Eliminating collinearity observed in chemical mass balance analysis using multi linear regression and synthetic receptor source profile dataset

• Published in: Sadhana (Bangalore) Vol. 49; no. 1
• Main Authors: Rejivas, V A, Praveen, A, Ajitha, T
• Format: Journal Article
• Language: English
• Published: New Delhi Springer India 28.01.2024
• Subjects: Engineering; source profile; synthetic receptor data sets; particulate matter emissions; chemical mass balance; multi linear regression model
• ISSN: 0973-7677; 0973-7677
• DOI: 10.1007/s12046-023-02403-w

Getting accurate source contribution results through Chemical Mass Balance (CMB) analysis requires unique and specific elemental species characterization within the source profiles. However, this uniqueness often gets compromised when the mixing of multiple sources occurs, leading to slightly skewed species characterizations due to the variability in the source representations. Consequently, the collinearity of species, leads to undesirable negative source apportionment outcomes, during the CMB analysis. To address this issue, demonstrating the effectiveness of Multi Linear Regression parameters, including tolerance and variance inflation factor, is crucial. Additionally, collinearity diagnostics, such as the condition index and variance decomposition proportions, are showcased as part of this effort to analyse the problem. These parameters were applied to identify the contributing sources of collinearity in the source apportionment result of particulate matter samples collected from a semi-urban road segment. For eliminating the collinearity in the source profiles, synthetic receptor source profile datasets are created by replacing the elemental species mass proportions of collinear source profiles with those from available literature and found as an effective tool that can be employed in CMB analysis. Furthermore, the synthetic receptor source profile data can enhance source contributions by determining more number of sources, which was achieved through the execution of CMB analysis by introducing a brake tyre source profile from the literature alongside the synthetic receptor source profile data sets.