Exploring farmers’ support needs for the adoption of agri-environmental practices: an application of the theoretical domains framework

Shifting agriculture onto a more sustainable trajectory requires significant changes in farmer behaviour in terms of implementing agri-environmental practices. Understanding the underlying factors influencing farmers’ behaviour can provide guidance when it comes to targeting agricultural policies and ensuring that they are effective. This study builds on the 14 domains of the Theoretical Domains Framework (TDF) to explore farmers’ support needs for adopting agri-environmental practices. TDF ratings from 29 interviews with German farmers reveal the importance of the TDF domains for three agri-environmental practices, namely agroforestry, biological pest control and controlled release fertilization. Farmers state support needs for all TDF domains, but the ratings reveal significant differences in support needs across the TDF domains as well as among the practices. Higher overall support needs for agroforestry and biological pest control compared to those for sustainable fertilization practices may reflect the additional challenges associated with more systemic shifts in farming practices. Applying the TDF to the agricultural context can be helpful for developing targeted and theoretically informed policy interventions.


Introduction
The role of agriculture in environmental outcomes is widely recognized (Campbell et al 2017;Pe'er et al 2020).Agricultural policy in Europe and beyond faces the challenge of establishing the institutional and regulatory conditions needed to shift the sector towards a more sustainable trajectory (Rockström et al 2017, Biffi et al 2021).The question of how to change farmers' behaviour in terms of increasing their willingness to adopt agrienvironmental practices is therefore paramount (Pe'er et al 2017, Chan et al 2020).Policy making aimed at changing farmers' behaviour can benefit from drawing on the theoretical and empirical knowledge of the behavioural sciences (Banerjee et al 2021, Colen et al 2016, Delaroche 2020, Dessart et al 2019).Indeed, an increasing body of applied research is investigating the drivers of farmers' behaviour as a basis for designing agrienvironmental policies (Lastra-Bravo et al 2015, Bartkowski andBartke 2018, Brown et al 2021).
There are many theories of behaviour, emphasizing different factors that influence individual decision making (Klöckner 2013, Michie et al 2011).So far, few studies on farmer behaviour incorporated behavioural theory in any systematic way (Bartkowski and Bartke 2018).Those that did, usually relied on single theories that emphasize a subset of relevant behavioural factors (Delaroche 2020).Dessart et al (2019) provide a review of factors affecting farmers' behaviour, in which they also systematized the identified factors in a framework.
However, their review focused on internal socio-psychological factors.The present study builds on the Theoretical Domains Framework (TDF) to systematically explore the barriers and enabling factors affecting the adoption of agri-environmental practices with a broader view on 'behavioural factors'.The TDF integrates constructs from diverse theories to facilitate the design of behavioural interventions for the purpose of practical implementation (Cane et al 2012, Michie et al 2005).Thereby the framework includes all factors that can enable behaviour change, including external and contextual factors.It is closely related to the Behaviour Change Wheel (BCW) framework, which has been developed by the same team of implementation-oriented psychology scholars around Susan Michie to link broader behavioural factors (capability, opportunity, motivation) to targeted interventions and policies.Whereas the BCW and the TDF have been extensively applied in the healthcare context (Spillane et al 2021), the BCW has so far rarely been used in sustainability research (Axon et al 2018, Gainforth et al 2016, Hedin et al 2019, Marselle et al 2021, Tensi and Ang 2022, van Asseldonk et al 2018) and, to our knowledge, the TDF has not been applied in sustainability or agriculture research at all.We operationalized the TDF domains by adapting the terminology of its constructs to reflect the specificities of the farming context.
We conducted interviews with farmers from the German region of North-Rhine Westphalia, in which.the farmers were asked to consider three particular agri-environmental practices: agroforestry, biological pest control4 , and the use of controlled-release fertilizer.For each of the three practices, the farmers rated the extent to which each of the TDF domains was already given (current status) and the extent to which they would require further support that would enable them to adopt the practice (support need).The data on support need is the most relevant information to guide policy design and implementation.The elicitation of current status data can help understand the general importance of a TDF.If the stated support need for a specific TDF is low, then a low current status indicates that this factor may generally be of low relevance for adoption, while a high current status indicates that the factor may indeed be relevant, but is already fulfilled.
Given the complexity of factors impacting farmers' behaviour, using the TDF framework may help in rendering visible and structuring the full spectrum of behavioural factors (Biesheuvel et al 2021).Due to the limited sample size, our study can only provide first indications on the relevance of the different TDFs for each of the three agri-environmental practices.Its main contribution is to pilot the applicability of the TDF to the agrienvironmental context and to highlight how such an analysis allows to derive policy implications.The discussion section explains how our analysis can serve as knowledge base for policy-oriented follow-up-studies and the conclusions section points to promising avenues for further research.

