This consortium launches the ambitious quest to utilize the hidden potential of terabytes of data, documents and metadata collected at all partners by developing data spaces linking data platforms, implementing AI methods to find genotype to phenotype links for breeding and projection towards cultivation practises, using crop models and unlock these insights via domain specific Large Language Models. Globally AI developments generate incredible results in an extreme pace, the Dutch companies are at risk to be left behinnd. Therefore, this unique public-private partnership will translate these rapid developments to the horti- and agri domain. The largest collection of phenomics data was captured in recent years in the greenhouses and fields of the Netherlands Plant Eco-phenotyping Centre (NPEC). This data links simulated drought and salinity responses to actual quinoa breeding fiel trials, where plant performance and yield estimates were recorded using NPEC UAVs. And that is only one of the ambitions of the DAS-CROPPER project. We will also link performance data of crop genotypes to the underlying genetic mechanisms on quinoa. And we will incorporate crop growth models for greenhouse varieties in state-of-art digital twin approaches that can be consulted in chatbot style using domain specific large language models trained on publications, reports and ontology-based datasets. Growers need to receive live information of the status of their crop, and as high-tech houses are getting larger, this demands a data-driven approach. A grower needs to schedule the amount of labour needed at a specific moment over the crop season, instruct personnel to make use of (natural) resources efficiently, and verify the impact of this on the crop quality and yield prognosis. This project will be the ICT/data bridge between the data domain and agriculture. SMART and FAIR data strategies are required to link multi-omics datasets to be able to train novel AI models for times series analysis of crop performances. Profits and higher yields will have an economic counterpart - practical agreements and ways of working need to be defined to assist growers as well as technology and data companies on how to share in costs and profits via software license agreements, data ownership and AI model ownership (ELSA).

This project will explore the potential of data spaces, linking data platforms to develop novel AI methods in two application fields:
1. AI driven decision support models in the cultivation of greenhouse crops, with tomato and lettuce cultivation as an example, as it has the commercial interest of the consortium partners, and available and new data from applied cultivation, including digital phenotypic data from NPEC, and cultivation management data and crop performance.
2. AI driven genomic/phenomic selection model in the selection of field crop parental outlines in breeding.

The project will develop key technologies for plant breeding towards climate adaptative crops and varieties (ST2 with impacts for MMIP C2 Climate-adaptive agricultural and horticultural production systems), and builds logical links from within the ambition of the KIA Digitalisering, such as digital connectivity tech, digital twinning and the ELSA-labs within AiNed.

Within eacht work package the following deliverables will be developed:
1. Data space development
2. Development of AI tools for trait extraction form highly dimensional digital phenotyping information
3. Establishment of an AI supported decision support model for greenhouse cultivation and management
4. Establishment of AI supported genomic/phenomic selection models for breeding of field crops
5. Implementation of Large Language Models for easy interrogation of the data space
6. Establishment of optimal governance of data spaces