Because herbs grow close to the ground and are not tall they are susceptible to contamination with microorganisms. In addition, they are exposed to numerous environmental influences, leading to great inconsistencies in quality and high processing costs.

The Austrian Bergkräutergenossenschaft, a cooperative of innovative farmers in the Mühlviertel district of Upper Austria that grows and sells herbs produced by certified organic farming methods, is carrying out a project with FFoQSI and the Bioinformatics Research Group at the University of Applied Sciences Upper Austria that uses applied statistics and machine learning to reduce these quality inconsistencies as far as possible. This experiment is based on data collected and supplied by the approximately 60 members of the cooperative.

This data included over 100 parameters relating to the cultivation, harvesting and drying of every batch (e.g. planting conditions, type and number of tillage steps, type and application of fertilizers, conditions during harvesting and drying parameters). Additionally, every batch was examined under laboratory conditions to determine the bioburden (e.g. presence of yeasts, moulds or pathogens such as salmonella). All this data was recorded for every batch of each crop.

To predict the risk of bacterial contamination in future batches, a number of different machine learning algorithms were used, such as random forests, gradient boosting trees, artificial neural networks and symbolic regression.

Additionally, applied statistics and hypothesis testing made it possible to identify the most relevant parameters for high levels of bacterial contamination. Consequently, specific measures for reducing the bioburden could be recommended to the farmers.

All of these aspects are communicated via a web application which can also be used to collect fresh data and present findings to the farmers. During its development, the application was continually adapted to the farmers’ needs following several feedback meetings.

Outcomes and effects

The information system developed on this project provides both the cooperative and its member farms with greater transparency thanks to improved monitoring of the data. The use of machine learning to analyse large amounts of data from different sources has made it possible to identify correlation and forecast models.

This leads to a better understanding of the processes, reduces the bioburden and consequently increases the product security and storage stability of the herbs produced.

(c)iStock / alle12