Greenhouses are nowadays the most sophisticated way to control plant environment to increase their production, reduce impact of climate uncertainty, provide physical barriers to diseases, enabling strong reduction of chemical pesticides. However, they require more and more parameters to be set by the grower (e.g., 200 in a standard soil-less glasshouse used for tomatoes). As a consequence, parameters are mostly set to default values, without adaptation to the location of the farm, the needs of the species and of the cultivar, their potential in yield and quality (dry matter and sugar content). Thus a more dynamic, online process should be pursued in order to gather collected data, model and constantly optimize parameter setting in the greenhouse .
Involvement of UC partners
CYBELETECH is a high tech SME dealing with some of the most critical issues for food production, sustainable development and protection of the environment through rationalization of agriculture and forest exploitation. In this context, CybeleTech probides tools, software and services for yield forecast, optimization of crop or forest management and breeding based on a strong expertise in digital technologies including dynamic modelling of plant growth, digital and statistical methods for data assimilation (climatic and agronomic data, satellite images) and software engineering. In previous works on greenhouse vegetables, developed greenhouse modelling solutions significantly improved crop management and yield estimation , including savings of 50-100€/ ha/day of CO2 (92% of CO2 cost) and a reduced emission of liquid CO2 of 90% on tomato crops, while yield and quality estimations on salad crops reached 90-95% of precision. This has already been done on field crops which do not require the management of so many management data and no need for quasi RT answer (an hour granularity is needed in the greenhouse compared to a few days in field crops). Thus applicability in greenhouses could be substantially increased by a more connected and more reactive “digital twin”, processing in quasi RT the meteorological data of the greenhouse (hourly to daily reactions depending on the actuators). The uncertainty would be strongly reduced by automated data assimilation and simulation iteration.