people on the planet by 2050
According to the latest figures from the United Nations, the world’s population will number 9.77 billion in 2050, compared with today’s figure of 7.55 billion, and by 2100 the planet will be playing host to 11.18 billion inhabitants. This is a huge challenge – both in environmental terms and for the agricultural sector. Obviously, it is simply not feasible to go on expanding the area of arable land for ever, but within the above timeframe the agriculture sector will be called upon to increase food production by 50% in order to feed a population that is both increasingly demanding in terms of daily calorie consumption and more and more concerned about the conditions under which food is produced. For instance, people in rich countries are now pushing farmers to move to ‘zero residue’ pesticides in food. To meet these two seemingly contradictory demands, AgTech companies are developing tools to help farmers transition to what has come to be called ‘precision agriculture’ – using connected sensors, drones and satellite imagery to make a detailed analysis of a farmer’s fields so as to optimise the use of fertilisers, pesticides and water. The notion of crops being monitored in close to real-time via satellite, and tractors steered to the right spot by data feeds is no longer the stuff of science fiction. The Big Data approach is now linking up with another area of technology: digital simulation.
Simulating plant growth (a relatively old research topic)
MODéLISATION DE PLANTES EN 3D
As long ago as the 1970s, scientists were trying to reproduce plant growth on a computer. At the time Philippe de Reffye – an agronomy PhD and a researcher at the French Agricultural Research Centre for International Development (CIRAD) – and his team were pioneers in this new field. In fact, it was only in the 1980s, when computing power and 3D started to become available, that the new discipline really took off. Explains Marc Jaeger, a research engineer at CIRAD: “The idea was to describe vegetable growth – plant structure – using statistics and to link that with computer-aided design and computer imagery. Our work gave us some spectacular visual results. It was only afterwards that we incorporated scientific input from biology and agronomy.” The growth models developed by the researchers from the French combined research group AMAP – ‘botAnique et bioinforMatique de l'Architecture des Plantes’ (botany and bio-computing for plant architecture) – have for instance been used in movie productions to generate realistic scenes of fields and forests and they are being used today by developers, architects and landscape gardeners. Montpellier, France-based startup Bionatics uses these digital models for its software packages, which simulate environments for developers, town planners and the like. They use the software to visualise the landscape, or an urban environment, ten or twenty years on and see what it will look like when the vegetation has grown.
Research engineer at CIRAD
"The idea was to describe vegetable growth – plant structure – using statistics and to link that with computer-aided design and computer imagery.
The growth models were originally purely graphic, but were later improved to incorporate plant behaviour linked to biology. The scientists focused on simulating the internal functioning of a plant’s organs and the growth of its fruit. Quite apart from the visual aspect, simulating with precision the plant’s biology, the entire metabolism of a plant, is a way of assessing exactly how much it is likely to produce in an open field or inside a greenhouse. However, achieving these simulations was a highly complex task for the researchers. Marc Jaeger underlines: “Modelling the yield of a plant might seem simple. The relationship between plant growth and temperature is a straight line. Similarly, water volume and plant production are also a straight line, but the plant itself is a dynamic system. When the temperature rises, the quantity of water usually diminishes. We need to see plants as complex dynamic systems, a bit like a self-driven electronic circuit, with new elements adding to the circuit throughout its growth. In this sense a plant, or a tree, is a dynamic system, and it’s fascinating to model.”
The many parameters impact each other, making it extremely difficult to calculate outcomes. Although the experts are able to model the growth and flowering of a sunflower plant with a high degree of precision, the problem becomes one of an entirely different scale when they need to work out how many hundred kilograms a certain field will produce, given that the farmer will sow 60,000 to 80,000 seeds per hectare. Simulating several hectares of crops hugely increases the number of calculations needed and raises many questions in terms of the variability between plants and as regards modelling the interactions between them.
Despite all these challenges, algorithms capable of supplying reliable data on the plot yields have been in existence since the last decade.
In addition, understanding the mechanics of plant development opens the door to other applications. It is also feasible to analyse a plant’s state of health just by looking at a photo. Marc Jaeger points out: “The cost of obtaining images, whether via satellite or drone, have come right down recently and we’re now receiving an increasing number of requests to diagnose the state of a tree in a city from a photo or estimate the level of tree dieback due for example to the advance of the armyworm.” This knowledge of plant taxonomy has led to the development of the PlantNet mobile app, which can recognise a plant just from a photo.
SIMULATING NATURAL SPACES
While nowadays the techniques of using simulation models for the main crop varieties have been thoroughly mastered, research into simulation still has some major challenges to overcome. Says Jaeger: “Simulating a natural space is still a real challenge and we’ll certainly need many more years of research in order to complete the modelling of all interactions between plants or the trees in a forest, if we ever do manage it!” Modelling the development of plants of various species that are in competition with each other is a highly complex task for researchers but it is going to be crucial. We know that in the future we will no longer see mono-crop agriculture. Farmers will increasingly grow a number of crops together in order to benefit from the insect-repellant properties of some species, which will in turn reduce the use of pesticides and other agrochemicals. In order to scale up the practice of growing this kind of crop combinations, it will be vital to have a deep understanding of the interactions between species and to be able to simulate them so as to choose the right species and decide where they should be planted in a given field so as to optimise plant interactions.
