Your access to this publication has been provided by Learn More
Treatment and Resource Recovery
Prediction of Oxidant Exposures and Micropollutant Abatement during Ozonation Using a Machine Learning Method
使用机器学习方法预测臭氧化过程中的氧化剂暴露和微污染物去除Click to copy article link
点击复制文章链接Article link copied!
点击复制文章链接Article link copied!
- Dongwon ChaDongwon ChaSchool of Chemical and Biological Engineering, Institute of Chemical Process (ICP), Seoul National University, 1 Gwanak-ro, Gwanak-gu, Seoul 08826, Republic of KoreaMore by Dongwon Cha
- Sanghun ParkSanghun ParkSchool of Urban and Environmental Engineering, Ulsan National Institute of Science and Technology (UNIST), 50 UNIST-gil, Ulju-gun, Ulsan 44919, Republic of KoreaMore by Sanghun Park
- Min Sik KimMin Sik KimSchool of Chemical and Biological Engineering, Institute of Chemical Process (ICP), Seoul National University, 1 Gwanak-ro, Gwanak-gu, Seoul 08826, Republic of KoreaDepartment of Chemical and Environmental Engineering, Yale University, New Haven, Connecticut 06520, United StatesMore by Min Sik Kim
- Taewan KimTaewan KimSchool of Chemical and Biological Engineering, Institute of Chemical Process (ICP), Seoul National University, 1 Gwanak-ro, Gwanak-gu, Seoul 08826, Republic of KoreaMore by Taewan Kim
- Seok Won HongSeok Won HongWater Cycle Research Center, Korea Institute of Science and Technology (KIST), 5 Hwarang-ro 14-gil, Seongbuk-gu, Seoul 02792, Republic of KoreaMore by Seok Won Hong
- Kyung Hwa Cho*Kyung Hwa Cho*Email: khcho@unist.ac.kr. Phone: +82-52-217-2829. Fax: +82-52-217-2819.School of Urban and Environmental Engineering, Ulsan National Institute of Science and Technology (UNIST), 50 UNIST-gil, Ulju-gun, Ulsan 44919, Republic of KoreaMore by Kyung Hwa Cho
- Changha Lee*Changha Lee*Email: leechangha@snu.ac.kr. Phone: +82-2-880-8630. Fax: +82-2-888-7295.School of Chemical and Biological Engineering, Institute of Chemical Process (ICP), Seoul National University, 1 Gwanak-ro, Gwanak-gu, Seoul 08826, Republic of KoreaMore by Changha Lee
Abstract 摘要
Click to copy section link
点击复制部分链接Section link copied!
点击复制部分链接Section link copied!
Oxidation of micropollutants (MPs) by ozonation proceeds via the reactions with molecular ozone (O3) and hydroxyl radicals (•OH). To predict MP abatement during ozonation, a model that can accurately predict oxidant exposures (i.e.,
微污染物（MPs）的氧化通过与分子臭氧（O3）和羟基自由基（•OH）的反应进行。为了预测臭氧化过程中的 MP 去除，需建立一个能够准确预测氧化剂暴露的模型。) needs to be developed. This study demonstrates machine learning models based on the random forest (RF) algorithm to output oxidant exposures from water quality parameters (input variables) that include pH, alkalinity, dissolved organic carbon concentration, and fluorescence excitation–emission matrix (FEEM) data (to characterize organic matter). To develop the models, 60 different samples of natural waters and wastewater effluents were collected and characterized, and the oxidant exposures in each sample were determined at a specific O3 dose (2.5 mg/L). Four RF models were developed depending on how FEEM data were utilized (i.e., one model free of FEEM data, and three other models that used FEEM data of different resolutions). The regression performance and Akaike information criterion (AIC) were evaluated for each model. The models using high-resolution FEEM data generally exhibited high prediction accuracy with reasonable AIC values, implying that organic matter characteristics quantified by FEEM can be important factors to improve the accuracy of the prediction model. The developed models can be applied to predict the abatement of MPs in drinking water and wastewater ozonation processes and to optimize the O3 dose for the intended removal of target MPs. The machine learning models using higher-resolution FEEM data offer more accurate prediction by better calculating the complex nonlinear relationship between organic characteristics and oxidant exposures.
需要开发。本研究展示了基于随机森林（RF）算法的机器学习模型，以从水质参数（输入变量）中输出氧化剂暴露量，这些参数包括 pH、碱度、溶解有机碳浓度和荧光激发-发射矩阵（FEEM）数据（用于表征有机物）。为了开发这些模型，收集并表征了 60 个不同的自然水样和废水排放样本，并在特定的臭氧剂量（2.5 mg/L）下确定了每个样本中的氧化剂暴露量。根据 FEEM 数据的使用方式，开发了四个 RF 模型（即一个不使用 FEEM 数据的模型，以及三个使用不同分辨率 FEEM 数据的模型）。评估了每个模型的回归性能和赤池信息量准则（AIC）。使用高分辨率 FEEM 数据的模型通常表现出较高的预测准确性和合理的 AIC 值，这意味着通过 FEEM 量化的有机物特征可能是提高预测模型准确性的关键因素。 所开发的模型可以用于预测饮用水和废水臭氧化过程中的微污染物（MPs）去除情况，并优化用于去除目标微污染物的臭氧（O3）剂量。使用高分辨率荧光增强光谱（FEEM）数据的机器学习模型通过更好地计算有机特征与氧化剂暴露之间复杂的非线性关系，提供了更准确的预测。
This publication is licensed under the terms of your
institutional subscription.
Request reuse permissions.
本出版物根据您的机构订阅条款获得许可。请求重用权限。
Copyright © 2020 American Chemical Society
版权 © 2020 美国化学学会
版权 © 2020 美国化学学会
Introduction 引言
Click to copy section link
点击复制部分链接Section link copied!
点击复制部分链接Section link copied!
Ozonation is an effective process for the degradation of refractory chemical pollutants and the inactivation of pathogenic microorganisms. This process has been extensively studied for the treatment of water and wastewater for the past few decades, targeting different contaminants that include taste and odor compounds, toxins, metal species, and natural organic matter, as well as pathogens. (1−3) Recent studies on ozonation have focused more on the oxidation of micropollutants (MPs), such as pharmaceuticals and personal care products, in drinking water and secondary wastewater. (4)
臭氧化是降解难降解化学污染物和灭活病原微生物的有效过程。在过去几十年中，该过程已被广泛研究用于水和废水的处理，针对不同的污染物，包括味道和气味化合物、毒素、金属物种、天然有机物以及病原体。(1−3) 最近关于臭氧化的研究更侧重于饮用水和二次废水中微污染物（MPs）的氧化，例如药物和个人护理产品。(4)
臭氧化是降解难降解化学污染物和灭活病原微生物的有效过程。在过去几十年中，该过程已被广泛研究用于水和废水的处理，针对不同的污染物，包括味道和气味化合物、毒素、金属物种、天然有机物以及病原体。(1−3) 最近关于臭氧化的研究更侧重于饮用水和二次废水中微污染物（MPs）的氧化，例如药物和个人护理产品。(4)
Determining the optimal dose of ozone (O3) in the ozonation process is crucial to prevent problems caused by underdose (e.g., incomplete removal of target contaminants) and overdose (e.g., formation of harmful byproducts and excessive operating cost). (1,5,6) In order to do this, the abatement of target contaminants during ozonation needs to be precisely predicted. The kinetics for the abatement of an MP during ozonation can be expressed by equations with rate constants and oxidant exposures (i.e., 
) (eq 1 and the subsequent eq 2, derived from the integration of eq 1), (2) where kO3 and k•OH are the second-order rate constants for the reactions of MP with O3 and hydroxyl radicals (•OH), respectively.
确定臭氧（O3）在臭氧化过程中的最佳剂量对于防止因剂量不足（例如，目标污染物去除不完全）和剂量过多（例如，形成有害副产物和过高的运营成本）而导致的问题至关重要。(1,5,6) 为此，需要精确预测臭氧化过程中目标污染物的去除情况。目标污染物在臭氧化过程中的去除动力学可以用包含速率常数和氧化剂暴露（即 ）的方程表示（方程 1 及其后续方程 2，源自方程 1 的积分），(2) 其中 kO3 和 kOH 分别是目标污染物与 O3 和羟基自由基（•OH）反应的二级速率常数。
) (eq 1 and the subsequent eq 2, derived from the integration of eq 1), (2) where kO3 and k•OH are the second-order rate constants for the reactions of MP with O3 and hydroxyl radicals (•OH), respectively.确定臭氧（O3）在臭氧化过程中的最佳剂量对于防止因剂量不足（例如，目标污染物去除不完全）和剂量过多（例如，形成有害副产物和过高的运营成本）而导致的问题至关重要。(1,5,6) 为此，需要精确预测臭氧化过程中目标污染物的去除情况。目标污染物在臭氧化过程中的去除动力学可以用包含速率常数和氧化剂暴露（即
）的方程表示（方程 1 及其后续方程 2，源自方程 1 的积分），(2) 其中 kO3 和 kOH 分别是目标污染物与 O3 和羟基自由基（•OH）反应的二级速率常数。(1)
(2)
The rate constant (kO3 and k•OH) values for a number of MPs are already well documented, (7,8) while unknown values, if any, can be readily determined from laboratory experiments. The oxidant exposure (
and 
) values vary depending on water quality parameters and O3 dose and thus are determined experimentally. (2) The 
value can be calculated by monitoring O3 decay over time, and the 
value is usually quantified by the decomposition kinetics of an externally supplied •OH probe compound, such as para-chlorobenzoic acid (pCBA). (9,10) The kinetic equation for the MP abatement (eq 2) can also be expressed using the Rct concept (usually the ratio of 
to 
); (10) there are also other modified expressions based on Rct. (11) However, determining the Rct value (or modified analogues) requires the measurements of decay (or demand) of O3 and the •OH probe compound during the ozonation process, and thus, the kinetic expressions based on Rct are not conceptually different from eq 2.
许多微污染物（MP）的速率常数（kO3 和 kOH）值已经有很好的文献记录（7,8），而未知值（如果有的话）可以通过实验室实验轻松确定。氧化剂暴露（ 和 ）值因水质参数和臭氧（O3）剂量而异，因此通过实验确定。（2） 值可以通过监测 O3 随时间的衰减来计算，而 值通常通过外部提供的•OH 探针化合物（如对氯苯甲酸（pCBA））的分解动力学来量化。（9,10）微污染物去除的动力学方程（方程 2）也可以使用 Rct 概念表示（通常是 与 的比率）；（10）还有其他基于 Rct 的修正表达式。（11）然而，确定 Rct 值（或修正的类似物）需要在臭氧化过程中测量 O3 和•OH 探针化合物的衰减（或需求），因此，基于 Rct 的动力学表达式在概念上与方程 2 并没有不同。
and ) values vary depending on water quality parameters and O3 dose and thus are determined experimentally. (2) The value can be calculated by monitoring O3 decay over time, and the value is usually quantified by the decomposition kinetics of an externally supplied •OH probe compound, such as para-chlorobenzoic acid (pCBA). (9,10) The kinetic equation for the MP abatement (eq 2) can also be expressed using the Rct concept (usually the ratio of to ); (10) there are also other modified expressions based on Rct. (11) However, determining the Rct value (or modified analogues) requires the measurements of decay (or demand) of O3 and the •OH probe compound during the ozonation process, and thus, the kinetic expressions based on Rct are not conceptually different from eq 2.许多微污染物（MP）的速率常数（kO3 和 kOH）值已经有很好的文献记录（7,8），而未知值（如果有的话）可以通过实验室实验轻松确定。氧化剂暴露（
和 ）值因水质参数和臭氧（O3）剂量而异，因此通过实验确定。（2） 值可以通过监测 O3 随时间的衰减来计算，而 值通常通过外部提供的•OH 探针化合物（如对氯苯甲酸（pCBA））的分解动力学来量化。（9,10）微污染物去除的动力学方程（方程 2）也可以使用 Rct 概念表示（通常是 与 的比率）；（10）还有其他基于 Rct 的修正表达式。（11）然而，确定 Rct 值（或修正的类似物）需要在臭氧化过程中测量 O3 和•OH 探针化合物的衰减（或需求），因此，基于 Rct 的动力学表达式在概念上与方程 2 并没有不同。For the full-scale ozonation process at treatment plants, it is not feasible to experimentally determine oxidant exposures. Therefore, for field application, the prediction of oxidant exposures using a mathematical model can be an alternative solution. In our recent study, empirical models were developed to predict oxidant exposures during the ozonation of natural waters using response surface methodology (RSM) with the O3 dose and water quality parameters (i.e., pH, alkalinity, dissolved organic carbon (DOC) concentration, and temperature) as independent variables. (12) These models successfully predicted oxidant exposures and MP abatement in the water source used to create the models. However, when applied to other natural waters, they did not show high accuracy, possibly because of the different characteristics of natural organic matter (NOM).
在处理厂的全规模臭氧化过程中，实验性地确定氧化剂暴露是不切实际的。因此，对于现场应用，使用数学模型预测氧化剂暴露可以作为一种替代解决方案。在我们最近的研究中，采用响应面方法（RSM）开发了经验模型，以预测自然水体臭氧化过程中的氧化剂暴露，独立变量包括 O3 剂量和水质参数（即 pH、碱度、溶解有机碳（DOC）浓度和温度）。这些模型成功地预测了用于创建模型的水源中的氧化剂暴露和微污染物去除。然而，当应用于其他自然水体时，它们并未显示出高准确性，这可能是由于自然有机物（NOM）的不同特性所致。
在处理厂的全规模臭氧化过程中，实验性地确定氧化剂暴露是不切实际的。因此，对于现场应用，使用数学模型预测氧化剂暴露可以作为一种替代解决方案。在我们最近的研究中，采用响应面方法（RSM）开发了经验模型，以预测自然水体臭氧化过程中的氧化剂暴露，独立变量包括 O3 剂量和水质参数（即 pH、碱度、溶解有机碳（DOC）浓度和温度）。这些模型成功地预测了用于创建模型的水源中的氧化剂暴露和微污染物去除。然而，当应用于其他自然水体时，它们并未显示出高准确性，这可能是由于自然有机物（NOM）的不同特性所致。
In order to develop more comprehensive models that are applicable to a wide range of water types, the fluorescence excitation–emission matrix (FEEM) data can be considered as input variables. The FEEM is a useful tool to analyze the characteristics of NOM and wastewater effluent organic matter (EfOM) in water samples. (13,14) The FEEM data may reflect the reactivity of NOM and EfOM with O3 and •OH; in fact, it has been reported that the FEEM spectra of natural water and wastewater changed after ozonation. (15,16) To the best of our knowledge, FEEM data have not been used for any purposes related to the prediction or interpretation of oxidant exposures and MP abatement during ozonation.
为了开发适用于多种水类型的更全面模型，荧光激发-发射矩阵（FEEM）数据可以作为输入变量。FEEM 是分析水样中天然有机物（NOM）和废水排放有机物（EfOM）特征的有用工具。FEEM 数据可能反映 NOM 和 EfOM 与臭氧（O3）和•OH 的反应性；事实上，有报道指出，天然水和废水的 FEEM 光谱在臭氧化后发生了变化。根据我们所知，FEEM 数据尚未用于与臭氧化过程中氧化剂暴露和微塑料去除相关的任何目的。
为了开发适用于多种水类型的更全面模型，荧光激发-发射矩阵（FEEM）数据可以作为输入变量。FEEM 是分析水样中天然有机物（NOM）和废水排放有机物（EfOM）特征的有用工具。FEEM 数据可能反映 NOM 和 EfOM 与臭氧（O3）和•OH 的反应性；事实上，有报道指出，天然水和废水的 FEEM 光谱在臭氧化后发生了变化。根据我们所知，FEEM 数据尚未用于与臭氧化过程中氧化剂暴露和微塑料去除相关的任何目的。
Meanwhile, a machine learning technique needs to be considered for the development of a model using FEEM data. This technique is suitable to subsume an enormous number of FEEM data points into the model; in contrast, the RSM-based polynomial model can accommodate only a limited number of independent variables. Furthermore, the machine learning model is able to find the complex nonlinear relationship between independent and dependent (i.e., input and output) variables. (17)
与此同时，在使用 FEEM 数据开发模型时，需要考虑一种机器学习技术。这种技术适合将大量的 FEEM 数据点纳入模型；相比之下，基于 RSM 的多项式模型只能容纳有限数量的自变量。此外，机器学习模型能够找到自变量和因变量（即输入和输出）之间复杂的非线性关系。
与此同时，在使用 FEEM 数据开发模型时，需要考虑一种机器学习技术。这种技术适合将大量的 FEEM 数据点纳入模型；相比之下，基于 RSM 的多项式模型只能容纳有限数量的自变量。此外，机器学习模型能够找到自变量和因变量（即输入和输出）之间复杂的非线性关系。
This study aims to develop machine learning models that can successfully predict O3 and •OH exposures and accordingly MP abatement during ozonation, using water parameters (i.e., pH, alkalinity, DOC concentration, and FEEM data) as input variables. The random forest (RF) algorithm, one of the most popular machine learning methods, was chosen for model development. Sixty different water samples (30 natural waters and 30 wastewater effluents) were collected, and their water quality parameters (input variables for the models) were analyzed. In addition, the oxidant exposures (output variables for the models) and the abatement of selected MPs in those water samples were measured at a specific O3 dose (2.5 mg/L). The machine learning models were developed and validated using training and validation subsets of variables, respectively.
本研究旨在开发能够成功预测臭氧（O3）和羟基自由基（•OH）暴露以及相应的微污染物（MP）去除的机器学习模型，使用水参数（即 pH、碱度、溶解有机碳（DOC）浓度和荧光增强光谱（FEEM）数据）作为输入变量。随机森林（RF）算法，作为最流行的机器学习方法之一，被选用于模型开发。收集了六十个不同的水样（30 个自然水样和 30 个废水排放样），并分析了它们的水质参数（模型的输入变量）。此外，在特定的臭氧剂量（2.5 mg/L）下，测量了这些水样中的氧化剂暴露（模型的输出变量）和选定微污染物的去除情况。机器学习模型分别使用变量的训练和验证子集进行开发和验证。
本研究旨在开发能够成功预测臭氧（O3）和羟基自由基（•OH）暴露以及相应的微污染物（MP）去除的机器学习模型，使用水参数（即 pH、碱度、溶解有机碳（DOC）浓度和荧光增强光谱（FEEM）数据）作为输入变量。随机森林（RF）算法，作为最流行的机器学习方法之一，被选用于模型开发。收集了六十个不同的水样（30 个自然水样和 30 个废水排放样），并分析了它们的水质参数（模型的输入变量）。此外，在特定的臭氧剂量（2.5 mg/L）下，测量了这些水样中的氧化剂暴露（模型的输出变量）和选定微污染物的去除情况。机器学习模型分别使用变量的训练和验证子集进行开发和验证。
Materials and Methods 材料与方法
Click to copy section link
点击复制部分链接Section link copied!
