1 Introduction

Widespread conventional agricultural practices that developed in Western countries after the Second World War relied on the high use of synthetic inputs to ensure high production and product quality levels. Yet they can cause major environmental and human health issues such as biodiversity loss, water quality degradation, and pesticide residues in food and water (Tilman et al. 2002; Plumecocq et al. 2018). Consequently, there is now a pressing demand for a transition toward more sustainable agriculture. Agroecological practices are farming practices “aiming to produce significant amounts of food, which valorize in the best way ecological processes and ecosystem services in integrating them as fundamental elements in the development of the practices” instead of synthetic inputs (Wezel et al. 2014). They are currently not implemented on a large scale even though they are central to sustainable agriculture. They have the potential to combine environmental benefits with a good economical balance (e.g., biological pest control, direct seeding in cover crops). This high level of inertia of conventional agriculture can be explained by the technological transition theory (e.g., Geels 2002, 2004): a stable structuring of agriculture and food actors’ relations through common rules and artifacts causes a lock-in effect, where self-reinforcing mechanisms hinder the implementation of alternative systems even if they are promising (Arthur 1989; Cowan and Gunby 1996). In French agriculture, sociotechnical analyses have revealed that the lock-in around conventional practices has limited the implementation of agroecological practices (e.g., Meynard et al. 2018; Della Rossa et al. 2020), despite numerous incentives from public policy over the last 10 years. The wished agroecological transition, from farming systems relying on high pesticide use to farming systems based on agroecological practices with a low use of pesticides, did not occur. Such change can be defined as a complex agricultural problem (Schut et al. 2015): an agricultural problem with multiple dimensions (e.g., political, economic, technological), embedded in interactions across multiple levels from production to consumption of agricultural products, and where multiple stakeholders are involved (Fig. 1). It therefore requires a systemic analysis.

Fig. 1
figure 1

a Sheltered crops of lettuce infected by root-knot nematodes (RKN) (Photograph by Henri Ernout). RKN root infection reduces lettuce growth. When there is too much heterogeneity among lettuces, farmers destroy the crop as too many of them do not meet marketing firms’ standard. b Fruit and vegetables section in a supermarket. The requirements of marketing firms, distributors, and consumers for standard product quality (e.g., caliber) are decisive in farming practice choices.

Full size image

Schut et al. (2014) report that system analysis of complex agricultural problems, especially those related to crop protection, is understudied and that methodological frameworks to achieve such analysis are rare. According to them, the existing literature pays insufficient attention to multi-level interactions and to multi-actor interactions, especially those that involve private actors and policymakers. Very few papers propose a transformation-oriented analysis that puts forward structural transformation improving crop protection. Few recent studies have proposed multi-level, multi-actor, or transformation-oriented analysis (Maxim and Spangenberg 2009; Schut et al. 2015; Meynard et al. 2018; Della Rossa et al. 2020), but none has proposed and tested a comprehensive analytical framework of the determinants of farming practice, which take simultaneously field, farm, territory, and supra-territorial levels into consideration, as a means to understand the barriers to change in farming practice and to foster the design of solutions to overcome those barriers. Therefore, in this article, we propose an original analytical framework that aims to characterize (i) the determinants hindering or facilitating change toward agroecological practices and their evolution; (ii) the underpinning sociotechnical processes; (iii) the stakeholders involved and their interactions; and (iv) the levers that can be used to facilitate change and unlock the agroecological transition.

Most studies on transition to agroecological systems were carried out on arable systems (Meynard et al., 2018; Mawois et al., 2019). The sheltered vegetable sector presents some particularities that must be understood from a transition perspective: high degree of specialization of farms and farming systems, intensity of soil occupation, fresh products, and close attention to zero default visual quality (Bernard de Raymond 2013; Gamliel and van Bruggen 2016). We therefore applied the framework in a comprehensive analysis on the ongoing agroecological transition of vegetable production systems in the south of France, with a focus on the management of root-knot nematodes (RKN), Meloidogyne spp., the current most problematic soilborne pest in the area (Fig. 1). We analyzed jointly organic and conventional systems, both strongly present on the territory, to deepen our understanding of the barriers and levers to farming practice change. The outcomes of the study are therefore twofold: a generic analytical framework, and a deep understanding of the lock-in process and levers for vegetable production in a specific context.

We first present the analytical framework and how we implemented it in the organic-conventional case study (Section 2). In Section 3, we describe the dominant and alternative farming practices, the impediments to change in farming practices, the resources supporting the change, the underpinning sociotechnical structure and processes, the stakeholders involved, and the possible levers for steering the agroecological transition. We end this section with a discussion on the methodological contribution of our study. Finally, in Section 4 we open new research perspectives.

2 Materials and methods

2.1 A framework to analyze changes in farming practices

Farming practices are implemented in the field by farmers and farm workers and are influenced by multiple stakeholders through a web of interactions. They are at the intersection between the biophysical world and the socio-economic world. As such, a set of farming practices can be seen as a technology, when it is consistent with an agricultural paradigm and serves a specific purpose in the farming system (Rip and Kemp 1998). In this section, we present our analytical framework that aims to support the characterization of (i) the determinants of farming practices and their evolution, impacting farmers’ decisions directly, hence the impediments and resources for change in farming practice (Section 2.1.1); (ii) the sociotechnical organizations and processes that underlie the evolution of the determinants and their interactions (Section 2.1.2); and (iii) the levers for the transition toward agroecological practices (Section 2.1.2).

2.1.1 Determinants of farming practices

This framework is based on studies from two broad sets of disciplines: (i) systemic agronomy explaining changes in farming practices and (ii) social sciences, including technological and sustainability transition studies, and agricultural innovation studies. We assume that a multi-level analytical framework, consisting of a classification of the determinants of farming practices at each level (Fig. 2, Table 1), will enable an understanding of what hinders or favors change in current farming practices toward an agroecological model. Four levels are considered: field, farm, sectoral and territorial agrifood system, and sociotechnical landscape.

