AI system monitoring soil biology

Until now, the analysis of soil life was a complex analytical process requiring a laboratory and trained specialists. Thus, it is dependent on experts, time-consuming, and too expensive for regular use. As a result, conventional commercially available analyses only depict biological soil fertility based on key indicators such as microbial biomass, microbial activity, and the fungal/bacterial ratio. Based on current knowledge, conventional analyses do not currently provide any assessment or recommendations for the composition of organic matter. Using new microscopy techniques, image recognition, AI and machine learning, existing analyses are being supplemented and automated with detailed microbial-level analysis and can be deployed by users in the field as part of the mobile SHAPE field laboratory. The newly developed product will detect the presence of microorganisms in soil and compost and evaluate them quantitatively and qualitatively.

The four main groups of soil microorganisms and fauna are bacteria, fungi, protozoa, and nematodes. These organisms form the basis of the soil ecosystem. A distinction is made between quantity (number of organisms per gram of soil) and quality (favourable and unfavourable organisms and diversity). Both the quantity of organisms and the ratio of favourable to unfavourable organisms provide valuable information about the overall state of soil biology in the sample under investigation.

This project will initially investigate the nematode population in the soil and its significance for ecological function(s)/ecosystem services (Source 12), using image analysis, AI and machine learning. Over the course of the project, the method will be expanded to include other microbes. The conceptual innovation lies in the ability to perform this evaluation directly and in near real time at the end user’s site, without requiring soil samples to be sent to the laboratory. This creates the opportunity to directly incorporate the on-site evaluation data into subsequent actions and to examine areas not only field by field, but also at fine-grained resolutions. The new measurement and analysis methodology enables diverse, ongoing documentation even after cultivation measures have been carried out. The presented data can be evaluated and transferred into a Decision Support System (DSS) for soil health, which supports the user in their crop management decisions.

Partial developments:

-Development of an AI-based microbiological object recognition system (Hal24K Agri, HSRW)

-Validation of object recognition using conventional analysis methods (Hal24K Agri, HSRW)

-Standardization of sampling protocols (Royal Eijkelkamp)

-Concept development for mobile use (Hal24K Agri)

-Market analysis, business development (RheWaTech, Royal Eikelkamp)

– Preparation of a certification based on the results of the working group on CEN/TC444 with a focus on soil health (Royal Eijkelkamp).