点击复制部分链接Section link copied!
Reagents 试剂
All chemicals were of reagent grade (Sigma-Aldrich) and were used as received without additional purification. All aqueous stock solutions used deionized (DI) water (>18.2 MΩ cm) provided by a water purification system (Millipore Milli-Q Integral 5). The stock solution of O3 (>20 mg/L) was prepared in a reactor placed in a cooling bath by sparging O3-containing gas generated from an O3 generator (OzoneTech Lab-II) into DI water.
所有化学品均为试剂级（Sigma-Aldrich），并在未经过额外纯化的情况下直接使用。所有使用的水相储备溶液均采用去离子水（DI 水，>18.2 MΩ cm），由水纯化系统（Millipore Milli-Q Integral 5）提供。O3 的储备溶液（>20 mg/L）是在冷却浴中的反应器内制备的，通过将从 O3 发生器（OzoneTech Lab-II）产生的含 O3 气体通入 DI 水中。
所有化学品均为试剂级（Sigma-Aldrich），并在未经过额外纯化的情况下直接使用。所有使用的水相储备溶液均采用去离子水（DI 水，>18.2 MΩ cm），由水纯化系统（Millipore Milli-Q Integral 5）提供。O3 的储备溶液（>20 mg/L）是在冷却浴中的反应器内制备的，通过将从 O3 发生器（OzoneTech Lab-II）产生的含 O3 气体通入 DI 水中。
Sampling and Characterization of Natural Water and Wastewater Effluent
自然水和废水排放的采样与表征
Thirty natural water samples were collected at different positions of five major rivers in the Republic of Korea (i.e., the Han, Geum, Nakdong, Yeongsan, and Sumjin Rivers) and their tributaries (refer to Figure S1 of the Supporting Information for sampling locations). Additionally, 30 wastewater effluent samples were collected from 30 different wastewater treatment plants located in Seoul, as well as in Gyeonggi, South Jeolla, and Gangwon provinces. The water samples were immediately filtered using glass microfiber filters (GE Whatman GF/C) and subsequently using nylon membrane filters (GE Whatman 0.45 μm) and stored at 4 °C until use.
在韩国共和国的五条主要河流（即汉江、金江、洛东江、荣山江和隋珍江）及其支流的不同位置收集了三十个自然水样（采样位置请参见支持信息的图 S1）。此外，还从位于首尔以及京畿道、全南和江原道的三十个不同污水处理厂收集了三十个废水排放样本。水样立即使用玻璃微纤维过滤器（GE Whatman GF/C）过滤，随后使用尼龙膜过滤器（GE Whatman 0.45 μm）过滤，并在 4°C 下保存直至使用。
在韩国共和国的五条主要河流（即汉江、金江、洛东江、荣山江和隋珍江）及其支流的不同位置收集了三十个自然水样（采样位置请参见支持信息的图 S1）。此外，还从位于首尔以及京畿道、全南和江原道的三十个不同污水处理厂收集了三十个废水排放样本。水样立即使用玻璃微纤维过滤器（GE Whatman GF/C）过滤，随后使用尼龙膜过滤器（GE Whatman 0.45 μm）过滤，并在 4°C 下保存直至使用。
Water quality parameters that included pH, alkalinity, DOC concentration, and FEEM data were analyzed for each sample. pH was measured using a pH meter (Thermo Scientific Orion Star A Series). DOC concentration was measured using a TOC analyzer (Sievers M5310 C); all water samples were filtered as mentioned above. Alkalinity was measured according to APHA Method 2320. (18) The FEEM data of water samples were obtained by fluorescence spectrophotometry (Hitachi F-4500). For FEEM analysis, excitation was performed from 200 to 400 nm at 5 nm intervals, while monitoring the emission spectra from 280 to 500 nm at 1 nm intervals. The scan speed was set at 1200 nm/min with both excitation and emission slits of 5 nm. The FEEM data of DI water were used as the baseline. (19)
水质参数包括 pH、碱度、DOC 浓度和 FEEM 数据，针对每个样本进行了分析。pH 值使用 pH 计（Thermo Scientific Orion Star A 系列）测量。DOC 浓度使用 TOC 分析仪（Sievers M5310 C）测量；所有水样均如上所述进行了过滤。碱度根据 APHA 方法 2320 进行测量。(18) 水样的 FEEM 数据通过荧光分光光度法（日立 F-4500）获得。FEEM 分析时，激发波长范围为 200 至 400 nm，间隔为 5 nm，同时监测 280 至 500 nm 的发射光谱，间隔为 1 nm。扫描速度设定为 1200 nm/min，激发和发射狭缝均为 5 nm。去离子水的 FEEM 数据用作基线。(19)
水质参数包括 pH、碱度、DOC 浓度和 FEEM 数据，针对每个样本进行了分析。pH 值使用 pH 计（Thermo Scientific Orion Star A 系列）测量。DOC 浓度使用 TOC 分析仪（Sievers M5310 C）测量；所有水样均如上所述进行了过滤。碱度根据 APHA 方法 2320 进行测量。(18) 水样的 FEEM 数据通过荧光分光光度法（日立 F-4500）获得。FEEM 分析时，激发波长范围为 200 至 400 nm，间隔为 5 nm，同时监测 280 至 500 nm 的发射光谱，间隔为 1 nm。扫描速度设定为 1200 nm/min，激发和发射狭缝均为 5 nm。去离子水的 FEEM 数据用作基线。(19)
Ozonation Experiments 臭氧化实验
All experiments were conducted using 50 mL of water samples spiked with pCBA (•OH probe compound for 
measurement, 1 μM) or the target MP compound (atrazine, caffeine, carbamazepine, and ibuprofen for MP abatement experiments, 0.1 μM). The experiments were conducted at room temperature (22 ± 1 °C) in conical flasks, with the caps closed. To initiate the reaction, an aliquot of O3 stock solution was injected into the reaction solution under stirring at 200 rpm ([O3]0 = 2.5 mg/L). Samples (2.5 mL) were withdrawn at predetermined time intervals and immediately added to indigo trisulfonate solution (0.28 mL) to quench the O3 residual (and at the same time, to determine the O3 concentration). For the MP abatement experiments, the final MP concentrations were measured after O3 is depleted in the solution; the time for the complete depletion of O3 varied from 1 to 30 min depending on the water sample.
所有实验均使用 50 毫升的水样，加入 pCBA（用于 测量的•OH 探针化合物，1 μM）或目标 MP 化合物（阿特拉津、咖啡因、卡马西平和布洛芬用于 MP 去除实验，0.1 μM）。实验在室温（22 ± 1 °C）下于锥形瓶中进行，瓶盖关闭。为启动反应，在搅拌 200 rpm 的条件下向反应溶液中注入一定量的 O3 储备溶液（[O3]0 = 2.5 mg/L）。在预定时间间隔内取出样品（2.5 mL），并立即加入靛蓝三磺酸盐溶液（0.28 mL）以消除 O3 残留（同时测定 O3 浓度）。对于 MP 去除实验，在 O3 在溶液中耗尽后测量最终的 MP 浓度；O3 完全耗尽的时间因水样而异，范围为 1 到 30 分钟。
measurement, 1 μM) or the target MP compound (atrazine, caffeine, carbamazepine, and ibuprofen for MP abatement experiments, 0.1 μM). The experiments were conducted at room temperature (22 ± 1 °C) in conical flasks, with the caps closed. To initiate the reaction, an aliquot of O3 stock solution was injected into the reaction solution under stirring at 200 rpm ([O3]0 = 2.5 mg/L). Samples (2.5 mL) were withdrawn at predetermined time intervals and immediately added to indigo trisulfonate solution (0.28 mL) to quench the O3 residual (and at the same time, to determine the O3 concentration). For the MP abatement experiments, the final MP concentrations were measured after O3 is depleted in the solution; the time for the complete depletion of O3 varied from 1 to 30 min depending on the water sample.所有实验均使用 50 毫升的水样，加入 pCBA（用于
测量的•OH 探针化合物，1 μM）或目标 MP 化合物（阿特拉津、咖啡因、卡马西平和布洛芬用于 MP 去除实验，0.1 μM）。实验在室温（22 ± 1 °C）下于锥形瓶中进行，瓶盖关闭。为启动反应，在搅拌 200 rpm 的条件下向反应溶液中注入一定量的 O3 储备溶液（[O3]0 = 2.5 mg/L）。在预定时间间隔内取出样品（2.5 mL），并立即加入靛蓝三磺酸盐溶液（0.28 mL）以消除 O3 残留（同时测定 O3 浓度）。对于 MP 去除实验，在 O3 在溶液中耗尽后测量最终的 MP 浓度；O3 完全耗尽的时间因水样而异，范围为 1 到 30 分钟。Analytical Methods 分析方法
The concentration of O3 was determined by monitoring the decolorization of indigo trisulfonate (the absorbance decrease at 600 nm) using UV/Vis spectrophotometry (PerkinElmer LAMBDA 465). (20) The concentrations of pCBA and MPs were analyzed by rapid separation liquid chromatography (RSLC) (Thermo Scientific UltiMate 3000) with UV absorbance detection. Phosphoric acid solution (0.1 wt %) and acetonitrile were used as eluents, and the separation was performed on a C18 column (Thermo Scientific Acclaim 120) (refer to Table S1 of the Supporting Information for detailed information about the RSLC analysis).