Theoretical domains framework (TDF)
The TDF is an implementation-oriented framework that integrates 84 theoretical constructs from 33 behavioural theories and proposes fourteen 'theoretical domains' as determinants of behaviour (Cane et al 2012, Michie et al 2014, Atkins et al 2017).The domains of the TDF cover the whole range of 'cognitive, affective, social and environmental influences on behaviour' (Atkins et al 2017).The TDF domains were empirically refined and validated in a three-step validation process using Discriminant Content Validation and Fuzzy Cluster Analysis (Cane et al 2012).Michie et al (2014) relate the TDF domains to three behavioural components that constitute the inner part of the Behaviour Change Wheel (BCW): Capability, Opportunity and Motivation together can enable or hinder a Behaviour (COM-B).Capability relates to peoples' underlying psychological and physical capabilities, opportunity relates to their social and physical environment, and motivation relates to both their underlying automatic and reflective processes.These factors interact over time and it is acknowledged that behaviour is part of a dynamic system, although the COM-B does not explicitly deal with these interactions.
The TDF framework complements many other methodologies and frameworks proposed in the literature to systematically explore the barriers and enabling factors affecting the adoption of agri-environmental practices (e.g.Dessart et al 2019, Schlüter et al 2017, van Valkenvoed et al 2022).While a systematic comparison of these approaches is beyond the scope of this study, the cross-disciplinary potential to assess implementation problems, design interventions and/or understand change processes has been emphasized as a unique feature of the TDF framework (Atkins et al 2017).The BCW framework can be used to relate the TDF domains to corresponding intervention functions and policy categories.
Previous studies have used the TDF mainly to understand the barriers and enablers to health-related behaviours and to assess the effectiveness of healthcare interventions (e.g., Alexander et al 2014, Griffith et al 2021).For this study, we adapted the TDF domains to the farming context (see left-hand column of table 1) in order specifically to explore the influences on farmers' uptake of environmentally friendly agricultural practices.
The context-specific adaptation of the TDF and its corresponding items are closely based on the suggested generic questionnaire of Michie et al (2014) and on discussions with fellow researchers.The TDF domains Material Resources and Time Resources were adapted from the original TDF domain 'Environmental context and resources'.The TDF domain Memory and attention was originally called 'Memory, attention and decision process'.These context-specific definitions (see right-hand column of table 1) were the basis for the interview questions.

