Using machine learning to trace the pollution sources of disinfection by-products precursors compared to receptor models.
Sci Total Environ
; 914: 169671, 2024 Mar 01.
Article
en En
| MEDLINE
| ID: mdl-38184251
ABSTRACT
To increase the efficiency of managing backup water resources, it is critical to identify and allocate pollution sources. Source apportionment of dissolved organic matter (DOM) was investigated in our work. Parallel factor analysis (PARAFAC) and the Spearman correlation analysis were used for source identification. After that, a newly hybrid model applying the fuzzy c-means and support vector regression (FCM-SVR) was employed for source apportionment compared to receptor models. The results demonstrated that the FCM-SVR model exhibited excellent generalization, and only required standardization and normalization as pre-processing steps for dataset. According to the results, microbial sources played a key role (28.1 %) in the formation potential of disinfection byproducts (DBPFPs). Additionally, shipping marine sources exhibited a substantial contribution (21.2 %) to DBPFPs. The prediction accuracy of DBPFPs was matched or exceeded receptor models, and the R2 of DOC (0.884) was significantly high. Therefore, we recommend the FCM-SVR model combined with PARAFAC to trace the source of DBPFPs as its significant effectiveness in source identification, source apportionment, and prediction accuracy, possessing the potential for further applicability in tracking more organic compounds. ENVIRONMENTAL IMPLICATION The disinfection byproducts precursors in water sources, which were thought to be hazardous materials in this study, are proved to be chlorinated into carcinogenic disinfection byproducts (DBPs) during drinking water treatment, However, the source apportionment methods of DBPs are not well developed compared to other inorganic matter, e.g., heavy metals and ammonia nitrogen. We proposed a new FCM-SVR model to trace the source of DBPs, which required easier pre-treatment and resulted a better source apportionment and prediction accuracy. As a result, it could provide a different prospect and useful management advices to trace the source of DBPs.
Palabras clave
Texto completo:
1
Bases de datos:
MEDLINE
Asunto principal:
Contaminantes Químicos del Agua
/
Purificación del Agua
/
Desinfectantes
Tipo de estudio:
Prognostic_studies
Idioma:
En
Revista:
Sci Total Environ
Año:
2024
Tipo del documento:
Article
País de afiliación:
China