Fig. 2
figure 2

Analytical framework of the determinants of farming practices, showing interactions between four levels: (i) field level, (ii) farm level, (iii) sectoral and territorial agrifood system (STAFS), and (iv) sociotechnical landscape. The colored arrows show how all determinants influence farming practices, whereas the black double arrows between two levels illustrate that all levels interact with one another. The STAFS level gathers stakeholders from the X food sector (e.g., vegetable, fruit, cereal sector) under study and from the territory influencing the farming practices of farmers who are part of this food sector. The darker blue area represents the territory. Actors who are represented at the frontier of the darker blue area can be part of the territory or beyond.

Full size image
Table 1 Detailed analytical framework of the multi-level determinants of farming practices.
Full size table

The systemic agronomy literature shows that farming practices result from processes at three interacting levels: field, farm, and territory (Belmin 2016). At the field level, farming practice choices depend on the interactions between human (farmer), biophysical, technical factors (at crop and rotation levels), and the presence of specific infrastructures for production (e.g., irrigation system, plastic shelters that enable the production of specific crops) (Meynard 2014) (Fig. 2, Table 1). At the farm level, farming practices depend on available material resources in farming systems; for instance, crop diversity depends partly on infrastructure and equipment availability at farm level such as a specific seeder and the share of land equipped with irrigation (Aubry et al. 2006; Mawois et al. 2019). At the farm level, farming practices also depend on farmers’ cognitive resources (Mawois et al. 2019), strategic decisions (Aubry et al. 2006), and their personality, preferences, and objectives that frame them (Mawois et al. 2019). Biophysical factors at the farm level (e.g., natural enemies hosted in hedges) also play a role (Duru et al. 2015) (Fig. 2, Table 1).

The third level refers to the agrifood system, the “system [that] gathers all the elements (environment, people, inputs, processes, infrastructures, institutions, etc.) and activities that relate to the production, processing, distribution, preparation and consumption of food and the outputs of these activities, including socio-economic and environmental outcomes” (HLPE 2014 cited by Meynard et al. 2017). As agriculture is embedded in physical areas, a large part of these elements, activities, and interactions are located in a territory. From an agronomic point of view, the territory is defined both as the place where actors’ networks are built and interact, influencing farming practices, and as a cultivated and natural landscape with biophysical and ecological dimensions, both elements interacting with farming practices (Caron 2005). Lamine et al. (2019) then define the concept of territorial agrifood system as a sociotechnical system anchored within a territory but also embedding elements of the global scale with which it interacts (e.g., food distribution in long value chains goes beyond the local territory). A sociotechnical system is a relatively stable configuration of stakeholder collectives, their networks, their knowledge and practices, the artifacts and technologies they use, and the rules that frame their interactions (Rip and Kemp 1998). Moreover, other authors have noted that complex agricultural problems related to changes in farming practices are often sector-specific (e.g., parasitic weed management in rice farming — Schut et al. 2015). Therefore, we propose the concept of a sectoral and territorial agrifood system (STAFS) and define it as all stakeholders from the agrifood system influencing food production, processing, distribution, preparation and consumption within the territory, and the food production sector(s) in which the farming practices are embedded and the interaction between these stakeholders. Compared to the concept “territorial agrifood system,” STAFS excludes the stakeholders that are not part of the specific food sector under study or in strong interaction with it.

STAFS is the third level on which farming practices are built (Fig. 2, Table 1). The technological transition literature defines three kinds of rules underlying stakeholders’ actions and interactions within a STAFS, all of which impact practices change: (i) regulatory rules such as food quality standards; (ii) normative rules such as shared social norms (e.g., the idea that a well-cultivated field has to be weed-free); and (iii) cognitive rules such as shared paradigms (e.g., only the conventional agricultural model could feed the world) (Fig. 2, Table 1) (Geels 2004). The agricultural innovation support system is a central part of STAFS and provides structural conditions that can facilitate change (when such conditions are present) or hinder it (when they are absent or malfunctioning). These structural conditions are (i) the three abovementioned rule types; (ii) STAFS stakeholder cognitive resources; (iii) the quality of knowledge infrastructure; and (iv) multi-stakeholder interactions (Fig. 2, Table 1). The crucial role of interactions between multi-stakeholders in farmers’ capacity to change has also been highlighted by agronomists (Chantre and Cardona 2014; Meynard et al. 2018; Mawois et al. 2019). Finally, available material resources, human factors, and biophysical factors are also important determinants at the STAFS level (Duru et al. 2015) (Fig. 2, Table 1). The last level with a strong influence on farming practices is what Geels (2002) called the sociotechnical landscape in the multi-level perspective. It is defined as all factors external to the STAFS “framing” the interaction between the stakeholders: global sociopolitical factors, market dynamics, and environmental dynamics.

2.1.2 The sociotechnical processes underpinning practices change

Chantre and Cardona (2014) and Mawois et al. (2019) have demonstrated that farming practices evolve under the influence of interacting changing determinants. The interaction between determinants at the STAFS level and their evolution are structured by sociotechnical organizations and processes. For instance, the structuration of the organic value chain provided a better paid outlet (On-farm availability of material resources, Table 1) and combined it with specific extension services (Quality of knowledge infrastructure), which facilitated the change toward alternative practices without the use of synthetic pesticides. In this section, we present the concepts from the theory of technological transition that support the understanding of these sociotechnical organizations and processes and complement the framework presented above (Section 2.1.1).

In the multi-level perspective framework (Geels 2002, 2004), three levels interact: the sociotechnical landscape (Section 2.1.1), the sociotechnical regime, and the technological niches. The sociotechnical regime is a stable sociotechnical system where actors, rules, and artifacts are aligned and well structured around a specific technology (i.e., for us, a consistent set of practices — see Section 2.1). The interdependence between sociotechnical regime actors, rules, artifacts, and technology, and their coevolution, lead to the lock-in of the regime around technology and lock out any possible alternatives. The so-called “dominant” regimes represent the predominant actor networks in a specific sector or market. In agriculture, conventional farming is often considered as the dominant regime, locked around conventional farming practices (Meynard et al. 2018; Plumecocq et al. 2018; Della Rossa et al. 2020). A locked regime allows alternative practices only if they are compatible with the current system. This phenomenon is called path dependency (David 1985). By contrast, a technological niche is a protected sociotechnical system where new configurations of actors, shared rules, artifacts, and techniques can emerge and develop around an alternative technology. The three levels coevolve in interaction. For instance, disruption from the sociotechnical landscape can destabilize the regime and open up opportunities for radical innovations from niches, thus provoking a restructuring of the regime around a new technology. This evolution from one sociotechnical regime to another is a “sociotechnical transition.” According to the multi-level perspective framework, such transition is necessary for alternative technologies, incompatible with the current dominant regime, to be massively implemented.

