Grapevine: reducing the environmental impact by optimising the use of phytosanitary products and increasing biodiversity

GRAPEVINE aimed to improve current pest prediction and control systems in the wine sector, thanks to Big Data and Artificial Intelligence technologies. One of its challenges was the improvement of existing Machine Learning predictive models to offer a more effective response to the treatment of pests and reduce the amount of fungicide and its number of treatments. The developed services were tested to forecast grapevine diseases in Aragon (Spain) and the Protected Denomination of Origin region of Goumenissa (Greece). Implementing these services required the execution of complex computational tasks, mainly the high-resolution calculation of meteorological models for forecasting the evolution of grape diseases, which are resource-intensive tasks. As these resources are quickly consumed, many High-Performance Computing (HPC) resources are needed.

The implementation required a modular and cross-border solution with components that can be distributed and coordinated by an HPC orchestration solution to become an intelligent decision support tool for grapevine farmers adaptable to other regions. A large amount of existing data was used to adapt the models to specific grape varieties, phenological specificities, climatic conditions, and the training of Deep Learning algorithms.

Additionally, the amount of data managed required the computational activities to be close to the data source to execute efficient data transfers and collection of results.

GRAPEVINE has used High-Performance Computing (HPC) resources from the EGI Federation. The collaboration between EGI and GRAPEVINE has successfully contributed to the following: