It’s a subject that crystallizes tensions: is cohabitation possible between agroecological systems, or those on the way to becoming so, and digital technologies?
For some, these are two totally opposed worldviews, so different are the epistemological foundations and value systems of agroecology and digital technologies. For others, on the contrary, there’s no reason why digital technologies can’t accompany the agroecological transition, in the sense that these two dimensions are not to be taken on the same level and are therefore not in phase opposition. And there are still others who unfortunately don’t even really ask the question…
The definitions of agroecology and digital technologies are actually quite blurred, with everyone having different representations and interpretations. Nor can we hide the fact that, in some cases, these – sometimes highly personal – choices of definition actually reflect the values and priorities of each individual. This makes it difficult to find common ground when the subject under discussion is not the same.
On controversial subjects, it is sometimes preferable to put the question differently, in an attempt to bring about reconciliation. Just as the paths to greening are many and varied, so too are the trajectories of digitization. Agroecological systems abound: organic farming, regenerative agriculture, permaculture, mixed farming, low-input systems, soil conservation agriculture, and so on. And, on the other hand, there is not ONE, but MANY digital pathways. How can you compare an electric robot for localized spraying with a push-button to open an irrigation valve remotely? The answer is simple: they can’t be compared.
What I’m trying to do in this work is to provide a bit of perspective. Through examples drawn from old blog entries, literature and interviews, I try to sketch out encounters between these two worlds that find it hard to talk to each other. The dossier also returns to the complexity of the agroecological transition for farmers, and the need to rehumanize and reincarnate the debate by ensuring that farmers are always at the heart of it, even if they are undoubtedly not the only ones concerned by its deployment.
This dossier on the coexistence of digital and agroecological trajectories is also an opportunity to capitalize on all the knowledge that is beginning to be capitalized on the directory of digital tools for agriculture (https://www.wiki-agri-tech.com/). In addition to serving as a collaborative watch, this platform can now be used to take a step back from existing digital tools and identify trends.
As usual, for readers of the blog, this article is based on video interviews with players in the sector (whose names you’ll find at the end of the article), whom I’d like to thank for the time they were able to spare. Several articles, reports and webinars have enabled me to complete the interview feedback. For reasons of confidentiality, some of the interviewees did not want their names to appear at the end of this article.
Enjoy your reading!
Soutenez Agriculture et numérique – Blog Aspexit sur TipeeeImportant preamble
Let’s get one thing straight right away. I’m by no means a specialist in agroecology. In fact, the subject is so broad that I wonder if anyone can really claim to have mastered all its ins and outs. So I wouldn’t dare to give a lecture on agroecology in this dossier. I’ll confine myself to sketching out its features in the light of what I’ve been able to read in the literature and discuss in the course of my interviews. What interests me most here is crossing the theme of agroecology with that of digital technology, to look for potential synergies, fractures, compatibilities or even oppositions.
I’d been wanting to tackle this subject in a blog entry for quite some time, but I’d never found a satisfactory entry point. Not mature enough, too broad, too boilerplate, too divisive – there were plenty of reasons to push this work to the bottom of the pile. It has to be said that the scientific literature – which is now beginning to expand – remains rather divided.
On the one hand, articles, mostly from the humanities and social sciences, approach the intersection of agroecology and digital technology from a highly techno-critical angle. On the other, articles extol the merits of digital technologies to support the agroecological transition, without really clarifying the type of agroecological systems considered or their degree of ambition. And research articles that look at the compatibility of agroecological and digital trajectories often do so by comparing their values and foundations, or the perception of the actors who use them. In the end, these articles pay little attention to the role of agricultural innovation systems (with technology) in the ecologisation of agriculture (Schnebelin et al., 2022).
France and Europe are currently funding projects to study this cohabitation. These include the Horizon Europe projects “Path2Dea” (https://www.path2dea.eu/index.html) and “D4AgEcol” (https://d4agecol.eu/), which started in early 2022. France has launched its priority research equipment program (PEPR) “Agroécologie et Numérique” (https://www.pepr-agroeconum.fr/) in 2023 with, within it, four major clusters designed to support, and even accelerate, the agroecological transition. The first will focus on the links between technology and society (digital transformation of professions, orientation of research trajectories for breeding programs, etc.). Secondly, robotics and agri-equipment. A third one revolving around genetic resources (characterization of plant and animal genomes, characterization of animal holobionts, etc.). And a last one around modeling and data science (modeling ecosystem states and dynamics, digital twins, etc.).
The debate on the coexistence of digital and agroecological trajectories is all the more important as there are many models for greening agriculture, and also many systems of technological innovation. In this battle for the future of agriculture, all agricultural models claim to be aligned with agroecological trajectories, even if no one is ultimately talking about the same thing.
Let me remind you that I write popularization reports, not scientific articles (although I have written some in the past). Nevertheless, these dossiers are extensively researched. They are a synthesis (sometimes only slightly revised) of what I have read and/or heard from my interviewees. For me, popularization is not an oversimplification of reality, but a way of making science more accessible. I try to make this work as objective as possible, even if I do remain committed to my writing.
Please keep this in mind as you read this work!
Introduction and definitions
Agroecology, a three-part problem: “scientific discipline – alternative practices – social movement”.
Agroecology is often defined by the triptych “scientific discipline – alternative practices – social movement” (Figure 1). A scientific discipline because it is a science of observation and knowledge, and seeks to apply ecological principles (biological processes and regulations, etc.) to agricultural systems. Agroecology is also seen as a range of alternative practices in the field (agroforestry, cover crops, metail, etc.), sometimes in strong opposition to what is proposed by the dominant industrial agricultural model. Finally, it is a social movement, with a desire to deliver a strong political message on how to produce our food and build our food system as a whole, while respecting the physical limits of our environment.
These three proposals are obviously linked and can co-evolve to form a global approach. Agroecology can thus be seen as a transdisciplinary, participatory and action-oriented approach to ecology, agriculture, food, nutrition and the social sciences. It raises questions about what needs to be mastered to preserve a “healthy” agro-ecosystem, and about the extent to which local and regional players need to be involved in this process.
Figure 1: Historical evolution of agroecology and its principles. A) Disciplinary basis of the principles articulated within agroecology. B) Spatial scales. C) Emergence of the three manifestations of agroecology (science, practice and social movement. Source: Wezel et al., 2020).
Of all the ways of representing agroecology, the Gliessman scale, the FAO (United Nations Food and Agriculture Organization) pillars and the HLPE (an acronym for high-level experts) principles [HLPE 2019] are perhaps the most exhaustive. Hence the importance of relying on them to avoid reinventing the wheel. These notions have also recently been correlated with each other to give them a little more scope (Figure 2). From this all-encompassing graph derive a number of concepts and keys to understanding agroecology. Applying these generic concepts opens up the possibility of either gradual or largely transformational change towards more sustainable agricultural and food systems.
Figure 2: Agroecology levels, elements and principles. Note that links between elements and principles are represented by connecting lines, while the color of the lines refers to the levels of the graph. Source: Ewert et al., 2023.
Let’s take a stroll through these modes of representation:
Agroecology is already seeking to make the most of heterogeneity, rather than erasing or homogenizing it. Thierry Caquet and his colleagues describe it particularly well: “Agroecology favors the use of diversity, whether genetic (association of varieties, search for hardiness…), specific (association of crops, diversity of species) or functional (agroforestry, crop-livestock association…)”. The intention behind this multi-scale heterogeneity (intraspecific genetic diversity at the scale of a species, specific diversity and interactions between species) is to move from a paradigm based on the ideal individual, which aims to obtain the elite (most efficient) individual, animal or plant, in an environment made optimal and which has forged current agricultural systems, to a new paradigm based on the efficiency of interactions between individuals and their integration into ecosystems, whether field or landscape” (Caquet et al., 2019). The underlying assumption is that a diversity of individuals, varieties/breeds or species will be better adapted to heterogeneous and changing environments due to the interactions they maintain, even if this assumption remains to be demonstrated in an operational context (Gascuel et al., 2022).
The notion of spatial scale is particularly decisive. First and foremost, it is the fact that knowledge and observations are situated in space (we speak of situated or contextualized knowledge) and that we need to identify and apply the best agricultural practices adapted to the local context of the farm. This means understanding both what the ecosystem offers and the function of each ecosystem inhabitant, in a logic of connectivity and strong inter-relationships. These best practices will not necessarily be identical, even for farms just a few kilometers apart. The generalization of agroecological practices is therefore not a miracle recipe that can be easily transposed to any production conditions.
The notion of spatial scale is also important, because the framework of agroecology goes beyond the farm (metapopulations, landscape, territory, region, value chains, etc.) and seeks to reach the food system as a whole (levels 4 and 5 of the Gliessman scale) with a broader vision of social, economic and political change, with food sovereignty as a central element. Insofar as agroecological systems will depend to a greater extent on neighborhood effects, landscape elements (hedges, ditches, trees, grass strips, ecological corridors) or common goods (state of the water table, absence of soil erosion, etc.), this broad spectrum of spatial scales appears fundamental.
Agroecology works on a time scale that includes long periods of time in the sense that all the concepts around resilience, robustness, cumulative effects, or adaptability of production systems are designed for the long term. It is on these time scales that the diversification of crop rotation (space) and rotation (time) bring interesting emerging properties.
Figure 2 emphasizes that the farmer must be at the center of the agroecological system, and that all forms of domination must be eliminated. Experiential, traditional, local and tacit knowledge – the fruit of the farmer’s experience and history – is considered particularly rich and must be fully integrated into the design of agroecological systems if they are to capitalize on locally present assets. The farmer’s role is thus reinforced in relation to all stakeholders, and his skills and participation are deemed indispensable to the design of agroecological itineraries.
Unsurprisingly, close attention to ongoing ecological processes and to the physical constraints and limits of the land give agroecology a strong environmental protection dimension in the broadest sense. The efficient use of resources, in limited quantities to respect ecological ceilings, with a committed view to the circularity of flows (as opposed to the classic linear logic) on a multi-scale and multi-actor basis to ensure maximum reuse of resources, is an integral part of this safeguarding ambition.
Closely linked to diversity issues are the notions of vulnerability and resilience, the latter being defined as the capacity to absorb disturbances and reorganize in a context of change. In view of current climate change and the volatility of input and crop prices (which can be directly linked to climate change), all these issues of anticipation and adaptation are at the heart of a long-term agroecology strategy.
A first entry into the agricultural digital ecosystem?
Let’s take a gentle look at the digital tools and technologies available in agriculture, to give you an initial overview of what’s out there – we’ll have plenty more to say about this later.
Our collaborative census of digital tools in agriculture (Wiki Agri Tech) has led us to categorize the digital ecosystem into 5 major functions.
Figure 3 : Digital technologies for upstream agricultural production
Without representing this discrimination in detail, you can roughly identify :
Observe and Measure – a function that brings together all the measuring instruments and sensors used to collect data to measure, describe and quantify a soil, a plant or a stand, an animal, a flock or a herd, a climate, a topography, and so on. This also includes geolocation data and services to support and/or improve geo-positioning on the farm. Scouting robots are also represented in this observation function – and are to be distinguished from those present in the last major function “Act and Apply in the field”.
Organize, Manage and Trade – this function brings together tools for centralizing and organizing data on and around the farm. This includes FMIS (farm management information systems), which structure and display data on plant and animal production, farm machinery and any other connected objects on the farm. We’ve also included data portals, whether open-source or not, that can be mobilized by agricultural players. Last but not least, we’ve included agricultural e-commerce sites (market places), which bring together structured information to enable the ecosystem to buy or sell production results or inputs. This “Organize, Manage and Trade” function may seem a bit of a catch-all, but in reality it brings together all the technologies needed to prepare, structure and manage data.
Advise and Support – this function mainly covers advisory tools (decision support tools come to mind, but the range is actually much wider, with tools for identification [of plants, diseases, animals, etc.] or assistance with machine settings, for example). Also included in this main function are training tools (still not very present on the market) and a category known as “Swiss Army Knives and Gadgets” (no disparagement intended), which often free up the farmer’s mental workload. A “Services” category has been added for requests and support that go beyond simple digital technology. Keep in mind that this category is more about companies than tools as such.
Exchanging, Sharing and Collaborating – this function covers tools that are concerned with data exchange and its derivatives (traceability, telemetry, exchange security, etc.), as well as collaborative technologies that enable users to work in teams, exchange information between peers, and meet up in groups.
Act and Apply in the field – a function that highlights digital technologies for acting in the field, once a decision to take action has been made. Unsurprisingly, we find here the vast majority of robotic technologies (as distinct from those dedicated solely to surveillance, which are present in the first major function), but also actuators of various kinds, mainly positioned to complement data collection by tools referenced in the “Observe and Measure” function. We have restricted ourselves here to digital tools up to the farm gate. In other words, we have not considered all the tools used in the value chain downstream of production (agri-food industry, distributors, consumers, etc.).
Our classification of digital tools offers a first classification so that everyone can visualize the digital ecosystem right up to the farm gate as a whole. Please understand that classification is complicated by the number and diversity of existing tools. Classifying a few dozen tools may seem reasonably accessible. But when we try to classify several thousand (the directory of digital tools for farmers currently counts over 1,500), the organization becomes far more complex.
Multiple forms of agricultural transition and coexisting agricultural models
In their work, Gaël Plumecocq and colleagues highlight the diversity of existing agricultural models, each based on different foundations and value systems. This diversity is embodied in a multiplicity of practices and supported by a variety of institutions, organizations and infrastructures (Plumecocq et al., 2018).
Figure 4 shows the so-called conventional model of agriculture (bottom left of the graph) in the sense that it corresponds to a mode of organization of the Western economy based on an industrial and capital-intensive system.
Figure 4 considers several ways of greening the conventional model, including two main ones with various ramifications. The first, more incremental approach, focuses on the means (technologies, practices – type 2 models in the figure) used to meet ecological challenges, and which, according to the authors, avoids any discussion of the very foundation or definition of the food system. A second, more radical and transformational path (type 3 models on the figure) involves a break with the values, modes of production and organization and close relationship with nature of the conventional model. In this voice of type 3 models, it should be understood that type 2 models are not called into question because they are not productive or cannot improve farm profitability, but because they do not support the value system that makes diversified-globalized (3a), relocalized (3b) and integrated (3c) landscape models coherent.
On a rather similar theme, Maywa De Wit considers that calls for compatibility between new genomic technologies (CRISPR technology – Plumecocq’s model 2b) and agroecology have almost always focused on technical aspects, because examining the epistemic and structural compatibility between digital and agroecological technologies would, in her view, reveal blockages and losses that techniques alone could not destabilize or repair (De Wit et al., 2021).
Figure 4. Main agricultural models (from 1 to 3b in blue) with farming systems identified according to their varying degree of ecosystem service use in relation to exogenous anthropogenic inputs (Y axis) and connected to globalized food systems or local dynamics (X axis). Emblematic examples are shown in grey. The number 1 corresponds to the current conventional agricultural model). The main alternative agricultural models have been grouped into two types of alternatives to reflect the paradigm shift between input-based (type 2) and biodiversity-based (type 3) agricultural systems. Sub-models a, b and c primarily reflect the relationships between agricultural systems, globalized food systems and local dynamics. (CA: conservation agriculture; FS: farming system; ICLS: integrated crop-livestock systems). Source: Plumecocq et al, (2018).
The study of value systems enables us to better understand the drivers of change in socio-technical systems and also to analyze the coexistence and coevolution of multiple models of agriculture. For Gaël Plumecocq and his colleagues, all these models seek both to legitimize themselves and to disqualify the others to a greater or lesser extent (and thus to respond to the criticisms sent out by the other models), which would help to explain the degree of stability and coherence of these different visions of the agricultural world when we try to take a step back. Basically, each needs a little of the others to exist…
The question of the coexistence of these agricultural models remains fundamental. We can indeed ask ourselves whether these models can coexist, or whether the presence of one or more models (dominant or not) prevents the others from existing (in which case we would be in a form of lock-in). Hybridization would nevertheless be possible, and would indeed be a kind of third proposal for a transition path (Plumecocq et al., 2018). Niche (or less dominant) agricultures would in fact play a part in making dominant models evolve because their existence, which is also legitimized in the field (or by a fringe of people willing to pay for this model), would push dominant models to commit to evolutions in practices or more radical changes.
As agricultural practices and perspectives are situated more in a continuum than in a binary schema, we might think that the agroecological transition of agricultural systems could thus be based on a combination of elements from different models. This complementarity opens the way to a new conceptualization of change in which models co-evolve based on arrangements of varying stability and duration (Plumecocq et al., 2018). It is therefore important for agricultural policies to ensure that marginal models are not prevented from emerging or developing, not only to help other, more established models evolve, but also because at some point we will have to come to a profound rethinking of the way we approach the food system.
There’s no need to reject digital technologies by default
Insofar as a technology in support of agroecology can be defined by its contribution to informing or controlling the processes that underpin the principles of agroecology (Caquet et al., 2019), we might question systematically rejecting any form of technological innovation.
It cannot be denied that there is a rather fierce opposition to the digital ecosystem, but this tension actually seems to be turned more towards a particular – but predominant – definition of digital technologies. Criticism is often levelled at a conception of the digital ecosystem that tends to be prescriptive, investment-intensive, power-concentrating and production-standardizing (Schnebelin et al., 2022). In light of the definition of agroecology we shared above, this format of digitizing agriculture would be particularly rejected in that it would create new forms of exclusion, standardization, lock-in, and dependency.
To speak of digital technology in the broad and singular sense invisibilizes the diversity of digital tools and fails to take into account the reality of the technologies involved (I refer you above to the categorization of digital tools). The umbrella term “digital agriculture”, apart from the fact that it in no way represents a given agricultural production system, masks the reality of the technological maturity and deployment of each technology taken independently. By way of example, the adoption of so-called “precision farming” technologies, which indeed seems to be in the majority on larger, industrial farms, says nothing about the adoption of another variety of more sober, minimalist digital technologies. And it is these first technologies that are mainly studied by the human and social sciences (Bellon Maurel et al., 2022).
There are many examples – many of them theoretical – of cohabitation around digital tools at the service of agroecology (we’ll come back to them later in this dossier). Hybridization seems possible, particularly if digital technologies can be appropriated and (re-)adapted by users in the field, and if they are used to support biology and enhance the heterogeneity of environments encountered at different scales. By way of example, the exchange and sharing of formalized knowledge, enabled by digital tools, seems to be an interesting option.
