The EMMAH joint research unit, “Modelling agricultural and hydrological systems in the Mediterranean environment”, joins #DigitAg

The EMMAH joint research unit, “Modelling agricultural and hydrological systems in the Mediterranean environment”, joined #DigitAg in September 2018. The unit includes teams from INRA and Avignon University (UAPV). It is situated in Avignon (AgroParc and Domaine Saint-Paul at the INRA PACA centre). Presentation by Stéphane Ruy, its Director, and Frédéric Baret, leader of the “CAPteurs et TElédétection” (CAPTE – Sensors and remote sensing) team.


EMMAH is a multidisciplinary joint research unit. Its teams are comprised of members of the Environment and Agronomy Division at INRA and members of the Hydrogeology and Physics departments at UAPV.

One distinctive feature of EMMAH is that its CAPTE team (“Sensors and remote sensing”) is also part of the INRA-Arvalis Institut du Végétal joint technical unit of the same name. In addition, this unit hosts Hiphen, a spin-off company created by a former doctoral student in the unit, specialising in high throughput phenotyping and monitoring of field variety trials. Other ACTA technical institutes are also associated (Terres Inovia, ITB, CTIFL), as well as GEVES.


Which disciplines are represented by EMMAH and its expertise?


The fields concerned by digital agriculture are, more specifically, agronomy, remote and proximity sensing, environmental science, with soil and transfer physics, and applied mathematics. The unit has joint expertise in crop modelling and spatialisation, heat transfer modelling [a1] (developing sensors and methods for remote and proximity sensing), and signal modelling (plant and soil properties). Beyond digital agriculture, the unit’s expertise extends to soil science (hydrogeology, geochemistry, microbiology and soil ecology) and, to a lesser extent, to human science.

The goal of our scientific project is to identify tools for action that can be used to adapt agricultural territories to the global changes underway (climate and land use), for example through the spatial allocation of agricultural practices at the territorial level and the adaptation of agrosystems. Our approaches are based on the analysis and modelling of the main biophysical processes involved in agricultural production, the water cycle and the functioning of aquifers. The analysis and modelling of land use change also integrate socio-economic drivers.

Our research focuses on the development of methods, sensors (interpretation of remote and proximity sensing images) and processing chains to monitor the biophysical variables of vegetation, land use and agricultural practices, from the m² level to that of the territory. We also seek to develop methods and instruments to estimate the properties of the soil and subsoil, and to quantify the water balance of soils, the state of aquifers and the interactions between agricultural practices, surface water and groundwater.

In terms of modelling, the areas we study are: the biophysical processes involved in aquifer recharge, the quality of groundwater and their evolution under the effect of global changes; land use change caused by social, economic, biotechnical and biophysical factors; the evaluation of regulation and provisioning services in cultivated soils; and the estimation of production at the territorial level, especially in connection with climate characteristics and agricultural practices.


Our main contribution to #DigitAg will be expertise in:

  • remote and proximity sensing: development of sensors, methods for interpreting signals, modelling of plants and crops and assimilation (processing and integration of data into models and model adjustment based on images), deep learning techniques;
  • modelling: modelling transfers in the aquifer-soil-plant-atmosphere system in soils (transfers in the soil, soil-plant, atmosphere), spatial modelling of agrosystems (functioning of an agricultural territory with its different plots, crops), development of decision support tools.


Examples of outcomes linked to digital agriculture

We have developed several automatic measurement systems using robots (in connection with the companies Robopec and Meca 3D), or drones (associated operating procedures and processing chains). The corresponding web services are currently being set up. We have also developed a multispectral camera in collaboration with the start-up Hiphen.


The Phénomobile at Toulouse


The Phénomobile is a first example of our achievements. Developed for high throughput phenotyping in the field and plant breeding, it is composed of a self-propelled robot fitted with a set of sensors (RGB camera, lidar, spectrometer) to automatically observe vegetation. The measurements and associated processing chains are applied to high throughput phenotyping.

Its first version for low crops is operational and is in use at the site in Gréoux-les-Bains (Alpes de Haute-Provence, France).

The second version, developed in the context of the Phenome project, is suited to all crops (maize, sunflower, rapeseed, etc.). It has been in use for a year at the site in Toulouse and two new machines will be available in Montpellier and Clermont-Ferrand next year.


