What is research infrastructure?

Research infrastructure refers to shared equipment and services that enable scientists to carry out their work.

This may include:

• scientific equipment or instruments;
• resources such as scientific collections, archives, and libraries;
• virtual infrastructure such as databases, computer systems, and communication networks.

In most disciplines, the use of research infrastructure has become an essential requirement for scientific competitiveness and international influence.

A different way of observing: phenotyping in the service of agroecology

Phenotyping consists of measuring the observable characteristics of a living being (size, color, yield, disease resistance, etc.). These characteristics depend both on its genetic makeup (genotype) and the environment in which it lives. They are measured at all levels of life, from the molecule to the entire organism, and even its ecosystem.

The combination of robotization and high-throughput measurements has made it possible to automate the data acquisition process using sensors, cameras, robots, and analysis software. This automation results in more data (known as high-throughput data), which can be exploited more quickly (sometimes in real time) and is more accurate than manual measurements.

But the ambition of infrastructure projects goes far beyond simply increasing throughput: it is about diversifying the characteristics observed to identify new agroecological indicators. The data collected contributes to research aimed at better understanding and predicting how plants and animals react to stress (water, heat, etc.), climatic hazards, disease, changes in diet, and interactions with other species. All this information is necessary to adapt agricultural practices and enhance the sustainability of systems (reducing greenhouse gas emissions, limiting the use of synthetic inputs, improving animal welfare, etc.).

PATASEL project: tools for more sustainable livestock farming

Livestock farming systems are evolving: they no longer focus solely on economic performance, but seek to balance efficiency, health, animal welfare, and a reduced environmental footprint.

The PATASEL project, supported by the LiPH4SAS research infrastructure, aims to provide researchers with equipment capable of measuring new key characteristics for agroecological transition: greenhouse gas emissions, feed efficiency, social behavior, health, and welfare.

Since the project began in 2023, around 15 multi-species pieces of equipment have been funded within INRAE experimental units. Depending on the case, the equipment is purchased from commercial suppliers or specially developed to meet the needs of the research.

Measuring greenhouse gas emissions to combat climate change

For small ruminants, portable chambers (or PACs, Portable Accumulation Chambers) can be used to measure the greenhouse gas emissions of 12 animals simultaneously. The principle consists of placing the animal in a closed chamber and calculating the total gases emitted by each animal. This is a portable system that can therefore be transported to different experimental sites.

GreenFeed equipment has been financed for cattle. This innovative tool measures methane and carbon dioxide emissions by capturing the breath of cattle at the feeding trough and is now considered the gold standard.

The data collected and analyzed helps improve estimates of methane emissions from animals and contributes in particular to selection strategies for reducing greenhouse gases.

Determining feed efficiency and water consumption to optimize resources

The PATASEL project funded automatic water dispensers for small ruminants, which measure both the animals' consumption and weight, in order to better take their needs into account and optimize the use of resources. These dispensers have been adapted to measure water consumption on pasture. In buildings, the dispensers are fixed and have a constant source of energy. A team of automation engineers therefore adapted the dispensers for outdoor use, making them removable and equipping them with solar panels.

Several other automatic feeders for measuring intake have been funded: milk intake by calves, concentrate intake by pigs, innovative devices for measuring individual fish feed consumption.

This equipment makes it possible to monitor animal consumption very accurately and better manage the use of food resources in order to reduce waste and the environmental footprint of food. It is also essential for studying feeding behavior.

Assessing behavior to measure animal welfare

Connected collars equipped with accelerometers—sensors that record animal movements such as displacement and posture—have been custom-designed for small ruminants. Several constraints had to be taken into account: a smaller neck circumference than in cattle, differences in behavior between sheep and goats, not to mention the well-known curiosity of goats, which do not hesitate to nibble on the collars of their fellow animals. A significant amount of adaptation work was therefore necessary. These connected collars make it possible to monitor behavior in the pasture and provide key indicators of health and well-being.

For cattle, funding was also provided for: around 100 accelerometers and two Farmbox systems, a system of connected collars that can detect heat, calving, eating disorders, and well-being.

Finally, 3D cameras combined with artificial intelligence now make it possible to automatically analyze animal behavior, a task that was previously carried out through direct observation and was very time-consuming. In practical terms, the ability to monitor the posture of sows during farrowing makes it possible to identify and select females with more pronounced maternal instincts, who are careful when lying down, thereby limiting the risk of accidentally crushing piglets (instead of doing so by preventing the sow from lying down).