And structures such as Atelier Paysan (French structure), a fervent advocate of self-repair and the appropriability of technologies, give courses in Arduino (an open-source prototyping board) and develop crop planning software (Qrop software). Does buying a computer board for a farm machine preclude de facto membership of an activist collective like Atelier Paysan?
Without considering agroecological and digital trajectories mutually exclusive, the ability of digital and agri-equipment to specifically carry an agroecological farming model remains a gamble, as these technological developments can also reinforce further industrialization of agriculture (Caquet et al., 2019). Eleonore Schnebelin and her research team, for example, highlight examples of the conventionalization of organic farming, with digital tools used to facilitate the implementation of organic modes of production on relatively industrialized farms (Schnebelin et al., 2021). Basically, a form of continuum with certain conventionalized organic farming players who might have a conventional vision of digitization. But more generally, we could think of digital technologies used to improve input use efficiency (in reference to level 1 of the Gliessman scale) in such a way as to extend existing farming itineraries and not call them into question. These structures in place certainly contribute to an increase in the size of farms, but the underlying question is in reality whether we can just as easily deploy agroecology when we are in charge of 100, 1000 or 10,000 hectares.
Could the milking robot be France’s first agroecological robot? While this gentle provocation may make some people howl, I’d like to take the liberty of throwing this first paving stone into the pond. For several years now, the milking robot has undoubtedly made it possible to maintain farming systems (family or otherwise) that would otherwise have given up animal production. Robots make work less physically demanding, even when compared with highly modernized milking parlors. To a certain extent, they can also motivate young people to set up in business and/or attract employees to the farms and hire labor (because otherwise the work would have been considered too demanding).
On the other hand, the compatibility of a milking robot with an agroecological itinerary may be questioned, in the sense that current robots limit the number of times cows are taken out to pasture. Some organic farmers have milking robots, but it has to be said that these robots tend to reduce the cows’ ability to eat grass. One way of getting the cows to the robot is by feeding them concentrates near the robot. Primholsteins, an extremely productive and over-selected dairy breed, have difficulty walking and will therefore benefit more from a barn-based system.
And farm buildings and plots of land are not always ideally organized. Highly fragmented parcels of land, with plots far apart, make it difficult to combine a milking robot with grazing. What’s more, highly (over)equipped farms are also difficult to transfer because the assets to be bought out are too large. And the return on investment, including the cost of servicing, maintaining and servicing the milking robot, is sometimes lower than originally imagined.
All this to say that we need to be able to bring a little nuance to the debate.
What are the criticisms of digital technologies?
Digital players reclaim agroecology definitions
Agroecology is a polysemous term. In spite of the broad meanings of agroecology, it is clear that the definitions proposed by different agricultural organizations, institutions and countries are not always consistent with each other, and in fact reflect the concerns, interpretations, values and priorities of each. These definitions even differ in some agricultural transition scenarios.
For example, when it comes to farming practices alone, the production systems considered to be remotely close to agroecology are extremely varied in the mouths of those involved in the field (organic farming, regenerative farming, permaculture, mixed farming, low-input systems, soil conservation agriculture, etc.). However, these include farming systems that break with the classic definition of conventional farming systems.
Is integrating precision enough to make a system agroecological or does it only reinforce the model without questioning it? Most digital technologies approach agriculture from the angle of optimizing and/or improving the efficiency of resource use. If we take the schematic representations of agroecology (Figure 2), digital technologies are mainly concerned with level 1 of the Gliessman scale (and to a lesser extent with level 2), the FAO’s efficiency pillar and the HLPE’s input reduction principle, i.e. a small part of all agroecological principles.
Before going any further, we could discuss this relatively broad concept of efficiency and optimization, which links inputs (water, seeds, plant protection products, etc.) with outputs (biomass), so that there are in fact many different efficiencies that could be explored. An increase in one efficiency ratio, such as yield per hectare or per unit of labour, has often been associated with a reduction in other efficiency ratios, such as yield per unit of fossil fuel or per unit of biodiversity (Wezel et al., 2020).
This “weak” understanding of agroecology, implying that it is part of a continuity of current agricultural systems without calling into question the model of socio-economic organization of the sectors (specialization, concentration, subcontracting, etc.), is seen by some as a reappropriation of the concept of agroecology by technologists and key players in the agricultural sector, having first emptied the basic concept of its substance. This distortion of the concept of agroecology is even being taken up, even conventionalized, by the dominant players in the sector, as we see the development of industrialized organic farming on large farms (with the potential use of agricultural contractors). Digital technology could thus be seen as a gas pedal of this conventionalization, to the detriment of more radical organic farming, carried out by small farms that don’t necessarily use the same inputs and that could have other strategies than just optimization. And we could end up with a divide between “classic organic farmers” and their fellow “digital organic farmers” (IFOAM, 2020).
The question is not so much whether digital technology can serve agroecology but whether, in a race against time, it cannot more easily or more quickly promote systems that go against agroecology. The reference is therefore not so much the current situation as the comparison between different futures that diverge from the current situation.
This conventionalization of organic farming should not, however, obscure the process of “silent agroecology”, a farmer-led movement for agricultural change that is largely unknown and misunderstood (Lucas, 2021).
For digital critics, Agritech companies would seek to remedy their socio-ecological impacts by adopting a model of sustainable agricultural intensification with agroecological nuances (IFOAM, 2020). By promoting a soft agroecological trajectory – some present it as a symbolic greening, even as a junk agroecology (“Junk Agroecology”, let’s not mince words) – the sector’s industries could maintain their business concepts and way of working. To “change everything so that nothing changes”, Agritech companies would have found in agroecology a range of extremely useful solutions that they would have decided to selectively integrate into their industrial model. Agroecology would thus be at a crossroads today, facing a major struggle for its eventual co-optation by the mainstream and subordination to conventional agriculture (Altieri et al., 2017).
Renewing the spirit of productivism and capitalism in the service of an “ecologized productivism” (with the capacity of the digital to quantify, rationalize, or even objectify the environment), for some we would have to distance ourselves from these digital technologies at the risk of encouraging ecological dystopias. Basically, digital technology would be the bearer of a value proposition based on greater resource efficiency, and it wouldn’t go much further than that. Digital tools would focus on optimizing inputs and not on their ability to facilitate agroecological farming practices.
This weak agroecology, which takes up many of the ideas of “low-input” agriculture, is opposed to “strong” agroecology, defined by its objective of coherence and sustainability through the use of biological processes (Gascuel et al., 2022). Strong agroecology moves away from incremental levers towards much more transformational ones (see Gliessman scale), calling for a renovation of the food system as a whole.
Apparent contradictions between digital technologies and agroecology
Current digitization shows several forms of opposition to an agroecological transformation of agriculture, whether in technical terms (e.g. unsuitability for organic inputs), objectives (maximization of yield versus maximization of ecosystem services), reasoning (reasoning on the scale of an annual crop versus multi-annual reasoning on the scale of a set of plots), temporal dynamics, but also political and social issues (limiting dependencies, favoring local economic players…). (Schnebelin, 2024).
Agricultural players have a different sensitivity and perception of digitization. In her field survey, Eleonore Schnebelin shows that the main differences between players in the organic and conventional sectors lie in the direction each expects digitization to take (Schnebelin et al., 2021). For example, organic farmers tend to expect digital technologies to help them design and analyze their production systems in a systemic way, and to support their experimentation practices. These same players see collaborative digital tools (social networks and the like) as an engine for emancipation and knowledge sharing. On the other hand, the more conventional players place greater emphasis on the creation of agricultural information through traceability technologies, in line with a value chain logic and the need for trust throughout the value chain. These players also see the technology as relatively neutral and suitable for all farming systems.
Some see the lack of attention paid to automation as an obstacle to the amplification of agroecology (e.g. Bellon Maurel & Huyghe, 2017), believing that where labor is a limiting factor, automation offers the opportunity to implement agroecological practices in new contexts and on larger scales. Painting a picture of an automated ecological farming utopia, Daum (2021) imagines, for example, that fleets of robots working 24/7 will enable farmers to adopt agroecological farming methods where high labor demand would otherwise be a constraint (Ditzler and Driessen, 2021) (Figure 5).
For others, particularly in small-scale farming systems and within less mechanized contexts, the manual labor requirements of doing agroecology are seen instead as opportunities to foster meaningful livelihoods and community involvement, connecting humans both to the land and to each other. From this perspective, there is concern that digital technologies could undermine the intrinsic value of the farming profession, displace workers or lock farmers into disadvantageous power asymmetries (Ditzler and Driessen, 2021).
Figure 5. Robots in ecological utopia. Illustration by Natalis Lorenz. Source: Daum et al. (2021).
The very fact that we are currently focusing only on incremental changes in practices would de facto prevent digital tools from being compatible with fairly fundamental principles of agroecology such as “Diversity”, “Synergy”, or even “Resilience”. Certain tools, perhaps the most technological and the least appropriable by farmers (because they are difficult to repair, for example), distance these same tools from the concepts of “responsible governance”. These technologies favour an analytical vision of systems, far removed from the systemic vision defended by agroecology (Hostiou et al., 2022).
Digital technologies are not neutral
Agricultural technologies are not neutral. To the argument of the knife that is usually brandished, with the implication that everything depends on how the knife is used, we must remember that every technology is part of an already well-established socio-technical system. We can’t and shouldn’t say much about technological forms when they are extracted from their networks, practices, affects and discourses. Technological devices are never simply independent objects; they are always relational in their essence. It’s time to stop asking what these technologies are, and instead focus on what they do, what they promote, what they imply and in which system they fit (The Shift Project, 2024).
The players involved in the technological ecosystem in agriculture need to take responsibility and bear in mind that they all have a role, at one time or another, in the landing of technologies on the agricultural field and, by extension, on the associated consequences and impacts, whether positive or negative (The Shift Project, 2024). Some authors even go so far as to speak of a kind of culture of disengagement among engineers working on agricultural technologies (Sullivan, 2023), because they do not question the values, perceptions or assumptions underlying the development of the technologies they support.
The ways of thinking associated with technologies continue to influence the way we ask questions and provide answers to problems. In other words, it’s not the technologies themselves that determine certain effects. Rather, it is the social relations and assumptions associated with these technologies that play an important role in structuring our thinking and actions (Avaria, 2020).
Agroecology and digital technologies are said to represent different ways of knowing and being in the world, giving rise to particular, often orthogonal, assessments of the nature of the underlying problems and the way in which social change occurs.
Digital tools, by promoting a desire to transform agricultural systems, focus on a certain form of agriculture (arable farming, livestock rearing in buildings, value-added crops). It has to be said, however, that some areas of agriculture have been left out in the cold by agricultural technologies, in the sense that they are not as well equipped as others (mixed farming, legumes, organic farming, low-input systems, etc.), for financial or regulatory reasons, or because of the need to organize supply chains or markets. We must therefore face up to the risk of a single digitization model that favors only one type of greening path, where we would sell only one tool for one form of agriculture.
The majority of robots developed for field crops are designed to operate within monospecific stands and to increase the efficiency of existing practices (Fountas et al., 2020). The main characteristics of robots that could make them particularly suitable for applications in diverse environments – their potential for lightness, modularity or multifunctionality, high mobility, autonomy and learning – are not used to embrace heterogeneity, but are more often called upon to further homogenize production environments (Ditzler, 2021). In this configuration of connected agriculture, there is a risk of conceiving of agroecological agriculture as nothing more than techno-centric precision farming, of transforming biological regulation and knowledge networks into an agriculture equipped entirely by information and communication technologies and seeking optimization at all costs (Leveau et al., 2019).
By focusing on what could be rather than what actually exists, abstraction strips away history and systems, leaving only tools. We cannot simply be satisfied with a theoretical approach to the use of digital tools. Our understanding of the use of digital technologies to move towards an agroecological transition is still at an embryonic stage (Rozenstein et al., 2023). Theoretically, digital tools can initiate and/or accompany profound transformations in agricultural systems – by supporting a high profusion of widely shared and disseminated knowledge, by facilitating relationships between the different food system chains. In practice, however, these tools remain largely focused on optimization.
This constant quest for efficiency traps us in a cult of economic performance. And this performance, in the words of Olivier Hamant, is waging war on life.
Digital technologies can create positions of dominance and dependence.
Our collective commitment to a particular technology (or technologies), sometimes at a relatively low level of maturity, sets us on a trajectory from which we cannot easily deviate (notions of technological lock-in – and path dependency). By making certain agricultural itineraries technology-dependent, new dependencies (on a manufacturer, a tool supplier, a breakdown manager or even a collector/storage organization) are also likely to emerge (The Shift Project, 2024). Moreover, these dependency phenomena may not be limited to farmers, but also apply to researchers (for example, depending on the scientific discipline) or policymakers (for example, depending on current political discourse) [Van Hulst et al., 2020].
Figure 2 on the levels, elements and principles of agroecology invites us to consider issues often ignored by digital technologies. These include power imbalances in food systems, and how knowledge and data are generated, transferred and by whom they are actually owned. This involves issues of responsible governance, human and social values, and the co-construction of knowledge and know-how.
Digital tools that depend on the control and ownership of data outside of exploitation, or that devalue the various sources of knowledge in data analysis, risk reinforcing the commodification and privatization of skills and knowledge, which is incompatible with a paradigm that seeks to improve participation, transparency and equity within food systems (IFOAM, 2020).
In this context, the ability of digitization to be compatible with agroecology would largely depend on the main agricultural and Agritech players. If these players have an ideal image of agriculture that corresponds to intensively managed conventional farms, digitization is unlikely to be compatible with an agroecological approach. Some dominant players are not necessarily opposed to changing farming practices, and would even be willing to add data around organic and agroecological itineraries to their empire as additional data points and potential revenue streams. But it will be necessary to analyze and evaluate the underlying intentions.
What criticisms are levelled at agroecology?
After the first volley of criticism levelled at digital technologies, it would be a little too easy to overlook the criticisms that could be levelled at agroecology.
Agroecological systems are sorely lacking in technical references. Perhaps because their definitions are still too vague or not shared, these systems lack assessable criteria. It is therefore difficult to express quantitatively and qualitatively the state of these agroecological systems and to follow their trajectories (Dumont, 2021). The effects, whether combined or not, of climate change, markets, farming practices or agricultural policies on the quantity and quality of agricultural production (at different spatial scales) are not really measured (Ewert et al., 2023). This creates new difficulties for reproducing these agroecological systems and comparing them with each other, since we don’t always know how we got here in the first place.
Generally speaking, it seems that the majority of publications focus on studying agroecological farms with a fairly narrow focus. The literature focuses mainly on specific alternative practices, or on a given spatial scale (territory, plot, etc.), whereas we have seen that agroecology attempts, on the contrary, to bridge the gap between the different food spaces we know, from the plot to the food system as a whole. In particular, research is focused on the plot level. This research does not take sufficient account of the systemic changes that occur at farm level, nor of the complexity of the interactions and trade-offs that farmers have to manage within their farm and between the farm and its environment (Prost et al., 2023).
To fully embrace a systems approach and holistic vision, work in agroecology needs to include much more interdisciplinary work and take into account multiple entry points and transition trajectories, including social, cultural, political and economic issues in particular. The fundamental principle of knowledge co-creation requires a very different approach to research: one that places farmers and stakeholders at the center of defining research questions and developing solutions, alongside scientists (Wezel et al., 2020).
With regard to the focus on alternative agronomic practices, the subject is all the more damaging as this can lead to innovative agroecological farms being overlooked, where newly developed practices do not easily fit into existing evaluation frameworks (Dumont, 2021). Focusing only on alternative practices also obscures the effects of interactions between multiple agroecological practices.
Agroecology also suffers from an outdated image – in the sense of a step backwards or counter-progress – and an imaginary of work linked to arduous and painful tasks. But we have to accept that agroecology doesn’t necessarily guarantee better working conditions. Time savings may not be achieved in all situations, as they depend on the initial state of the farm’s equipment and the effectiveness of preventive measures taken (to enable the system to self-balance via its own regulations), which may then allow curative action to be dispensed with.
By way of example, some farmers have reported an increase in their working time by switching from chemical to mechanical weeding, and have expressed that herd monitoring was lengthened after implementing new, more agroecological grazing practices (Hostiou et al., 2023). On the other hand, other farmers would have seen their working time reduced following the cessation of systematic animal treatments or reorientation towards grassland systems (less time spent on harvesting, simplification of feed distribution in winter as it was no longer necessary to mix different types of feed).
The fact remains that taking greater heterogeneity and diversity into account in one’s agricultural production system will, almost by definition, lead to a more complex system, which may lead to an additional amount of work (which could quite easily be supported by increasing the number of people working in the fields, or by delegating the responsibility for doing so to technology). Nevertheless, as discussed above, the lack of objective criteria for agroecological systems leaves some doubt as to the level of on-farm work that will be required in these systems.
Agroecology could also be criticized for not pushing much for the development of tools, whereas some proponents of the agroecological discipline consider this to be a barrier to commitment, all the more so if working conditions are deemed difficult. The challenge of translating the agronomic, ecological and social requirements of agroecological systems into farm tool designs (not to mention automated machinery) is often peripheral and secondary to the design of the farming system itself (Ditzler, et al., 2021; Salembier, 2020). Can we really imagine an agroecological system so resilient that it will need no technology at all, and can be self-sufficient in all autonomy? Perhaps it makes more sense to imagine risk-taking covered by an insurance system. In this context, digital technology would be there to provide proof that good practices have been deployed, enabling insurance coverage to be applied (https://www.aspexit.com/lassurance-climatique-agricole-en-pleine-reforme/).
These fears linked to the use of – sometimes digital – tools can be fully justified, as we discussed in the previous section (appropriation of data and farmers’ knowledge, non-adaptability to agroecological itineraries, etc.). The question of what automation and digital technologies might look like, be like or do in transition trajectories remains largely unexplored (Ditzler, 2021). Regularly repeated examples (e.g. Atelier Paysan) show that we have difficulty going much further than covering niche situations with equipment adapted to particular situations. The anti-technology argument can also be dangerous to handle, whatever the production context, particularly when yield gaps are significant (Cote et al., 2022).
