Automated and AI-supported peronospora infestation forecast for the further development of integrated plant protection in Bavarian hop cultivation
PeroHop4.0 is a collaborative research project of LfL, TUM and HSWT, which bundles competencies across institutes in order to create added value for agricultural practice with modern technologies. The „Peronospora Warning Service“, operated by project partner LfL, has been a central pillar of integrated plant protection management for permanent hop crops for 40 years, enabling the targeted and needs-based use of plant protection products (PSM) to control downy mildew (Pseudoperonospora humuli). In light of the increasing challenges for Bavarian hop production, modernization of the service, including the automation of spore evaluation and an area-specific forecasting is urgently needed in order to ensure a resource-efficient and effective plant protection and management strategy for hop cultivation and to make it sustainable for the future.
Objectives
The aim of the project is modernization of the „Peronospora Warning Service“ through the use of digital microscopy, automated spore traps and artificial intelligence (AI). This is intended to increase quality and efficiency and secure it in the long term. The current warning service requires a significant amount of manual work and does not provide forecasts. The combination of modern technologies with the LfL’s existing expertise and long-standing database enables automation, improved regional resolution and short-term infestation forecasts, thereby ensuring the long-term safeguarding and further development of the service.
Procedure
In a first step, AI-based image processing will be used to reliably identify and automatically count zoosporangia of Pseudoperonospora humuli based on digital microscope images. This reduces the personnel, increases reliability and objectivity, and preserves many years of expert knowledge. The next step will be to fully automate the warning service by means of AI-supported automated spore traps. For this purpose, commercially available real-time spore traps will be adapted, and an AI model for automated analysis will be developed, eliminating the need for manual sampling. This automation enables a significant reduction in personnel effort and thus a shorter response time as well as improved scalability. In addition, a forecasting model based on historical data and other agronomic parameters is to be developed to predict the infestation pressure on a variety-specific and region-specific basis for the coming days. Thus an area-specific early-warning system is created, which enables more targeted applications and reduces the use of pesticides, as warnings can be issued on a more area-specific basis rather than for the entire Hallertau region as has been the case to date.
Publications
-
Publications