O3 的浓度通过监测靛蓝三磺酸盐的脱色（在 600 nm 处的吸光度降低）来确定，使用 UV/Vis 分光光度法（PerkinElmer LAMBDA 465）。(20) pCBA 和微塑料的浓度通过快速分离液相色谱（RSLC）（Thermo Scientific UltiMate 3000）进行分析，采用 UV 吸光度检测。磷酸溶液（0.1 wt%）和乙腈作为洗脱剂，分离在 C18 柱（Thermo Scientific Acclaim 120）上进行（有关 RSLC 分析的详细信息，请参见支持信息的表 S1）。
O3 的浓度通过监测靛蓝三磺酸盐的脱色（在 600 nm 处的吸光度降低）来确定，使用 UV/Vis 分光光度法（PerkinElmer LAMBDA 465）。(20) pCBA 和微塑料的浓度通过快速分离液相色谱（RSLC）（Thermo Scientific UltiMate 3000）进行分析，采用 UV 吸光度检测。磷酸溶液（0.1 wt%）和乙腈作为洗脱剂，分离在 C18 柱（Thermo Scientific Acclaim 120）上进行（有关 RSLC 分析的详细信息，请参见支持信息的表 S1）。
Machine Learning Process 机器学习过程
The regression tree (RT) method creates a model in the form of a tree structure to correlate the input and output variables. While the model is being developed, it sequentially breaks down a large data set into smaller and homogenous subsets. (21,22) The RT method contains two major steps; the first step allocates the entire data into the first two subsets by determining the optimal split that minimizes the sum of the squared deviations from the mean values of the two separated subsets. This splitting rule is further applied to each of the two subsets and is continued until it reaches the terminal nodes or the minimum size of a subset. The second step removes tree sections or nodes that have little influence on the target variables to achieve a more accurate prediction with less model complexity. This second step is conducted to avoid overfitting, which leads to poor results with new data sets.
回归树（RT）方法以树结构的形式创建模型，以关联输入和输出变量。在模型开发过程中，它将大型数据集逐步分解为更小且同质的子集。（21,22）RT 方法包含两个主要步骤；第一步通过确定最优分割来将整个数据分配到前两个子集中，以最小化两个分离子集均值的平方偏差之和。这个分割规则进一步应用于两个子集，并持续进行，直到达到终端节点或子集的最小大小。第二步移除对目标变量影响较小的树部分或节点，以实现更准确的预测并降低模型复杂性。此第二步的实施是为了避免过拟合，这会导致在新数据集上表现不佳。
回归树（RT）方法以树结构的形式创建模型，以关联输入和输出变量。在模型开发过程中，它将大型数据集逐步分解为更小且同质的子集。（21,22）RT 方法包含两个主要步骤；第一步通过确定最优分割来将整个数据分配到前两个子集中，以最小化两个分离子集均值的平方偏差之和。这个分割规则进一步应用于两个子集，并持续进行，直到达到终端节点或子集的最小大小。第二步移除对目标变量影响较小的树部分或节点，以实现更准确的预测并降低模型复杂性。此第二步的实施是为了避免过拟合，这会导致在新数据集上表现不佳。
RF is an ensemble and supervised learning method for classification or regression by developing multiple RTs with randomly sampled data (i.e., bootstrapped method) from the original data set. (23) The RF model can be developed through a much simpler training or optimization process compared to other machine learning algorithms, and the RF model has demonstrated reliable predictions in various environmental areas. (24−26) In brief, the RF model builds multiple trees to produce multiple predictions on target variables and averages them into a final output (eq 3):
随机森林（RF）是一种集成和监督学习方法，通过从原始数据集中随机抽样数据（即自助法）开发多个回归树（RT）进行分类或回归。与其他机器学习算法相比，RF 模型可以通过更简单的训练或优化过程进行开发，并且 RF 模型在各种环境领域中已显示出可靠的预测能力。简而言之，RF 模型构建多个树以对目标变量产生多个预测，并将其平均为最终输出（公式 3）：
where f̂(x) is the final output from the RF model, K is the number of trees, and T(x) is the output of each RT. MATLAB software (R2018b, MathWorks) was used to explore the RF algorithms. Further details, hyperparameters, and development procedures are described in Text S1 of the Supporting Information.
其中 f̂(x) 是 RF 模型的最终输出，K 是树的数量，T(x) 是每棵回归树的输出。使用 MATLAB 软件（R2018b，MathWorks）来探索 RF 算法。更多细节、超参数和开发过程在支持信息的文本 S1 中进行了描述。
随机森林（RF）是一种集成和监督学习方法，通过从原始数据集中随机抽样数据（即自助法）开发多个回归树（RT）进行分类或回归。与其他机器学习算法相比，RF 模型可以通过更简单的训练或优化过程进行开发，并且 RF 模型在各种环境领域中已显示出可靠的预测能力。简而言之，RF 模型构建多个树以对目标变量产生多个预测，并将其平均为最终输出（公式 3）：
(3)
其中 f̂(x) 是 RF 模型的最终输出，K 是树的数量，T(x) 是每棵回归树的输出。使用 MATLAB 软件（R2018b，MathWorks）来探索 RF 算法。更多细节、超参数和开发过程在支持信息的文本 S1 中进行了描述。
Model Development and Evaluation
模型开发与评估
Four RF models (i.e., FEEM-Free, FEEM-LowRes, FEEM-HighRes, and FEEM-FullRes) were devised with different input variables regarding the FEEM data; the use of higher resolution FEEM data (greater number of data points) leads to longer computation time. The three variables (pH, alkalinity, and DOC concentration) were used in common for the four RF models. FEEM-Free did not use the FEEM data as input variables. FEEM-LowRes and FEEM-HighRes used percent fluorescence response (P(i,n)) data obtained by the fluorescence regional integration (FRI) with low and high resolutions, respectively. Details of the FRI calculation are described elsewhere. (19) FEEM-FullRes used all FEEM data points as input variables.
四个 RF 模型（即 FEEM-Free、FEEM-LowRes、FEEM-HighRes 和 FEEM-FullRes）是根据 FEEM 数据的不同输入变量设计的；使用更高分辨率的 FEEM 数据（数据点数量更多）会导致更长的计算时间。这三个变量（pH、碱度和 DOC 浓度）在四个 RF 模型中是共同使用的。FEEM-Free 没有使用 FEEM 数据作为输入变量。FEEM-LowRes 和 FEEM-HighRes 分别使用了通过荧光区域积分（FRI）获得的低分辨率和高分辨率的荧光响应百分比（P(,n)）数据。FRI 计算的详细信息在其他地方有描述。(19) FEEM-FullRes 使用所有 FEEM 数据点作为输入变量。
四个 RF 模型（即 FEEM-Free、FEEM-LowRes、FEEM-HighRes 和 FEEM-FullRes）是根据 FEEM 数据的不同输入变量设计的；使用更高分辨率的 FEEM 数据（数据点数量更多）会导致更长的计算时间。这三个变量（pH、碱度和 DOC 浓度）在四个 RF 模型中是共同使用的。FEEM-Free 没有使用 FEEM 数据作为输入变量。FEEM-LowRes 和 FEEM-HighRes 分别使用了通过荧光区域积分（FRI）获得的低分辨率和高分辨率的荧光响应百分比（P(,n)）数据。FRI 计算的详细信息在其他地方有描述。(19) FEEM-FullRes 使用所有 FEEM 数据点作为输入变量。
The RF trees were trained for 70% (42 out of 60) of the water samples, while the remaining 30% of the samples (out-of-bag samples) were used for validation. (21) The prediction accuracy of the models was evaluated by calculating the coefficient of determination (R2) and the root mean square error (RMSE) for the linear regression of measured versus predicted values. Variable importance (VI) was assessed by measuring the extent of the decrease in prediction accuracy after reordering the variable; a single variable in out-of-bag samples was randomly reordered without changing the other variables. In addition, the Akaike information criterion (AIC), derived from information theory, was used to evaluate the performance of the RF models, where a smaller AIC value indicates better performance. (27,28) The residual sum of squares (RSS) and AIC were calculated by the following equations (eqs 4 and 5)
RF 树模型使用 70%的水样（60 个样本中的 42 个）进行训练，而剩余 30%的样本（袋外样本）用于验证。（21）通过计算测量值与预测值的线性回归的决定系数（R²）和均方根误差（RMSE）来评估模型的预测准确性。变量重要性（VI）通过测量在重新排序变量后预测准确性的下降程度来评估；在袋外样本中，单个变量被随机重新排序，而不改变其他变量。此外，基于信息理论的赤池信息量准则（AIC）用于评估 RF 模型的性能，其中较小的 AIC 值表示更好的性能。（27,28）残差平方和（RSS）和 AIC 通过以下方程计算（方程 4 和 5）。
where n is the number of observations, Yiobs is the ith observation, Yisim is the ith modeled value, and K̂ is the number of variables.
其中 n 是观察值的数量，Yobs 是第 th 个观察值，Ysim 是第 th 个模型值，K̂ 是变量的数量。
RF 树模型使用 70%的水样（60 个样本中的 42 个）进行训练，而剩余 30%的样本（袋外样本）用于验证。（21）通过计算测量值与预测值的线性回归的决定系数（R²）和均方根误差（RMSE）来评估模型的预测准确性。变量重要性（VI）通过测量在重新排序变量后预测准确性的下降程度来评估；在袋外样本中，单个变量被随机重新排序，而不改变其他变量。此外，基于信息理论的赤池信息量准则（AIC）用于评估 RF 模型的性能，其中较小的 AIC 值表示更好的性能。（27,28）残差平方和（RSS）和 AIC 通过以下方程计算（方程 4 和 5）。
(4)
(5)
其中 n 是观察值的数量，Yobs 是第 th 个观察值，Ysim 是第 th 个模型值，K̂ 是变量的数量。
Results and Discussion 结果与讨论
Click to copy section link
点击复制部分链接Section link copied!