How can we operationalize the transition to diversified cropping systems? Without going so far as to talk about scaling up (we’ll come back to this later), what does commitment to an agroecological trajectory mean in concrete terms? In their interviews with agricultural actors, Ditzler and his team testify that none of the discussions in which they participated focused on how agroecological values and approaches could or should be put into practice when the participants in their survey were talking about agroecology. In fact, their group reportedly didn’t talk about work at all, and never discussed the physical implementation of agroecological practices (Ditzler et al., 2021). Some would even go so far as to say that the adoption of agroecological itineraries has always been slow because agroecology could only offer general principles (e.g. crop diversification), not least because there was a lack of technical references (Duff et al., 2022).
While the desire to evaluate an agroecological system in its entirety is laudable, the reality is that it will most certainly be necessary to simplify the evaluation of these systems to make them operational. Some evaluation grids are so imposing and multi-criteria that it’s hard to imagine how they can be completed, even if we could imagine digital technologies supporting the feedback process. If the agroecological protocols are too heavy and don’t enable concrete decisions to be taken in the field, there is relatively little chance that they can be deployed in the field. It’s important to understand that the implementation of agroecological systems will also require the training of employees and seasonal workers – and consequently, asking them to follow protocols as well. An interesting way of thinking about this would be to turn the problem on its head and ask what it would be reasonable to ask a farmer to collect and follow, so that the agricultural players can work on it. While ensuring that this information is robust and of high quality, of course.
Taking a position on transition trajectories is particularly challenging, and all the more so when the change is complicated. The dynamics of change at work are complex and multifactorial (objectives, values, work organization or the farmer’s professional network, etc.), making it all the more difficult to scale up agroecological systems. More on this later in this issue.
What kind of cohabitation can we imagine between digital and agroecological trajectories?
The intention of this section is not to take a frontal position on the relevance of digital tools to support an agroecological transition. Rather, the aim here is to attempt to establish a link between the potential advantages of digital tools, which we consider to be facilitating technologies (in the sense of facilitating agricultural practices), and the limitations identified around the development of agroecological systems.
Here, we take the major functions of digital technologies we identified in the introduction to this blog entry (those we use in our Wiki Agri Tech platform) and apply them to agroecological systems.
This section proposes case studies – often theoretical because they have not yet been put into practice – in an attempt to visualize the compatibility between agroecological and digital trajectories. I’d like to draw your attention once again to the diversity of digital tools and the need to make a clear distinction between different technologies. A social network is not comparable to a measuring instrument, a plot sensor or a robotic tool.
New case studies will have to be proposed by sourcing technological innovations (or the combination of innovation systems) directly from farmers, supporting their own problem-solving processes. Insofar as technologies will have to be adapted to local contexts and conditions, it will be interesting to identify systems in transition, in the form of an innovation tracker, in which agricultural technologies are used at a regular pace and integrated into the course of the farm (Salembier, 2021). This innovation-tracking format makes it possible to flush out atypical practices, describe and analyze the underlying logics, evaluate practices and, ultimately, offer support for the (co-)design of innovation systems (Paget et al., 2022). More broadly, this work calls for a search for innovations in agroecological systems, innovations that also encompass technological innovations (Ewert et al., 2023).
In this section, I focus on one form of innovation among others: technological innovations. Other approaches to innovation, whether agronomic (relay-cropping, direct seeding under cover, associated crops and service plants, etc.) or organizational (direct supply and marketing circuits, pooling of tools via collective organizations, etc.), are perfectly capable of facilitating agroecological trajectories.
We must bear in mind that these innovations are entirely compatible with each other, and that it is above all systems of innovations, combining different techniques and modes of organization, that will be able to respond both to the different challenges and to the diversity of specific local situations. Technologies, with their highly varied compositions, can play a part in supporting these other forms of non-technological innovation. These coupled innovations (couplings between different forms of innovation and couplings at several levels of food systems) can help to overcome current system constraints or create new opportunities for innovation.
These innovation systems, or coupled innovations (Salembier et al., 2020), enable us to think about systems in which the design process is no longer aimed at achieving a single end goal (how do we automate a system?) but rather a feedback process driven by the underlying ethos of the desired system (how do we facilitate the processes and outcomes we want?) for which the right tool may or may not be a robot, or may involve combinations of humans, manual tools and forms of automation (Ditzler et al., 2021). The case studies presented here are drawn from a variety of sources: interviews I may have conducted, blog entries I may have written in the past, or more simply from the literature (scientific, technical report, etc.).
Objectivize & Measure agroecological systems
Envisioning and considering the complexity of agroecological systems may call for generating a significant amount of data that digital sensors and instrumentation systems – both offline and machine-embedded, can chaperone (Caquet et al., 2019, Inrae-Inria, 2022). We need to :
- better understand the dynamics of biotic/abiotic interactions and biological regulation in agroecosystems,
- make visible and integrate the reality and mechanisms that trigger the stimulation of internal defenses
- evaluate the occupation of ecological niches and their preservation,
- characterize multiple stresses (mixed infections or combined biotic and abiotic stresses),
- better understand and assess soil conditions. Note that there is no reliable high-throughput method to measure roots non-invasively or to quantify root distribution in the field (Storm et al., 2024), or to
- change scale for both experimentation and observation.
These data can be used to create models of the complex mechanisms specific to agroecology, to capture the interactions between genetics, the environment and agricultural practices, because agroecological systems are difficult to model using deterministic approaches (Bellon-Maurel et al., 2022). Generally speaking, the acquisition of such data, via on-farm high-throughput phenotyping systems under a variety of conditions, could help to overcome the lack of technical references and reference points in agroecology. Multi-source data, cross-referenced and combined, could help to gain perspective on the phenomena involved.
Digital tools can thus be seen as a way of objectifying agroecology, making it easier to describe, quantify and put into practice (with a view to supporting the transition). Farmers in agroecological systems, for example, can take advantage of on-farm experiments (OFEs) – with technological monitoring systems (even if these experiments don’t necessarily need digital tools to see the light of day) – to rapidly understand patterns of spatial and temporal variation of production factors in plots, and thus manage them more effectively (Lacoste et al., 2021; Duff et al., 2022). This situated agronomy, in the form of on-farm experimentation, could considerably increase references on agroecological practices. The question then arises as to how to characterize what is being experimentally monitored and how to assess what has changed over time.
This use of measuring instruments thus contributes to a continuous improvement of the production system, and even to reassuring farmers about the practices they have started to implement, in the sense that they can realize that the classic indicators they follow (for example, yield initially), are not necessarily impacted by the actions undertaken. These regular measurements thus promote an extension of practices.
It is important to remember that these field data are supplemented by genomic and biomarker data obtained on analytical platforms (Rogel-Gaillard and Sainte-Marie, 2024). These data can be generated at various scales (single cells, organs and tissues, whole organisms, biological fluids, environmental samples) and apply to both visible and invisible living organisms (micro-organisms and microbial ecosystems such as microbiota). Genomics and metagenomics have undergone considerable development over the last twenty years, thanks in particular to new high-throughput sequencing technologies, the development of bioinformatics, and the miniaturization and parallelization of measurements. For example, there are “DNA chips” capable of detecting tens of thousands of DNA variations, and technologies that produce data characterizing gene expression at RNA (transcriptomics), protein (proteomics) and metabolite (metabolomics) levels. When these multi-omics data are compared with measurements obtained in the field for the same individuals, they enable the identification of biomarkers (DNA variations, epigenetic marks, genes and gene networks) used for diagnosis, genetic and genomic prediction (Rogel-Gaillard and Sainte-Marie, 2024).
Agroecological processes are inherently dynamic: they involve flows of matter and energy and changes of state (Caquet et al., 2020). It is therefore the very unfolding of the ecological process that must be emphasized, rather than an absolute state: are we storing or destocking carbon? Can the potential of regulations be mobilized quickly? Digital technologies and measuring instruments could thus be used to monitor temporal trends at varying degrees of resolution, with an emphasis on relative measurements, to understand the deltas or variations in value between two states of a system.
While field data entry is already tedious in conventional systems, it becomes even more time-consuming the more diversified and complex the agricultural systems monitored. Automatic data entry therefore appears to be a way of meeting part of the information needs of agroecology.
However, this data is not always accessible (open source or open data), understandable, legible or sufficiently well structured. They are sometimes pre-interpreted through the prism of the person who acquired them, and the transparency of the acquisition and pre-processing assumptions is not always explicit. Some databases are theoretically accessible – they exist – but nobody actually uses them. Consolidating existing databases and making them accessible is really something to consider closely, so as not to start from scratch every time.
Example: Landscape and agroecological infrastructure monitoring
The landscape is a major scale for agroecology, due to its structuring elements (hedges, woodlands, flower strips, buffer zones, roadsides, etc.) and the spatio-temporal organization of crop rotations, grasslands and cropping and livestock practices (i.e., the “landscape of practices”). Knowledge is needed on the composition, spatial organization and management types of multifunctional and resilient landscapes, as well as on the coexistence and complementarity of production systems on a territory (Gascuel et al., 2022).
Satellite imagery remains a preferred tool for observing large spatial areas. Even if the images from the Sentinel-2 constellation, favored for agriculture, do not necessarily have the spatial resolution to track ecological infrastructures in detail – they can provide some initial answers. Other, much more spatially resolved images (but with a longer revisit time) could be used for this large-scale monitoring, on the assumption that some of these ecological infrastructures evolve over several weeks, months or years. Note that Europe is currently working on the implementation of a processing chain (SEN4CAP: https://www.esa-sen4cap.org/) based on Sentinel-2 data to improve the efficiency and traceability of CAP controls. To support the agroecological transition, this processing chain will need to focus on the principles and practices underpinning ecology, and not just on the silly and nasty controls of current agricultural practices.
Some players in the remote sensing field are imagining very fine temporal monitoring of these ecological infrastructures with the next generations of satellites, or at least future constellations, with, for example, the desire to go as far as tracking the flowering of hedges to better characterize the functioning of the environment, or even to go as far as limiting interventions in the field by considering the state of these ecological infrastructures. But we’re not there yet…
Every year, France has a graphical parcel register (RPG) that is the envy of all countries. The RPG, which is used as a reference for CAP (Common Agricultural Policy) subsidies, covers a large part of France’s land parcels and crop islands, and references the crops grown. The RPG has evolved considerably since its creation in 2006. Since 2015, when IGN took over the RPG data from the Agence de Service et de Paiement (ASP), IGN has simplified the RPG output data and offered RPG data at the crop block scale, this time only providing the majority crop. Even if data is generally available 1 or 2 years late, the graphical parcel register remains an unrivalled tool for monitoring the dynamics of agricultural land use, and assessing in particular how crop rotations are evolving across France. For many years, INRAE (French Research Institute) has been working on the development of a tool that takes advantage of the RPG for this purpose: RPG Explorer. Several companies have also taken up the subject (e.g. Kermap).
A number of players are also working on making land-use maps available in France (to compensate for the fact that the RPG often arrives one or two years after the current year). This is particularly true of the OSO maps generated by the Théia cluster. OSO maps are based on high-resolution, multi-temporal optical image series (Sentinel-2, but also SPOT-6/7 and even Pleiades in the future), and auxiliary reference data to calibrate methods and validate maps.
All regions have land cover data, more or less interoperable with the Corine Land Cover classification. These data can be accessed at https://geo.data.gouv.fr/fr/ (type ‘Land Cover’ in the search bar, for example) (Figure 3), or on a whole host of regional sites such as OpenIG, CrigePACA, Pigma and GeoBretagne. These data are often presented at the cadastral parcel scale. Some operators of territorial coherence schemes or public establishments for inter-municipal cooperation have developed their own land-use layer to manage their territory. These are often more precise than regional or departmental land use maps. Please note, however, that the names of these information sources are not the same everywhere. To go further : https://www.aspexit.com/ou-recuperer-des-sources-de-donnees-en-agriculture/
Example: Monitoring grazed resources on agro-pastoral farms
The farming systems we want to see must be supported, not least because some can be time-consuming to manage. For example, some agro-pastoral farms with controlled-release livestock were very early adopters of GPS tracking to find out where their animals were located, especially in hilly areas.
GPS tracking technologies can be used in conjunction with remote sensing technologies (satellite, airplane, drone) to cross-reference the passage of animals in agro-pastoral systems with geomatic indicators of grazed resources. This cross-referencing could, for example, be used to discriminate between certain patches of vegetation to be protected and others where grazing pressure could be increased. These technologies also make it possible to label grass milk or pasture milk produced by cows grazing in meadows, rather than cows eating silage and soya in stalls.
Grass meters are also available to measure grass cover on soils and more broadly support changes in livestock management and rotational grazing practices. While this data is relevant at the farmer level, aggregating data from connected grass meters (or a network of grass meters) can reveal interesting spatio-temporal landscape dynamics to consider at the local policy level (in the sense that a global assessment is important for policy makers). Behind this logic, which may seem like a win-win situation, the farmer cannot be the only one to bear the cost of capturing this vegetation cover data since we can clearly see here that the entire territory also benefits from it. Multi-scale agroecology therefore requires thinking about data sharing models and consultation models where everyone benefits (we will talk about this again). By ensuring that this data sharing is not used simply as an additional source of control over farmer practices.
Example: Towards environmental biomonitoring supported by digital tools
As a flagship species (attracting public support), an umbrella species (whose conservation needs incidentally protect other species), an indicator species (sensitive to change/degradation) and a keystone species (whose ecological impact is disproportionate to its abundance), pollinating insects, including the honey bee, offer an unparalleled showcase of living things.
As an important link in the ecosystem, the bee – as a sentinel of the environment or marker of the living world – will represent a variety of impacts on fauna, flora and humans, which we can try to approach by translating what the bee experiences in its daily life, in the difficulties of living and adapting to its environment. Taking a step back from the search envelope covered by bees (a few km in diameter around a hive), in reality several hundred million plants are foraged by the bees of a single hive, giving access to a rather impressive quantity of information on the state of the surrounding environment.
Data gathered on a much larger scale than that of an apiary could be of interest in developing a more detailed understanding of honey production by type of territory and honeyflow period, or in helping to set up regional and national honey production observatories. In fact, we should stop talking about biodiversity on a site-by-site basis, and instead look at how the site fits into its territory, to try and get a slightly more macro view of the subject at hand.
In addition to bees, there are in fact many different matrices to explore: pollen, honey, nectar, wax – all of which provide a wealth of information on the context surrounding the apiary. Pollen is particularly interesting because it can be collected in traps even before the bee has entered the hive (a kind of raw information at your disposal). Pollen binds pollutants and also contains information about the plants foraged by the bees, even in trace form.
Digital tools could be used to support broader environmental biomonitoring, even if the fine-scale monitoring of insects and pollinator decline is particularly complex to grasp. Between multifactorial, additive, collinear and confounding issues, digital tools could be useful to navigate and see things more clearly, but reality will remain complex. Digital tools should help us to move away from the prism of the bee as an individual towards broader landscape and territorial approaches, and in particular to monitor sanitary practices in a coordinated way, or to discriminate between territories and the factors influencing the state of health of colonies. But are we prepared to spend money on this, or to accept what it would teach us about the (poor) state of the environment? To go further : https://www.aspexit.com/numerique-apiculture-des-promesses-pour-la-production-apicole-et-la-bio-surveillance-de-lenvironnement/
Example: Large-scale insect health monitoring
The majority of current numerical tools focus on observation rather than prediction, and even these observations are lacking. Modeling will be hard to achieve without the right level of information, good meshing (with good localization, close to the areas where contamination outbreaks are most likely to start) and the ability to react quickly in the event of detection or prediction of the arrival of insect flights. The challenge remains to correctly estimate pest density in the field, without over-sampling. Shared alert systems, with collaborative trapping networks, could make sense. In addition to the financial benefits for growers who would share these trapping networks, they would also enable better monitoring of migratory insect dynamics, for example. When modelling is carried out, it is often based on related climatic data. Auxiliary landscape data (surface area of nearby rapeseed and/or forest, surface area and weed rate, biomass indices of nearby crops, etc.) could also help explain observed dynamics.
If we think of monitoring from an agroecological perspective, we could even give priority to monitoring crop beneficials rather than pests, to ensure that their introduction has worked or that their presence in numbers has been confirmed. It is conceivable that part of biocontrol actions could consist in quantifying predation potential in real time, and coupling observation with the eventual release of resources capable of maintaining predators at high densities. What is currently lacking is the entire chain of equipment needed to carry out a large part of the auxiliary breeding on the farm, assess the needs of the auxiliaries, package and release food resources likely to stabilize the auxiliary population on the plot, covering a period of dearth which would otherwise lead to its death or mass departure from the plot (Caquet et al., 2020).
If equipped, this sanitary surveillance will require the use of combinations of digital tools (we can speak of technological systems). A study by The Shift Project sheds light on the value of digital technologies for dynamic, spatialized monitoring of bio-aggressors on a territorial scale, notably by combining connected traps, participatory approaches, data from connected weather stations or satellite data (The Shift Project, 2024b).
To go further : https://www.aspexit.com/using-digital-technology-to-manage-pests-and-diseases-in-agriculture/
Example: Understanding animals’ perception of the environment in relation to farming practices
How does biodiversity react to agricultural practices? If the example of bees, as sentinels of the environment, could be one way of approaching this indirectly, a more direct approach could be to install sensors directly on certain animals. The intention being to tool a biological component of the environment and see how this component reacts, so as to identify the minimal factors (or minimal conditions) for the biological to thrive.
For example, we might ask ourselves whether a 5-metre-wide plough generates an impassable zone for a toad wishing to reach a pond. If this is the case, perhaps we need to put in place an architecture to facilitate access to the pond. Or simply stop ploughing on the busiest roads. Without going so far as to install miniaturized sensors on pollinating insects or birds (even if this already exists), connected hives and/or nest boxes are able to monitor egg-laying frequencies, reproduction dynamics, or more broadly the state of populations in relation to their environment. Significant drops in the number of pollinating insects measured at the end of the day in a connected hive would suggest a one-off or structural inability of these insects to navigate in their environment (weather, pollution, clumsy practice such as treatment during the day or mowing a very visited plot)
Example: Information on soil condition by analyzing vegetation cover maps
Plant cover, through the development of certain species rather than others, or through the speed of their development, is an excellent indicator of soil condition. However, biases can be significant in the case of cultivated plants, particularly when it comes to fertilization, which attenuates some of the variability in the plot. On the other hand, when sowing cover crops, we regularly find ourselves in more complicated climatic conditions, without fertilization, and the patterns stand out. Crucifers and grasses are more likely to appear in areas with high mineralization, while legumes are more likely to be found in nitrogen-limited soils. Mapping plant cover using aerial vectors (drones, airplanes, etc.) could be a way of rethinking and spatializing PPFs (manuring forecast plans) – in the sense of spatial variability of inputs to plots. Once the patterns have been identified, there’s nothing to prevent us from thinking in the opposite direction and spatially modulating the cover crops introduced into the plots.
