The Evolve Project, funded by the European Union’s Horizon 2020 program under grant agreement No. 825061, aims to tackle the ‘significant challenges’ posed by agricultural production for data processing.
The project sees a ‘data deluge’ opening up a ‘new frontier’ for the food and other industries, which are struggling to leverage existing tools to manage the huge amounts of data becoming more available.
Evolve’s ultimate aim is to be able to automatically identify a crop in a region based on satellite images and then predict crop growth for accurate in-season forecasting of agricultural production. The project is leveraging a centralised ‘testbed’ to optimise agri-production using yield and numerical models as well as ‘massive historic data’. This will allow ‘easy’ deployment and use of the testbed by domain experts in agri analytics.
“There are 11 countries involved and 19 partners,” Pierre-Aimé Agnel, Evolve’s technical project manager told FoodNavigator. "Partners in the project bring their technologies together on a prototype in order to bring high performance computing (HPC) techniques to companies while maintaining the simplicity of use of cloud platforms. The environment is designed to be versatile and performant, easy to use, easy to access and secure.”
Better, faster data interpretation
In agri-food, the project aims to reduce computing time while increasing dataset size and accuracy, Agnel explained. “We intend to increase the area covered from 400 sq. km to 100 000 sq. km and the error margin from 4% to less than 1%.”
Evolve partner CybeleTech is a ‘use case’ that leverages satellite images and machine learning algorithms to identify crops and predict yields.
Founded in 2011, the company is focused on developing numerical technologies in agriculture. The core products of CybeleTech are based on either numerical simulations of plant growth through dedicated biophysical models or machine learning methods extracting knowledge on processes through large databases, CybeleTech’s Teddy Debroutelle explained.
This approach can bring benefits at various stages of the production process. In plant breeding, simulating plant growth can help reduce the amount of field trials by approximately 50%; HPC models can optimise on-farm practices; forecast yields to anticipate storage and market requirements; and strengthen quality control.
“We develop technologies on the big crops such as cereals, corn, sugar beet, sunflower and rapeseed, as well as some tropical crops,” CybeleTech director Marie Joseph Lambert added.
Within the Evolve project, the tech company is designing and implementing the agri analytics workflows, starting from the current prototype. The end game is to develop ‘realistic massive dataset processing’ using the HPC features of Evolve’s application.
“The exact data we extract from satellite imagery relates to industrial property. However, they are only a small part of all data that go into yield prediction. Our algorithms also include a large part of plant mechanical modelisation to achieve accurate predictions. The output of crop models is accurate to some percent of the actual field value,” Lambert claimed.
The partners are seven months in to the 36-month project, which has already secured funding of €14m from Horizon 2020. By bringing together partners in markets where data capability is already a source of disruption, Evolve aims to address issues around data ownership and processing capabilities, leveraging data as a force for progress.
“As these markets are socially critical for European citizens, Evolve is not a pure technology project but frames itself in the more global perspective of data ownership in an open society.”
