Comprehensive characterisation of landfill leachate odours by using an electronic nose with multiple qualitative and quantitative modelling

Odorants released from landfills have been widely reported as a significant source of odour nuisance, causing discomfort to the neighbourhood and decreasing air quality in the surrounding area. Therefore, to attenuate the negative impact of these pollutants, an extensive analysis of their emissions is required. This paper presents an integrated framework based on a dedicated electronic nose (e-nose) for odour assessments emanating from a landfill. Odour samples were collected from four operational sources within the facility: Leachate pond, Leachate treatment pond, Fresh waste, and Covered soil, totalling 40 sampling bags for laboratory analysis. Consequently, odour emissions from both solid and liquid waste surfaces were considered. Multiple data processing techniques were performed for qualitative and quantitative assessments. Initially, boxplots and outlier detection were applied. Then, odour fingerprints, Principal Component Analysis (PCA), and a Multi-Layer Perceptron (MLP) classifier were used to discriminate odours from their sources and to detect relationships among the samples. Qualitative results showed that the e-nose could reliably distinguish landfill odours, with the MLP achieving an overall accuracy greater than 84% across repeated runs. Quantitative prediction of odour concentration, expressed in odour units per cubic meter ( ou E m − 3 ), was computed. Using an MLP regression model, R adj values were 90.35, 85.80 and 83.30 for the training, validation and test datasets, respectively. Also, low prediction errors were obtained. Overall, these results demonstrate that the proposed e-nose, when combined with multivariate analysis and targeted calibration, represents a feasible, robust, and promising tool for both qualitative source identification and quantitative estimation of landfill odours.