Behavioral analysis also makes it possible to study social interactions (positive or negative) within a group and prevent problems. For example, pigs may bite the tails of their fellow pigs as a sign of frustration, which can lead to serious infections. To prevent this, some farmers still cut pigs' tails. With video analysis, farmers can spot the first attempts at biting and adapt the environment to prevent this behavior from becoming widespread.

AgroEcoPhen project: tools for more resilient crops

On the plant side, the AgroEcoPhen project is part of the national PHENOME-EMPHASIS infrastructure, which provides the scientific community with a range of tools (sensors, image analysis tools) for phenotyping hundreds of varieties or genotypes.

The AgroEcoPhen project aims to expand the network of experimental sites for testing new varieties and cultivation practices under a variety of environmental conditions. This will provide insight into how plants respond to the environment, enabling the development of more resistant cultivars that require fewer inputs (water, fertilizers, pesticides) and the adaptation of practices.

The originality of the AgroEcoPhen project lies in the use of sensors installed on high-speed vectors. This approach makes it possible to collect a very large number of measurements in a short time, compared to manual counting. It also guarantees that images are taken under standardized conditions (same light, same angles), ensuring greater data reliability.

The project, which began in 2023 and will run for five years, has already led to the development of a prototype connected stake that can autonomously acquire images and environmental data in field crops and transmit them to a remote server. Around 20 connected stakes have been manufactured for various INRAE units and GEVES (Groupe d'étude et de contrôle des variétés et des semences, or Group for the Study and Control of Varieties and Seeds). The aim is to implement the stakes at different sites to increase data flows and sources, thereby improving the reliability and accuracy of data processing models.

The connected stakes combine:

• cameras, which analyze plants (flowering stage, signs of disease, presence of pests),
• and sensors that measure environmental conditions (wind, humidity, temperature).

Unlike a traditional weather station, connected stakes combine environmental data with the observed condition of plants (flowering, infestations, diseases). Artificial intelligence is then used to link the two and analyze the environmental conditions that are favorable and unfavorable to plant growth and disease development. The major advantage is that it provides real-time monitoring of crops without having to visit the field. The data collected will thus make it possible to propose appropriate agricultural practices to optimize growth while limiting pathogens and pests. In addition to the connected stakes, the project also acquired two drones to provide information on plant cover and crop implementation at the beginning of the cycle.

The AgroEcoPhen project also led to the design of a mobile phenotyping robot, or phenomobile, specially adapted to row crops (arboriculture and viticulture), which will soon be deployed. Phenomobiles are autonomous robots equipped with sensors, cameras, and a guidance system that travel across plots to scan plants. They operate at crop height, offering finer resolution and greater precision than drones for detecting pests or weeds.

Each phenomobile is adapted to the specific characteristics of the crop being studied. For example, they are equipped with a straddle frame for low crops such as wheat, while they are equipped with an offset arm for tall crops such as corn. For arboriculture and viticulture, a dedicated model has been developed with a vertical imaging head. Researchers are still working on perfecting the onboard artificial intelligence in order to analyze images more finely and distinguish trees from the background.

Indicators for transforming practices

These systems illustrate how digital technology and robotics are transforming phenotyping in order to collect new data that is essential for agroecology. Beyond the equipment itself, these infrastructures represent a strategic investment for the entire scientific community. They enable the accumulation of data that is used to generate knowledge that contributes to providing concrete solutions to major agricultural challenges.

The infrastructures promote collaboration between disciplines and projects. The PATASEL project collaborates with other projects in the PEPR Agroecology and Digital Technology program, notably the WAIT4, HOLOBIONTS, and ADAAPT projects. The AgroEcoPhen project works in conjunction with the Pl@ntAgroEco project to improve species and disease recognition.

These new tools generate a considerable amount of data. One of the challenges ahead is to manage and analyze this data in a more sober and sustainable manner. This will involve strengthening analysis chains while integrating artificial intelligence and machine learning approaches in order to fully exploit the potential of the data while controlling its environmental impact.

By supporting the development of cutting-edge tools in research infrastructures, the PEPR Agroecology and Digital Strategy is contributing to the national strategy to accelerate the development of sustainable agricultural systems and equipment.

Thanks for their help in writing this article: Jean-Pierre Bidanel and Yvon Billon for the PATASEL project, and Tania Rougier and Bertrand Muller for the AgroEcoPhen project.