Due to increasingly dry summers as a result of climate change and at the same time decreasing water availability (more difficult approvals for new wells by the counties), questions from nurseries about precise and effective irrigation are increasing. In this context, the control of the most important production factors "water" and "nutrients" is today almost exclusively carried out according to the current situation of the crops and in the short term according to estimates or inaccurate measured values. Even today, the only useful forecast is the weather report adapted to the agricultural sector, from which the farms have to draw their conclusions independently.

Monitor environmental conditions

The goal of the Predictive Plant Production project is to first develop a system for the two model plants, the tree of life (Thuja) and the rhododendron, which monitors the environmental conditions of these plants. Subsequently, the data obtained can be used to determine the maintenance measures for plant growth that is as resource-efficient as possible or can be scheduled. For this purpose, substrate moisture, fertilizer concentration, soil temperature, local weather conditions as well as the addition of water and fertilizer and the shading and ventilation of greenhouses in a nursery are recorded by special sensors. Thus, local conditions and influencing variables are learned individually for each location using artificial intelligence (AI) methods. With the learned knowledge, accurate prediction models of the water, fertilizer and temperature balance and, based on this, of plant growth are configured and then used to assist in the control of irrigation, fertilization and temperature regulation and even to automate these.

Avoid damage to the plants

Predictive plant production can be a valuable aid in this respect by using the given data on site, weather forecasts, experience of the release of the depot fertilizers administered, and the measured values on the plant to make a timely prediction of water and nutrient requirements that is as precise as possible and an automated control of all components based on this. This would ensure more uniform quality in the plant stand and significantly increase the prevention of plant damage. Farms would therefore benefit in the long term from larger, more beautiful, higher-yielding and faster-growing plants.

In general, the project will develop a self-learning method that answers the following four questions for each farm:

•     Which maintenance measures ensure optimal plant growth?
•     How can I ensure normal plant growth with minimal use of inputs?
•     How can I schedule crop production with the least possible input use?
•     How can I support pest and disease control through the choice of environmental conditions and thus reduce the use of pesticides?

Currently, there is no self-configuring - water, fertilizer, and temperature control that adapts to local conditions - nor any approaches to predictive plant production or related techniques from which a comparable solution could be derived.

Partners of the project are besides OFFIS as coordinator: Baumschulberatungsring Weser-Ems,  Baumschulen Johann Bruns and Hellwig,  Landwirtschaftskammer Niedersachsen and the company Communicate 2 Integrate. Predictive Plant Production is funded by the German Federal Ministry of Food and Agriculture as part of the European Innovation Partnership "Productivity and Sustainability in Agriculture" EIP-Agri.