To conclude, this original analytical framework mobilizes the analysis of multi-level determinants of farming practice to understand the logic behind the current practices implementation and identify the main STAFS stakeholders, the impediments to their change, resources to favor it, and levers to overcome impediments. It then relies on technological transition’s concepts to understand the interactions and evolution of the practices’ determinants at STAFS, hence of the impediments and resources to practices’ change. In that way, it provides material to clarify levers for overcoming the impediments and the resulting lock-in. Impediments to change are considered as practices’ determinants blocking the change. Resources are seen as practices’ determinant favoring the change. Finally, levers are seen as actions that modify the determinants of farming practices and contribute to overcoming the impediments to their change. We tested this analytical framework on the analysis of the change of root-knot nematode management practices in Provençal sheltered vegetable farming systems.

2.2 Case study

At the time of the study (2017–2020), the Provençal territory (located in the south of France, around Avignon) was a major French fresh vegetable production basin. It was mainly composed of numerous small farms (3.2ha in average), specialized in a very limited number of crop species cropped under plastic shelters (65% of the farms), and involved in long food value chains (at national and international levels) (RGA agricultural census 2010: http://www.agreste.agriculture.gouv.fr). Organic farming accounted for more than 810 farms and 12% of vegetable farms in Provence (RGA agricultural census 2010), with a well-structured value chain involving specific and mixed actors (organic and non-organic). In 2009, RKN was identified as the largest soil health issue in Provence, with 40% of the farms bearing losses of up to 100% of the yield, depending on the conditions (crop, infestation level) (Djian-Caporalino 2010). RKN contamination was particularly high under plastic shelters, due to the intensity of cropping and favorable climate for RKN (Gamliel and van Bruggen 2016) and concerned both organic and non-organic farming. Non-organic farming was dominated by a conventional model, although alternatives were developing. They will be presented in Section 3.2.3. Major crops cropped under shelters are vulnerable to RKN damage: tomato, eggplant, bell pepper, cucumber, melon, and lettuce.

Despite a progressive ban of synthetic nematicides resulting in increasing RKN issues, Provençal farmers sparsely implemented the available agroecological alternatives. Previous studies showed that their implementation for RKN control was impeded by numerous factors, for example, competition between commercial and cover crops at field level, availability of the workforce at farm level, and lack of outlets for diversification crops at value-chain levels (Navarrete et al. 2015; Navarrete et al. 2016). Yet, no comprehensive analysis of these factors and their interactions was available. Therefore, we hypothesized that the application of our multi-level analytical framework would be relevant to understand this apparent lock-in of RKN management practices in Provençal market gardening systems and identify levers to overcome it.

2.3 Methodology for sociotechnical analysis of change in farming practices

We built a participatory methodological approach to collect data and test our analytical framework on our case study. We considered eight analytical dimensions, explored in parallel, using six data collection methods (A–F in Table 2).

Table 2 Data sources for each analytical dimension and corresponding stakeholder groups.
Full size table

First, we analyzed the problem context (1) to obtain a global picture of the sheltered vegetable production systems in the territory, which was necessary to lead a comprehensive analysis and start to identify relevant stakeholders. Second, we identified agroecological and non-agroecological RKN management practices (2), according to scientific evidence and farmers’ or agronomists’ empirical knowledge, and characterized their degree of implementation by Provençal farmers. Third, using the analytical framework developed (Section 2.1), we characterized jointly the stakeholders of the STAFS (3), impediments to practices change toward agroecological ones and resources supporting the change (4), STAFS configuration (5), processes underlying impediments (6), and levers to overcome them (7). Levers could be divided into two categories using the distinction proposed by Schut et al. (2015). Specific levers concerned innovations devoted to change of RKN management practices toward more sustainable ones. Generic levers addressed innovations tackling structural problems, likely to have spillover effects on issues other than soilborne pests and diseases within the STAFS. Finally, we checked the analysis resulting from these data with the STAFS stakeholders (8).

The 8 dimensions were analyzed using 6 data collection methods involving all stakeholder categories we identified: farmers, upstream value chain and downstream value chain actors, policymakers, researchers, extension agents, and NGO and food sector animation structures (Table 2).

We led three types of semi-structured interviews. First, we conducted exploratory interviews with 10 stakeholders within the researcher team’s network (end 2017–early 2018) (A). We then conducted 11 interviews with new stakeholders (2018–2019) based on a topic list adapted to each interviewed stakeholder, in order to collect complementary information (C). Finally, we held 12 comprehensive interviews of farmers (end 2018–early 2019) (E) to establish an in-depth farm analysis, with a focus on the RKN management implemented and the impediments encountered in implementing agroecological practices (Table 2). We used the same guide for organic and conventional farmers. All the stakeholders interviewed were selected on a purposive basis aiming to get a diversity of stakeholders’ profiles and completed with snowball sampling, where interviewees made suggestions for who else should be included in the sample.

Additionally, we conducted a literature review of 18 documents (scientific and technical documents and statistic reports) (B) (end 2017–2020) and carried out participant observations of 22 scientific and professional meetings (2017–2019) (D) (Table 2). This provided us with complementary data on RKN management techniques, determinants of farming practices, the STAFS’ structure, and stakeholder interactions. Participant observation consisted in retrieving information with detailed notes to inform elements of the analytical framework (Fig. 2, Table 1).

Finally, 5 workshops with key STAFS stakeholders were held (F, Table 2, Boulestreau et al. 2019) to check if the analysis was consistent with their points of view, to complete data, and to co-design innovative solutions capable of fostering the implementation of agroecological RKN management in our case study. Note that the innovative solutions designed are not detailed in this article as it exceeds its scope. As the workshops took place between December 2018 and February 2020, it enabled us to progressively refine and stabilize the analysis. In particular, the third and fourth workshops (August and September 2019) confronted the results of our analysis to a diversity of stakeholders of the problem (Table 2, Supplementary information), who validated most of it and provided complementary elements for the rest. Detailed notes were taken by observers for 4 of the workshops, and all of them were recorded.

