Microplastic contamination in high-altitude soils of Sagarmatha National Park: a spatial assessment with deep learning-supported detection

Microplastic contamination is a pervasive environmental concern, yet its study in remote high-altitude regions remains limited. This study provides the first comprehensive assessment of microplastics in the soils of Sagarmatha National Park (SNP) in Nepal by integrating field sampling, laboratory extraction, and deep learning. 50 soil samples representing diverse land use types and elevation zones were collected in May 2023. Microplastic particles were extracted and categorized by morphology (fragments, fibers, films) with polymer identities verified using optical photothermal infrared spectroscopy (O-PTIR). To accelerate detection and classification from microscopic images and reduce human-induced bias, a deep learning instance segmentation model (YOLOv11n-seg) was trained on annotated datasets. The model achieved high precision and efficiency in identifying morphological categories, significantly reducing analysis time compared to the manual method. Results revealed widespread microplastic pollution across the park, with the highest concentrations in surface soils and near human settlements. By combining environmental assessment with machine learning innovation, this study establishes a critical baseline for microplastic pollution in Himalayan soil and highlights the potential of advanced computational tools to enhance environmental monitoring and inform mitigation strategies.