Mathematical Modelling and Artificial Intelligence in Agriculture

Agricultural issues have been critical for mankind since Neolithic times. The current period is no exception, and is even revealing new issues and new vulnerabilities that societies must face. The need to feed a fast-growing population is now accompanied by the need to develop farming methods that are more respectful of the environment, the constraints of climate change and the demand for quality and relocalized production. The recent health crisis has reminded us that even in the most advanced societies, tensions on food markets can quickly re-emerge, and that agriculture remains one of France's key strategic areas.

To meet these challenges, a new agricultural revolution is underway. It benefits greatly from the contribution of new digital technologies and artificial intelligence. The first objective of this article is to present the mathematical foundations, methodologies and techniques involved in the digital transformation of the sector. The second objective is to illustrate the implementation of these technologies and how they affect or will affect agriculture. We can identify two main scientific axes on which new digital technologies for agriculture are being developed: mathematical modeling of processes and artificial intelligence based on agro-environmental data.

In the first part of this article, we show how understanding how plants and crops interact with their environment enables the development of integrative models describing growth dynamics and yield constitution. The scientific issues are those typically encountered in modelling: what are the right scales of description (temporal and spatial), which processes need to be taken into account depending on the target objective, what is the right compromise between precision and robustness, how to parameterize the model so that it corresponds to the situation of interest? Once the model has been built and parameterized, it can be simulated to produce yield forecasts, or optimized to determine the best cultivation practices.

In the second part of this article, we'll see that agriculture is no exception to the trend towards an explosion of data: more and more sensors in the field or on intelligent tractors, satellite or drone imagery, pedoclimatic data, agricultural statistics... This wide availability of data paves the way for statistical learning and new artificial intelligence methods.

Finally, we'll conclude by outlining the major trends ahead, and highlighting some of the many scientific and technological challenges still to be met.