点击复制部分链接Section link copied!
Determination of Oxidant Exposures
氧化剂暴露的测定
The exposures of O3 and •OH (i.e., 
) during ozonation were determined for the 60 water samples by monitoring the decay of O3 and pCBA, respectively. Figure 1 presents the time–concentration profiles of O3 and pCBA for two selected water samples (no. 2 for natural water and no. 5 for wastewater effluent; refer to Tables S2 and S3 of the Supporting Information, respectively). Upon adding O3, a portion of O3 is instantaneously consumed because of the direct reactions of O3 with NOM and EfOM [defined as instantaneous ozone demand (IOD)]. In the second phase, O3 decays slowly, following pseudo-first-order kinetics. (1,5,29) These typical patterns of O3 decay were clearly shown in our results (Figure 1a); in particular, the O3 decay in wastewater effluent was faster than that in natural water because of the higher organic loading. Meanwhile, the decomposition of pCBA during ozonation was faster in natural water than in wastewater effluent (Figure 1b), indicating that the •OH concentration is higher in natural water; organics usually scavenge •OH.
在臭氧化过程中，通过监测 O3 和 pCBA 的衰减，确定了 60 个水样中 O3 和•OH 的暴露量（即 ）。图 1 展示了两个选定水样（自然水样编号 2 和废水排放样编号 5）的 O3 和 pCBA 的时间-浓度曲线（具体参见支持信息中的表 S2 和 S3）。在添加 O3 后，由于 O3 与天然有机物（NOM）和效应有机物（EfOM）的直接反应，部分 O3 瞬时被消耗[定义为瞬时臭氧需求（IOD）]。在第二阶段，O3 缓慢衰减，遵循伪一级动力学（1,5,29）。这些 O3 衰减的典型模式在我们的结果中得到了清晰的展示（图 1a）；特别是，废水排放中的 O3 衰减速度快于自然水样，原因是有机负荷更高。同时，臭氧化过程中 pCBA 的分解在自然水中比在废水排放中更快（图 1b），这表明自然水中的•OH 浓度更高；有机物通常会清除•OH。
) during ozonation were determined for the 60 water samples by monitoring the decay of O3 and pCBA, respectively. Figure 1 presents the time–concentration profiles of O3 and pCBA for two selected water samples (no. 2 for natural water and no. 5 for wastewater effluent; refer to Tables S2 and S3 of the Supporting Information, respectively). Upon adding O3, a portion of O3 is instantaneously consumed because of the direct reactions of O3 with NOM and EfOM [defined as instantaneous ozone demand (IOD)]. In the second phase, O3 decays slowly, following pseudo-first-order kinetics. (1,5,29) These typical patterns of O3 decay were clearly shown in our results (Figure 1a); in particular, the O3 decay in wastewater effluent was faster than that in natural water because of the higher organic loading. Meanwhile, the decomposition of pCBA during ozonation was faster in natural water than in wastewater effluent (Figure 1b), indicating that the •OH concentration is higher in natural water; organics usually scavenge •OH.在臭氧化过程中，通过监测 O3 和 pCBA 的衰减，确定了 60 个水样中 O3 和•OH 的暴露量（即
）。图 1 展示了两个选定水样（自然水样编号 2 和废水排放样编号 5）的 O3 和 pCBA 的时间-浓度曲线（具体参见支持信息中的表 S2 和 S3）。在添加 O3 后，由于 O3 与天然有机物（NOM）和效应有机物（EfOM）的直接反应，部分 O3 瞬时被消耗[定义为瞬时臭氧需求（IOD）]。在第二阶段，O3 缓慢衰减，遵循伪一级动力学（1,5,29）。这些 O3 衰减的典型模式在我们的结果中得到了清晰的展示（图 1a）；特别是，废水排放中的 O3 衰减速度快于自然水样，原因是有机负荷更高。同时，臭氧化过程中 pCBA 的分解在自然水中比在废水排放中更快（图 1b），这表明自然水中的•OH 浓度更高；有机物通常会清除•OH。Figure 1 图 1
Figure 1. (a) Time–concentration profiles of O3. Inset: time-dependent exposures of O3 during ozonation. (b) Time–concentration profiles of pCBA. Inset: time-dependent exposures of •OH during ozonation (Data shown for two selected water samples, no. 2 in Table S2 of the Supporting Information for natural water and no. 5 in Table S3 of the Supporting Information for wastewater effluent) ([O3]0 = 2.5 mg/L).
图 1. (a) O3 的时间-浓度曲线。插图：臭氧化过程中 O3 的时间依赖性暴露。(b) pCBA 的时间-浓度曲线。插图：臭氧化过程中•OH 的时间依赖性暴露（数据来自两个选定的水样，自然水的表 S2 中编号 2，废水排放的表 S3 中编号 5）([O3]0 = 2.5 mg/L)。
The time–concentration profiles of O3 and pCBA (Figure 1a,b) were converted into the exposures of O3 and •OH (see the inset of Figure 1a,b, respectively). For the calculation of 
, the data points for O3 decay were fitted to two-phase exponential association equations (ExpAssoc) using Origin software (Origin 2020, OriginLab) (refer to the dashed lines as shown in Figure 1a), and the resultant curves were integrated over time to obtain 
(see the inset of Figure 1a). The 
value as a function of time (see the inset of Figure 1b) was calculated from the decomposition kinetics of pCBA (Figure 1b) using the following equation (eq 6)
O3 和 pCBA 的时间-浓度曲线（图 1a,b）被转换为 O3 和•OH 的暴露量（分别见图 1a,b 的插图）。为了计算 ，O3 衰减的数据点使用 Origin 软件（Origin 2020, OriginLab）拟合为双相指数关联方程（ExpAssoc）（参见图 1a 中所示的虚线），并对得到的曲线进行时间积分以获得 （见图 1a 的插图）。 值作为时间的函数（见图 1b 的插图）是通过以下方程（方程 6）从 pCBA 的分解动力学（图 1b）计算得出的。
where k•OH,pCBA is the second-order rate constant for the reaction of pCBA with •OH (5 × 109 M–1 s–1). (30) The fit curves of pCBA decomposition to two-phase exponential association equations (refer to the dashed lines as shown in Figure 1b) were used for the 
calculation.
kOH,pCBA 是 pCBA 与•OH 反应的二级速率常数（5 × 10^9 M–1 s–1）。(30) pCBA 分解的拟合曲线采用了两相指数关联方程（参见图 1b 中所示的虚线）用于 计算。
, the data points for O3 decay were fitted to two-phase exponential association equations (ExpAssoc) using Origin software (Origin 2020, OriginLab) (refer to the dashed lines as shown in Figure 1a), and the resultant curves were integrated over time to obtain (see the inset of Figure 1a). The value as a function of time (see the inset of Figure 1b) was calculated from the decomposition kinetics of pCBA (Figure 1b) using the following equation (eq 6)O3 和 pCBA 的时间-浓度曲线（图 1a,b）被转换为 O3 和•OH 的暴露量（分别见图 1a,b 的插图）。为了计算
，O3 衰减的数据点使用 Origin 软件（Origin 2020, OriginLab）拟合为双相指数关联方程（ExpAssoc）（参见图 1a 中所示的虚线），并对得到的曲线进行时间积分以获得 （见图 1a 的插图）。 值作为时间的函数（见图 1b 的插图）是通过以下方程（方程 6）从 pCBA 的分解动力学（图 1b）计算得出的。(6)
calculation.kOH,pCBA 是 pCBA 与•OH 反应的二级速率常数（5 × 10^9 M–1 s–1）。(30) pCBA 分解的拟合曲线采用了两相指数关联方程（参见图 1b 中所示的虚线）用于
计算。Figures S2–S9 and Tables S2 and S3 of the Supporting Information summarize the time–concentration profiles of O3 and pCBA and the calculated oxidant exposures as functions of time during the ozonation of all 60 water samples, respectively.
支持信息中的图 S2–S9 和表 S2、S3 总结了在臭氧化所有 60 个水样品过程中 O3 和 pCBA 的时间-浓度曲线以及计算的氧化剂暴露随时间变化的情况。
支持信息中的图 S2–S9 和表 S2、S3 总结了在臭氧化所有 60 个水样品过程中 O3 和 pCBA 的时间-浓度曲线以及计算的氧化剂暴露随时间变化的情况。
Effects of DOC Concentration, pH, and Alkalinity on Oxidant Exposures
DOC 浓度、pH 值和碱度对氧化剂暴露的影响
The concentration of DOC, pH, and alkalinity were measured for the 60 water samples. In general, wastewater effluents showed higher DOC concentration (3–6 mg/L) and lower pH (7.0–7.5) values, compared to natural waters (DOC concentration of 1–5 mg/L and pH 7.5–8.5). Wastewater effluents showed a wider range of alkalinity (10–200 mg/L as CaCO3), compared to natural waters (25–75 mg/L as CaCO3). Tables S4 and S5 of the Supporting Information summarize the measured values of DOC concentration, pH, and alkalinity for all samples.