However, we must bear in mind that modulation or spatialization carries the risk of ultimately wiping out atypical systems if the aim is to homogenize everything, whereas it could be seen as a way of making the most of observed heterogeneities.
Managing & Organizing agroecological production
Agroecological systems are complex: multiple production workshops, alternative practices, diversity of production on the farm, etc. Digital tools can help to manage these heterogeneous structures, lightening the mental load, reducing the drudgery of work and facilitating the organization of work on the farm.
We need powerful visualization tools. By storing, organizing and reshaping the information provided, in the form of dashboards or data aggregates, these digital technologies can provide a panoramic view of the farm, cross-referencing all the workshops involved. These visualization charts are also useful for comparing with neighboring farms as a source of inspiration and improvement, but also to help project oneself towards farming systems that have succeeded in transforming themselves. Digital tools can thus facilitate the delegation and distribution of tasks, shed light on and simplify the organization of worksites to ensure that these farms, richer in field labor forces, are efficient. All the recurring tasks – such as administrative tasks – that can be automated by digital technologies are free time for farmers to explore new opportunities and improve their system (as long as it’s not just an excuse to get bigger and more fragile).
Management tools dedicated to diversified systems are still lacking. Time-consuming initial settings and long maintenance times make them difficult to use on a large scale. These tools need to remain as flexible as possible, so as not to impose a particular way of thinking on farmers.
Advise & Train on agroecological systems
Agroecological systems, which are complex by nature, need to be approached from a systemic perspective, which current models have difficulty taking into account.
Today’s models, for example, take very little account of biological interactions – such as competition between species, the relationship between above- and below-ground biodiversity, or the relationship between biotic and abiotic processes in landscapes (Gascuel et al., 2022). More generally, these models need to make the transition from cropping systems to agricultural production systems as a whole.
Agroecological transition mobilizes multiple levels of organization (both animal and plant) that interact strongly with each other: molecular and cellular scale, organism scale, physiology and phenotypic expression, plot and farm scale, landscape and community scale mobilizing ecology and dynamics. Spatial and temporal components, which are all the more important as agroecological systems will depend more and more on neighborhood effects or landscape elements, are still given too little consideration and remain a challenge in agroecology. The role of interfaces between cultivated and semi-natural environments (refuge, exchange of organisms between the two habitats, etc.) remains poorly documented, whether for disease management, biological control or pollination (Caquet et al., 2020). The effects of agricultural diversification on the intensity and stability of ecological processes of interest in agroecology are little known at the landscape scale. The effects of concentration and dilution, connectivity and flow regulation are still poorly understood.
Farmers need to be supported in their decision-making, not only because the knowledge and skills to be mobilized in an agroecological transition are extremely varied, but also more profoundly because in the agricultural world, it is often the last proponent of advice who wins or who is right. Weed recognition tools, for example, make sense because to learn to recognize plants, you need to be accompanied by a botanist. And putting botanists’ books (such as the Flore Bonnier or others) in the hands of farmers is quite complicated. In this case, a recognition tool enables farmers to teach themselves how to identify weeds.
By moving to a broader vision of the farm, agroecological systems need to move from a tactical tool logic (vision in reactive mode, for year-round management or in crisis situations) to a strategic tool logic (vision in reflexive mode, for more long-term management). This second, more complex, logic enables us to prioritize and arbitrate actions on a global scale, and to move away from siloed actions in response to a specific problem to thinking in terms of overall resilience and robustness. Farmers will need user-centered decision-support tools, which will have to integrate the farmer’s strategy and available local expert knowledge all the more, as solutions to problems will be spatially located. We could even go so far as to speak of design support tools (rather than decision support tools) that would support farmers’ evaluation, creativity, observation and action (for example, a new wheat management method based on monitoring its nitrogen nutritional status) (Prost et al., 2023).
We could also add that risk and uncertainty modeling, particularly in transitional systems, deserves even more attention. Digital data and local agronomic models, for example, can be used to quantify risk and minimize insurance costs. This notion of risk needs to be properly understood, because if a health risk, however small, is detected by a modeling tool, there’s every reason to believe that farmers will want to treat when they might not have done so if they hadn’t had this information. This knowledge could, moreover, be monetized by adjusting the amount of the insurance policy upwards or downwards (which could be considered an acceptable adjustment to the efforts made not to mobilize the insurance).
The full range of data science tools (spatio-temporal modeling, multi-criteria and multi-performance approaches, constraint programming, fuzzy logic, etc.) fed by more massively collected data could support the construction of these complex models and make up for the lack of knowledge about the agroecological processes at play. Integrating farmer strategy into models, taking uncertainty into account and being able to define optimal models in spatio-temporal agroecological systems with several levels of organization remains very much a challenge (Bellon-Maurel et al., 2022).
Modeling the agroecological transition of the farm requires, first and foremost, the ability to model the farm. This modeling is a way for farmers to project and test agroecological or other innovation systems (with agronomic, organizational innovations, etc.) and to represent these complex systems. Farmers will need to visualize and understand the radical changes that could take place on their farm, and try to better appropriate them over the long term. Modeling a farm requires attention to a very wide range of criteria: the farmer’s cognitive structures (goals, plans, preferences), physical structures (the land parcel in place) and social structures, diversity of production workshops, etc. And this modeling will be made even more difficult by the fact that the farmer’s cognitive structures are not always the same. And this modeling will be made even more difficult as agroecological systems bring new management objects onto the farm.
Here again, digital tools (digital simulation methods, multi-agent modeling, serious games, digital twins, etc.) can be recomposed and remobilized to support an agroecological transition (De Graeuwe et al., 2025). Some modeling tools, such as digital twins (by which we mean a digital twin of a farm), make it possible to overcome the learning difficulties of artificial intelligence tools that would have required large amounts of data to operate. The level of complexity of these models will increase with the agroecological components to be considered, as will the problems linked to their interpretability and associated uncertainty.
Artificial intelligence tools, and in particular large-scale language models (LLMs), could help farmers improve their skills. For farmers who have not had the opportunity to attend training courses, artificial intelligence tools, by screening large databases, could help provide simplified and contextualized information, or provide specific decision-making support for operational itineraries (which crops to grow, which fertilizers to use). These tools could also be capable of translating complex information into spoken language (for indigenous communities, for example).
Example : Supporting the spatio-temporal organization of crop rotation through constrained modeling
Drawing up a crop rotation plan is a multi-criteria, highly combinatorial problem, all the more so in diversified agroecological systems. It is difficult for the human mind to take into account all the constraints of the land. To reduce the complexity of the process, agroecological systems tend to use crop groups predefined on the basis of criteria they have chosen, these criteria often revolving around the botanical family. This strategy only imperfectly integrates the agroecological constraints that favor natural regulation processes.
Taking constraints into account in these crop rotation plans is also complicated by the fact that there is as yet no established consensus on how to organize crop rotation. The lack of technical references (which we have already mentioned) is underlined by the fact that questions of inter-species interactions or rotation duration are answered in particular contexts and not necessarily generalizable.
Examples of constraints include: rotation constraints before putting back the same plant or a plant from the same family, interaction constraints to characterize potential competition between adjacent plants, proximity constraints to facilitate logistics and worksites, placement constraints to consider shade or adjacent trees, etc. (Challand et al., 2024).
The role of modeling tools may therefore be more to provide a relatively flexible framework for expressing constraints (rather than fixed constraints), so that everyone can evolve within a framework they control. These tools are therefore more suited to a simulation and/or group facilitation logic, to share experiences and learn together.
In addition to these constraints, we also need to define an “objective function”, i.e. the criteria on which the modeling function should be optimized. In an agroecological context, the intention is to move away from a monocriteria logic to a multi-objective logic, and not just consider economic attributes.
Some work is inspired by pixel cropping, implying that each pixel of a plot (if cut into a mosaic) could represent a different crop. Pixel cropping relies heavily on the principles of established agroecological methods that take advantage of diversity, notably companion cropping and indigenous intercropping practices (milpa or others).
Perhaps putting aside the notion of modeling under constraints of this example, let us also note that there are online platforms to facilitate the exchange of plots between farmers. This exchange mechanism can accompany the logic of organizing crop rotations between neighbors.
Example: Rethinking the usefulness of weed recognition for crop protection
The acknowledged decline in chemical control solutions and the desire to reduce the use of phytosanitary products are driving the development of technological tools dedicated to spraying. As a general rule, cameras are mounted on agricultural equipment and/or robots to ensure weed control – selective or otherwise – which is generally localized. Weeds are thus seen primarily as a problem – the plots are dirty – affecting yield. The main objective is then to discriminate them so as to be able to remove them from the field (by chemical, mechanical, electrical or thermal means, etc.).
Using weed mapping as a basis for assessing whether plots have a high weed pressure or not, and thus prioritizing which weeds need to be removed to protect yield, seems an interesting idea. The practical implementation of this strategy is potentially complicated by the fact that weeding has to be carried out at an early stage, when weeds are not yet sufficiently developed to cover large areas. The question therefore arises as to whether information on weed cover at the seedling stage is sufficiently significant to predict where the competitive effect between the weeds and the cash crop will begin.
To date, most work has focused on the economic potential of weed control robots and agri-equipment, but there are no studies on the conditions under which the ecological potential of weed control robots can be exploited. Deciding whether an individual weed should be eliminated requires specifying which weed species and how many weeds are economically acceptable and ecologically desirable. The working perimeter is not obvious to consider here: who wins: the farmer or a pollinator? What time step should be considered: the cropping season or all the following seasons?
While weeds can certainly be detrimental to agricultural production, they also provide breeding sites and shelter for wildlife (pollinators, natural enemies of pests). They form the food base for herbivores and therefore for higher trophic levels such as birds. Weeds also encourage mycorrhizae and can thus contribute to increasing soil fertility. Developed weeds can prevent other weeds from germinating. If the weeds selected are not too damaging, they form a natural vegetation cover that feeds the soil, limits erosion and limits infestations by more problematic species. Given the great functionality of weeds in the agroecosystem, it could be relevant to integrate the associated biodiversity into the decision to remove them from the plot or not. And localized application of plant protection products is not enough to reduce the trade-off between yield gains and biodiversity conservation.
Marie Zingsheim and Thomas Doring explain that for a defined yield loss, gamma diversity (number of weed species over the whole study area) can be maintained to a large extent, even in the absence of information on weed or crop heterogeneity in the field to decide where to weed (Zingsheim and Doring, 2024). However, to maintain alpha diversity (average number of weed species per plot), more spatially explicit input information is required, such as the number of species per plot, weed quantity (weed cover per species) and the average potential competitiveness of the weed species present. Consequently, a weeding robot should be technically capable of distinguishing between different weed species, measuring weed cover, processing the information captured in real time and eliminating weeds at the level of individual plants.
Weed detection capabilities (“green on brown”, and “green on green”) are advancing, and some players are able to discriminate between major weed families: dicotyledons, grasses and perennials. Insofar as grass and broadleaf weed control products exist and are different, distinguishing these major weed families makes sense for avoiding or limiting phytosanitary treatments. These agri-equipments should also consider the functional characteristics of the weed, by targeting the ecological functions of the weed community.
There are still many unanswered questions regarding ecological weed management. We don’t know, for example, how this logic affects the soil weed seed bank, or what localized weed selection will do to the weed population in the long term. The question of the effects of herbicide reduction on yield depends essentially on whether or not other compensatory weed control methods are introduced when reducing the intensity of chemical weeding. Robotic systems could be used to apply ultra-localized treatments of herbicides to manage perennials (dock, thistles…) before the farmer then releases beneficials. Robots could be used to remove the few resistant weeds after a localized herbicide treatment, to limit the problems of resistant populations.
We still lack knowledge of the competitive effects of individual species in weed communities at field level, as well as the response effects of weeds remaining in the field after selective weed control. The number of species for which we know the competitive effects of individual species is currently limited, with significant gaps, particularly for rarer species. These competitive effects are also site-dependent, i.e. they interact with site properties and with the weed community in which the species is present due to niche differentiation, so that transferability to new sites may be questioned. Competition is also quite dependent on the speed of establishment of the crop or well-accepted species, in the sense that the earliness ratio between weeds and crop plays an important role.
To go further : Zingsheim, M.L., & Doring, T.F. (2024). What weeding robots need to know about ecology. Agriculture, Ecosystems, and Environment, 364
Example: Facilitating cycle looping on a regional scale
The territorial scale is certainly where the notion of the circular economy is most relevant, particularly with the exchange of organic matter, as one person’s agricultural waste is transformed into a valuable resource for another. Agriculture, which plays a central role in these territories, is the object of tensions linked to the use of resources (land, water) or its role in ecological services. This territorial scale would also make it possible to reintegrate agricultural production at regional level (both in terms of material flows and from a social point of view).
Working on a territorial scale requires bringing together a wide range of stakeholders and ensuring coherence between different levels of organization (territory, department, regions, etc.), particularly when it comes to federating political proposals. And it’s often in the design and facilitation of workshops that the key lies. Digital tools can, on their own scale, support this regional engineering work by :
- collect data on a territorial scale to better identify material flows (by managing the compromise between specificity, measurement scope, resolution and heterogeneity of multisource data),
- visualize these data and post-processing results for non-specialists (by restoring complex notions such as uncertainty, incompleteness, etc.),
- enriching territorial engineering methods to facilitate participation and open innovation (need for accompaniment models, gamification, analysis tools for participatory sessions), collective decision-making (digital tools for deliberation, negotiation and voting processes) and mediation (creation of digital “boundary objects”, such as accompaniment models, to encourage dialogue between stakeholders).
Another important subject is to work on better quantification and analysis of material flows on different territorial scales, in order to facilitate the implementation of alternative economic models based, for example, on the bioeconomy or the biophysical economy.
Further information: Inria – Inrae (2022). Agriculture et Numérique. Making the most of digital technology to contribute to the transition towards sustainable agriculture and food systems.
Exchanging & Sharing agroecological knowledge
“Success is a little knowledge, a little know-how and a lot of faire-savoir”.
Farmers’ experiential, traditional, local and tacit knowledge needs to be shared and mutualized as much as possible if the agroecological movement is to spread. Add to this the lack of technical references we mentioned earlier, and it remains difficult to initiate an agroecological transition, especially when you don’t know where to start.
Models for disseminating knowledge and practices are currently very territorial, and depend very much on local players (advisors, cooperatives, agricultural chambers, etc.). When practices are much more specific, it’s difficult to find farmers who do exactly the same thing, and sharing knowledge through third parties is not easy.
Given the diversity of knowledge to be consolidated and concatenated, basic exchanges of information will certainly no longer suffice. Knowledge will have to be stored, organized, sorted, formatted and made accessible if it is to survive viral expansion. It will certainly be necessary to formalize and document this knowledge, using ontologies or shared representations for example, in order to feed it into computer architectures, and to ensure that this information is relevant to farmers who do not necessarily produce under the conditions in which this information was generated (management of mutli-lingual translation, inter-farming context comparisons, etc.).
Insofar as farming practices are as much about know-how as they are about knowledge, encoding information in these databases will certainly be a challenge (Colliaux et al., 2022). The use of speech synthesis and natural language processing tools could facilitate access, structuring and combination of agricultural information.
In agroecology, it is important to focus on different spatial scales (plot, farm, territory, food system). These different levels of organization will need to be ever more closely linked, and farmers will need to be even more closely integrated into their ecosystems (value chains, supply chains, territories, etc.), as well as with end consumers. Building a territorial transition requires the coordination of a wide variety of players who generally do not communicate with each other (Leveau et al., 2019). The diversity of actors involved needs to be made visible (especially those who have successfully transitioned) and these actors need to be able to meet to accelerate the agroecological transition. Digital technologies, with online advertising systems (market place or others) or dynamic mapping, are one way of responding to the sharing of information and knowledge between large and previously isolated communities.
Online videos (on Youtube or other sites) are a great way of disseminating knowledge (the Ver de terre production and Maraichage en sol vivant channels alone have several million views). Video formats need to be carefully thought out beforehand (immersive tutorials, games, current affairs webinars, vlogging, etc.) to provide knowledge that makes people react and want to know more. In these videos, farm immersion strategies give farmers the keys to reasoning and learning how to make decisions. Are these videos more for popularization or real training? Perhaps the question isn’t as important as all that. At some point, these media help to question, to open minds, to self-educate, and many farmers who go through training sessions have actually been through these videos at one time or another.
Social networks, via asynchronous communications that can take place anywhere on the farm or in the field, open up additional avenues for knowledge exchange. Social media engage the responsibility of professionals for what they write, and encourage them to provide quality responses. At least four distinct uses could be made of them (Prost et al., 2022): self-training, emotional or identity reassurance, the desire to pass on knowledge, expanding a community.
Social media make it possible to reach communities that are sometimes outside traditional information systems, and facilitate the transmission of more targeted messages tailored to the people concerned. They are virtual places for people who don’t necessarily dare share their feelings and problems in real life, and the questions asked online are not the same as those that would have been asked in real life. It’s true that on the networks, members tend to share a more attractive view of themselves, so as not to damage their image. On the other hand, sharing problems and feedback, even negative ones, is an incredible source of knowledge. These networks are therefore an opportunity to consolidate knowledge, in the form of surveys for example, anonymized or not.
Social media structure communities. We all need points of reference, peers, people who share our visions and can support us. Social networks can provide comfort to someone facing a challenge, such as a farmer in transition, who has encountered problems and become discouraged. By being present, community members can act both as a form of technical assistance and as resource people who boost farmers’ morale (Soulignac et al., 2023).