Combining these different data collection methods provided us with the different stakeholders’ individual and collective representations of problems and solutions. This enabled us to compare insiders’ (stakeholders’) and outsiders’ (researchers’) analyses. They provided sufficient detail and allowed us to triangulate and validate data, enhancing the credibility, validity, and quality of the analysis, even though the total sample of stakeholders directly involved in the study cannot be considered as representative. As such, they satisfied most of the selection criteria proposed by Schut et al. (2015).

The data collection and analysis were conducted jointly in organic and conventional agriculture. Some organic farmers are seen as “agroecology pioneers,” while other organic farmers are seen to be following a conventionalization path (e.g., Ramos García et al. 2017). We therefore assume that analyzing organic and conventional systems jointly could facilitate the identification of levers in conventional systems and of arising impediments for organic systems regarding the agroecological transition.

3 Results and discussion

In this section, we focus on the results of the problem analysis. The context is provided in Section 2.2; therefore, the context analysis (Dimension 1, Table 2) is not presented in the following sections. We first describe two RKN management technologies coexisting in Provençal sheltered vegetable production systems (Section 3.1). On this basis, we explore the impediments and resources for farmers to adopt agroecological RKN management, for each level (field, farm, STAFS, and sociotechnical landscape), as well as the stakeholders and their roles (Section 3.2). We then highlight the main underpinning sociotechnical processes explaining the impediments and resources and the levers to overcome them (Section 3.3). Finally, we discuss the contribution of our study to sociotechnical analysis of complex agricultural problems (Section 3.4). Data sources are indicated in brackets with the letters referring to Table 2.

3.1 Two competing soil health management technologies in the Provençal sheltered vegetable sector

In this section, we first present two technologies: “drastic soil disinfection” or “agroecological management of soil health” (B, C), referring to the concept of technology presented at the start of Section 2.1. We then present how the two sets of practices were used in Provençal sheltered vegetable systems (C, D, E) (Dimension 2, Table 2).

The RKN management techniques are based on three interacting mechanisms (Hoefferlin et al. 2018) (Fig. 3): (1) directly killing RKN larvae present in the soil, (2) preventing their reproduction and spreading, and (3) enhancing in-soil competition. Based on the above mechanisms that the techniques involve, we identified two consistent sets of techniques: those relating to drastic soil disinfection and those relating to soil health management with agroecological techniques. The former mobilizes only the first mechanism (1), whereas the latter synergistically mobilizes all three mechanisms.

Fig. 3
figure 3

Root-knot nematode management techniques and technologies. The techniques (in black) are sorted along the sides of the triangle, depending on the mechanisms of action they rely on (named in the red, purple, blue, and green-colored boxes). When underlined, techniques were identified in the scientific literature. [f] in superscript means that the technique was identified thanks to farmers’ empirical knowledge. [ag] in superscript means that the technique was identified thanks to agronomist empirical knowledge (in unpublished experiments or on-farm observations). Techniques were identified based on a combination of sources. The red and green boxes represent, respectively, drastic soil disinfection and agroecological management of soil health technologies. The red arrow highlights the incompatibility at field level between the techniques of drastic soil disinfection technology and the stimulation of in-soil competition. Adapted from Hoefferlin et al. (2018).

Full size image

The “drastic soil disinfection” technology consists of the implementation of the following consistent set of techniques: synthetic nematicides application, acting by fumigation or drip-irrigation, and soil disinfection by steam. It is hereafter referred to as “disinfection technology.” Disinfection technology drastically reduces soilborne pests and disease inoculum and occupies shelter space for a short time (<1 month). However, as it is unspecific, it also drastically reduces neutral and beneficial organism populations (e.g., arbuscular mycorrhizal fungi, parasitic bacteria), which seriously jeopardizes the natural regulation of pests and diseases (Gamliel and van Bruggen 2016), and therefore agroecological management of soil health. This technology is unsustainable, as it requires expensive applications that are repeated each year with a decreasing efficiency on RKN, due to resistances. Moreover, it can cause important direct or indirect human health (Pruett et al. 2001) or environmental damage (Hoefferlin et al. 2018). It is focused on short-term control. This set of techniques thus appears consistent with the conventional farming paradigm characterized by the preventive use of synthetic inputs to drastically reduce any risk of crop damages (Plumecocq et al. 2018).

The technology of agroecological management of soil health consists of the synergistic implementation of several techniques, each of which is insufficiently effective in RKN management when applied separately. It is hereafter referred to as “agroecological soil health technology.” It is embedded in the approach of agroecological crop protection defined by Deguine et al. (2020). This technology stems from a more holistic definition of soil health, where “the soil’s continued capacity to function as a vital living system, within ecosystem and land-use boundaries” is promoted (Gamliel and van Bruggen 2016). It is focused on long-term control. For example, the agroecological techniques involved are the use of resistant cultivars, organic amendment, and biofumigation (Fig. 3, Collange et al. 2011; Gamliel and van Bruggen 2016; Hoefferlin et al. 2018). They are based on ecological processes instead of synthetic inputs, for example, the application of organic amendment stimulates biological activity and in-soil competition. Unlike disinfection technology, they do not drastically reduce the natural enemies’ population in the soil (Gamliel and van Bruggen 2016). Effects on crop yield of agroecological soil health technology compared to drastic soil disinfection varied tremendously depending on the practices implemented and the context (e.g., soil temperature increase during solarization is generally insufficient when cloudy weather, decreasing the total effectiveness of the combination of agroecological techniques).

The two technologies (Fig. 3), which are incompatible at the field level, coexisted within the Provençal sheltered vegetable STAFS. In 2018, two synthetic fumigants (dazomet, metam sodium) before plantation and another synthetic nematicide (fluopyram) applied through drip-irrigation maximum twice during cultivation were still allowed for non-organic farming (Hoefferlin et al. 2018). Metam sodium and fluopyram were used extensively. Conversely, soil disinfection by steam and dazomet were seldom used, mainly because of their high cost. After the banning of metam sodium (end 2018), non-organic farmers went mostly for techniques that were compatible with their systems: techniques from disinfection technology (synthetic nematicides, ozonated water) but also from agroecological soil health technology (e.g., solarization for short spring crops’ producers, short trap cropping, biopesticides, resistant varieties). Organic farmers were using mainly agroecological techniques. For instance, non-organic farmers typically grew 3 species vulnerable to RKN (tomatoes, melon, lettuce), whereas organic farmers grew 5 to 8 species, including poor host crops (e.g., scallion, lamb’s lettuce). Agroecological techniques had been experimented in Provence for more than 10 years (Hoefferlin et al. 2018). However, at the time of the study, only a few farmers had successfully implemented a synergistic combination of agroecological techniques (Fig. 3), such as the combination of a diversified crop sequence and solarization. Therefore, in 2019, in Provençal market gardening systems, disinfection technology was predominant, and many farmers were still heavily impacted by RKN.