对 60 个水样的溶解有机碳（DOC）浓度、pH 值和碱度进行了测量。一般而言，废水排放物的 DOC 浓度较高（3–6 mg/L），pH 值较低（7.0–7.5），与自然水体（DOC 浓度为 1–5 mg/L，pH 值为 7.5–8.5）相比。废水排放物的碱度范围更广（10–200 mg/L，以 CaCO3 计），而自然水体的碱度为 25–75 mg/L（以 CaCO3 计）。支持信息中的表 S4 和 S5 总结了所有样本的 DOC 浓度、pH 值和碱度的测量值。
对 60 个水样的溶解有机碳（DOC）浓度、pH 值和碱度进行了测量。一般而言，废水排放物的 DOC 浓度较高（3–6 mg/L），pH 值较低（7.0–7.5），与自然水体（DOC 浓度为 1–5 mg/L，pH 值为 7.5–8.5）相比。废水排放物的碱度范围更广（10–200 mg/L，以 CaCO3 计），而自然水体的碱度为 25–75 mg/L（以 CaCO3 计）。支持信息中的表 S4 和 S5 总结了所有样本的 DOC 浓度、pH 值和碱度的测量值。
The limit values of oxidant exposures during the ozonation of water samples (i.e., the oxidant exposures until the complete depletion of O3) were plotted as functions of DOC concentration, pH, and alkalinity, respectively. Oxidant exposures tended to decrease with increasing DOC concentration because organic substances serve as major oxidant sinks (Figure 2). However, pH and alkalinity did not exhibit clear correlation with the oxidant exposures (refer to Figure S10 of the Supporting Information).
在水样臭氧化过程中，氧化剂暴露的限值（即，直到 O3 完全耗竭的氧化剂暴露）分别作为 DOC 浓度、pH 和碱度的函数进行绘制。随着 DOC 浓度的增加，氧化剂暴露趋向于减少，因为有机物质作为主要的氧化剂汇（见图 2）。然而，pH 和碱度与氧化剂暴露之间并未表现出明显的相关性（参见支持信息的图 S10）。
在水样臭氧化过程中，氧化剂暴露的限值（即，直到 O3 完全耗竭的氧化剂暴露）分别作为 DOC 浓度、pH 和碱度的函数进行绘制。随着 DOC 浓度的增加，氧化剂暴露趋向于减少，因为有机物质作为主要的氧化剂汇（见图 2）。然而，pH 和碱度与氧化剂暴露之间并未表现出明显的相关性（参见支持信息的图 S10）。
Figure 2 图 2
Figure 2. (a) O3 exposures and (b) •OH exposures (for the 60 water samples) as functions of DOC concentration ([O3]0 = 2.5 mg/L).
图 2. (a) O3 暴露和(b) •OH 暴露（针对 60 个水样）作为 DOC 浓度的函数（[O3]0 = 2.5 mg/L）。
Fluorescence Excitation–Emission Matrix
荧光激发-发射矩阵
The FEEM spectra were obtained for the 60 water samples. Figure 3 shows the FEEM contour plots of two selected water samples (no. 4 for natural water and no. 5 for wastewater effluent); Figures S11 and S12 of the Supporting Information show the respective plots for all 60 samples. It is noteworthy that the two samples show different FEEM patterns (compare Figure 3a,b) even though both samples have similar levels of DOC concentration (4.77 and 4.99 ppm, respectively), indicating that FEEM reflects different characteristics of organic substances.
FEEM 光谱是针对 60 个水样获得的。图 3 显示了两个选定水样的 FEEM 等高线图（编号 4 为自然水，编号 5 为废水排放）；支持信息中的图 S11 和 S12 显示了所有 60 个样本的相应图。值得注意的是，尽管两个样本的 DOC 浓度相似（分别为 4.77 和 4.99 ppm），但这两个样本显示出不同的 FEEM 模式（比较图 3a 和 b），这表明 FEEM 反映了有机物质的不同特征。
FEEM 光谱是针对 60 个水样获得的。图 3 显示了两个选定水样的 FEEM 等高线图（编号 4 为自然水，编号 5 为废水排放）；支持信息中的图 S11 和 S12 显示了所有 60 个样本的相应图。值得注意的是，尽管两个样本的 DOC 浓度相似（分别为 4.77 和 4.99 ppm），但这两个样本显示出不同的 FEEM 模式（比较图 3a 和 b），这表明 FEEM 反映了有机物质的不同特征。
Figure 3 图 3
Figure 3. FEEM contour plots of (a) natural water and (b) wastewater effluent (the data shown for two selected water samples, no. 4 in Table S2 of the Supporting Information for natural water and no. 5 in Table S3 of the Supporting Information for wastewater effluent).
图 3. 自然水和废水排放的 FEEM 等高线图 (a) 自然水和 (b) 废水排放（所示数据为两个选定水样的结果，支持信息中自然水的表 S2 中的第 4 号样本和废水排放的表 S3 中的第 5 号样本）。
According to the literature, (31−34) the FEEM spectra are divided into five regions that consecutively indicate aromatic proteins, such as tyrosine (aromatic protein I, Region 1); aromatic proteins, such as tryptophan or compounds attributed to biochemical oxygen demand (aromatic protein II, Region 2); fulvic acid-like substances (Region 3); soluble microbial byproduct-like compounds (Region 4); and humic acid-like substances (Region 5) (refer to Figure S13a of the Supporting Information for the division of the five regions). The FEEM spectra of natural waters show at least two characteristic peaks: a low intensity peak at excitation wavelength (EX) 330 nm/emission wavelength (EM) 430 nm in Region 5 and a high intensity peak at EX 240 nm/EM 440 nm in Region 3 (Figures 3a and S11 of the Supporting Information). This result is consistent with the observation that fulvic- and humic acid-like substances (representing Regions 3 and 5, respectively) account for up to 80% of NOM in natural water. (35) For some natural water samples (particularly those from the lower reaches of the Nakdong River, see plots 4–7 in Figure S11 of the Supporting Information), two additional peaks in the FEEM spectra were observed at EX 240 nm/EM 350 nm in Region 2, and at EX 280 nm/EM 320 nm in Region 4. These peaks may result from residual chemical contaminants and some protein-like soluble microbial products (SMPs) from nearby wastewater effluents (36,37) because the Nakdong River is highly impacted by municipal and industrial wastewater discharges from the Daegu metropolitan area and the Gumi National Industrial Complex.
根据文献（31−34），荧光增强光谱（FEEM）被分为五个区域，依次指示芳香族蛋白质，例如酪氨酸（芳香族蛋白质 I，区域 1）；芳香族蛋白质，例如色氨酸或归因于生化需氧量的化合物（芳香族蛋白质 II，区域 2）；类腐殖酸物质（区域 3）；可溶性微生物副产物类化合物（区域 4）；以及类腐殖酸物质（区域 5）（有关五个区域的划分，请参见支持信息的图 S13a）。自然水体的 FEEM 光谱显示至少有两个特征峰：在区域 5 的激发波长（EX）330 nm/发射波长（EM）430 nm 处有一个低强度峰，在区域 3 的 EX 240 nm/EM 440 nm 处有一个高强度峰（支持信息的图 3a 和 S11）。这一结果与观察到的类腐殖酸和类腐殖酸物质（分别代表区域 3 和 5）占自然水体中有机物质（NOM）高达 80% 的现象一致。 对于一些自然水样（特别是来自 Nakdong 河下游的样本，参见支持信息中图 S11 的图 4-7），在区域 2 的 EX 240 nm/EM 350 nm 和区域 4 的 EX 280 nm/EM 320 nm 处观察到了 FEEM 光谱中的两个额外峰。这些峰可能是由于附近废水排放中的残留化学污染物和某些类蛋白质的可溶性微生物产品（SMPs）所致，因为 Nakdong 河受到大邱都市区和龟尾国家工业园区的市政和工业废水排放的严重影响。
根据文献（31−34），荧光增强光谱（FEEM）被分为五个区域，依次指示芳香族蛋白质，例如酪氨酸（芳香族蛋白质 I，区域 1）；芳香族蛋白质，例如色氨酸或归因于生化需氧量的化合物（芳香族蛋白质 II，区域 2）；类腐殖酸物质（区域 3）；可溶性微生物副产物类化合物（区域 4）；以及类腐殖酸物质（区域 5）（有关五个区域的划分，请参见支持信息的图 S13a）。自然水体的 FEEM 光谱显示至少有两个特征峰：在区域 5 的激发波长（EX）330 nm/发射波长（EM）430 nm 处有一个低强度峰，在区域 3 的 EX 240 nm/EM 440 nm 处有一个高强度峰（支持信息的图 3a 和 S11）。这一结果与观察到的类腐殖酸和类腐殖酸物质（分别代表区域 3 和 5）占自然水体中有机物质（NOM）高达 80% 的现象一致。 对于一些自然水样（特别是来自 Nakdong 河下游的样本，参见支持信息中图 S11 的图 4-7），在区域 2 的 EX 240 nm/EM 350 nm 和区域 4 的 EX 280 nm/EM 320 nm 处观察到了 FEEM 光谱中的两个额外峰。这些峰可能是由于附近废水排放中的残留化学污染物和某些类蛋白质的可溶性微生物产品（SMPs）所致，因为 Nakdong 河受到大邱都市区和龟尾国家工业园区的市政和工业废水排放的严重影响。
The FEEM spectra of wastewater effluents generally show two distinctive peaks at EX 330 nm/EM 430 nm in Region 5 and at EX 240 nm/EM 440 nm in Region 3 (Figures 3b, and S12 of the Supporting Information). In some cases, the aforementioned peak in Region 5 shifted to EX 300 nm and EM 380 nm in Region 4 (see plots 10, 17, and 25 in Figure S12 of the Supporting Information). The fluorescence in regions 3 and 5 is attributed to fulvic- and humic acid-like SMPs derived from the biological treatment processes in wastewater treatment plants; a similar EEM pattern was observed for laboratory-generated SMP. (37)
废水排放的 FEEM 光谱通常在区域 5 的 EX 330 nm/EM 430 nm 和区域 3 的 EX 240 nm/EM 440 nm 处显示出两个明显的峰（见支持信息的图 3b 和 S12）。在某些情况下，区域 5 中上述峰值在区域 4 中移至 EX 300 nm 和 EM 380 nm（见支持信息的图 S12 中的图 10、17 和 25）。区域 3 和 5 中的荧光归因于来自废水处理厂生物处理过程的类腐殖酸和类腐殖质的 SMP；实验室生成的 SMP 也观察到了类似的 EEM 模式。(37)
废水排放的 FEEM 光谱通常在区域 5 的 EX 330 nm/EM 430 nm 和区域 3 的 EX 240 nm/EM 440 nm 处显示出两个明显的峰（见支持信息的图 3b 和 S12）。在某些情况下，区域 5 中上述峰值在区域 4 中移至 EX 300 nm 和 EM 380 nm（见支持信息的图 S12 中的图 10、17 和 25）。区域 3 和 5 中的荧光归因于来自废水处理厂生物处理过程的类腐殖酸和类腐殖质的 SMP；实验室生成的 SMP 也观察到了类似的 EEM 模式。(37)
The FEEM data were converted into the P(i,n) values (input variables) using FRI for two RF models (i.e., FEEM-LowRes and FEEM-HighRes). For the FEEM-LowRes model, five P(i,n) values were calculated for the five regions in the FEEM contour plot (see Figure S13a of the Supporting Information); Tables S6 and S7 of the Supporting Information provide the values of all 60 water samples. For the FEEM-HighRes model, a total of 83 P(i,n) values were obtained for each FEEM plot divided into smaller regions with the grid dimension of 20 nm × 20 nm (see Figure S13b of the Supporting Information). The P(i,n) values for the FEEM-HighRes model were calculated for all 60 water samples (data not shown). For the FEEM-FullRes model, all 9724 data points of the FEEM plot were used as input variables.
