Agriculture connectée ni arnaque ni panacée !

The size of the investments required, and their uncertain profitability. On this theme, the author’s argument is mainly based on the example of milking robots and expensive equipment for geolocalised precision agriculture. However, this question already arises for many traditional “unconnected” equipment. And the author seems to forget that many of the most innovative tools rely on the use of low-cost sensors, thanks to the popularization of Internet of Things technologies already well amortized on their consumer applications. Even on purely software offers, available on annual subscription without any long-term commitment or heavy initial investment. The majority of offers based on agronomic models, such as those of itk, only have a recurring cost of a few hundred euros per year and per farm, and an initial investment cost that is only slightly higher. These are easily bearable costs for any type of economically sustainable farm, and can be waived by the farmer at no cost from one day to the next in case of dissatisfaction.

The dependence into which these new techniques could lead their users. This is indisputable, but the best answer was given in the same chapter: B. Tisseyre, head of the AgroTIC specialization at Montpellier Supagro, recalls that his grandfather, a winegrower, had been very hesitant to replace the horse by the tractor for this reason… forgetting all the dependencies that the breeding of draught horses generated! Is it better to depend on a new technique or on the constraints of one’s natural environment? This is a choice that only farmers themselves can make with full knowledge of the facts, and the author only examines one side of the issue.

Finally, V. Tardieu wonders a great deal about the risk that these innovations will lead agriculture into a “productivist” spiral (the swear word that is enough to discredit any agricultural technique without any other form of process).  This is understandable for the most expensive among them, such as milking robots, which effectively encourage the increase in the size of farms in order to amortize the equipment over greater production. But many of the recent techniques of precision farming are inexpensive, and potentially beneficial for all forms of agriculture. On this subject, we should therefore ask ourselves where is the chicken and where is the egg: are new techniques intrinsically conducive to intensification? Or are “alternative” farmers too reticent towards innovation to appropriate it and adapt it to their uses? A debate dodged by the author, whose critical spirit we would like to see exercised with the same acuity towards more traditional techniques, such as simplified tillage or extensive livestock rearing.

He regrets the opacity of the agronomic models on which these ADOs are based, when they are produced by private companies. That is true, but this is a false debate. In the case of statistical models, derived from Big Data techniques, the algorithms have no agronomic significance that an expert can assess. Only by comparing the results of this model with actual data sets can the validity of the ADO really be judged. In the case of so-called mechanistic (or process-based) models, the algorithms do have an interpretable meaning, since they describe the fundamental physiological mechanisms of the plant (or its diseases and pests). However, examining these algorithms would be of little value to a potential user. Given the complexity of these mechanisms, a model containing an error or an abusive approximation detectable by an expert would have no chance of yielding acceptable results for marketing. Again, it is only by comparing the model simulations with actual data that the quality of the ADO can be judged. In this field, private companies need industrial secrecy to defend their R&D investments, because a model is not patentable. Requiring the publication of algorithms would therefore be tantamount to excluding them from the game, without any benefit to the end user. On the other hand, certification of these ADOs by a public body would be a real step forward. Curiously, V. Tardieu mentions this subject, without being particularly favourable to it. However, it would be the best solution to improve the transparency of these tools. It could even be a “win-win” operation for companies producing ADOs, if this certification is an opportunity for them to test their models on the databases of Research Institutes and Technical Institutes.

This chapter is an opportunity to insinuate once again that ADOs are part of an intensive agriculture logic. But curiously, the example chosen to illustrate this problem clearly shows that the blocking factor is not at the level of ADO producers, but of research itself. The researcher questioned regrets that current fertilization tools assume that crop nitrogen requirements must be met continuously, whereas it would be possible to reduce nitrogen doses on wheat by accepting temporary deficiencies during the least critical periods. But this claim is based on research not yet finalized at the time of writing, so it is not surprising that ADOs do not yet take this into account. Arvalis is planning to incorporate this new knowledge into its future tools… and private companies will be happy to do the same, if they have access to all this work, including experimental data. More generally, this chapter could have been an opportunity to question the effectiveness of the French agricultural R&D model, which relies heavily on the complementarity between basic public research and the applied R&D of the Technical Institutes: a system in which private companies sometimes have difficulty integrating.

V. Tardieu is right to put forward the question of the benefit to the farmer, but it is regrettable that the question of the benefit to the consumer never arises. On this point, he remains in the classical bias of the urban vision of agriculture, which often forgets that agriculture is not an end in itself, but a production activity that aims to provide a basic necessity: food. In his “Tale No. 3”, he contrasts connected farmers with a Hérault shepherd, who does not need any sensors to monitor his animals remotely, as he prefers to follow them continuously in the garrigue. Very well, but apart from the fact that not all breeders necessarily want to adopt his way of life, we would like to know at what price he sells his sheep, and for what clientele. Moreover, the tools of connected farming are not only aimed at improving farmers’ income and living comfort. They also make it easier to respond to strong consumer demand for traceability of the origin and production method of their food. V. Tardieu’s analysis is mainly based on economics. He seems to forget that farmers’ work is also largely conditioned by consumer expectations and regulatory demands, and that any tool that can facilitate the response to these demands is welcome, even if it does not bring in any financial return.

It echoes concerns about the appropriation of agricultural data by agribusiness groups, or global Big Data luminaries such as Google or IBM. Of course, the risk exists, but in the agricultural field it is not unavoidable. Innovation in this field is mainly driven by start-ups, and French companies are particularly well placed in this field. Large groups not specializing in agriculture are aware of the difficulty of penetrating this very particular market, as demonstrated by itk’s partnership with the American telephone operator Verizon, for the development of its agricultural offer. Even majors in the sector such as John Deere have given up the fantasy of a private platform that would centralize all their customers’ data and the service offer that can be associated with them: they now open up (with the farmer’s consent) their data to producers of connected services. Precision agriculture is therefore moving towards a much more decentralized model than that of consumers’ personal data, where the famous GAFA (Google Amazon Facebook Apple) have constituted impregnable citadels. In this networked mode of innovation that is beginning to take shape, it is much easier to control the circulation and use of data, and to leave control to the farmer. The American Farm Bureau’s American charter, and in Europe the RGDP (General Personal Data Regulations), are moving in this direction. In France, the Ministry, the Chambers of Agriculture, and the technical institutes are also very active on this subject. Of course, we must remain vigilant, but conversely, too many restrictions could be counterproductive. Overly restrictive regulations or case law would primarily penalize SMEs, which are more fragile than large groups… and could therefore encourage a recentralization of connected agriculture among the major global players.