The following sections identify the factors and processes explaining why Provençal vegetable farmers did not massively shift to agroecological soil health technology, as recommended by policymakers and public R&D actors.

3.2 Impediments, resources, and stakeholders of practices change toward agroecological management of soil health

In the following sections, we present how the practice determinants constrained or enabled the change of RKN management practice and the role of the stakeholders involved. These results are summarized in Fig. 4. The phrases below in italics refer to the farming practice determinants presented in Fig. 2 and Table 1. When it is not specified, the results presented hold for both conventional and organic farming.

Fig. 4
figure 4

Sociotechnical analysis of change in root-knot nematode (RKN) management practices within Provençal sheltered vegetable production systems. a Coevolution of regimes and niches within the sheltered vegetable STAFS of Provence. The dotted double arrow represents the interaction between the organic and the conventional regime. (STAFS: sectoral and territorial agrifood system). b Historical lock-in of the dominant sociotechnical regime. c Evolution of dominant sociotechnical regime at study time. The dotted blue arrows represent the relations in-the-making, and the bold arrows represent stabilized interactions between niches and dominant regime stakeholders. d Organic sociotechnical regime at study time. The yellow boxes represent elements favorable to a lock-in around drastic soil disinfection. The green boxes represent elements favorable to a structuring around agroecological management of soil health. The blue and bold double arrows represent the alignment of rules between stakeholder groups. The rectangular boxes at the center of the subfigures or next to niches’ boxes represent technologies. Upstream and downstream are references to upstream and downstream value chains.

Full size image

3.2.1 Field level

At field level, interaction between cropping practices was the main determinant for the practices change toward agroecological soil health technology. The principal corresponding impediment identified was a time-space incompatibility between major commercial crop production and agroecological techniques. Diversification with service crops or poor RKN-host commercial crops (e.g., Allium sp.), soil solarization, and biological soil disinfection all competed with the main commercial crop production and, among them, for land and workforce (B, C, D, E, F). The substitution of long summer crops (tomato, pepper) for short spring crops (melon, cucumber, or zucchini) was a way of overcoming this issue. It freed up space in summer to use the previous agroecological techniques. Presence of infrastructure played a role mainly through the shelter type used for the plot; for example, double walls were said to be incompatible with solarization (E). A still poorly understood resource related to field’s biophysical factors was “suppressive” soils (Gamliel and van Bruggen 2016), showing very limited RKN pressure. This was likely due to the higher clay percentage and biological activity within RKN suppressive soils (A, C, E; Mateille et al. 2008; Gamliel and van Bruggen 2016).

3.2.2 Farm level

On-farm availability of material resources and farmer’s cognitive resources were the main determinants, at farm level, for the change of farming practices toward agroecological soil health technology.

Regarding on-farm availability of material resources, the capital was very limiting. A large proportion of Provençal sheltered vegetable farms were under intense economic pressure: high debt levels and low income, especially in conventional farming compared to organic. This drove farmers toward specialization on the few cash crops that allowed them to recoup their investments (shelter installation, plastic cover replacement) but prevented them from using agroecological practices that compete for shelter space (see Section 3.2.1). Furthermore, the high economic pressure reinforced risk aversion regarding alternative practices that required extra investments (e.g., service crop seeds) but could lower their sales (e.g., uncertain vegetable quality of a new resistant variety) (A, B, C, D, E, F). Limited time and mental load reduced farmers’ ability to acquire new knowledge on agroecological techniques, stand back from the situation, and organize new forms of production (e.g., sanitation measures) and commercialization (e.g., for crop diversification) (A, B, C, D, E, F). Lack of local outlets was the main issue raised with regard to commercial crop diversification (B, C, E, F). Moreover, specific equipment was often lacking to implement agroecological practices, such as a specific seeder for poor RKN-host radish, or a manure spreader (A, C, D, E, F). The seasonal unskilled workforce made it difficult to implement new agroecological practices, as it would require time to train workers (A, C, D, F). Finally, the cultivation area was also a determining factor, as farms with small cultivation areas under shelter (typically under 1 ha) had limited leeway to free up space for alternative soil occupation to cash crops (C, D, F).

The main limitation to change due to farmer’s cognitive resources was farmers’ poor knowledge of agroecological soil health technology (D, E, F). This main impediment was reinforced by a low risk awareness of production losses due to soilborne pests and diseases, as the damage is less visible than on the aboveground part of plants (A, C). Moreover, farmers did not think in a systemic way, which caused them to reject “expensive techniques” without comparing them with their current practices at the cropping system level (C). Such knowledge limitation was revealed both for organic and for conventional farming systems but was stronger in the latter (D, E, F).

Farmer’s personality, preferences, and objectives were important factors influencing practices change, according to our interviews with farmers and extension agents (A, C, E) and from our observation of meetings (D) and workshops (F). The farmers least motivated to implement the agroecological soil health technology were those mainly focusing on the need to produce cheap food in large quantities to feed the maximum number of people or with the objective to maximize short-term economic outcomes, with little attention for the middle- and long-term side effects on environment and human health (conventional paradigm). These farmers were a part of the conventional or mixed farmers in Provence. Conversely, farmers strongly motivated to implement agroecological soil health technology considered the soil as a living organism (agroecological paradigm). They set environmental and human health care as one of their main farming goals that they aim to conciliate with good production levels. They encompass most organic farmers and part of the conventional or mixed farmers.

Farming strategic choices clearly set back change in farming practices toward agroecology, for instance, by limiting commercial crop diversification (focus on specialized outlets, investment in specialized tools) (D, E, F).

