Soil biological quality index (QBS) determination using machine learning (QBS)

Understanding of the state of soil biodiversity worldwide is relatively low. This creates difficulties in determining soil ecological status categories (what constitutes healthy or unhealthy soil) and developing guidelines for effective soil health monitoring. Although detailed research on soil biodiversity would provide essential information, it is often complex and costly. Therefore, as a simpler and more accessible solution, the biological quality index QBS-ar has been developed, based on the analysis of soil arthropods. The main goal of this project is to develop machine learning models that can determine QBS-ar by analyzing images of microarthropod populations obtained from soil samples. These models will be suitable for large-scale studies, including soil monitoring programs.

In the project, EDI develops machine learning models for detecting various soil arthropods and researches methods to improve the performance of these models. This includes studying explainable artificial intelligence methods to reveal which image parameters most significantly affect the capabilities of the trained detectors.