Another achievement is the Pastis sensor. This is a simple, robust system for measuring transmittance under cover to assess the leaf surface. Pastis is now sold by Hiphen.

The Plant Phenotyping Processing Platform (4P) using drone and Phénomobile is currently being developed in the context of the Phenome project. It should help to transform the raw data acquired by phenotyping systems into plant characteristics that are useful for plant breeders or for modelling crop functioning.

In addition, we have developed the Virtual Soil (or VSoil) modelling platform. Its design and development were driven and supported by the Environment and Agronomy Division at INRA. Its goals include first facilitating the multiphysics coupling of the different processes (physical, biological, geochemical, etc.) occurring in soils, which generally have a high degree of interaction, and, second, fostering the use of models within the community of experts in soils, plants and soil-plant-atmosphere interactions, and thereby encouraging multidisciplinary approaches to the use, behaviour and future of soils.


Partnerships underway with other #DigitAg units

We already have a number of partnerships with the member units of #DigitAg. All of the developments concerning high throughput phenotyping in the field were achieved in close collaboration with the #DigitAg units involved in the Phenome project.

We are currently participating in the ANR ROSE Challenge (Robotics and Sensors for EcoPhyto). We are involved in the WeedElec project, which brings together scientists from Montpellier, members of #DigitAg (IRSTEA – ITAP, INRIA – ZENITH, CIRAD – AMAP and AIDA, and INRA – EMMAH). The project offers an alternative to chemical weed control. It combines aerial detection of infested zones, the differentiation of weeds and crops thanks to the use of digital vision and deep learning techniques, and a non-chemical high-voltage weeding process. This electrical process destroys both the aerial parts and the roots of weeds.


Our role is to identify the specific electrical signature of the weeds targeted by the weeding system, and to understand the electrical conduction in the plant-root-soil system in order to optimise the design of the high-voltage discharge system, in terms of both energy and destruction efficiency.


What is your perspective on digital agriculture?

The first important point is data collection and open access to this data and to the associated metadata. In research, there are still problems of standards, but we are getting there. For CAPTE, we work from data from the phenotyping platform, experimental stations (INRA, Arvalis, etc.) and the databases of our colleagues. We also work with farmers, where the issue of data is even more complex. Data is obtained through surveys or remote sensing, via institutional databases or collected by us.

The focus should be on modelling, through which we will have access to plant functioning and we will be able to make decisions. Often, decisions are made during crop development, with a past that may or may not be well-known, and a future that we only know statistically, especially for climate data. The challenge is to determine the cultivation techniques to be applied in order to adapt to this future climate. Modelling is an approach that is particularly suited to this challenge: it can be used to test different scenarios and multiple assumptions, and to identify the most appropriate agronomic strategies. And in order to model correctly, we need access to a large amount of data and the ability to easily integrate this data into models.

Finally, the soil compartment is within the scope of digital agriculture. It is not simply a support interface for crops with static or homogenous properties: it is an extremely heterogeneous environment that evolves under the actions of humans, cropping systems and the climate. Developing sensors, models and data-model integration chains to continuously monitor and predict the evolution of its properties, of the different variables of agronomic interest (water content, organic matter content, etc.), or of indicators of physical, chemical or biological quality is a key challenge that is firmly placed in the context of digital agriculture.


EMMAH à l’international

The EMMAH joint research unit works on the Mediterranean environment, but not exclusively. It has developed partnerships with Lebanon and North Africa, the universities of Boston (US), Leuven (Belgium), Valencia (Spain) and Pisa (Italy, for human science, economics, changes in production systems, etc.). The unit has a visiting professor at the Universities of Barcelona and Valencia (remote sensing).

The CAPTE technical unit works with CSIRO and the Universities of Queensland (Australia), Valencia (Spain), Tokyo (Japan) and Nanjing (China), where a member of the unit is a visiting professor.


More information:

  • Website
  • Contact: Stéphane Ruy, Directeur de l’UMR EMMAH – 04 32 72 22 42 – stephane.ruy [AT]
  • UMT CAPTE: Frédéric Baret (frederic.baret [AT] et Benoit de Solan (b.desolan [AT]
  • Hiphen : Alexis Comar (acomar [AT]