Integration of nontarget analysis with machine learning modeling for prioritization of odorous volatile organic compounds in surface water

Abstract

Assessing the odor risk caused by volatile organic compounds (VOCs) in water has been a big challenge for water quality evaluation due to the abundance of odorants in water and the inherent difficulty in obtaining the corresponding odor sensory attributes. Here, a novel odor risk assessment approach has been established, incorporating nontarget screening for odorous VOC identification and machine learning (ML) modeling for odor threshold prediction. Twenty-nine odorous VOCs were identified using two-dimensional gas chromatography-time of flight mass spectrometry from four surface water sampling sites. These identified odorants primarily fell into the categories of ketones and ethers, and originated mainly from biological production. To obtain the odor threshold of these odorants, we trained an ML model for odor threshold prediction, which displayed good performance with accuracy of 79%. Further, an odor threshold-based prioritization approach was developed to rank the identified odorants. 2-Methylisoborneol and nonanal were identified as the main odorants contributing to water odor issues at the four sampling sites. This study provides an accessible method for accurate and quick determination of key odorants in source water, aiding in odor control and improved water quality management. Environmental ImplicationWater odor episodes have been persistent and significant issues worldwide, posing severe challenges to water treatment plants. Unpleasant odors in aquatic environments are predominantly caused by the occurrence of a wide range of volatile organic chemicals (VOCs). Given the vast number of newly-detected VOCs, experimental identification of the key odorants becomes difficult, making water odor issues complex to control. Herein, we propose a novel approach integrating nontarget analysis with machine learning models to accurate and quick determine the key odorants in waterbodies. We use the approach to analyze four samples with odor issues in Changsha, and prioritized the potential odorants.