To conclude, these results show how on-farm material resources, farmers’ and workers’ cognitive resources, mindset, and farmers’ strategic choices can impede or support change in farming practices. These results are consistent with the findings of Navarrete et al. (2016) on the same case study, which showed that the main factors influencing the implementation of a RKN-trap crop on farms were the compatibility of the cropping calendar, the cost, the working time required, and the interest of farmers in agroecological techniques and innovations. Mawois et al. (2019) showed on arable systems that thinking in a systemic way was key if farmers were to introduce legumes. Della Rossa et al. (2020) and Belmin (2016) have shown on the cases of Galion watershed farmers (French West Indies) and Corsican clementine producers that aversion to risk when it comes to production losses and farmers’ conventional agricultural paradigm both hinder practices change toward more sustainable choices.

3.2.3 Sectoral and territorial agrifood system level

The analysis of determinants of farming practice at STAFS level revealed that STAFS stakeholders other than farmers and workers were playing an important role, mostly impeding the change of farmers toward agroecological practices.

Regarding multi-stakeholder interactions, we distinguished the downstream and upstream value chain actors. First, most wholesalers and retailers had many restrictive demands to farmers, which hindered the change of soil management practices toward agroecology. They offered low and fluctuating prices, required large volumes of a few species, refused any visual defect on vegetables, and required high organoleptic value, technological quality (transport, conservation), and a precise crop production calendar (A, B, C, D, E, F). Marketing firms and especially supermarkets’ purchasing division could force farmers to meet their demands due to higher bargaining power. They claimed that all these requirements reflect consumer demands. Yet, their requirements greatly constrained the development of resistant varieties for commercial crops; for example, an eggplant variety was refused because of red spots and a promising rootstock (Solanum torvum) because it delayed the start of harvesting (C, D). Low constancy and dialogues between farmers and marketing companies made it difficult for both to initiate change in their commercial and farming practices (E, C, F).

At the time of our study, marketing firms specialized in organic left generally more leeway to their suppliers compared to the conventional ones (A, B, D, F). This has been reported in other case studies as well (e.g., Belmin 2016 on the Corsican clementine). It could be partially explained by alternative governance of the value chain organized by organic marketing firms, as shown by Lamine (2014) for the French commercial network Biocoop. They offered higher prices, as well as outlets for a wider diversity of crops, lower visual quality specifications, and better communication. On the other hand, large conventional farmers and marketing companies entering the organic market were fostering the evolution of the organic sector toward stricter requirements and the prevalence of short-term economic issues over long-term agronomic ones (A, F, D). This was driving the sector toward what was described as a “conventionalization” process (Ramos García et al. 2017).

Second, stakeholders from upstream value chain were not offering adequate knowledge and inputs for agroecological RKN management to farmers. Farmers and extension service agents criticized input suppliers for giving little information on the inputs available for agroecological practices, for example, RKN-resistant traits of commercial and service crops for seeds, rootstocks, and plantlets, despite tests being run (D, F). They were also criticized for offering low-quality plantlets or seeds for minor crops and service plants or ones that differed from what the farmer had ordered (D, E). Regarding the interaction between farmers and extension services, two thirds of Provençal vegetable farmers were relying only on free recommendations provided by input suppliers, who had little knowledge on RKN management techniques. Only one third were using independent, better trained but paid public services (A, C, D, E, F). This result echoes others on arable crops; for example, Chantre and Cardona (2014) highlighted the fact that the diversification of knowledge sources helped farmers in their transition to highly sustainable farming practices.

When farmers’ organizations could be a resource to overcome the above-cited impediments, many stakeholders highlighted that farmers were poorly organized on the territory and that the lack of an equipment cooperative or a marketing cooperative was limiting the access to specific equipment and new outlets (A, B, C, D, E, F). However, despite being rare, they were not completely absent. For instance, two collective initiatives to value production resulting from agroecological management of soil (including soil health) were being developed during the period under study and could become resources for practice change in the future (C, D). The first was an NGO founded at national level between marketing firms and agroecology pioneer farmers or extension agents. In Provence, they were fostering the structuring of a value chain offering better prices to vegetable farmers, which are produced with conservation agriculture techniques, based on a certification process. They organized technical support on agroecological management of the soil with public and private extension services and raised awareness of marketing firm on the enabling conditions to allow farmers to change their practices toward agroecology. The second initiative was led by the regional market platform cluster (a public institution), which involved a group of vegetable farmers in the adaptation of a national certification standard to certify soil management practices favorable to the environment.

The multi-stakeholder interaction issues mentioned above arose partly from limited stakeholder cognitive resources concerning the problem under study. Upstream and downstream value chain actors were lacking awareness on the issues encountered by vegetable farmers and especially their responsibilities with regard to them. This resulted in miscommunications (F). Moreover, all the stakeholders lacked knowledge on agroecological techniques for soil health management (C, D, E, F). Researchers, extension services, and farmers lacked actionable knowledge on the origin of RKN contamination, on how to implement agroecological techniques with the available inputs (e.g., how to apply biocontrol effectively), and on how to combine them. Input suppliers likewise lacked basic knowledge on RKN-plant interactions that would enable them to develop specific input for agroecological practices (F). The quality of knowledge infrastructure was both an impediment and a resource in our case study (A, C, D, E, F). At national level, low investment by public and private research in vegetable production focused only on a few major species, limited varietal development, and knowledge on RKN-crop interactions. On the other hand, in the Provençal territory, a tight network of research, organic and mixed extension services, and experimentation platforms, directly involving some farmers, was a strong resource to speed up the development and implementation of agroecological management of soil health. This network was already engaged in several research projects in that direction.