FEEM 数据通过 FRI 转换为 P(,n)值（输入变量），用于两个 RF 模型（即 FEEM-LowRes 和 FEEM-HighRes）。对于 FEEM-LowRes 模型，计算了 FEEM 等高线图中五个区域的五个 P(,n)值（见支持信息的图 S13a）；支持信息的表 S6 和 S7 提供了所有 60 个水样的值。对于 FEEM-HighRes 模型，总共获得了 83 个 P(,n)值，针对每个 FEEM 图将其划分为更小的区域，网格尺寸为 20 nm × 20 nm（见支持信息的图 S13b）。FEEM-HighRes 模型的 P(,n)值是为所有 60 个水样计算的（数据未显示）。对于 FEEM-FullRes 模型，使用了 FEEM 图的所有 9724 个数据点作为输入变量。
FEEM 数据通过 FRI 转换为 P(,n)值（输入变量），用于两个 RF 模型（即 FEEM-LowRes 和 FEEM-HighRes）。对于 FEEM-LowRes 模型，计算了 FEEM 等高线图中五个区域的五个 P(,n)值（见支持信息的图 S13a）；支持信息的表 S6 和 S7 提供了所有 60 个水样的值。对于 FEEM-HighRes 模型，总共获得了 83 个 P(,n)值，针对每个 FEEM 图将其划分为更小的区域，网格尺寸为 20 nm × 20 nm（见支持信息的图 S13b）。FEEM-HighRes 模型的 P(,n)值是为所有 60 个水样计算的（数据未显示）。对于 FEEM-FullRes 模型，使用了 FEEM 图的所有 9724 个数据点作为输入变量。
Prediction of Oxidant Exposures by RF Models
通过 RF 模型预测氧化剂暴露
The measured oxidant exposures were compared to those predicted by the four RF models, and the R2 and RMSE values were calculated (refer to Figure 4 for 
and Figure 5 for 
). For the 
prediction (Figure 4), the FEEM-FullRes model showed the highest accuracy among the four RF models with the highest R2 (0.874 and 0.798 for training and validation, respectively) and the lowest RMSE values (1.65 × 10–3 and 1.61 × 10–3 M s for training and validation, respectively). The prediction accuracy (assessed by R2 and RMSE) decreased in the order: FEEM-HighRes > FEEM-LowRes > FEEM-Free. For the 
prediction (Figure 5), the FEEM-FullRes and FEEM-HighRes models showed the highest prediction accuracy, followed by FEEM-LowRes and FEEM-Free. The FEEM-FullRes model showed the highest R2 and the lowest RMSE values for training (0.931 and 4.98 × 10–11 M s, respectively). However, the FEEM-FullRes and FEEM-HighRes models showed similar R2 (0.766 and 0.772) and RMSE (5.53 × 10–11 and 5.51 × 10–11 M s) values for validation, respectively. The comparison of the four RF models confirms that the incorporation of FEEM data into the model leads to more accurate prediction of oxidant exposures, implying that FEEM can reflect the different characteristics of NOM and EfOM regarding their reactions with oxidants. Additionally, a direct comparison with simpler ridge regression models was also conducted (Figures S14 and S15 of the Supporting Information); detailed discussions were included in Text S2 of the Supporting Information.
测量的氧化剂暴露量与四个随机森林模型预测的结果进行了比较，并计算了 R²和 RMSE 值（参见图 4 中的 和图 5 中的 ）。对于 预测（图 4），FEEM-FullRes 模型在四个随机森林模型中显示出最高的准确性，R²值最高（训练和验证分别为 0.874 和 0.798），RMSE 值最低（训练和验证分别为 1.65 × 10⁻³和 1.61 × 10⁻³ M s）。预测准确性（通过 R²和 RMSE 评估）按顺序递减：FEEM-HighRes > FEEM-LowRes > FEEM-Free。对于 预测（图 5），FEEM-FullRes 和 FEEM-HighRes 模型显示出最高的预测准确性，其次是 FEEM-LowRes 和 FEEM-Free。FEEM-FullRes 模型在训练中显示出最高的 R²和最低的 RMSE 值（分别为 0.931 和 4.98 × 10⁻¹¹ M s）。然而，FEEM-FullRes 和 FEEM-HighRes 模型在验证中显示出相似的 R²（分别为 0.766 和 0.772）和 RMSE（分别为 5.53 × 10⁻¹¹和 5.51 × 10⁻¹¹ M s）值。 四个 RF 模型的比较确认，将 FEEM 数据纳入模型可以更准确地预测氧化剂暴露，这意味着 FEEM 能够反映 NOM 和 EfOM 在与氧化剂反应时的不同特征。此外，还进行了与更简单的脊回归模型的直接比较（见支持信息的图 S14 和 S15）；详细讨论已包含在支持信息的文本 S2 中。
and Figure 5 for ). For the prediction (Figure 4), the FEEM-FullRes model showed the highest accuracy among the four RF models with the highest R2 (0.874 and 0.798 for training and validation, respectively) and the lowest RMSE values (1.65 × 10–3 and 1.61 × 10–3 M s for training and validation, respectively). The prediction accuracy (assessed by R2 and RMSE) decreased in the order: FEEM-HighRes > FEEM-LowRes > FEEM-Free. For the prediction (Figure 5), the FEEM-FullRes and FEEM-HighRes models showed the highest prediction accuracy, followed by FEEM-LowRes and FEEM-Free. The FEEM-FullRes model showed the highest R2 and the lowest RMSE values for training (0.931 and 4.98 × 10–11 M s, respectively). However, the FEEM-FullRes and FEEM-HighRes models showed similar R2 (0.766 and 0.772) and RMSE (5.53 × 10–11 and 5.51 × 10–11 M s) values for validation, respectively. The comparison of the four RF models confirms that the incorporation of FEEM data into the model leads to more accurate prediction of oxidant exposures, implying that FEEM can reflect the different characteristics of NOM and EfOM regarding their reactions with oxidants. Additionally, a direct comparison with simpler ridge regression models was also conducted (Figures S14 and S15 of the Supporting Information); detailed discussions were included in Text S2 of the Supporting Information.测量的氧化剂暴露量与四个随机森林模型预测的结果进行了比较，并计算了 R²和 RMSE 值（参见图 4 中的
和图 5 中的 ）。对于 预测（图 4），FEEM-FullRes 模型在四个随机森林模型中显示出最高的准确性，R²值最高（训练和验证分别为 0.874 和 0.798），RMSE 值最低（训练和验证分别为 1.65 × 10⁻³和 1.61 × 10⁻³ M s）。预测准确性（通过 R²和 RMSE 评估）按顺序递减：FEEM-HighRes > FEEM-LowRes > FEEM-Free。对于 预测（图 5），FEEM-FullRes 和 FEEM-HighRes 模型显示出最高的预测准确性，其次是 FEEM-LowRes 和 FEEM-Free。FEEM-FullRes 模型在训练中显示出最高的 R²和最低的 RMSE 值（分别为 0.931 和 4.98 × 10⁻¹¹ M s）。然而，FEEM-FullRes 和 FEEM-HighRes 模型在验证中显示出相似的 R²（分别为 0.766 和 0.772）和 RMSE（分别为 5.53 × 10⁻¹¹和 5.51 × 10⁻¹¹ M s）值。 四个 RF 模型的比较确认，将 FEEM 数据纳入模型可以更准确地预测氧化剂暴露，这意味着 FEEM 能够反映 NOM 和 EfOM 在与氧化剂反应时的不同特征。此外，还进行了与更简单的脊回归模型的直接比较（见支持信息的图 S14 和 S15）；详细讨论已包含在支持信息的文本 S2 中。Figure 4 图 4
Figure 4. Plots of measured vs predicted O3 exposures for four RF models, (a) FEEM-Free, (b) FEEM-LowRes, (c) FEEM-HighRes, and (d) FEEM-FullRes ([O3]0 = 2.5 mg/L).