Simulation allied with Big Data
Chairman of CYBELETECH
With machine learning we are now able to develop forecasting models that incorporate highly random complex phenomena which we can’t model.
One of the most avid fans of simulation in the farming sector in France is Christian Saguez. A former director of industrial relations at the French National Centre for Space Studies (CNES), and then at the French Institute for Research in Computer Science and Automation (INRIA), he founded Simulog, a software publishing company producing simulation software for industry. Since then, as President of the Teratec association, a hub of over seventy players in France in the field of high performance calculation, this strong proponent of simulation firmly believes in the benefits of computer simulation in agriculture. Vice-President of AgreenTech Valley – both a federation of sector players and a physical campus designed to test out new ideas – he is also Chairman of CybeleTech, a startup which markets the simulation models that have been developed by Digiplante, the joint research team comprising scientists from INRIA and the École Centrale Paris (ECP) – a French institute for research and higher education in engineering and science – which he co-founded with Philippe de Reffye in 2003. “We want to exploit the full potential of digital technologies all along the plant chain, from seed companies to farmers, cooperatives and food processors, all the way through to consumers,” explains Saguez. This simulation expert believes that the synergy between simulation and Big Data is what will generate the greatest value for farmers. He explains: “Our expertise is based on two facets of technology: on the one hand modelling and simulation of plant growth, with mechanical models that can simulate plant growth, taking into account their environment – soil, climate etc. Today we have tried and tested models for the main cereal crops such as wheat, maize, soya, market gardening plants such as tomatoes and cucumbers, and also crops such as rice and coffee. The second aspect is what is generally known nowadays as ‘data’. Today we have sensors everywhere, we’re beginning collect sets of historical data and we have a lot of data from drones and satellites. With machine learning we are now able to develop forecasting models that incorporate highly random complex phenomena which we can’t model.”
simulation AND big data
The Magestan project is a perfect illustration of this two-pronged approach spearheaded by French cooperative organisations working in the fresh fruit and vegetables sector. The basic idea is to develop a fully automated greenhouse. The objective is to optimise the effectiveness of inputs and reduce losses so as to subsequently produce more tomatoes and cucumbers, and at the same time improve the taste of the produce. Christian Saguez points out just how difficult it is to model the entire operations of a greenhouse: “A greenhouse is a highly complex environment. You have temperature, nutrition, water level, ripeness – a lot of interwoven systems. We’ve managed to reach an overall optimal system by linking up equation models which on the one hand simulate the growth of tomato plants and on the other use sensor data to provide information on the greenhouse environment.”
Combining different types of digital models enables the startup to provide tools that aid decision-making to a range of players in the agriculture sector so that they can select the seed or plant varieties they use. The company can also meet the needs of precision agriculture, as simulation tools can be used on a smaller scale, within a given plot. They can tell the farmer which areas of a given field most need fertiliser, where the plants most at risk from water stress are, and so on. Big Data and digital simulation provide ways and means of getting close to ‘zero residue’ farming, which is today one of the greatest areas of consumer demand in developing countries. These techniques can also be used as forecasting tools for producers to help them deal with volatility in the agricultural markets. “It’s very important to have information on the protein content of a cereal because that’s what decides the market value of the harvest. If farmers know the quantities they’ll be able to produce and the protein level their harvest will yield, they’ll be able to take the right decisions,” argues Saguet. CybeleTech is also involved in the Smart Agriculture System (SAS) project, whose aim is to develop a forecasting system for wheat crop yields at field level.
Technology serving developing countries
OPTIMISING CROP PRODUCTION IN DEVELOPING COUNTRIES
Tools that help to grow better crops enable farmers to obtain the best production quality and yields and also provide a means for farmers in developing countries to increase the volumes and yields of their agricultural crops so as to meet the increasing needs of the population. “We’re starting to work in Africa, in Senegal for example, and we have contacts in South America,” reveals Christian Saguez, underlining: “These countries need to improve their agricultural techniques, they need to move towards sustainable agriculture so as to avoid depleting the soil by growing the same crops for five years in succession. On top of that, as we’ve seen recently in the coffee market, where market value is closely linked to the quality of the harvest, agricultural production has to cope with highly fluctuating global markets. Producers need some guarantee as to the quality of their expected crop. Africa is moving very fast towards using digital technology, and agriculture is one of the sectors that will benefit most. Producers in those countries are crying out for solutions that will help them take the right decisions.”
Simulation is now at last coming out from the research laboratories and, in combination with Big Data, now finding a market. As an example, CybeleTech posted growth of 50% in 2017 and aims to double its revenues in 2018.