-Wostl, C., Mostert, E., Tabara, D., 2010. The growing importance of social
learning in water resources management and sustainability science. Ecology &
References Society 13, 24 (online).
Potter, C., Gasson, R., 1988. Conservation through land diversion: results from
Barreteau, O., Bots, P.W.G., Daniell, K.A., 2010. A framework for clarifying ‘partici- a survey. Journal of Agricultural Economics 39, 340e351.
pation’ in participatory research to prevent its rejection for the wrong reasons. Quinn, P.F., Hewett, C.J.M., Muste, M., Popescu, I., 2010. Towards new types of water-
Ecology & Society 15 (online). centric collaboration. Proceedings of the Institution of Civil Engineers e Water
Chilvers, J., 2009. Deliberative and participatory approaches in environmental Management 163, 39e51.
geography. In: Castree, N., Demeritt, D., Liverman, D., Rhoads, B. (Eds.), Raymond, C.M., Fazey, I., Reed, M.S., Stringer, L.C., Robinson, G.M., Evely, A.C., 2010.
A Companion to Environmental Geography. Blackwell, Oxford, pp. 400e417. Integrating local and scientific knowledge for environmental management.
Daniell, K.A., Mazri, C., Tsoukiàs, A., 2010. Real world decision-aiding: a case of Journal of Environmental Management 91, 1766e1777.
participatory water management. In: Rios Insua, D., French, S. (Eds.), e- Selfa, T., Fish, R., Winter, M., 2010. Farming livelihoods and landscapes: tensions in
Democracy: A Group Decision and Negotiation Perspective. Springer, Dordrecht, rural development and environmental regulation. Landscape Research 35,
pp. 125e150. 595e612.
D.M. Oliver et al. / Environmental Modelling & Software 36 (2012) 76e85 85

Sterk, B., Leeuwis, C., van Ittersum, M.K., 2009. Land use models in complex societal Voinov, A., Bousquet, F., 2010. Modelling with stakeholders. Environmental
problem solving: plug and play or networking? Environmental Modelling and Modelling and Software 25, 1268e1281.
Software 24, 165e172. Voinov, A., Brown Gaddis, E.J., 2008. Lessons for successful participatory watershed
Ticehurst, J.L., Curtis, A., Merritt, W.S., 2011. Using Bayesian networks to comple- modelling: a perspective from modelling practitioners. Ecological Modelling
ment conventional analyses to explore landowner management of native 219, 197e207.
vegetation. Environmental Modelling and Software 26, 52e65. Winsten, J.R., Kerchner, C.D., Richardson, A., Lichau, A., Hyman, J.M., 2010. Trends in
Van Delden, H., Seppelt, R., White, R., Jakeman, A.J., 2011. A methodology for the the Northeast dairy industry: large-scale modern confinement feeding and
design and development of integrated models for policy support. Environ- management-intensive grazing. Journal of Dairy Science 93, 1759e1769.
mental Modelling and Software, 266e279. Winter, M., Oliver, D.M., Fish, R., Heathwaite, A.L., Chadwick, D., Hodgson, C., 2011.
Vayssieres, J., Guerrin, F., Paillat, J.M., Lecomte, P., 2009. GAMEDE: a global activity Catchments, subcatchments and private spaces: scale and process in managing
model for evaluating the sustainability of dairy enterprises part I e whole-farm microbial pollution from source to sea. Environmental Science & Policy 14, 315e326.
dynamic model. Agricultural Systems 101, 128e138. Worm, G.I.M., van der Helm, A.W.C., Lapikas, T., van schagen, K.M., Rietveld, L.C.,
Vignola, R., Koellner, T., Scholz, R.W., McDaniels, T.L., 2010. Decision making by 2010. Integration of models, data management interfaces and training support
farmers regarding ecosystem services: factors affecting soil conservation efforts in a drinking water treatment plant simulator. Environmental Modelling and
in Costa Rica. Land Use Policy 27, 1132e1142. Software 25, 677e683.

You might also like