These communities may or may not be led by advisors, whose role will thus be more that of a network facilitator or moderator, guaranteeing trust within the group, which will call for the development of new skills and know-how. One of the roles of these facilitators could be to identify the most read and liked topics on social media, or to examine expectations around these topics of interest, or even to identify reference persons able to answer questions from the community.
lIt’s important to bear in mind that communication from institutions or certain structures is not always welcome in these peer networks or user clubs. These communities are not necessarily just for farmers, but will also be ideal for advisors, technicians and other agricultural players, to exchange ideas and share best practices.
Example: Digitization of agricultural archives
Even if there is a lack of technical references for agroecological systems, the fact remains that our elders had nevertheless aggregated a certain amount of knowledge and know-how in old books, documentaries or maps. Artificial intelligence tools could be dedicated to recovering these historical data (tables, analog records, digital maps, etc.) and redigesting them in a usable quantitative and qualitative format. The reconstruction of agricultural systems in history and the associated ecological processes are a means of better understanding the conditions of existence and trends in past agroecological systems.
Pour aller plus loin : Viana, C.M., & Carvalho, D. (2024). Unveiling historical agroecological patterns through artificial intelligence (AI) and Geographic Information Systems (GIS). Agile : GIScience Series, 5.
Example: Creating agroecological communities through social networks and shared knowledge systems
Social networks can help bring farmers together and connect them. In addition to inter-personal networking, which could help to increase the silent agroecological dynamic underway (Lucas, 2021), the informational link provided by social networks through the simple sharing of messages, sounds or videos is already a first step towards the propagation of knowledge. These exchanges facilitate individual and collective learning as a source of innovation. In fact, digital tools have contributed to the rise of new emancipatory and revolutionary social movements, giving certain farmers new ways of communicating and sharing their practices.
Setting up a social network isn’t enough – you need to create a community. Moderators and community animators need to be on hand to fluidify and (re-)activate exchanges as needed, and create the conditions for farmers to discuss and share information: simple, clear channels, moderation and/or rethinking of information, setting up useful links, celebrating results, etc. The example of the Centipede RTK community, which promotes the self-installation of RTK beacons on territories through a dynamic and user-friendly community, is particularly enlightening (https://docs.centipede.fr/).
Agritech structures such as Triple Performance (https://wiki.tripleperformance.fr/wiki/Triple_Performance) are seeking to become a kind of agroecology wiki. Their intention would be to have the most accurate on-line literary documentation possible (agris tell what they do), supplemented by semantically exploitable technical itineraries, technical figures and performance indicators (agroecology flora, IDEA4, regeneration index, etc.). This wiki would enable us to characterize examples that stand out from the crowd, and to extract practices whose impact had not been suspected, but which demonstrate an interesting effect.
It would be the communities themselves who could validate the farm portraits, rather like the classic scientific peer-review mechanism. Mechanisms will have to be put in place to encourage farmers to declare themselves, but in a transparent way, with objective criteria (e.g. % soil cover, consumption of average IFT, etc.) that will then enable them to compare themselves with members of the community. These communities will be all the more powerful if farm locations are known (to visit them, for example), and if farmer portraits are concrete, simple and effective (e.g.: how much the farmer earns, what practices are implemented). Mapping agroecology players can be interesting, unless it’s simply a matter of listing the players in agriculture.
Example: Towards participatory research frameworks and the creation of networks promoting public debate and the dissemination of knowledge
Agroecology needs open innovation systems. This is a way of designing plausible future scenarios and transition scenarios, instantiated according to previously identified future scenarios. These foresight mechanisms are a good way of representing phenomena occurring at different levels (biological processes, farm management, territories), projecting them and, above all, discussing them. These scenarios can be energized by digital tools, in the form of gamified projects (serious games), or even augmented reality to help stakeholders visualize future diversified landscapes when designing cropping systems.
Digital systems can also facilitate the exchange and cross-fertilization of information through knowledge-sharing platforms: wikis, videos on agroecological practices produced by mediators working with farmers. Digital tools also offer media for collaborative work, facilitating collective decision-making and mediation: digital collaboration platforms, videoconferencing systems, digital tools for deliberation processes, etc.
Orienting equipment towards agroecological practices
This is perhaps the area where the link with agroecology is least concretized, even if some form of intention is present.
Agroecological systems will not have access to the same range of curative measures as conventional farms, all the more so as the means of chemical control available (available molecules, products on the market, etc.) will tend to diminish over time for regulatory reasons. It will then be necessary to become a master in the art of anticipation, and to position interventions more finely to reduce the risks to agricultural production as much as possible. Early detection of problems – which measuring systems and instruments can facilitate – means we can be better prepared and potentially take curative action when needed (detection of mildew and oidium spores, connected traps for game damage or the arrival of insect flights, etc.).
It’s perhaps when we need to move towards more expensive products, or products with a narrower spectrum of action (and greater precision of application), that digital tools may prove interesting. Even when it comes to releasing crop protection agents, it doesn’t seem completely idiotic to release these biological control methods willy-nilly.
Much of the knowledge generated in an agroecological logic can only be taken into account if dedicated agroequipment exploits it to modulate their action (Caquet et al., 2019). Sowing in an existing crop, harvesting variety by variety to be able to use crop mixtures of asynchronous maturity, are all practices that can be activated by agricultural equipment in an agroecological logic. These practices are particularly recognized for raising the level of productivity of low-input systems that are considered by their detractors as leading to yield reductions.Technological systems can be integrated into existing equipment (in the form of retrofits, for example), or new agroequipment will have to be designed.
One of the challenges will be to ensure that agro-equipment does more than just what is currently referred to as “precision agriculture”, i.e. the modulated management of inputs and practices. Some would even go so far as to say that providing farmers with a direct-seeding drill (with tines or discs) remains the most basic requirement, as the first step towards an agroecological trajectory, at least for field crops. Continuously covered soils, on the other hand, will have difficulty being worked with localized alternative weeding tools (thermal, mechanical or electric), because it will not be easy to detect weeds.
Fixed and on-board actuator systems (e.g. automatic greenhouse opening tools) can help reduce the workload involved in setting up complex agroecological systems.
Example: Support for the introduction of alternative practices using yield and profitability maps
The use of within-field yield maps, combined with farm cash flow monitoring (cost of selling production, cost of purchasing inputs), offers the possibility of constructing within-field profitability maps. These maps could be used to discriminate between profitable and unprofitable production zones, and to implement conservation strategies in the latter (wildflowers, red clover, white clover, alfalfa, phacelia, peas and oats…) or, more broadly, to establish agroecological infrastructures.
Figure 6. Spatial distribution of benefits in identified potentially fallow land. Source: Capmourteres et al. 2018. Precision conservation meets precision agriculture: A case study from southern Ontario. Agricultural Systems. 167, 176-185
These agroecological zones, in addition to helping to consolidate biological regulation within the plot, could be economically enhanced by current dynamics: carbon premiums, sector premiums, additional CAP premiums, etc. The economic models remain to be found, however. However, economic models have yet to be found, as the current voluntary carbon market is already insufficient to finance changes in agronomic practices.
To find out more :
- https://www.aspexit.com/quand-lagriculture-de-conservation-des-sols-rencontre-le-numerique-et-lagriculture-de-precision/
- https://www.aspexit.com/la-course-au-carbone-en-agriculture/
Example: Review sowing patterns and propose original plot arrangements
If we consider that robots have their raison d’être in the farming ecosystem, is it then up to the robot to adapt to the plot and the farming itineraries, or is it the plot that must facilitate the action of the robots? In reality, the cursor may lie somewhere in between: the tool must co-evolve with the technical itinerary. And this co-evolution is necessary if we are to avoid becoming locked into a particular type of itinerary, and thus creating a barrier to change. In any case, it’s difficult to reason with the current agricultural system, which is not set in stone.
Agroecological practices require a high degree of frequency and precision in both farming practices and the number of passes made, which robots could help to meet. The use of robots with highly precise geo-positioning systems could make it possible to redesign agricultural itineraries in the sense that the robots would have in memory the exact positions of all the seeds sown (and therefore no longer need an on-board vision system) and thus perform extremely fine mechanical actions such as hoeing crops in all directions in space.
This high level of positioning precision also makes it possible to envisage sowing patterns that are totally different from those currently used (grid, rhombus, etc.), or even to imagine that prior soil maps could be used to modulate sowing density and depth according to soil characteristics. These sowing patterns could then be used to integrate a whole complexity of companion plants, plant diversity and different phenological stages, mapped very precisely as soon as they are sown, and monitored throughout the crop cycle. We could then imagine having plants at various phenological stages within the same plot, each of which could be considered and taken into account to lead to dedicated actions.
Let’s think even more broadly about placing annual crops in perennial cover, leaving strips to resow beets or sensitive plants, leaving strips without cover, organizing the cover in strips, or even relay-cropping. Arvalis (French Technical Institute) trials have shown that a well-regulated alfalfa cover, i.e. one that is regularly cut, works very well in association with soft wheat. As soon as the alfalfa is the same size as the wheat, it can be regularly cultivated using small robotized units to manage one of the two elements of the plot (in this case, the alfalfa), with a deliberate differentiation between the row and the inter-row.
Given their complexity, agroecological systems will be increasingly difficult to observe and control in their entirety. The division of elementary actions (sowing, weeding, harvesting…) by robots – in swarms or not – with a dynamic cartography of the plot in mind could be made available for punctual and precise tasks: weeding an area to transplant a seedling, removing senescent or diseased organs from a plant, or harvesting what is ripe. The division of elementary tasks is also interesting in that it is not blocking if one of the systems fails and there is only a part of the itinerary to intervene on.
By thinking about robotics in a way that is completely interwoven with other technologies or innovative systems, robotized units make it possible to envisage heterogeneity and diversity (of companion plants, cover crops, varieties, phenological stages, etc.). Bear in mind, though, that we’re not there yet…
To find out more : https://www.aspexit.com/la-robotique-est-dans-le-pre-ou-sommes-nous-et-ou-allons-nous/
Example: Developing crop associations with legumes and the agri-equipment needed to make the most of them
The strong development of leguminous crops and crop associations (wheat/pea, wheat/lentils, etc.) may require technological efforts upstream of the agri-food industries to ensure both that varieties suitable for interspecific mixtures are available and that no leguminous residues are found in cereal stocks sent to processors after sorting. Developments in optical sorters can thus be expected to enable the practice of species and variety associations adapted to local conditions in the majority of plots, so as to better assess product heterogeneity, both upstream and downstream of the farm. This support for change is realistic in the sense that, just a few decades ago, there wasn’t a single farm without a sorter.
This development of legumes will have to be accompanied by major efforts in varietal selection and genetic improvement of legumes, which are still too few and far between at present. Two key moments are of interest: prophylactic operations before farm-saved seed sowing, and post-harvest sorting to achieve the homogeneity required by the market (Caquet et al., 2020). In particular, this results in material shortcomings for geneticists and breeders working on legumes (ability to access molecular markers and other technologies for identifying alleles of agronomic interest).
We could nevertheless question the legitimacy of developing densimetric and optical sorters to avoid legume residues in crop associations at the end of the field. It is indeed conceivable that the agri-food industries will call into question the norms and standards of products entering their chain, or evolve their product ranges. Citizens can also be expected to change their consumption expectations to facilitate and accompany the landing of these associated crops, without necessarily needing to develop massive sorting infrastructures. Time windows for agricultural itineraries are becoming shorter, yet downstream players are increasingly demanding in terms of harvest quality (in the past, for example, they allowed themselves to harvest wetter products). Storage facilities are closing earlier in the night, which means that the size of the machines has to be increased to compensate for the limited human resources of these cooperatives.
Example: Agro-equipment to reduce the drudgery of agroecological systems
The robot could be designed primarily to reduce the drudgery of work and carry out strenuous agricultural activities. Melon picking, for example, is a particularly exhausting task for humans, who have to bend or stoop to pick up melons from the ground. Humans think fast and their backs hurt, so how can we imagine a collaborative activity? The human could lie or recline on an autonomous harvesting system. We could also imagine that the human is only there to identify or tag the walled melons that will then be harvested by a robotic unit, thus avoiding the need to install complex vision and recognition algorithms on the robots.
The robotics sector could even offer exoskeletons without electronics, with harness and spring systems to support gestures, as used in other industrial sectors. Robot followers could also play an important role, whether for carrying heavy loads and leaving the operator’s hands free, or so that the robot can do exactly the same thing as an operator in parallel (based on the leader-follower principle). These devices would double work rates and compensate for regulations, since an operator would de facto be present, but also and above all help to carry out work that the farmer cannot do alone (e.g. beet work). However, we need to distinguish between the automation of agricultural functions (harvesting, sowing, etc.) and the automation of safety functions (anti-collision detection, rollover protection, work zone exit detection, etc.), and it is mainly on these second functions that manufacturers will have to set the cursor to know how to mobilize an operator who would be doing something else at the same time.
The robotic system may or may not perform its own surveillance and safety functions. If the robot relies on humans, it may be considered a highly automated machine, but not an autonomous system. If the robot is not autonomous, it will be difficult to imagine a strategy based on a nearby worker performing another task.
It is questionable whether current robots will be able to respond to the search for economies of scale, given that they were developed with electric motor logic (with electrical storage) that is not compatible with high-power, long-term operation. And even if this were the case, would it really be a problem? A system that doesn’t respond to economies of scale must respond to different logics, and this could be the added value of robotized units, with modular autonomous machines capable of carrying out different tasks on different crops – which could prove relevant in agroecological contexts. Systems aiming for economy of scale can be diversified and compatible with satisfactory production on a plot scale, while at the same time being highly efficient in environmental terms.
To go further :
- https://www.aspexit.com/la-robotique-est-dans-le-pre-ou-sommes-nous-et-ou-allons-nous/
- Ditzler, L., & Driessen, C. (2022). Automating Agroecology: How to Design a Farming Robot Without a Monocultural Mindset? Journal of Agricultural and Environmental Ethics.
- Colliaux, D., et al. (2022). Computational agroecology: should we bet the microfarm on it? Eighth Workshop on Computing within Limits
Example: Low-tech systems to facilitate the work of agroecological systems
“Low-tech” digital technologies lack visibility. In addition to the fact that these technologies are by nature more resilient and less energy-intensive (we will discuss this later), many sober technological systems can support the deployment of agroecological practices, quite simply by facilitating the daily lives of farmers through remote control or the detection of failures and alerts on the operation of the farm’s equipment. Simple fixed and embedded actuator systems can be set up to open and/or close irrigation hammers, open valves, activate irrigation pumps, automatically open and close market garden greenhouse doors. All with a view to avoiding or at least limiting recurring travel. GPS trackers on irrigation guns can be used to detect watering problems, for example if the location of the irrigation system does not change sufficiently in a given period of time. These tools will raise the alarm in the event of winding problems (of the reels) or water leaks (via connected flow meters or others) in the irrigation infrastructures in place.
Networks of temperature sensors in Lora communication can be used to light gel candles at the right time in the plots. In a more complex agroecologization dynamic, these sensors could be used to analyze temperature differences between plots with or without plant cover to report or not the interest in switching to these agricultural cover practices.
And the possibilities of these low-tech sensors are quite endless: bending sensors, pressure sensors, ultrasonic sensors, LiDar sensors (to measure distances or depths)
To go further : https://www.agrotic.org/mobilab/
Example: Sanitary monitoring of orchards and the environment using robotized tools
At present, the health status of fruit crops (perennial or otherwise) is monitored using meteorological data and an intervention schedule. Could robots be used to postpone intervention times, or even avoid some of them? Harvesting fruit trees is a particularly difficult task for robots. Some are able to do it, especially in greenhouses, but the work rate is particularly slow. During the interviews I conducted as part of my report on agricultural robotics, some slightly teasing interviewees told me that today’s robots are capable of picking up fruit while damaging it and moving slowly… Rather than seeing robots only for harvesting, couldn’t we imagine robots intervening to pick up damaged fruit? This activity, entrusted to robotized systems, would not only limit contamination (and thus delay the first treatments and limit the doses applied) but also reduce the need for sorting at the end of the season. We need to find situations where the action to be carried out is very important, but not too complex for the robot.
We could even imagine using robots to respond to public health issues, by removing toxic or allergenic plants from paths or crossings for which no-one is responsible. The challenge here will be to see how to make robots profitable in areas where the return on investment is not necessarily immediate.
To go further : https://www.aspexit.com/la-robotique-est-dans-le-pre-ou-sommes-nous-et-ou-allons-nous/
Let’s take a step back
Not ONE but MANY digital pathways
Digital technologies are extremely numerous, and I hope I’ve made it clear that it’s nonsense to talk about a single digital technology, which seems to be implied in the concept of digital agriculture. This ecosystem is all the more heterogeneous in that not all these tools have the same maturity, the same state of adoption in the field, or the same objectives. The same could be said of Agritech companies.
The trajectories of digitization are manifold, even within communities that share strong agroecological values. Some strong agroecology movements advocate minimal use of digital technology, while others rely on the commons approach to develop non-hegemonic initiatives for sharing resources and knowledge, which can be supported by digital projects (Aboueldahab, 2023). The Mobilab (https://www.agrotic.org/mobilab/) of the Chaire AgroTIC, by allowing everyone to develop their own low-tech sensors autonomously, is also helping to bring farmers closer to the digital ecosystem than they had been at the outset, sometimes simply because their sector (low-input system, market gardening, organic farming, etc.) was not equipped.
There have also been cases of farmers abandoning their robots (milking robots, for example) because there was already a robot on the farm when they took over, or because the tool kept breaking down. These abandonment trajectories remain largely uninvestigated today, even though they also demonstrate the diversity of tool adoption and non-adoption paths.
The diversity of tools also suggests that not all digital technologies are considered suitable for agroecological systems, whether for technical or socio-economic reasons. These tools may not be suited to farmers’ business models, ways of thinking and decision-making (Schnebelin et al., 2021). The key issue is to differentiate technological proposals that will promote agroecological and organic principles and support a transformation agenda from those that will undermine them if introduced without thought and encounter a community unprepared to participate in the development of digital technology products (IFOAM, 2020).