Selection of farmer sample and agri-environmental practices
The data for this study were elicited between December 2020 and February 2021 by means of interviews with farmers from the extended Münsterland region in North Rhine-Westphalia (Germany).The region is considered a centre of the livestock industry (Lammers and Becker 2014).The target sample size of the study (n = 20) was set in the initial research design phase using four key considerations (i.e. the breadth, depth, and nature of the current research topic, the heterogeneity of the farming population; the level of interpretation required; practical parameters such as the availability of and access to the respondents, time constraints, logistics Table 1.TDF domains and their specification for the context of agri-environmental practices etc) as proposed by Roller and Lavrakas (2015).Ultimately, 29 farmers were interviewed.Interview participants were recruited through e-mails, phone calls and mobile phone messages.The farmers were either personally known to the lead author or were recommended by an acquaintance.This relationship reduced response biases such as social desirability as well as non-responses, while at the same time increasing question-answer validity (Roller and Lavrakas 2015).While we acknowledge that this may introduce a selection bias, in the sense that participants who volunteered to take part in the study may be more motivated to adopt agri-environmental practices or differ in other relevant characteristics from the overall population, the approach ensured the availability of respondents and economizing on scarce time and money resources.
We used the following criteria to select the three agri-environmental practices: 1. Practice has positive impacts on the four major environmental dimensions of biodiversity, climate, soil health and water (Baaken 2022); this criterion aims to capture practices whose sustainability contributions are relatively unequivocal and which in that sense represent particularly potent leverage points.
2. Practice is not covered by German agricultural regulations; the logic behind this criterion being that the practices are likely to be particularly relevant for future policy interventions (especially in combination with criterion 1).
3. Practice is not widely applied, yet not too exceptional or specific; this criterion implies all farmers are a priori equally (un)likely to have implemented it, which builds on criteria 1 and 2, and that farmers should be able to imagine its implementation.
4. Each practice covers a different aspect of farming (or 'type of farming practices' as defined in Baaken 2022); this criterion ensures some heterogeneity which was expected to allow us to see the context-specificity of the TDFs' relevance.
Based on these criteria, the following three practices were chosen: agroforestry, biological pest control and controlled release fertilizer.Agroforestry is understood as a land-use system that involves woody perennials (i.e.trees, shrubs, hedges, etc) on the same land as agricultural crops and/or animals.Biological pest control is a form of pest control that uses living organisms such as herbivorous arthropods, parasitoids, pathogens or other predators to suppress, decrease and control pests.Controlled-release fertilizer is a granulated fertilizer which releases its nutrients gradually into the soil because of its special coating (Baaken 2022).

Interview procedure
The first three interviews were conducted face-to-face.Due to the outbreak of the COVID-19 pandemic the remaining interviews were conducted virtually, using the GoToMeeting software (n = 24) and by telephone (n = 2).The interviews lasted between 35 and 105 min.A general principle behind the interview procedure was to create a trusting relationship where all responses are honoured and 'candid revelations can thrive' because it is understood that the identity of the interviewee will remain confidential (Roller and Lavrakas 2015).We carefully followed the steps of the interview method proposed by Roller and Lavrakas (2015), which ranged from determining research questions to conducting mock interviews and obtaining consent forms from respondents.
After a general introduction and explanation of the purpose of the study, the interviews first consisted in semi-structured questions for a related qualitative study (Baaken and Vollan, in preparation).Subsequently, we elicited semi-quantitative data based on farmers' ratings of the relevance of particular TDF domains for the uptake of the three agri-environmental practices.For each practice, the respondents were asked to rate each of the TDF domains on a 6-point scale from zero to five with respect to both i) current status, that is, the extent to which the particular TDF domain for adopting specific agri-environmental practices is already met, and ii) support need, that is, the extent to which they would need support regarding this TDF domain in order to implement the practices.
The interviews ended with socio-demographic questions about the personal characteristics of the farmer (e.g. level of education, age) and his or her role on the farm (e.g.ownership), as well as questions about the farm (e.g.size of farm) (Supplementary Material 5: Interview questions in German and English).

Analysis
The main part of the analysis is based on the TDF domain ratings.The ratings on 'current status' and on 'support needs' were first analysed with standard descriptive statistics to depict the distribution of answers.The Kruskal-Wallis (KW) test was then used to check for each TDF domain and each of the two rated aspects (i.e., current status and support needs) whether the ratings among the three agri-environmental practices differed significantly from one another.Subsequently, we looked at the heterogeneity of the answers among the TDF domains and practices (Supplementary Material 3: Sample statistics).

Sample characteristics
Table 2 provides an overview of the sample statistics.Only farmers with a certain amount of decision-making power on the farm were included in the sample; most were farm owners and tenants, but also the other interviewees were either family members with decision-making power or employed farm managers.Respondents listed as employees were mostly family members of the farm owner.The majority of respondents were well educated, young, male, and worked on conventional farms.In comparison to the farmer census data the farmers interviewed are on average younger and more educated.The age and education patterns reflect the sampling approach, as respondents were personally known to the lead author or recruited through a snowballing process.A large number of the interviewees kept livestock and cultivated crops.The size of the farms ranged from 8 to 2000 hectares.

