Read the full-text of this article for free

Title: Monitoring and sustainable management of soil microbial environmental quality based on machine learning

Authors: Miao Wang; Yu Zhang; Zeyu Wang; Yuxin Zhang; Yankui Chen; Jiao Zhang; Yu Zhang

Addresses: College of Agricultural Engineering, Xinjiang Agricultural Vocational and Technical University, Changji, 831100, Xinjiang, China ' College of Architectural Engineering, Xinjiang Vocational University of Science and Engineering, Kashgar, 844100, Xinjiang, China; Faculty of Engineering, UCSI University, Kuala Lumpur, 56000, Malaysia; Guangdong Digital Orchard Engineering Technology Research Center, Meizhou, 514015, Guangdong, China; College of Geographical Science and Tourism, Jiaying University, Meizhou 514015, Guangdong, China ' Institute of Soil Fertilizer and Agricultural Water Saving, Xinjiang Academy of Agricultural Sciences, Urumqi, 830091, Xinjiang, China ' Faculty of Agricultural, Putra University, Selangor, 43400, Malaysia ' Faculty of Engineering, UCSI University, Kuala Lumpur, 56000, Malaysia; Guangdong Digital Orchard Engineering Technology Research Center, Meizhou, 514015, Guangdong, China; College of Geographical Science and Tourism, Jiaying University, Meizhou, 514015, Guangdong, China ' Faculty of Engineering, UCSI University, Kuala Lumpur, 56000, Malaysia ' International College, Krirk University, Bangkok, Khet Bang Khen, 10220, Thailand

Abstract: Soil microbial environmental quality is an essential indicator of ecosystem health and agricultural productivity. However, current monitoring and management face challenges, including difficulty in integrating multi-source data and the poor sustainability of management measures. To address these issues, this article investigated a comprehensive method based on multimodal learning and graph neural network (GNN). By utilising multimodal learning models, multiple data sources - including microbial sequencing, soil physicochemical properties, and climate data - were integrated, and the features of each modality were extracted and fused. Using a GNN, the complex relationships between microorganisms and environmental factors were modelled to generate reliable predictions of ecological quality. Based on the predicted results, an adaptive management framework was designed to adjust management measures using real-time monitoring data dynamically. Finally, automatic optimisation of management strategies was achieved by applying a dynamic management system.

Keywords: soil microorganisms; environmental quality monitoring; multimodal learning; sustainable management; adaptive management framework; machine learning; graph neural networks.

DOI: 10.1504/IJEP.2026.153596

International Journal of Environment and Pollution, 2026 Vol.76 No.6, pp.1 - 22

Received: 07 Jul 2025
Accepted: 03 Dec 2025
Published online: 18 May 2026 *