The introduction of digital and/or automated tools and services in the service of agroecology will require the same range of situated and diversified approaches and actors that underpin the ethos of agroecology, and therefore new thinking about what automation might mean in different contexts. In a robotic context, Marie Zingsheim observes that her robotic field case study broadened the range of possible approaches and imagined relationships between humans and robots, and highlighted different aspects of a set of interconnected dialogues (Zingsheim et al., 2024).
Like Lenora Diztler and Clemens Drissens, I invite here to approach the notion of automation for agroecology as a dynamic range of context-dependent options and orientations, rather than as a binary all-or-nothing logic (Ditzler et al., 2021). More generally, we need to foster heterogeneity in agronomic innovation systems to enable the development of technologies adapted to different greening trajectories (Schnebelin et al., 2021).
How can digital technology meet the demands of agroecology?
The challenge is to take the transition a step further by focusing digital development on supporting the redesign of cropping systems in line with the principles of agroecology, i.e. taking advantage of functional interactions between ecosystem components (from plot to landscape), seeking to recycle nutrients (looping cycles) and moving from a curative to a preventive approach (Annales des Mines, 2022).
Cultivating diversity and increasing farm resilience.
Two visions of how to take heterogeneity into account currently coexist. In the first, knowledge of heterogeneities calls for differentiated action to better absorb them, level them out and offer the most homogeneous production possible in line with market standards (Caquet et al., 2020).
In the second view, heterogeneity signals the underlying existence of variation, giving rise to a differentiated response. This second heterogeneity is seen as a guarantee of better risk distribution, less competition between individuals, complementarities and offers of a biological response adapted to environmental conditions. In agroecological systems, heterogeneity increases at all levels of organization (genotypes, species raised/cultivated, intra- and inter-plot scale, lengthening and diversification of rotations, crop management methods, processing systems, food systems, etc.).
Agroecology starts from a voluntary refusal of uniformity, to take advantage of heterogeneity and make the most of it. Agroecological systems seek to put observation and biology back at the heart of decision-making. For agroecology to benefit from these technologies, it is vital that it harnesses biology and makes the most of the heterogeneity of environments encountered at different scales.
This heterogeneity is even more pronounced in southern countries, with contrasts between agroecosystems within the same country (Bellon Maurel et al., 2022). Production systems are also more complex: the high prevalence of integrated, multifunctional multi-species systems, such as agropastoral systems (in dry regions) or agroforestry systems (cocoa and coffee in humid regions), generates complex landscapes and organizational frameworks.
This culture of heterogeneity calls for the proper equipping of all agroecological production systems, some of which have been largely forgotten or at least sidelined (organic farming, conservation agriculture, low-input systems, etc.). It would be appropriate to rethink the technology transfer of what already exists in certain sectors, and to remobilize existing technologies and make them available and accessible to others. This adaptation is not obvious, and will require a powerful rethink of existing technology developers and institutions (The Shift Project, 2024).
Conventional technologies do not seem suited to these itineraries which, unlike the relatively homogeneous systems of the dominant agricultural model, seek to cultivate their heterogeneity. The design of digital tools must be sufficiently flexible to take into account the specific configuration of the farm. It must be adapted to dense, diversified crops planted in a specific way.
Herd management cannot be based solely on individual management, because a herd is not just a collection of individuals, but a complex system. By only looking at certain animals and not others, we can only observe problem animals (changes in physiological stage, etc.). Some animals may fall through the cracks of digital tools, and we may end up with animals that are invisible.
Embodying digital technology with farmers
Farmers must be at the heart of agroecological dynamics. And yet, one might be surprised to find that in the literature, and even a little in this dossier, not that much is said about them, or at least not everything revolves around them. It’s as if work on digital technologies and even agroecology were disembodied.
Technology, and here specifically digital technology, depends on contexts and uses. It is directly linked to users. The concept of appropriate technology, in the dual sense of appropriation by the user who would be in a position to use or maintain the technology, and appropriation of the technology to the use it is going to have, is particularly apt.
And it is indeed the farmer who can put digital technology at the service of his agroecological project. This use of digital technology will also depend on the maturity of the farmer’s agroecological project, in the sense that digital bricks may arrive at different stages of reflection on this transition project. Depending on the why, the how, or even the farmer’s desire to embark on an agroecological trajectory, the use of digital tools may enter into phase opposition or synergy with this transition project.
A soil moisture sensor, for example, can be used either to optimize irrigation, or as part of a more in-depth study to support the withdrawal of irrigation from plots. On social networks, some farmers who are reluctant to embrace agroecology will use these digital media to confirm that the transition is difficult, and select communities in this direction. Other farmers, highly motivated by the logic of transition, will interpret the same messages as a powerful source of inspiration.
As the master of his digital tool, the farmer plays an important role in the way he uses it, and can thus more or less divert its initial use, which is not necessarily a bad thing. These tools need to be made sufficiently flexible and appropriable so that the farmer has room for manoeuvre to make intimate use of the tool, and so that this personalized use can also, in the long term, help to improve digital technology more generally.
Developing low-carbon, resilient technologies
Digital technology is a catalyst. Where it is deployed, it enables optimization, acceleration, fluidity, parallelization… Deploying it without a strategy therefore leads to the acceleration of all dynamics, including those furthest removed from the resilience objectives dear to the founding principles of agroecology.
Although it is more relevant to look at farming systems, their interactions and their overall evolution than at technologies in isolation, it will certainly be necessary to go through a phase of precise quantification of technological effects, all things being equal, in order to assess the place of digital technologies in the sector’s transition. Reflection on the future of digital technologies will necessarily have to invest the field of sobriety (The Shift Project, 2024), especially if we consider that these technologies can accompany or are necessary for the agroecological transition.
These sobriety efforts must be considered on several scales: individual sobriety, collective sobriety and structural sobriety. For example, it is clear that French agriculture is over-mechanized in terms of agricultural equipment (FNCUMA, 2024). Tractors are often overpriced in relation to the implements they are supposed to attach. A portion of the fleet is largely under-utilized. On these forms of sobriety:
- For a farmer, individual sobriety means, for example, thinking about the act of purchasing his or her agricultural equipment, making more detailed diagnoses of the suitability of the tractor and implements in relation to the farming practices to be carried out, and making better use of the fleet (checking tire inflation, using the tractor in the right ranges, etc.).
- Collective sobriety will take the form of a reorientation of tax support to avoid individual over-mechanization, to avoid the logic of unnecessarily frequent renewal of the machinery fleet, or to support the use of alternative fuels (both for the farmer and for manufacturers).
- Structural sobriety, on the other hand, will call for different modes of organization, taking advantage, for example, of the sharing and pooling of agricultural equipment (via Agricultural Equipment Use Cooperatives).
Insofar as the deployment of agricultural technologies necessarily depends on the context in which they are inserted, we can’t do without a case-by-case approach to judging the interest of a particular digital technology. We need an assessment of the energy-climate relevance of digital technologies that is systematic (for each technology, each particular case, each given application, etc.) and exhaustive (taking into account direct impacts as well as indirect and systemic impacts).
The environmental footprint of digital tools is far from easy to measure. Clémence Huck and her team’s review of scientific articles on the subject shows that these studies highlight environmental gains resulting from efficiency gains and fossil fuel substitution (for agricultural machines and robots that have switched to electric mode) [Huck et al., 2024]. Nevertheless, this state of the art reveals several limitations
- digital tool evaluation methods are heterogeneous, making it difficult to compare results
- the life cycle of the digital tool itself is very rarely considered, which ultimately leads to an overestimation of the value of digital tools
- the functional units considered are different, which means that it is necessary to reason with several functional units (per kg of product and per cultivated area at least) to avoid over-simplistic conclusions
- the studies do not include all the farming operations impacted by digital tools, which means that not all digital impacts are considered.
A number of future difficulties have also been identified: the Agritech ecosystem moves very fast, digital tools can serve several functions at the same time, and the rebound effects of these digital agricultural technologies are never far away.
As agroecological dynamics require attention to be paid to numerous spatial scales (plot, farm, landscape, value chain, etc.), we might wonder what would happen if digital technologies were deployed on large enough scales. In 2024, Pierre La Rocca proposed some initial, fairly simplified case studies (one in crop production, and one in animal production), but which have the merit of laying the foundations for an initial quantification, in terms of both energy consumption and greenhouse gas emissions.
These case studies seek to assess the impact in terms of energy consumption and greenhouse gas emissions of different technological systems (computers + RFID tags, computers + RFID tags + connected collars, computers + RFID tags + connected cameras, simple robotics, advanced robotics, etc.). These studies show that behind the same concept of digital agriculture lie very different systems with orders of magnitude different GHG emissions (by a factor of 10 to 20 between their two case studies). The authors therefore stress the need to distinguish between the use cases and technological systems studied when assessing the carbon footprint of digital technologies in agriculture (La Rocca, et al., 2024).
More broadly, The Shift Project recommends adopting a cautious stance and not taking risky bets on the use and deployment of agricultural technologies, especially as we are not fully aware of the likely changes in our world, beyond a certain decline in the physical flow of materials and energy (The Shift Project 2024). Every new development needs to be questioned in terms of what surplus technology really brings. All the more so as the risks of non-deployment are numerous and not necessarily linked to a technological issue: infrastructural obstacles, obstacles to physical flows, organizational and/or skills obstacles, regulatory obstacles, economic obstacles, ethical obstacles, etc.
I invite interested readers to delve further into this subject in another Aspexit blog entry: the environmental footprint of AgriTech :https://www.aspexit.com/lempreinte-environnementale-de-lagritech/
Thinking about the post-automation of agricultural systems
The apparent lack of consensus on the role automation should play in the transition to more sustainable agriculture could hinder progress in automating diversified cropping systems (Ditzler et al., 2021).
Are digital tools supposed to do everything? It cannot be denied that the dream of total automation of agricultural practices persists in the brains of some technologists. Some authors propose to rethink the context of automation in the context of agroecology, namely automation not to replace humans, but to make way for analog and hybrid forms of agricultural work (Ditzler et al., 2021). What they would call a kind of “post-automatization”.
This new logic of thinking will certainly only come about by implementing design processes that include a diversity of actors, involve iterative design cycles and incorporate feedback between designers, practitioners, tools and cropping systems. Diversified design processes adapted to the local context could help prevent the limited values and norms of homogeneous designers from perpetuating the status quo, important human or non-human actors from being sidelined or forgotten, or technological development from fuelling a trajectory leading to ecological dystopia.
Digital tools need to harmonize bottom-up (farmers to experts), top-down (experts to farmers) and horizontal (peer-to-peer) modes of communication, co-production and dissemination of knowledge, where farmers are fully recognized as authors and co-creators of knowledge (Wittman et al., 2020). Technologies developed using transdisciplinary and participatory research methods could meet real user needs. The exchange of knowledge between farmers and the sharing of information from open sources can be used to democratize technology development and data use.
This co-construction between players, not to mention farmers, is easier said than done. Mixing agronomy, ecology, social sciences, engineering and automation takes time and profiles at the interface. What’s missing is a methodological framework for working with skills from different disciplines.
All these players also have different bases for understanding what constitutes proof. While it is clear that agroecology is also highly experimental and hypothesis-driven, it tends to be place-based, rooted in culture and immersed in everyday life. Much agroecological evidence is produced using methods and theories that intentionally trouble the divisions between layperson and expert, including oral history, citizen- and community-based science (De Wit, 2021).
For digital technologies to be compatible with an agroecological system, we may have to come to terms with the fact that agroecology is not reducible to mixed farming or crop protection. At its core, agroecology is a process for achieving food sovereignty and technological sovereignty. It is the process of establishing collectively negotiated community agreements and standards of responsibility that enables sovereignty to be forged (de Wit, 2021).
The notion of use allows us to take into account the materiality associated with a technology, and the difficulties and constraints of use. This notion allows us to go beyond a vision of passive users and integrate their role in the layout, tinkering and adaptation of technologies. By focusing on usage profiles, we can integrate the interactions and combinations of technologies (Schnebelin, 2024). The concept of appropriate technology, in the dual sense of appropriation by the user who would be able to use or maintain the technology, and appropriation of the technology to the use it is going to have, seems highly appropriate.
Access to technology should be considered from the angle of technological sovereignty. Maywa De Wit sees this as a much bolder challenge – one based on the rights of people to make decisions about and co-create technologies that reflect, respond to and mobilize the collective knowledge and power of communities (De Wit, 2021). According to this author, this sovereignty is possible and will effectively show what farming communities, in all their specific local diversities, want and choose. Digital technologies will be selected if it is proven that these solutions will serve them and give them priority.
Making digital technologies appropriate and accessible
Whatever their paradigm, farmers’ organizations play an important role in digitization, acting as intermediaries between digital companies and farmers, but also as proactive players in digital development and collecting, analyzing and transferring information. Digitization does not reduce the role of intermediaries, but can even strengthen it (Schnebelin et al., 2021).
While the private sector can contribute to a just digital transition, the public sector has a crucial role to play in imposing strict conditions on the distribution of research and investment funds for the development of any technology. If public money is to be spent for the public good, the role of the public sector is not to make the digital transition possible at any cost, but to ensure that it is fair, equitable and goal-oriented, which in this context means supporting a process of transformation towards more agroecological systems (IFOAM, 2020).
The diversity of definitions of agroecology reflects a diversity of paradigms for greening agriculture and ways of thinking about progress in agriculture. This diversity materializes in a heterogeneity of institutions, actors and knowledge, which accompany the different agricultural systems (Schnebelin, 2024). All these environments are not neutral in the greening trajectory they support. These structures will influence the conditions under which knowledge is created and disseminated, support funding for certain farming systems or research dynamics, and favor certain innovations (technological innovations, for example) over others. At farm level, the micro socio-economic environment, such as relationships with advisory service providers or cooperatives, also influences the use of these technologies (Schebelin et al., 2022).
For some authors, there would be an obvious interest in aggregating data from several farms, so that farmers can contribute to the construction of broader agroecological dynamics via open or participatory science frameworks (Colliaux et al. 2022). For these same authors, however, it should be clear that data must be kept on the farm and calculations carried out locally wherever possible. It would therefore be interesting to find application cases where the data is directly useful to the farmer in managing the farm, without requiring the use of a dematerialized service outside the farm.
Compatibility between digital and agroecological trajectories becomes both more possible and more equitable if changes occur in the control, design, accessibility, values and ownership of digital tools. This compatibility depends largely on the structures that develop digital tools and the purpose with which these tools are designed. How is complementarity understood? Who asks and defines this question? What are the political stakes of such a debate? (De Wit, 2021). These questions are important because it would be naïve to believe that small (and not so small) agricultural structures (family farming, small cooperatives, etc.) can play on equal terms in the agricultural data economy when we look at who is on the other side (Shelton et al., 2022).
While open data platforms and open-source protocols and standards can give farmers greater access to and control over the digital aspects of their farm equipment, to generate this data, they still need to invest in physical technologies (sensors, agro-equipment, etc.) developed by an increasingly centralized and corporate-controlled agricultural sector (Rotz et al., 2019).
While digital technologies are often presented as being adaptable to all farm sizes – although this could be debated – it must nevertheless be acknowledged that small farms do not have access to the same financial resources as larger ones and therefore potentially take more risks (financial, insurance, etc.) when equipping themselves. This is what the nice expression “scale neutrality does not mean resource neutrality” (Sullivan, 2023) seeks to convey. In her article, Summer Sullivan brings in agroecologists who say they have fundamentally fewer resources than their technologist colleagues to work at the University of California, Santa Cruz.
It is important for the key players involved in the deployment of agricultural technologies to have a framework for assessing the relevance of a technology to support the sector’s transition, or even simply to identify its negative externalities or rebound effects. Technology assessment must be systematic (for each technology, each particular case, each given application, etc.) and exhaustive (taking into account direct impacts as well as indirect and systemic impacts) to guarantee their relevance to the agroecosystems, contexts and communities in which they are to be applied, and their integration into food and farming systems.
Supporting the agroecological transition
Transition is difficult
The transition to agroecology is extremely polarized. It is caught between, on the one hand, a conservatism resistant to any change and, on the other, a naive simplicity that denies the difficulty of this change for farmers who are, most of the time, the most constrained players in the food value chain. (Terra Nova, 2023). The effects of climate deregulation and the biodiversity crisis could also be at the root of a potentially value-destroying scissors effect for farmers, with increasingly high input costs and falling yields. Organizing the diversity of farm transition paths to identify types or styles of transition dynamics remains a major challenge (Prost et al., 2023).
Some voices express the view that it remains extremely difficult to achieve a high-performance agroecological system without pesticides altogether (the example of glyphosate – even in low volumes – for conservation agriculture is quite telling). Farmers using simplified cultivation techniques and organic conservation agriculture will certainly have succeeded in doing without them, but only after many years’ work and having reached a very high level of technical expertise. The effort required is therefore extremely considerable, with a potentially long desert crossing before a possible economic equilibrium can be achieved, and is made even more complex by climate change and the current poor state of much of France’s soil (there is virtually no soil-plant immunity in soil that is too poor).
Agroecology can be seen as the equivalent of margin insurance. With classic conventional practices, a farmer will go out to treat, not knowing in advance whether the result of the treatment will necessarily be positive, but he will in any case have attacked his operating margin more or less severely. In an agroecological system, with far fewer treatments, pesticides and fertilizers have been greatly reduced, which means that in a good year, the results will be particularly interesting, and in a bad year, the yield will certainly be low (but it will be secure because the soil-plant system is resilient), but operating costs will also be very low.
The agroecological transition of the farm poses methodological problems, as the aim is to account for the dynamics of a complex system in a changing environment with a high degree of ambiguity and uncertainty. Ambiguity is linked to the fact that values change during the transition. What was acceptable yesterday becomes unacceptable today or tomorrow. Uncertainty, on the other hand, is a consequence of strategic activity that projects into the future, and of the lack of knowledge about the transition phenomena involved (Caquet et al., 2020).
Another promising approach is to assume that farm behavior is not purely deterministic. Thus, uncertainty is no longer considered as marginal noise around a predictable dynamic, but rather as a central element of that dynamic (Figure 7). By adopting this perspective, properties such as resilience, vulnerability, viability, or robustness become as important as productivity when it comes to assessing a system’s performance (Prost et al., 2023).
Figure 7. Conceptualization of farm transitions to agroecology and research activities that focus on and support these transitions. Figures in brackets are references to sections of the article from which this graph is taken: Prost et al., 2023.