Descriptive analysis
None of the practices were already implemented by any of the farmers.The interview questions, corresponding TDF domains as well as the response means and standard deviations can be found in the Supplementary Material 1.
The spider diagram in figure 1(a) shows that some TDF domains are, to a large extent, already met at the farm level, including Physical skills (i.e.farmers have the physical skills to adopt the practices) and Beliefs about capabilities (i.e.farmers possess professional self-esteem and expectations of a positive outcome).TDF domains that are not yet met include Reinforcement and, to some extent, Social influences.
The mean ratings in figure 1(a) illustrate that for some TDF domains the current status ratings are relatively similar across the agri-environmental practices (e.g.Reinforcement, Physical skills).The responses differ significantly among the three practices for Knowledge (KW test, p < 0.01), Time resources (p < 0.01), Social influences (p < 0.01), Optimism (p = 0.01), Social/professional role and identity (p = 0.01) as well as Material resources (p = 0.03) (see also Supplementary Material 4).For instance, time as a resource is already available for controlled-release fertilization and not so much for the other two practices, which are more time-consuming.Whereas farmers know of peers who have already implemented the fertilizer and the pest control practices (hence high ratings for Social influences), they seem to have no peers who apply agroforestry.
Figure 1(b) depicts the farmers' support needs for each of the TDF domains and agri-environmental practices.Mean support needs were positive for all TDF domains.The stated support needs are generally rather low for Emotions, Social/professional role and identity, and Physical skills.For some TDF domains (e.g.Reinforcement, Emotions, Goals, Intentions) the stated support needs are similar among agri-environmental practices.The support needs differ significantly between the practices for Time resources (KW test, p < 0.01), Material resources (p < 0.01), Knowledge (p = 0.01), Memory and attention (p = 0.03) as well as Social/professional role and identity (p = 0.03).The spider diagram shows that the mean stated support needs for the use of controlled-release fertilizer tend to be generally lower across almost all TDF domains compared to the other two practices.
Overall, we observe a considerable heterogeneity of support need ratings.More than 80% of the 45 TDF distributions (i.e. three practices with 15 TDF domains each) cover the entire spectrum of the rating scale (ranging from zero to five), and in the remaining 20% only one extreme value is not represented (either zero or five, see also Supplementary Material 3).
Comparing the lines with the same colours in figures 1(a) and (b) allows to compare the relationship between current status and support needs for each practice.In order to make the visual comparison easier, figure 2 depicts the current status and support needs for the TDF domains together, separately for the agri-environmental practice agroforestry (a), no or biological pest control (b) and controlled release fertilizer (c). 5or agroforestry, there are seven TDF domains where the support need is high and the current status fairly low: Knowledge, Memory and attention, Material resources, Time resources, Social opportunity, Optimism as well as Reinforcement.This reflects gaps for which agricultural policy could consider adequate response options.For biological pest control, there are six TDF domains for which the current status is already fairly high but the support need is even higher.For two TDF domains the current status and support need are nearly identical (Beliefs about consequences, Intentions).The results are different for the use of controlled-release fertilizer.Here, the current status is high for all 14 TDF domains, and support needs are rather high only for Reinforcement, Optimism, and Social influences.Hence, compared to the other two practices, the behavioural pre-conditions for the adoption of controlled-release fertilizer are already given to a fairly large extent, yet still insufficient for uptake of the practice.In that sense, controlled-release fertilizer should be a relatively lowhanging fruit for supportive policy interventions.

