This is a cross-over call for Public-Private Partnership (PPP) in the context of the Research and Innovation Agenda for Agriculture, Water and Food (KIA-LWV). The topsectors ICT, Horticulture & Starting Materials, Agri & Food, and Water & Maritime are working together on this call. The awarded project concerns a public-private partnership investigating the implementation of AI methods to make a projection to cultivation practices and crop models. The ultimate goal is to develop new, adaptable crops that are more resistant to changing environmental conditions and diseases. The research project starts this month and will run for four years.
Dr. Andres Torres Salvador of Radicle Crops BV is coordinator of the project. In addition to Radicle Crops B.V., the project's other participants include Growy Group, Daedong Kioti Europe, Omron Europe, KIST Europe Forschungsgesellschaft, Agr Information Partners, NAKtuinbouw and Wageningen Plant Research. Together, they will be putting in the other half of the required funding. This will bring the total budget to €1.9 million.
Data spaces, AI methods and LLMs
The consortium aims to exploit the hidden potential of terabytes of data, documents and metadata collected from all partners by developing data spaces which connect those data platforms. In addition, the parties want to implement AI methods to find links between genotype and phenotype for breeding and projection to cultivation practices, using crop models. The insights will be made accessible through domain-specific Large Language Models (LLMs).
This project will explore the potential of data spaces and link data platforms together to develop new AI methods in two application areas:
- AI-driven decision support models in greenhouse crop cultivation; Tomato cultivation is taken as an example, as it has the commercial interest of the consortium partners, and due to the large amount of data available from applied tomato cultivation, including digital phenotypic data from NPEC (Netherlands Plant Eco-phenotyping Centre), and crop management data and crop performance.
- An AI-driven genomic/phenomic selection model in the selection of parental lines of field crops in breeding; Quinoa has been chosen in this case as a model species because of the large datasets on digital phenotyping available in the NPEC greenhouse. Existing and new data from field trials and NPEC trials with crop performance measurements and digital phenotyping (UAVs, 3D and chlorophyll fluorescence) will be used.
GenAI solutions
The AI tools developed using these large datasets will also be generalisable for other greenhouse production systems and breeding in other crops. For improved and automated selection of genotypes for breeders and to help growers choose between management options, generative AI (GenAI) solutions will be implemented that rely on domain-specific trained LLMs (Large Language Models). Those models will be driven by Retrieval-augmented generation (RAG) and knowledge graphs linked to crop growth models.
In addition, the researchers want to develop user interfaces to enable the LLM to provide clear insights into the complex data. This will make use of technologies such as Open WebUI and BionicGPT. Both use cases will benefit from this GenAI/ crop model solution, resulting in two different applications.
The intended result
The DAS-CROPPER project is aimed at developing AI-driven technologies for plant breeding, with a focus on climate-adaptive crops and resource-efficient production systems. Various initiatives exist in the field of AI and agri-food, but none of them integrate AI for decision-making and genomic prediction models. This project seeks to fill this gap by improving data sharing strategies, developing AI-based phenotyping tools and improving predictive models for crop performance.
The DAS-CROPPER project offers significant benefits for various sectors. For businesses, it will create new commercial opportunities, including sales of digital phenotyping equipment, decision support systems and improved breeding efficiency. Farmers and society will benefit from access to more resilient crop varieties, reducing risks and growing costs and this will contribute to a more sustainable and safer food chain. The food industry will have a more stable and high-quality supply of agricultural raw materials, improving the overall reliability of production. Finally, the Dutch agricultural sector will see its competitiveness improved thanks to AI-driven decision models and genomic selection, promoting business expansion and new export opportunities.