Supporting agroecological transitions will require analysis of the joint evolution of the context, practices, properties and performances of systems in transition (resilience, vulnerability, efficiency, viability in particular), in order to characterize the required conditions and favorable factors, or on the contrary, the locking-in effects in farm and farmer activities (technical changes, integration into networks, learning processes, changes in marketing methods and work organization) (Caquet et al., 2020).
Deficits, accumulations and other dynamics must therefore be better anticipated, taking into account short- and medium-term dynamics. The management of agricultural activities is no longer predetermined, but adaptive. Objectives and decision rules can evolve in line with available information, both in the short and long term, as the state of the system changes, and by arbitrating between ecosystem services (Caquet et al., 2020).
There is a wide range of models for a transition that is built “along the way”. Given the complexity of the transition, we must not overlook the trial-and-error and fine-tuning phase that is inevitable when switching to an agroecological system, the design of which is based on specific pedoclimatic features. This phase is easier to get through when it takes place in a collective setting (CUMA, research/innovation carried out collectively) (The Shift Project, 2024). For some, this journey will never be over, because the context (soil and climate, agricultural players, etc.) will continue to evolve over time.
Embarking on a process of transition and becoming an ecologist of one’s own plot of land requires consideration of a number of variables:
- the commitment and perseverance of players (motivation, learning, risk management, etc.). Switching to agroecological practices can also lead to a change in the perception of income-related risks. The reduction of external inputs (e.g., less dependence on pesticides) often leads to an increase in perceived production risks (Ewert et al., 2023).
- confrontation with technical, organizational, cognitive and ideological obstacles, in relation to the farm’s social, technical and ecological environment
- redefining what is valuable (types of performance, expected properties); new management methods (information, intervention thresholds) [Caquet et al., 2020].
There is a strong case for making visible the farmers who have succeeded in changing, or those who are in the process of doing so, so as to motivate change and empower the actors around them. Some farmers will only redirect their trajectories if they are sure that these new systems are viable. And all farmers should have the means, if they so wish, to embark on these transition paths. In these complex transformations, we need to multiply the number of feedbacks and use cases to demonstrate the diversity of transition trajectories.
Agroecological transitions involve many social elements, values and emotions, and these personal and subjective elements are often not brought to the fore, or even ignored and rendered invisible. Farmers themselves and their advisors are not necessarily comfortable (or accustomed) to sharing such information (Prost et al., 2022).
Farmers can face many obstacles in their transition to an agroecological ideal – which underlines the need for better recognition of systems in transition. Once the gap between the normative principles of agroecology and their application in real life is recognized, the question of whether a farming system can be considered agroecological has two different meanings (Dumont et al., 2021). Antoinette Dumont proposes a three-step approach:
- assessing whether farmers justify their daily practices by referring to agroecological principles when faced with dilemma situations where they have to choose between several agroecological practices that seem incompatible in their personal context
- assess whether a system, such as a farm, implements sufficient principles to be considered agroecological
- examine whether the poor implementation of a particular principle is due to a lack of personal interest in setting up an agroecological system, or to external obstacles or barriers.
The article does warn, however, that the different stages may lead to different results. Agroecological systems identified on the basis of their orientation towards agroecological principles may be poorly evaluated when measuring the performance of farming practices. Conversely, systems that perform well in terms of applied practices may not be oriented towards agroecological principles.
Transition won’t happen on its own
The farm is embedded in a socio-ecological and socio-technical system. While changes at farm level are essential, they can be hampered by the multiple organizations, strategies and representations of other actors upstream and downstream of the farm (Prost et al., 2023). Technical choices in the field are thus made in coherence and interaction with all the dimensions of the production system. This also implies that farmers must have access to competent, independent structures for agroecological transition. Trained personalities (technicians, advisors, etc.) do exist – perhaps not in large enough numbers – but they are not sufficiently accessible.
Farm transitions to agroecology also require navigation between different levels of organization (from the field to the food system) in order to take account of overlaps, synergies and antagonisms when combining innovations in the context of specific socio-technical systems, each with its own collective rules.
If, for technological solutions, the challenge is to produce them at an affordable cost for farmers, then, for changes in farming practices (systematic cultivation of legumes, significant increase in the number of crops in rotations per farm, culturax mixes, reintegration of polyculture, etc.), the issue is more closely linked to the presence of suitable industrial machinery (specific silo for legumes, small local abattoirs, hemp processing plant, etc.).
Nor can the greening of the food supply be thought of without the links and constraints linked to downstream agro-industrial production and territorial organization. Veterinary and insemination services, slaughterhouses, livestock equipment suppliers and repairers, and dairies are all dependent on the density of livestock breeders in a given region. The same is true for many plant-based industries, which rely on processing units within regional units (crushing plants, dehydration plants, storage, sorting, drying capacities, etc.) and which function thanks to well-established logistical circuits (The Shift Project 2024).
These threshold effects, counterbalancing a purely linear effect of agricultural and agro-industrial dynamics, also call into question the need for local agro-industrial tools. This complementarity between an affordable technological offering and a coherent primary processing industrial fabric will consolidate the two legs of the agroecological transition. It will be essential for players in the collection and processing industries to be coordinated on a local scale.
Insofar as everyone is expected to do their bit, we might expect para-agricultural companies to be judged on their ability to support their members’ initiatives, and to be sufficiently agile and adaptable to consider transitional approaches. In most cases, the regulatory requirements imposed on farmers do not apply to the players in the chain who buy their products or sell them inputs. As a result, a processing company is entitled to refuse to source from a farm that has taken the risk of transition, as it has less visibility over the exact volume of product that this farm can supply, the exact biochemical composition of the product, etc.
It is important to support farmers in de-risking their transition. Managing this risk is all the more important in the field of agricultural transition, as farming is by its very nature particularly exposed to risks (climatic risks, disease risks, market risks, geopolitical risks, etc.).
The organization of these risk reduction mechanisms will not happen spontaneously. It will require all the players involved to develop them, and public authorities to ensure that these processes accompany the higher risks taken by farmers. And these mechanisms will have to be thought through in a systemic way, because it is above all systems of innovations, combining different techniques and modes of organization, that will be able to respond both to the different challenges and to the diversity of specific local situations. The ecological transition needs to be organized, planned in the sense that it is thought out collectively, and negotiated and contractualized in order to remove obstacles one by one.
Current agricultural policy is concerned with managing crisis after crisis. The fact that we become agitated at every crisis shows a structural lack of resilience. Food resilience can be defined as the ability of food system players to feed their population in quantity and quality, whatever the hazards (and food resilience includes food sovereignty). Future developments (climatic, geopolitical, etc.) are bound to create new crises in the future, and past years have shown the extent to which crises are recurrent phenomena. A genuine political project for the future needs to be proposed and courageously carried forward, while navigating between regulatory constraints and the lobbying powers that are blowing headwinds.
The difficulties of scaling up
Digital tools can be used as encapacitors or facilitators of changes in practices. These digital tools need to be differentiated according to the stage of the farmer’s agroecological transition project, in the following way: Which digital tool(s) at which stage(s) of the agroecological transition? At the start of the transition, the farmer may be using generalist tools, whereas during the transition, he may be looking for more refined tools for multi-criteria evaluation of his farm.
Nevertheless, we can legitimately ask ourselves whether digital technologies trigger locking or unlocking mechanisms by enabling the scaling-up of agroecological practices, and whether or not they prevent a farmer from moving on to the next stage of his agroecological transition project.
Developing agroecology on a large scale is no easy task. Its ecosystems are more complex than those of traditional agriculture (Inrae, 2022). Agroecology is a context-specific approach. Management must be adapted to local climates, markets, soil types and species. This means that the knowledge, techniques and crop combinations developed in one region may not be viable elsewhere. The development of resilient agroecological systems requires considerable and varied knowledge. And, sometimes, we simply don’t have the necessary knowledge. It can take decades to gather and apply this knowledge to create functional agroecosystems in a specific context.
This is a major bottleneck for the agroecological transition. The sensors and measurement tools to collect data and move in this direction do exist, but although some form of mechanization (and possibly automation) is required if diversified, agroecological approaches are to be translated and amplified in the industrialized contexts where they are most needed, technology developers have not yet paid much attention to automating these systems (Ditzler et al., 2021).
The question of to what extent and how agroecology can be developed to become the main approach to a sustainable and resilient agri-food system remains important. Research agendas have been developed, but there is a need to better understand the economic viability of agroecological practices and possible policy interventions for a wide range of farming systems (Ewert et al., 2023). If organic farming is to develop in a way that avoids the many disadvantages of conventional production, it will be important to better understand the effects of conventionalization on organic farms. Moreover, many of these experiences are difficult to scale up, as they cannot resolve all the obstacles faced by smallholders, who do not have the same investment capacity or incentive to take risks as others (Avaria, 2020).
Scaling up in agroecology differs from scaling up in the Green Revolution. Scaling up in agroecology relies on a plurality of actions and pathways. The classic notion of scaling up, which assumes that the same innovative systems developed in a small area can be reproduced in larger areas by many other farmers simply by increasing the resources allocated, is therefore not compatible with agroecology. For agroecology, scaling up should be replaced by the idea of agroecological transition: a contextualized, gradual and diversified transformation of systems (Cote et al., 2022).
Scaling up also requires an environment conducive to change, with technical support and advisors able to accompany this transition. The number of employees trained in these subjects (in chambers, cooperatives and elsewhere) seems difficult to reconcile with massive growth in commitment to agroecology.
Digital technologies (especially the most technically advanced) and agroecology both have steep learning curves that make them difficult to implement. Complex technologies (such as so-called precision farming services) generally require technological expertise, with large devices and datasets requiring an understanding of GPS, mapping and data management tools. Agroecology generally requires thinking about complex systems involving plants, integrated weed and pest management, and practices such as longer rotations and cover crops. The combination of these moves in precision agroecology therefore inherits high barriers to adoption in terms of the new learning required (Duff et al., 2022).
Experimental devices (living labs or other types) will not suffice, as they do not offer enough combinations to test all levels of agroecological system interactions. There will always be a variability, even a gap, between the practices implemented in experimental set-ups and those in real farm conditions (Caquet et al., 2020).
Research that analyzes and supports farms in transition should focus more on the dynamics of change processes, valuing what happens on the farm. In particular, research should give more credence to on-farm experiments carried out by farmers (Prost et al., 2023).
On Farm Experiments (OFE) is a response to the inability of the microplot trials commonly used in on-farm research to provide sufficiently usable information to farmers, and to the fact that new solutions encompassing agroecological scales are needed to better guide their practices. These forms of experimentation have several advantages:
- agroecological research takes place in farmers’ fields and at scales that are useful to them, rather than on small experimental plots designed off-farm.
- the private interests of farmers and other participants in the dynamics of “On Farm Experiments” are explicitly recognized as a prerequisite for negotiating their alignment and establishing productive relationships.
- experimentation in OFE research is understood as a deliberate process of joint exploration whereby researchers and other actors engage closely with agricultural realities in order to align themselves with farmers’ ways of learning (Lacoste et al., 2021). The role of environmental actors will be to listen to the specifications and say what is feasible
Adaptability is an essential characteristic of social innovations that achieve scale and impact. OFE can be undertaken in multiple ways and in a wide range of institutional contexts, even when resources are limited. Diversity galvanizes the OFE community, as it shows that much can be learned by understanding the solutions that others have found in specific contexts.
On-farm experimentation (OFE) does not require digital technologies, but the increasing investment in and supply of digital technologies is a strong incentive to use them. On the one hand, digital technologies facilitate OFE. Not only do they make implementation and analysis much easier, but they also make it possible to ask new or different questions by collecting and recording vast quantities of information that would otherwise be impossible to access, even in marginal environments. On the other hand, OFE initiatives are catalysts for digital technologies. The OFE process can be used to test the usefulness of data-driven advice, adapting tools to actual rather than anticipated needs (Lacoste et al., 2021).
How can agroecology be quantified?
The big challenge of agroecology is its quantification and visualization. Evaluations are crucial in transition processes: they allow to frame the problem, adjust the changes undertaken and identify the paths actually taken. This is a real challenge since it involves defining a framework that brings together several dimensions in order to evaluate the farm performance patterns that need to be monitored over time during the transition period to agroecology (Prost et al., 2023).
Evaluations are more complex because they must take into account the increasing expression of specific local characteristics (e.g. pedoclimatic contexts) and individual farmer characteristics (e.g. personal values) that accompany agroecology. This requires flexible frameworks in which a subset of indicators can be selected according to the farm context
Agroecology, based on the principles of resilience and robustness, is part of a gentle and anticipated management of certain dysfunctions (uncontrolled infestations, incidence of weather accidents). Farmers need indicators to measure the impact and change of the system. These indicators would allow farmers to adjust their practices iteratively and more easily during transitions. And digital tools (in particular the measuring instruments we have discussed) have a role to play. Collective trials and errors also contribute to the transition and technology, via measurement and description, to facilitate feedback on small-scale projects.
This quantification is also a way to move away from the operation of the obligation of means (often used due to a lack of data) to move towards a logic of obligation of results to prove what agroecology does and that actions have indeed been carried out in satisfactory conditions (trusted third party). This opens up avenues for the economic development of agroecological practices (carbon premium, sector premium, CAP aid, etc.) and systems for mutualized coverage of the risk of agroecological practices via an insurance system. Let’s be crazy, we could propose CAP aid proportional to the farm’s self-sufficiency capacity (ratio between the inputs used and what it can market).
There is currently no agroecological label or specification that would allow us to state that we are at a particular point in the transition process. The HVE level 3 label is perhaps the closest thing to it, even if it could go further (I am talking about level 3 here and not levels 1 and 2, which are generally about compliance with regulations and what allows access to CAP aid). We are rather dealing with a whole gradation in agroecological practices. Agroecology is more of a results book than a specification, but it is not very well objectified. If we could quantify and project the results of agroecological practices, for example to know if the agricultural system is consistent with planetary limits, we could start to really change the scale of work. Even if the orientation towards a results objective makes sense for agroecology, one can still wonder whether, given the time available, we should not also take advantage of what already exists in terms of obligation of means.
Until now, the evaluation of agricultural systems has mainly been based on instantaneous measurements of sustainability indicators. It is therefore more difficult to confirm the presumed potential of agroecological agriculture to increase the resilience of farms. The added value of agroecological systems is much more relevant to calculate at the level of the farm as a whole than in financial terms in gross per hectare for example. By considering the farm as a whole, we take into account the balances and rotations at the farm level and we can thus integrate the savings in fertilization or phytosanitary products linked to the rotations and the diversity of plant production as well as the landscape mesh (The Shift Project, 2024). It will also be more relevant to take into account average yields by aiming for stability of production rather than productive extremes which are difficult to maintain in the medium or long term.
Many indicators developed by researchers reflect precise and quantified objectives, but most of them require too many measurements, are too specific or are calculated at time intervals that are unsuitable for farmers’ practices. These indicators have been proposed by many structures: the FAO’s TAPE tool, the PADV regeneration indicator, the IAE4 index developed by INRAE, the RITA indicator at European level, the IndicIADes platform created by the Institute of Sustainable Agriculture, the flora of Agroecology, the Decide or Klimrek tools around the carbon footprint. It could be relevant to have software bricks to be able to output performance graphs on the fly by outputting the results on each of these indicators. It is therefore necessary to find a compromise between the ambitions of research and the operational activities of farmers to ensure that data can be uploaded. Behind the 30,000 farms targeted to support the achievement of Ecophyto, there was initially the idea of having large databases (Ecophyto-PIC) to be able to infer which practices work. Let’s keep in mind that indicators of success and progress will evolve over time and policies will have to adapt.
The transition evolves slowly and digital tools can have difficulty measuring indicators that do not change quickly. This is particularly the case for variables with potentially slow evolution that are not easily detectable (e.g. soil carbon, soil compaction, biodiversity and soil health, changes in the composition of plant spices in pastures), but which can have significant impacts on the long-term sustainability of agricultural systems (Rosenstein et al., 2023). Perhaps then we should focus on a comparative assessment of mature systems (at cruising speed), and not systems in the process of transitioning to agroecological practices.
These indicators must also consider threshold effects. For example, diversity is always considered a factor of resilience, whatever its level, even if, beyond a certain threshold, diversity could complicate management too much for most farmers.
Finding a business model
Who will be ready to assume the cost of the transition? Who would be ready to support the derisking of this transition? There might be interest in looking at the cost of inaction or the hidden costs of agriculture. If we cannot demonstrate that what we pay on one side (a higher cost of a diet based on agroecological systems), we save on the other (by reducing the costs of soil and water decontamination, the number of fossil calories consumed for 1 calorie produced, health costs, etc.), it will be difficult to move forward.
In November 2023, the Food and Agriculture Organization of the United Nations (FAO) estimated these costs for France at 177 billion dollars (FAO, 2023). In 2024, a French report commissioned by several associative structures (including Secours Catholique, Solidarité Paysans, CIVAM and the French Federation of Diabetics) considers that the dysfunctions of our food system are compensated to the tune of 19 billion euros in France (Secours Catholique et al., 2024), almost double the budget allocated for ecological planning in 2024. This work also puts into perspective the figures for the agri-food, distribution and catering sectors, which certainly generate 31.5 billion euros in net profits, in France and for export, but which benefit directly or indirectly from 48 billion euros in public support, to which must be added the 19 billion in repairs, all at the expense of the community.
The difference between the figures put forward by these two studies is linked to different working methods. The FAO calculations are based on the concept of negative externalities, with the monetization of losses attributable to productivity declines (for example, the loss of productivity of an individual who falls ill) in relation to a gross domestic product (GDP) considered theoretically, with a projection up to 2100. In the French report, the negative impacts of the agri-food system are calculated using an accounting approach, i.e. public expenditure actually incurred to compensate or repair them. The cost of the impacts is calculated in accounting and collectively – due to taxes, contributions and duties – paid at the time the report was written. The authors argue that their figures are largely underestimated because several costs could not be quantified due to lack of available data. Several blind spots (not taken into account) were identified:
- On waste: the cost of collecting plastic waste on French territory, the health cost of exposing populations to microplastics or the impacts of microplastics on the environment.
- On air quality: only the agricultural link is taken into account, and the costs of care for other diseases linked to air pollution could not be counted due to lack of available data.
