Machine Learning-Enhanced Air Quality Management via QFIELD and QGIS: Mapping and Bio-cover Materials at Ojota Landfill

Abstract Landfill emissions represent a major environmental challenge in rapidly urbanizing cities within developing countries. Methane, particulate matter, and odor-generating compounds released from poorly managed landfill sites contribute to air pollution and pose risks to surrounding communities. This study evaluates the effectiveness of a laboratory-simulated bio-cover system formulated using landfill topsoil, compost, rice husk and earthworm castings in a 6:2:1:1 mix ratio for mitigating microbial and gaseous emissions at the Ojota landfill in Lagos, Nigeria. Spatially informed field measurements are conducted at six sampling points established at 15m intervals across the landfill, coordinates recorded and mapped via QFIELD and QGIS to provide accurate spatial representation of emission gradients across the study area. Air quality parameters including particulate matter (PM2.5 and PM10), sulfur dioxide (SO₂), and carbon monoxide (CO) are monitored to establish baseline environmental conditions. To extend this point-based assessment, six machine learning approaches — Random Forest, Gradient Boosting (XGBoost/LightGBM), Support Vector Machine, K-Nearest Neighbours, Gaussian Process Regression/Regression Kriging, and an Artificial Neural Network — are applied to the geo-referenced air quality dataset to classify AQI status by pollutant driver and to generate a continuous, uncertainty-aware pollutant risk surface across the landfill footprint. Microbial indicators including total bacterial count (TBC), total coliform count (TCC), and fecal coliform (FC) are determined, while physicochemical characterization involves pH measurements and Fourier transform infrared spectroscopy (FTIR) over soil samples collected. Results show significant reductions in microbial indicators following bio-cover application, improved biological regulation within the cover material, pH values remain within near-neutral conditions favorable for microbial oxidation processes. FTIR spectra further reveal functional groups associated with organic matter stabilization and biodegradation. The novelty of this study lies in integrating geospatial air quality mapping and machine learning-based prediction with microbiological and FTIR-based biochemical characterization to evaluate a locally formulated bio-cover system under tropical landfill conditions, providing a sustainable and cost-effective strategy for improving landfill air quality management.