Discussion
The TDF framework provides a structure for analysing the broad range of factors that may be relevant for the implementation of agri-environmental practices.As a general result in support of the TDF as a useful analytical framework, the ratings indicate that, for all three practices, none of the TDF domains are regarded as irrelevant, in the sense of being neither given nor needed.On the contrary, most TDF domains exhibit a high or moderate support need.The low level of support needs for Emotions may result from the fact that various farmers could not imagine what form such support might take.The TDF domain Emotions may also have been rather difficult for the respondents to grasp, as it requires considerable introspection to fully recognize one's own emotions (Lane and Smith 2021).The combination of high current status and low support needs of Physical skills indicates that farmers believe they have the necessary skill set to implement the practices.Finally, Social/professional role and identity is an aspect that farmers view as difficult to change, especially externally.Farmers commented that a change in identity is more up to them, so they do not wish to have any interference from outside.This does not necessarily mean, however, that no attempt can or should be made in policy interventions to change emotions or identities (see Riley 2016).The results indicate instead that questions put directly to farmers may not be the best way to identify support needs with respect to this TDF domain.Furthermore, farmers' comments regarding Social/professional role and identity suggest that if they were to be addressed by interventions, this should be done in a careful manner in order to ensure legitimacy and transparency.In the longer run, supporting these factors might even be most consequential for a sustainability transformation in the agriculture sector (Burton et al 2008, Westerink et al 2021).
To a large extent the findings presented here match and complement the findings from previous literature on behavioural barriers and enablers of environmentally friendly farming.For instance, Gütschow et al (2021) highlight the importance of increasing policy coherence, providing additional financial incentives (including, possibly, investment support) and reducing the bureaucratic burden in efforts to increase the uptake of agrienvironmental practices.The TDF analysis suggests that, in addition, training activities and a better demonstration of the benefits of agri-environmental practices could increase their uptake.This would address the stated knowledge needs for less common practices (such as, in our case, agroforestry) and leverage the role of peers as multipliers.Advisory services, whose potential has been demonstrated in the context of sustainable soil management by Mills et al (2020), could play a major role here.Another key point is to strengthen the existing networks of 'demonstration farms' as well as projects that build on the positive experiences of the German Future Resources, Agriculture & Nature Conservation (FRANZ) initiative (Reiter 2021).At the same time, the differences in responses across the three practices show that some support needs are practice-specific.The higher support needs for agroforestry and biological pest control compared to those for the more 'incremental' fertilization practice suggest that shifting to the former may be more challenging and demanding.Both of these practices are likely to involve a more systemic perspective and entail changes to multiple auxiliary practices.
The fact that the farmers' support needs differ significantly among the three agri-environmental practices warrants targeted, context-specific policies instead of a one-size-fits-all approach-payments alone may not be sufficient.The farmers' support needs for the TDF domains identified in this pilot study can complement previous research to support the choice and design of policy interventions aimed at enhancing the uptake of agrienvironmental practices.Prior to designing any intervention, it is crucial to understand the specific needs governing the uptake of an agri-environmental practice.The TDF can help structure such an investigation by providing a comprehensive perspective on these needs and enabling factors.For instance, for agroforestry the results section reported high support needs regarding the TDF domains Knowledge, Memory and attention, Material resources, Time resources, Social opportunity, Optimism, as well as Reinforcement.A follow-up analysis specifically on agroforestry practices could build on these findings and apply the aforementioned Behaviour Change Wheel (BCW) framework to identify suitable interventions and policies to target the respective TDF domains (Michie et al 2014, PHE 2020).
An additional insight from our application of the TDF in the agri-environmental context is that any analysis that applies the framework outside its original domain in health can likely benefit from a contextspecific adaptation of the TDF domains, based on pre-testing with the target group in pilot studies.For the present study, we stayed relatively close to the TDF domains as proposed in Michie et al (2014), merely splitting the resource domain into Time resources and Material resources.During the interviews we learned that certain domains could be further differentiated in order to capture important elements, such as the distinction between social influences deriving from the farm setting itself and those deriving from peers and from the wider society.