As highlighted by Geels (2004), shared regulatory, normative, and cognitive rules structured multi-stakeholder interactions and as such, played an essential role in impeding or fostering practice change toward agroecological soil health technology. Regarding regulatory rules, the requirements of marketing firms regarding standard vegetable quality (visual aspect, caliber) were enforced by EU regulations (A, B). Input-supply firms had difficulties to obtain a marketing authorization for agroecological inputs (e.g., biocontrol agent) or to communicate on their effect on RKN inoculum, because of their partial effects and EU regulations on the efficiency evaluation protocol (C, D, F). On the other hand, organic certification was a resource to foster crop diversification by requiring that the cropping sequence should have at least three different consecutive crops (Gamliel and van Bruggen 2016). Normative rules, and especially “role expectations,” were key. For instance, during our study, several marketing firms declined our offer to get involved in a work on soil health management on vegetable farms. Their justifications, completed with stakeholder interviews (C), showed that the procurement and sales managers of these companies considered that technical aspects of the production were not their responsibility. Shared cognitive rules reinforced weak cooperation between farmers and between them and the marketing firms. Interviews and meetings revealed a shared belief, based on past events on the territory, that no cooperation was possible without some participants taking advantage of it at the expense of the others (A, C, D, E). Our data also showed that the conventional paradigm (see Sections 3.1 and 3.2.2), historically shared between most researchers, extension services, and farmers, was still structuring part of farmers’ and extension services’ mindsets (B, D, E). However, a growing number of public extension services and research, and some farmers, shared the alternative agroecological paradigm (A, C, D, E, F), constituting an important resource for practice change.

The availability of material resources at the STAFS level was another major determinant constraining practice change. Limited time prevented service providers (e.g., for tilling) from cleaning their tools between two farms (C, D, E). They were pointed to as potential major sources of initial RKN contamination on farms. Farmers and extension services pointed out a local lack of inputs that were needed for agroecological techniques, and that could be explained by the lack of development (see above), production, and distribution of these inputs (C, D, F). For instance, despite the existence of resistant rootstocks for melon (Cucumis metuliferus) (Expósito et al. 2019), they are not commercialized in France. On the other hand, financial infrastructure was strong, as public policy stakeholders pointed out, stating that various sources of public subsidies were available from the local to the EU level, to foster agroecological soil health technology (F).

To conclude, the implementation of all agroecological techniques mentioned in Section 3.1 encountered hindrances resulting from interactions between downstream and upstream value chains, policymakers, R&D, service providers, workers, and farmers (Fig. 3). Yet, the local resources showed high potential for agroecological transition of the Provençal sheltered vegetable systems, regarding RKN management.

3.2.4 Sociotechnical landscape level

Sociotechnical landscape level determinants reinforced impediments or resources at the previous levels. Global market dynamics, with national and international competition, were pointed out as responsible for low and fluctuating prices for farmers. They were also considered to be the cause behind the lack of local outlets for minor and poor RKN-host crops, produced at a lower cost on other territories (A, C, E, F). The analysis of French fruit and vegetable markets by Bernard de Raymond (2013) confirms the reality of these factors. Global environmental dynamics, namely, current global warming, has narrowed the cold period when RKN are inactive in the soil (C). As a consequence, escape cropping and crop calendar requirements became even more difficult to implement. Regarding global sociopolitical factors, an increase in societal and political awareness of the environmental and human health harm caused by pesticides led to the banning of synthetic nematicides and indirectly to the evolution of farmer’s practices toward agroecology (B, E).

The previous results show that numerous impediments, interacting from field to sociotechnical landscape level, block the change of farmers’ practices toward agroecological soil health technology. We have shown that numerous impediments to change toward more sustainable practices, which we observed, were also present in other case studies. However, our case study revealed specific features, especially compared to arable cropping (Chantre and Cardona 2014; Meynard et al. 2018; Mawois et al. 2019). The pressures of time and mental load are higher than in other production systems as vegetable farmers cultivate all year round and need to manage a large workforce. Additionally, working with fresh products tightens the requirements on production calendars, volumes, and visual quality.

3.3 From lock-in to levers for driving the sociotechnical transition toward agroecology

In this section we show the underlying sociotechnical processes responsible for the identified impediments (Section 3.2), using the technological transition theory. We then present the levers that were identified to overcome them.

Disinfection technology was still the predominant technology at the beginning of our study (Section 3.1). The prevailing conventional regime had been historically structured around this technology (Fig. 4b). Bernard de Raymond (2013) shows that the fruit and vegetable value chain had been structured to comply with the constraints of mass retailing: conservation, homogeneity, and zero visual defects. This structuring had gone hand in hand with massification and standardization of production and consumption, supported by public regulation on product quality standards and innovations from R&D and input suppliers (e.g., pesticides, varieties, new cropping systems). It enabled massive scale economies for the value chain actors. However, it resulted in economic and product constraints that made it necessary for farmers to use pesticides (Bernard de Raymond 2013; Belmin 2016), i.e., in our case study to use disinfection technology. In the previous section, we showed that this structuring was still present at the time of the study, resulting in interrelated impediments and lock-in around disinfection technology. It led to path dependency: only techniques compatible with the current system were largely implemented (Section 3.1). Diversification, for instance, was locked out. We also showed that the organic regime in Provence, evolving from a niche to a mass market, risked structuring itself around a similar lock-in (Fig. 4d).

We nevertheless identified that landscape pressures related to environmental awareness (Section 3.2.4) and interactions with the organic sector were destabilizing the dominant regime (Fig. 4a). As a result, an increasing number of non-organic stakeholders were structuring their beliefs and activities around the agroecological paradigm (Section 3.2.3, Fig. 4c) and participating in the creation of resources for unlocking (e.g., good knowledge infrastructure, alternative value chains). Technological niches were emerging, such as the value chain structuring around soil conservation farming practices (Section 3.2.3, Fig. 4c). As is often the case in agriculture, they were close to dominant regime actors, then interacting strongly with the dominant regime (Belmin et al. 2018). Following Geels and Schot (2007), this could be seen as an ongoing sociotechnical transition process. Nevertheless, landscape pressure from global market dynamics was not changing, which could hinder a full agroecological transition. The organic sector was converging with the conventional one, especially regarding downstream value chain structuring (Fig. 4a, d). However, it could find an alternative path.

To orientate the future transition pathway, stakeholders (during the workshops — F) and researchers identified several levers based on the previous analyses. Some were peculiar to RKN management and others were generic (Table 3). The main specific levers consisted in generating and sharing both basic and applied knowledge on agroecological soil health technology to foster the development of agroecological inputs and ease the development of agroecological practices by extension services and farmers (Table 3). As generic levers, it appeared crucial to give farmers more leeway in decision-making by freeing up time, increasing their income, and fostering organization with marketing firms (Table 3). This could foster a change of management practices regarding soil health or other farming practices and give rise to new marketing structures, increasing value chain sustainability. These levers are entry points in the complex agricultural problem under study to explore innovations that could foster practice change. This exploration could be supported by knowledge produced in other case studies. Indeed, many of the levers we identified have already been studied in other contexts; e.g., “fostering participation of farmers in peer-exchange groups on agroecological management” was explored in the case of campesino-to-campesino movement in Cuba by Rosset et al. (2011).

