This project's research activities officially ended in March 2021. Legacy in SmartAgriHubs Portal

Precision crop management


Mehdi Siné
Florence Leprince
Benoît de Solan


Thierry Milin
Borgne Danièle


Arvalis Orange

This use case is organic!

Precision crop management

Introducing smart wheat crop management by sensor data embedded in a low-power, long-range network infrastructure, mainly in wheat.


If we ask you to picture the European countryside, chances are that you will envision large wheat fields and lush meadows as far as the eye can see. Wheat is the most cultivated cereal worldwide and one of the most common cultivated crops in Western Europe. This is exactly the context of this use case, which has its test farms in France. One of those testing sites is located in the South East of France, about a hundred kilometers from Marseille. We met with the coordinator of the use case, Benoît de Solan in the ‘Digiferme’ of Boigneville about 60km from Paris.  Since it is crucial to evaluate new technologies in real-life conditions, having test farms in different climatic areas helps the use case to validate their solution.

Combining IoT technologies, drones and satellite images

By utilising sensors embedded in a low-power, long-range network infrastructure and computer modelling for precision crop management, farmers can greatly improve the efficiency of nitrogen and water use in wheat production.

Therefore, field sensors developed by HIPHEN and BOSCH are installed in wheat fields. The connected stakes are white, about 1,5 m tall, equipped with a camera RGB and a spectrometer. You barely notice them from afar, but they are gathering essential information and subsequently providing farmers with insights regarding their fields.

The sensors allow to monitor the crop along its phenological evolution, monitoring parameters that indicate the onset of any stress or potential disease. They do so by collecting pictures of the field, and through RGB images and spectral measurements the use case team is able to automatically detect wheat heads and determine the crop development stage. Furthermore, it provides valuable insights to help farmers predict the final yield. Therefore, in 2020, a deep learning model has been developed to accurately count wheat heads through high resolution RGB images. In addition to that, the field sensors provide the same state variables as satellites, such as green leaf area index (GLAI) and chlorophyll content (Cab), at a daily time step.

But the use case’s innovation does not stop there. The IoT devices collect a wide range of information, from the composition of the soil to the humidity in the air, as well as image-based data. All acquired data is transferred, processed and made available daily through a web platform supported by Orange. Moreover, this field data is combined with images from the SENTINEL 2 satellites which systematically acquires optical imagery at high spatial resolution over land and coastal waters.


Irrigation cost


Yield & quality


Greenhouse gas emissions

The main technological challenge is the energy needed for image transfer. The concept of this set-up was a low-energy, long-range network infrastructure. This way, the set-up could be installed in remote areas. Therefore, the energy required for image transfer is constantly improved by the researchers. With the proliferation of the 5G network, the team expects significant improvements in these regards in the upcoming years.

During the project phase the focus was put on nitrogen and water management. However, the information collected by the IoT set-up can also be used in many other ways. This way the end-users can harness the full potential of the installed system without the need of further investments. Deep learning analysis can further process these images: a machine learning algorithm could be trained with thousands of crop pictures collected by the current set-up to recognise accurately the first signs of water stress or disease. In order to develop accurate image classifiers for the purpose of plant disease diagnosis and potentially train deep learning algorithms, the use case needs large, verified datasets of images depicting infected as well as healthy plants. Fortunately, the current solution serves as an excellent source of such datasets, giving the team the chance to explore this option in the future.

A flourishing crop in a healthy environment

Nitrogen and water availability are the two main limiting factors in wheat production, and essential determinants to plant growth.

Without technological decision support farmers might use more fertilisers than they actually need to, and not all nitrogen applied is absorbed by plants. Some of the nitrogen goes back into the atmosphere as nitrous oxide - a potent greenhouse gas - or leaches into the water. Applying nitrogen at specific times of the growing cycle of the plant can help reduce its dosage, reducing the costs for the farmers and the impact on the environment. Therefore, the use case focused on field experiments aimed at chlorophyll and nitrogen estimation during the last year of the project. The result is a predictive model which is capable of accurately estimating the crop nitrogen content from the spectral measurements provided by the IoT sensors and satellites.

Data, data, data… and then?

As we elaborated in the previous chapters, the sensors combined with the satellite and drone imagery collect a lot of data about different meteorological factors and other environmental conditions.

One of the main challenges is the integration of all data coming from different sources and in different formats. As a first step the field sensor measurements are merged with the data from the Sentinel 2 satellite. Once the data is fused and cleaned up, the results are high resolution observations. But the process does not stop here. This data can only provide decision-making support to the farmer if it is presented in an actionable way. Hence, the team developed a dashboard interface that gives meaning to the data and presents it in a user-friendly way easily at one glance. As Benoît de Solan, engineer at Arvalis l’Institut Végétal and Use Case Coordinator puts it: ‘The sensor alone is not enough, it is necessary to add models and apply decision rules in order to transform these measurements into actual decisions and insights.’


Labour duration


Nitrogen (kg per ha)


Energy use

Therefore, all the collected information is integrated into ARVALIS agronomic models to provide precise advices on crop management. Only then can the data help to complete yield prediction models, from detecting the flowering date to determining the harvest time and yield to help farmers optimise their logistics. For example, the data is processed by algorithms which will transform the images collected by the sensors into the foliage coverage of the plant or detect any problems with potential disease and identify the stages of development of crops. Moreover, soil moisture information provides information about the status of the plant and if it is experiencing any stress. Developing data analysis algorithms based on imagery was one of the ambitious goals defined for the last year of this use case. A test campaign was initiated to evaluate the sensors and the proposed method. This made it possible to compare the service with traditional methods of fertilisation management and to study the impact of the assimilated data on the nitrogen recommendations proposed by the dashboard. The first results show that the sensors used allow a good estimation of the green area index and the heading stage of the canopy. This allows farmers to get more information, more frequently, on the condition of their fields and ultimately helps them to make informed decisions on fertilisation and irrigation. That this ambitious goal has been achieved, was officially confirmed in March 2019 when these field sensors received the SIMA Innovation Awards.

With and without IoT

This use case’s innovation highlights the need to develop more applications based on the data recorded from IoT systems, which brings additional value to the solution and increase user-acceptance by making the data easily accessible.

Therefore, the main objective of each experiment was to propose a viable solution to farmers. Consequently, they have been involved at several stages of the experiment and in the last period the use case officially rolled out the service.

In order to reduce costs, the researchers also developed the service without implementing the fixed sensors in the field using IoT technology. The analysis is then supported by the same software leveraging only the data collected via satellite. The crop observation is then less accurate, but this option offers more flexibility to farmers.  This ultimately makes precision crop management also possible for smaller farms as it significantly reduces the initial acquisition costs. This option is also independent from any hardware component, which has potential for more flexible applications.

The comparisons of the CHN model corrected without and with IoT data assimilation reveal a real gain of these last ones on the simulations and exits of the model and thus on the recommendations of contributions. This model calculates the daily flow of carbon (C), water (H), and nitrogen (N) between the soil, atmosphere and plant compartments on a daily basis during a cropping season. Ultimately, the proposed method leads to a lower total nitrogen input compared to the balance sheet.

Smart Agri Hubs

Get in touch with the use case team via the SmartAgriHubs Innovation Portal

Go to portal

Achievements, products & services

Simultaneous measuring of the vegetation growing status

Info on main meteorological variables and the soil water potential

Data are transferred and made available on data platform and combined with satellite images

Data integration of agronomic models for accurate advices on crop management

Use case partners

IOT Catalogue


Go to IOT catalogue