Integrating Remote Sensing and AI for precision Monitoring of Soil and Vegetation Contamination

Abstract. The paper presents recent advancements in monitoring pesticides and heavy metals using remote sensing techniques. The study is based on the pilot project “Support for Ecological Agricultural Production in Mazovia” (WEPR), which provided a scientific basis for the more comprehensive, larger-scale project “Support for Monitoring the Distribution of Pollutants in Agriculture” (WORZ). Addressing the needs of end-users, the research focuses on the development of reliable methods to assess environmental contaminants that impact soil quality. A combination of high-resolution UAV imagery, multispectral satellite data, and a suite of vegetation indices was utilized to correlate spatial variations in soil and crop conditions with pollutant concentrations. Machine learning algorithms, including Random Forest, were applied to classify contamination levels, while laboratory analyses validated the spectral findings. Despite challenges such as limited sample sizes and class imbalances, the integration of multi-source remote sensing data demonstrated promising results. The future of the study will be focused on the multi-temporal analyses, improving prediction accuracy and dataset and environmental risk mapping. Ultimately, the outcomes of this research show the potential for scalable monitoring systems that align with European sustainability and agricultural precision practices.