Conclusion and further research
In the study presented in this article, we used interviews with farmers to pilot the application of the Theoretical Domain Framework (TDF) as a means of identifying the factors that influence farmers' behaviour regarding the adoption of agri-enviromental practices.The results show that all TDF domains are important at the farm level.The behavioural dimensions in which farmers see a need for support if they are to adopt agri-environmental practices differ significantly across the three practices considered (agroforestry, biological pest control, controlled-release fertilization).Our study demonstrates the applicability of the TDF to the agricultural context, especially if the framework is adapted to the specific context prior to data collection.The findings can inform the quest towards selecting and designing theoretically grounded (policy) interventions to support the adoption of agri-environmental practices by farmers.
Our findings point towards five additional avenues for further research.First, they suggest the need for additional methods to assess support needs with respect to those TDF domains that were difficult to elicit with our rating exercise: Social/professional role and identity and Emotions.Second, the TDF domains could be further adapted to the agricultural context, possibly using participatory research methods.This might mean merely rephrasing their descriptions, but also splitting domains into useful sub-domains, or else removing certain domains that are less relevant.Third, it should be tested whether the 'support need' variable correlates with uptake of agri-environmental practices, for instance by testing whether targeted interventions that meet-and thereby decrease-the support needs actually lead to more uptake.Fourth, additional research could use data from farms that already implemented agri-environmental practices to investigate if there are certain threshold levels for the TDF domains that would be sufficient for the farmer to adopt a practice.This might mean that specific thresholds for each TDF domain exist or that the TDF domains can (partly) compensate one another.Fifth, it would be interesting to assess (e.g., using quantitative surveys) if farm and farm characteristics can explain the stated support need and current status as well as any differences between the two.

Figure 1 .
Figure 1.Respondents' mean ratings on (a) the current status of the three agri-environmental practices agroforestry, biological pest control and use of controlled-release fertilizer and (b) the support need for the three agri-environmental practices categorized according to the TDF domains.The ratings were made on a 6-point scale (ranging from zero to five).KW test: *** p < 0.01, ** p < 0.05, * p < 0.1.The colours of the TDF domains indicate the BCW behavioural components (green: motivation, blue: capability, orange: opportunity).Phys = Physical skills, Know = Knowledge, Mem = Memory and attention, Beh Reg = Behavioural regulation, Env (Material) = Material resources), Env (Time) = Time resources, Soc = Social influences, Opt = Optimism, Bel Cap = Beliefs about capabilities, Id = Social/professional role and identity, Bel Cons = Beliefs about consequences, Int = Intentions, Goals = Goals, Em = Emotions, Reinf = Reinforcement.A figure showing each agri-environmental practice individually can be found in the Supplementary Material 2: Support needs and current status for agri-environmental practices (ranked by mean).

Figure 2 .
Figure 2. Comparisons of the mean ratings of current status and support needs for all TDF domains, separately for the agrienvironmental practice agroforestry (a), no or biological pest control (b) and controlled release fertilizer (c).See figure 1 for an explanation of the acronyms.
Tensi A F., Ang F and van der Fels-Klerx H J 2022 Behavioural drivers and barriers for adopting microbial applications in arable farms : evidence from the Netherlands and Germany Technological Forecasting & Social Change 182 121825 van Asseldonk M A P M, Malaguti L, Breukers M L H and Van Der Fels-Klerx H J 2018 Understanding preferences for interventions to reduce microbiological contamination in Dutch vegetable production J. Food Prot.81 892-7 van Valkenvoed A M, Abrahamse W and Steg L 2022 To select effective interventions for pro-environmental behaviour change, we need to consider determinants of behaviour Nature Human Behaviour 6 1482-92 Westerink J, Pleijte M, Schrijver R, van Dam R, de Krom M and de Boer T 2021 Can a 'good farmer' be nature-inclusive?shifting cultural norms in farming in the netherlands Journal of Rural Studies 88 60-70

Table 2 .
Selected sample statistics and comparison to German farmer census data (where available).
hectares Size of the farm in hectares Mean (sd) : 169.9 (376.2) min < med < max: 8 < 100 < 2100 63 (DESTATIS, 2020)A full list of the variables can be found in the Supplementary Material 3: Sample statistics.