Table 3 Specific and generic levers for agroecological management of soil health.
Full size table

3.4 Contributions to sociotechnical analysis of practices change

In this section, we discuss the methodological contributions of our study to sociotechnical analysis applied to complex agricultural problems and especially to change in farming practices.

Our study contributes to system agronomy (Meynard 2014) and sociotechnical analysis literature applied to farming practices (Belmin 2016; Meynard et al. 2018) by proposing a clear operational multi-level analytical framework and demonstrating its relevance on a case study. The system agronomy literature has generally analyzed practices at one or two levels only (Belmin 2016). In our framework, all levels impacting farming practices were taken into consideration: field, farm, STAFS, and sociotechnical landscape. Moreover, building on studies from Belmin (2016) and Meynard et al. (2018), we have proposed a framework that operationalizes the concepts from technological transition theory (Geels 2002, 2004) on the analysis of practices change.

Our new framework complements other recent frameworks for systemic analysis of complex agricultural problems. Schut et al. (2015) proposed the RAAIS framework focused on agricultural innovation systems, to analyze constraints to and opportunities for innovation, for solving complex agricultural problems. Della Rossa et al. (2020) proposed a framework adapted to study the hindrances to sustainable management of shared natural resources in multiple agricultural production sectors. We built on the multi-level approach of farming practices from Belmin (2016), with proposing for the first time (i) the STAFS level to represent the level where stakeholders’ interactions influencing farming practices occur and (ii) a classification of determinants of practice at their different levels to facilitate the analysis.

Focusing on determinants of practices enables us to jointly characterize the different farming practice technologies available, their effects on the agroecosystem (including productivity, human health, and environmental effects), and the factors enabling or constraining change in practices toward more desirable ones. It thus provides all the elements required to foster the design of innovations “coupling” innovative practices at field level with technological, organizational, or institutional innovations at farm and STAFS levels. Coupled innovations are innovations designed “in raw production, exchange, processing, and consumption, while taking into account synergies or antagonisms between upstream and downstream. Thus, the innovations are not only technological – e.g., concerning cropping systems or processing – but also organizational and institutional” (Meynard et al. 2017). Our analytical framework is then specifically adapted to prepare the design of coupled innovation linking innovative practices at field level with enabling innovations at upper levels. In our case study, we showed that the lock-in of farming practice change was systemic. Implementing simultaneously coupled innovations is then needed to unlock farming practice change toward agroecological practices. Therefore, not one but various levers presented in Table 3 should be activated, together with the implementation of agroecological soil health technology within Provençal vegetable farms.

In our case study, the STAFS concept allowed us to break out of the local territorial level (Della Rossa et al. 2020) or the focus only on the actors of “production, processing and commercialization of agricultural commodities” (Schut et al. 2015), to consider all the actors that were influencing farming practices in the sheltered vegetable system in Provence (e.g., consumers).

Note that not all determinants identified in the literature were playing a role in our case study: farmer-field relation, human factors, and farm’s and STAFS’ biophysical factors (Table 1). The first two seem to be more adapted to study practices change for individual trajectories of farms, as in Chantre and Cardona (2014) and Mawois et al. (2019). The latter two were likely not present because RKN management depends only on very local biophysical conditions, as a result of its low mobility. We therefore posit that farm’s and STAFS’ biophysical factors would play a larger role in practice change regarding aboveground pest management. Therefore, the genericity of this framework should be assessed by applying it to other problems of change in farming practices (related to crop protection or not), in other production systems (e.g., cereals), in other territories, and on individual farm trajectories of practice change.

Finally, our methodological proposition to operationalize the analytical framework of multi-level determinants of farming practice could be improved as regards quantitative data collection and representation of the different stakeholders. Additional quantitative data could have been collected through a web survey of numerous stakeholders at the territorial scale (Schut et al., 2015). Although all stakeholder categories were invited to participate in the study and were involved in several data collection methods, their representation was unbalanced. Organic stakeholders responded more than conventional ones, and farmers, researchers, and extension service agents were overrepresented. Therefore, confronting the results of this paper to a larger number of the underrepresented stakeholders could help increase confidence in its conclusions. Moreover, it would have been interesting to involve NGOs representing consumers to discuss impediments related to purchasing and consumption of vegetables. Other disciplinary approaches such as economics and sociology would enhance the understanding of the sociotechnical processes, especially regarding downstream and upstream value chains, public policy roles, and sociotechnical landscapes. When applying our framework to another study, the methodology for data collection and analysis should be adapted to its specific context (e.g., territory characteristics, researchers’ resources).

4 Conclusion

The use of our new analytical framework has enabled the characterization of resources and impediments with regard to change from current RKN management to agroecological practices. It has likewise facilitated the characterization of the underpinning sociotechnical processes, of the stakeholders involved, and, finally, of the levers to overcome the impediments. In this study, we have shown for the first time that in Provençal vegetable farming systems, RKN management practices were locked around drastic soil disinfection technology. Yet, numerous stakeholders were restructuring their activities and beliefs around the agroecological soil health technology.

The joint analysis of organic and conventional farming was a powerful way to reveal resources, current and future impediments, and levers for both regimes. For instance, the characterization of impediments arising from the downstream value chain in the conventional regime showed the risks the organic regime was taking in the conventionalization process. Conversely, the better value chain governance by organic regime stakeholders is both a local resource and an example of lever for the organic and conventional regimes.

This analysis has revealed specific and generic levers that could be used (i) to identify relevant existing innovation (innovation tracking) and (ii) to design coupled innovation with local stakeholders adapted to their context. To support these actions, it would be valuable to evaluate the potential for supporting the implementation of agroecological management practices of the identified levers and of the way they are implemented. It would enable hierarchizing the levers and prioritizing their implementation. Developing a method to achieve this is a future avenue of research.

Finally, this paper lays a conceptual and methodological ground on which systemic analyses of complex problems regarding agricultural practice change could be built in other production and territorial contexts.