- On water: there are few decontamination or prevention measures outside the catchment area; we have not identified any expenses associated with the overexploitation of water resources.
- On biodiversity: only the agricultural link was taken into account in the calculation, but the other links in the food system also generate pressure on biodiversity (urbanization, transport, pollution, energy production, etc.).
- On soils: The impacts linked to erosion are also mainly assumed by private actors, by insurance and not by public spending.
- On society: several impacts do not represent costs, like the feeling of helplessness and anxiety of not being able to feed one’s children properly.
Other avenues need to be explored to transform business models, such as those approaching the functional economy. BASF is starting to deploy a new economic model around its Xarvio digital tool for plant protection. Farmers thus have the possibility of putting hectares of crops under an outsourced contract, the crop protection itinerary of which will be managed by BASF and its service providers agricultural work delegation companies for example). BASF guarantees a level of results and offers insurance to farmers if the result is not there. BASF commissions agricultural work delegation companies to intervene on the farmer’s plots based on the results of health risk prediction models. This new economic model allows both:
- use of the best possible equipment since it is shared by the agricultural work delegation companies
- a reduction in the use of plant protection products since they constitute a financial burden for the agricultural work delegation companies and BASF
- the possibility of taking a risk and not intervening for BASF by following the recommendations of its health models. As long as the amount to be reimbursed is lower than the cost of the phytosanitary products, it is interesting not to treat.
The transition to more agroecological ecosystems has a cost. Private actors must also get involved financially, for example with public-private co-financing mechanisms, so as not to leave public actors alone to support the agricultural economic model.
What level of agroecology do we expect?
When can a farm be considered agroecological? How can we compare agroecological systems in light of their multiple components and dimensions of interest? One farm may have significantly reduced the use of inputs while another may have completely rethought its routes. A farm may have saved water but to do so, may have increased its energy consumption. How can we weigh all these factors?
These questions call for putting the definition of agroecology on the table (we provided several elements at the beginning of the file) and discussing an agroecological cursor, meaning what we are prepared to accept in order to consider that a system is agroecological. Every system necessarily has impacts (direct, indirect, systemic). Precision agroecology, as those involved in digital agriculture might call it, improves the efficiency and optimization of agricultural systems but can perpetuate a model that uses mineral and phytosanitary inputs. Soil conservation agriculture offers a possible form of greening and brings important virtues in terms of biomass production and diversification but can use products such as glyphosate. Organic farming rejects the use of phytosanitary products and mineral fertilizers but authorizes the application of sulfur and copper components that can damage the state of the soil.
From the moment we measure the results brought by agroecological practices (at all the levels of organizations that we have discussed), perhaps we will be able to more easily set an acceptability cursor. Some indicators already do this in part. The PADV regeneration indicator requires a minimum score to be considered as in agroecological transition, and a higher score to be seen as in an agroecological system.
As a conclusion
A number of tensions are still embedded in the imaginary of our future agricultural systems. Making agroecological and digital dynamics coexist may be easier said than done given their divergent epistemological foundations. Besides that, the definitions of agroecology and digital technologies are themselves uncertain, with different interpretations leading to different results. It is therefore difficult to find common ground when the object(s) of discussion are not the same.
We must be careful not to advocate a systemic digitalization of agriculture as a miracle solution. The mixed legacy of past technological revolutions should not, however, be used as an excuse to put an end to innovation. On the contrary, it is an additional reason for all stakeholders involved to adopt a cautious approach to digital technologies. The compatibility of digital technologies with agroecological principles must therefore be assessed on a case-by-case basis to ensure their relevance to the agroecosystems, contexts and communities in which they will be applied (IFOAM, 2020; The Shift Project, 2024).
We must ensure that we understand the risks, so as not to embark on paths from which we cannot return. However, anticipating risks is essential to avoid poorly configured innovation (Bellon-Maurel et al., 2022). The four pillars of a responsible research and innovation approach must be mobilized: [1] anticipation (of risks), [2] inclusion (many actors around the table), [3] reflexivity (to assess whether mutually beneficial trajectories are followed) and [4] responsiveness (ability to respond quickly to problems caused) must be regularly questioned. This approach and the results that emerge from it must be made transparent and must seek to mobilize as much as possible diverse colleges of actors (in terms of skills and work discipline). These collaborations can be facilitated with common experimental platforms and central databases, interdisciplinary training and institutional cooperation and networks (Storm et al., 2024).
Once again, there is not ONE digital but MANY digital. Focusing only on one form of digitalization would amount to supporting only one form of greening of agriculture (Schnebelin et al., 2021). The question of whether digital and agroecological trajectories are compatible is perhaps more relevant to ask in the form: Which digital(s) to support agroecology?
An answer is certainly to be sought on the side of the diversity of solutions. That is to say via the development of a digitalization that preserves a balance between monitored/controlled systems and more disordered systems, that is to say including a great diversity of biotechnical conditions and farmers’ strategies. A digitalization oriented towards a constant search for balance and resilience rather than as a simple means of optimizing these complex systems. It remains nevertheless important to continually question whether it is really possible to have a diversified agricultural system with agroecological practices on one side and on the other a massified conventional system.
We must assume that at present, it is difficult to really know what all digital technologies do, in particular because many offer varied functions. There are no real digital tools dedicated to agroecology, but they all have an agroecological component, if only because they allow for optimization and more efficiency in inputs.
Innovation dynamics remain unbalanced. The push for digital technological solutions is disproportionately greater and much more unified than agroecology (IFOAM, 2020). We have mainly discussed technological innovation in this issue, but I would like to remind you that other forms of innovation exist (organizational, social, agronomic) and that it may well be the combination of these innovations, in the form of innovation systems, that will allow us to move forward.
Digital technology is mainly questioned on its ability to support the transition to agroecological systems. The question of whether these technologies will still have a place in an agriculture that has successfully transformed itself is also legitimate (we could talk here about the place of digital technologies in cruising mode). If technological dependencies are still too strong in future agricultural systems, crises that could impact the functioning of technologies (flow disruption, energy limits, etc.) would have a cascading effect on our relationship with agro-ecosystems. We must therefore ensure that the resilience capacity of the agricultural system is at the heart of any decision to deploy technological innovation
In some cases, digital technology will remain a tool for managing these new agroecological practices, but we can imagine others where digital technology is seen as a tool for waiting for the transition. If the systematization of agroecological infrastructures (such as hedges) is a good thing, sensors could be useful for monitoring the evolution of the hedge’s autonomy in terms of its ability to be correctly integrated into its ecosystem and to have reappeared and stabilized a whole bunch of biological regulations. Once autonomy is achieved, the sensor could be removed.
Another example with plots in which we would seek to remove datura. Today, this suppression can be done by selectively using glyphosate on datura with prior mapping of datura infestation in the plots by drone. Tomorrow, if we are able to multiply the caterpillar of the butterfly that eats datura, we may only need to organize localized releases of these caterpillars in the plots (by drone for example). The day after tomorrow, the evolution of genetic selection may have ensured that datura no longer has any poison and we would then no longer need to intervene. During the intermediate phases, digital technology is supposed to reduce the degree of intervention.
Scaling up agroecological systems is complex because the solutions are very much located in time and space, and therefore very dependent on local conditions, necessarily falling within the framework of adaptive management. The genericity in the implementation of agroecological transitions may not lie in technical solutions, but rather in the frameworks and tools to promote the adaptability of stakeholders.
Finally, and above all, we fundamentally need to give farmers back their space and power to act. In the end, we talk a lot about digital technology and agroecology, but in the end, little about farmers. The forms of digital technology and the forms of agroecology are very diverse. But the work is also very diverse, juggling between health issues, physical workload, relationships with animals and crops, and mental workload. It is the farmers who will be able (or not) to involve digital technology in the context of their agroecological project. The user has an important place in the way he uses the tool, and he can divert the tool to his advantage.
As long as a clear direction has not been given on the state of agricultural systems and the production levels expected in France in the years to come, it will be difficult for agricultural stakeholders to position themselves clearly and initiate a transition, possibly equipped with digital tools.
Additional bibliography for interviews
Aboueldabad, G. (2023). Mettre en commun par le partage numérique des connaissances : enjeux et potentialités pour les mouvements de l’agroécologie. Université de Montpellier
Annales des mines (2022). Enjeux numériques. L’agriculture Numérique.
Altieri, M., et al. (2017). Technological Approaches to Sustainable Agriculture at a Crossroads: An Agroecological Perspective. Sustainability.
Avaria, R.W.C. (2020). Searching for sustainability in the digital agriculture debate: an alternative approach for a systemic transition. Teknokultura. Revista de Cultura Digital y Movimientos Sociales.
Bellon-Maurel, V., et al. (2022). Digital revolution for the agroecological transition of food systems: a responsible research and innovation perspective. Agricultural Systems.
Bellon-Maurel, V., et al. (2022). Digital technology and agroecology: opportunities to explore, challenges to overcome. Agriculture and Digital Technology: Getting the most out of digital technology to contribute to the transition to sustainable agriculture and food systems, INRIA
Burns, S., et al. (2022). Exemplary features of digital tools for agroecology: A global review. Agroecological Transitions Programme. Policy Brief.
Caquet et al. (2021). Agroécologie – Des recherches pour la transition des filières et des territoires. Chapitre 6 : Contribution des agro-équipements et du numérique à l’agroécologiee. Renforcer la prise en considération du vivant. Edition Quae.
Centre National d’Agroécologie (2023). Livre Blanc de l’Agroécologie.
Challand, M. et al., (2024). Supporting Sustainable Agroecological Initiatives for Small Farmers through Constraint Programming. Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence (IJCAI-23)
Colliaux, D., et al. (2022). Computational agroecology: should we bet the microfarm on it? Eighth Workshop on Computing within Limits
Cote, F.X., et al. (2022). Levers for the agroecological transition of tropical agriculture. Agronomy for Sustainable Development.
Cumulus Consultant (2021). AgroEcoTech: How can Technology Accelerate a Transition to Agroecology? Report for the Soil Association
Daum, T. (2021). Farm robots : Ecological utopia or dystopia ? Trends in Ecology and Evolution.
De Graeuwe, M. et al., (2025). Serious Game for AgroEcology (SEGAE) : How much can be delivered with a 4-hour lesson. Agricultural Systems, 224.
De Wit, M.M. (2021). Can agroecology and CRISPR mix? The politics of complementarity and moving toward technology sovereignty. Agriculture and Human Values.
Ditzler, L., & Driessen, C. (2022). Automating Agroecology: How to Design a Farming Robot Without a Monocultural Mindset? Journal of Agricultural and Environmental Ethics.
Dumont, A., (2021). Bridging the gap between the agroecological ideal and its implementation into practice. A review. Agronomy for Sustainable Development.
Duff., H., et al. (2022). Precision Agroecology. Sustainability.
Ewert, F., et al. (2023). Agroecology for a Sustainable Agriculture and Food System: From Local Solutions to Large-Scale Adoption. Annual Review of Resource Economics.
FNCUMA (2024). Plaidoyer pour une mécanisation responsable, durable et vivable de l’agriculture française. https://www.cuma.fr/resource/plaidoyer-pour-une-mecanisation-responsable-durable-et-vivable-de-lagriculture-francaise/
Gross, H., et al. (2024). Agor@gri 1 : Les médias sociaux numériques au service de la transition agroécologique. Innovations Agronomiques.
Hilbeck, A., et al. (2022). Aligning digitalization with agroecological principles to support a transformation agenda. Einstein Center. Digital Future.
Gascuel-Odoux, C., et al. (2022). A research agenda for scaling up agroecology in European countries. Agronomy for Sustainable Development.
HCIAS (2023). A Sustainable Future for Small-Scale Farming? Qualitative Research on Agroecology and Digitalization
HLPE (2019). Agroecological and other innovative approaches
Hostiou, N., et al. (2023). Contributions des technologies de précision pour améliorer les conditions de travail dans les élevages agroécologiques : questionnements et perspectives. Natures Sciences Sociétés.
Huck, C., et al. (2024). Environmental assessment of digitalisation in agriculture : A systematic review. Journal of Cleaner Production, 472.
IFOAM (2020). Agroecology & Digitalisation. Traps and opportunities to transform the food system.
Ingrand (2018). Opinion paper: ‘monitoring te salutant:’ combining digital sciences and agroecology to design multi-performant livestock farming systems. Animal.
Inrae (2022). Digital Routes to sustainability in Agriculture. Society.
Inria – Inrae (2022). Agriculture et Numérique. Tirer le meilleur du numérique pour contribuer à la transition vers des agricultures et des systèmes alimentaires durables.
Inria (2023). Le numérique est-il un progrès durable ? Pour la science.
Klerkx, L., & Villalobos, P. (2024). Are AgriFoodTech start-ups the new drivers of food systems transformation? An overview of the state of the art and a research agenda. Global Food Security, 40.
La Rocca, et al., (2024). Estimating The Carbon Footprint Of Digital Agriculture Deployment: A Parametric Bottom-Up Modelling Approach. Journal of Industrial Ecology.
Lacoste, M., et al. (2021). On-Farm Experimentation to transform global Agriculture. Nature Food.
Leveau, L., et al. (2019). Information and Communication Technology (ICT) and the Agroecological Transition. In Agroecological Transition : From Theory to Practice in Local Participatory Design.
Liebert, F., et al. (2022). Farm size affects the use of agroecological practices on organic farms in the United States. Nature Plants.
Lucas, V. (2021). A “silent” agroecology: the significance of unrecognized sociotechnical changes made by French farmers. Review of Agricultural, Food and Environmental studies.
Martin, T., & Schnebelin, E. (2023). Agriculture numérique : une promesse au service d’un nouvel esprit du productivisme. Natures Sciences Sociétés.
Paget, N., et al. (2022). Traque des innovations numériques au service de la transition agroécologique au Bénin. Agricultures.
Plumecocq, G., et al. (2018). The plurality of values in sustainable agriculture models: diverse lock-in and coevolution patterns. Ecology and Society.
Prost, M. et al., (2022). How could social media support farmers concerned with sustainability issues? The Journal of Agricultural Education and Extension.
Prost, L., et al. (2023). Key research challenges to supporting farm transitions to agroecology in advanced economies. A review. Agronomy for Sustainable Development.
Rogel-Gaillard, C., and Sainte-Marie, J., (2024). Agroécologie et numérique pour répondre à la transition écologique. Techniques de l’ingénieur.
Rotz, et al. (2019). The politics of digital agricultural technologies : a preliminary review. Sociologia Ruralis.
Rozenstein, O., et al. (2023). Data‑driven agriculture and sustainable farming: friends or foes? Precision Agriculture.
Secours Catholique et al., (2024). L’injuste prix de notre alimentation. Quels coûts pour la société et la planète.
Schnebelin, Éléonore, Pierre Labarthe, and Jean-Marc Touzard. 2021. “How Digitalisation Interacts with Ecologisation? Perspectives from Actors of the French Agricultural Innovation System.” Journal of Rural Studies. https://linkinghub.elsevier.com/retrieve/pii/S0743016721002205.
Schnebelin, E. (2022). Linking the diversity of ecologisation models to farmers’ digital use profiles. Ecological Economics.
Schnebelin, E. (2024). Le développement du numérique en agriculture : vers une transformation agroécologique ? ISTE.
Shelton, S., et al. (2022). Critiques of digital tools in agriculture. Challenges & opportunities for using digital tools to scale agroecology by smallholders. Agroecological Transitions Programme. Policy Brief.
Soulignac, V. et al. (2023). What Lessons Can Be Learned for the Agroecological Transition From the Use of Social Media in Preventive Medicine? International Journal of Agricultural and Environmental Information Systems
Storm, H. et al. (2024). Research priorities to leverage smart digital technologies for sustainable crop production. European Journal of Agronomy, 156.
Sullivan, S. (2023). Ag‑tech, agroecology, and the politics of alternative farming futures: The challenges of bringing together diverse agricultural epistemologies. Agriculture and Human Values.
Terra Nova (2023). How to achieve the agroecological transition?
The Shift Project (2024a). Quelles technologies pour une agriculture bas carbone, résiliente et prospère ?
The Shift Project (2024b). Pour une agriculture bas carbone, résiliente et prospère – Note sur les technologies agricoles
Van Der Velden, D., et al. (2023). Cyborg farmers: Embodied understandings of precision agriculture. Sociologia Ruralis.
Van Hulst, F.J., et al. (2020). Using co-constructed mental models to understand stakeholder perspectives on agro- ecology. International Journal of Agricultural Sustainability.
Viana, C.M., & Carvalho, D. (2024). Unveiling historical agroecological patterns through artificial intelligence (AI) and Geographic Information Systems (GIS). Agile : GIScience Series, 5.
Wepner, B. et al. (2025). How can digitalisation support transformation towards sustainable agri-food systems? Scenario development in Lower Austria. Agricultural systems, 224.
Wezel, A., et al. (2020). Agroecological principles and elements and their implications for transitioning to sustainable food systems. A review. Agronomy for Sustainable Development.
Wittman, H., et al. (2020). Advancing food sovereignty through farmer-driven digital agroecology. Environmental and ecology.
Zingsheim, M.L., & Doring, T.F. (2024). What weeding robots need to know about ecology. Agriculture, Ecosystems, and Environment, 364
People Interviewed
| Name | Structure |
| Florence Amardeilh | Elzeard |
| Véronique Bellon Maurel | Inrae |
| Hermine Chombart de Lauwe | CNRA |
| Alban Bouvy | MicrofarmMap |
| Rémi Dumery | Agriculteur |
| Bertrand Gorge | Triple Performance |
| Héléne Gross | ACTA |
| Nathalie Hostiou | Inrae |
| Oriane Lafon | Landfiles |
| Simon Moinard & Basile Ploteau | Mobilab – AgroTIC |
| Ionna Mouratidiou | Isara Lyon – Path2dea Project |
| Xavier Reboud | Inrae |
| Claire Rogel-Gaillard | Inrae |
| Vincent Levavasseur | Ver de terre Production |
| Evelyn Reinmuth | Université de Hohenheim – Path2dea Project |
| Eleonore Schnebelin | Inrae |
| Philippe Vismara | Inrae |
| Stéphane Volant et Manon Lebrun | CUMA Ouest – projet Agroop |
| Paule Yacoub | Science PO – CNRS |