图 4. 四个 RF 模型的测量与预测 O3 暴露的图表，(a) FEEM-Free，(b) FEEM-LowRes，(c) FEEM-HighRes，(d) FEEM-FullRes ([O3]0 = 2.5 mg/L)。
Figure 5 图 5
Figure 5. Plots of measured vs predicted •OH exposures for four RF models, (a) FEEM-Free, (b) FEEM-LowRes, (c) FEEM-HighRes, and (d) FEEM-FullRes ([O3]0 = 2.5 mg/L).
图 5. 四个 RF 模型的测量与预测•OH 暴露的图表，(a) FEEM-Free，(b) FEEM-LowRes，(c) FEEM-HighRes，(d) FEEM-FullRes ([O3]0 = 2.5 mg/L)。
VI of the FEEM FRI data was assessed for 
and 
predictions of the FEEM-HighRes model, which provides insights into which FEEM regions have a greater impact on oxidant exposures; the contour plots featuring VI were presented for 
and 
predictions (refer to Figure S16 of the Supporting Information). For the 
prediction (refer to Figure S16a of the Supporting Information), the most important spots were observed in the upper edge of Region 5 (humic acid-like chromophores with relatively high molecular weight), and in the lower left side of Region 4 (protein-like SMPs). The organic substances corresponding to these spots appear to be sensitive to the reaction with O3; indeed, previous studies showed that the FEEM intensities in these areas were preferentially removed by ozonation. (15,16) For the 
prediction (refer to Figure S16b of the Supporting Information), a noticeably important spot was found in the upper edge of Region 5 (at a similar position to the one for 
). Considering the nonselective reactivity of •OH relative to O3, the organic substances for the common spot in Region 5 may be associated with the generation of •OH via the reactions with O3, rather than the consumption of •OH. Additional VI assessments (including non-FEEM variables) were conducted for FEEM-Free and FEEM-HighRes models (refer to Text S3 and Figure S17 of the Supporting Information for details).
FEEM FRI 数据的 VI 被评估用于 和 的 FEEM-HighRes 模型预测，这提供了对哪些 FEEM 区域对氧化剂暴露影响更大的洞察；包含 VI 的等高线图展示了 和 的预测（请参见支持信息的图 S16）。对于 预测（请参见支持信息的图 S16a），在区域 5 的上边缘（相对高分子量的腐殖酸类色素）和区域 4 的左下侧（蛋白质类 SMPs）观察到了最重要的点。这些点对应的有机物质似乎对与 O3 的反应敏感；实际上，之前的研究表明，这些区域的 FEEM 强度在臭氧化过程中被优先去除。(15,16) 对于 预测（请参见支持信息的图 S16b），在区域 5 的上边缘发现了一个显著重要的点（与 的位置相似）。 考虑到•OH 相对于 O3 的非选择性反应性，第 5 区的常见点的有机物质可能与通过与 O3 反应生成•OH 有关，而不是•OH 的消耗。对 FEEM-Free 和 FEEM-HighRes 模型进行了额外的 VI 评估（包括非 FEEM 变量）（详细信息请参见支持信息中的文本 S3 和图 S17）。
and predictions of the FEEM-HighRes model, which provides insights into which FEEM regions have a greater impact on oxidant exposures; the contour plots featuring VI were presented for and predictions (refer to Figure S16 of the Supporting Information). For the prediction (refer to Figure S16a of the Supporting Information), the most important spots were observed in the upper edge of Region 5 (humic acid-like chromophores with relatively high molecular weight), and in the lower left side of Region 4 (protein-like SMPs). The organic substances corresponding to these spots appear to be sensitive to the reaction with O3; indeed, previous studies showed that the FEEM intensities in these areas were preferentially removed by ozonation. (15,16) For the prediction (refer to Figure S16b of the Supporting Information), a noticeably important spot was found in the upper edge of Region 5 (at a similar position to the one for ). Considering the nonselective reactivity of •OH relative to O3, the organic substances for the common spot in Region 5 may be associated with the generation of •OH via the reactions with O3, rather than the consumption of •OH. Additional VI assessments (including non-FEEM variables) were conducted for FEEM-Free and FEEM-HighRes models (refer to Text S3 and Figure S17 of the Supporting Information for details).FEEM FRI 数据的 VI 被评估用于
和 的 FEEM-HighRes 模型预测，这提供了对哪些 FEEM 区域对氧化剂暴露影响更大的洞察；包含 VI 的等高线图展示了 和 的预测（请参见支持信息的图 S16）。对于 预测（请参见支持信息的图 S16a），在区域 5 的上边缘（相对高分子量的腐殖酸类色素）和区域 4 的左下侧（蛋白质类 SMPs）观察到了最重要的点。这些点对应的有机物质似乎对与 O3 的反应敏感；实际上，之前的研究表明，这些区域的 FEEM 强度在臭氧化过程中被优先去除。(15,16) 对于 预测（请参见支持信息的图 S16b），在区域 5 的上边缘发现了一个显著重要的点（与 的位置相似）。 考虑到•OH 相对于 O3 的非选择性反应性，第 5 区的常见点的有机物质可能与通过与 O3 反应生成•OH 有关，而不是•OH 的消耗。对 FEEM-Free 和 FEEM-HighRes 模型进行了额外的 VI 评估（包括非 FEEM 变量）（详细信息请参见支持信息中的文本 S3 和图 S17）。The RSS and AIC values were calculated for the prediction of 
and 
by four RF models (refer to Table S8 of the Supporting Information). The FEEM-FullRes model was found to exhibit extremely high AIC values (19,299.6 and 18,887.2 for 
and 
, respectively) compared to other models. Such high AIC values mainly result from the immense number of input parameters (9724 FEEM data points for the FEEM-FullRes model), which can cause overfitting (high accuracy for training, but relatively low accuracy for validation). Indeed, the gap between R2 values for training and validation was greater for the FEEM-FullRes model than for the FEEM-HighRes, especially for 
(Figures 4 and 5).
RSS 和 AIC 值是通过四个 RF 模型计算得出的，用于预测 和 （参见支持信息的表 S8）。与其他模型相比，FEEM-FullRes 模型的 AIC 值极高（ 和 分别为 19,299.6 和 18,887.2）。如此高的 AIC 值主要是由于输入参数的数量庞大（FEEM-FullRes 模型有 9724 个 FEEM 数据点），这可能导致过拟合（训练时准确度高，但验证时相对准确度低）。实际上，FEEM-FullRes 模型的训练和验证的 R2 值之间的差距大于 FEEM-HighRes 模型，尤其是在 时（图 4 和图 5）。
and by four RF models (refer to Table S8 of the Supporting Information). The FEEM-FullRes model was found to exhibit extremely high AIC values (19,299.6 and 18,887.2 for and , respectively) compared to other models. Such high AIC values mainly result from the immense number of input parameters (9724 FEEM data points for the FEEM-FullRes model), which can cause overfitting (high accuracy for training, but relatively low accuracy for validation). Indeed, the gap between R2 values for training and validation was greater for the FEEM-FullRes model than for the FEEM-HighRes, especially for (Figures 4 and 5).RSS 和 AIC 值是通过四个 RF 模型计算得出的，用于预测
和 （参见支持信息的表 S8）。与其他模型相比，FEEM-FullRes 模型的 AIC 值极高（ 和 分别为 19,299.6 和 18,887.2）。如此高的 AIC 值主要是由于输入参数的数量庞大（FEEM-FullRes 模型有 9724 个 FEEM 数据点），这可能导致过拟合（训练时准确度高，但验证时相对准确度低）。实际上，FEEM-FullRes 模型的训练和验证的 R2 值之间的差距大于 FEEM-HighRes 模型，尤其是在 时（图 4 和图 5）。 Prediction of MP Abatement
MP 减排预测
Four MPs (i.e., atrazine, caffeine, carbamazepine, and ibuprofen) were selected, considering the rate constants for their reactions with oxidants. Table S9 of the Supporting Information presents the kO3 and k•OH values of selected MPs. Atrazine and ibuprofen have low reactivity with O3, and thus their abatement during ozonation proceeds mainly by •OH. However, ibuprofen is anticipated to be more degraded than atrazine because it has approximately threefold higher k•OH. Caffeine has moderate reactivity with O3; therefore, the contributions of both O3 and •OH could be important. Carbamazepine is highly reactive with O3, and the direct reaction with O3 could be the dominant route for its abatement during ozonation.
选择了四种微污染物（即，阿特拉津、咖啡因、卡马西平和布洛芬），考虑到它们与氧化剂反应的速率常数。支持信息的表 S9 展示了所选微污染物的 kO3 和 kOH 值。阿特拉津和布洛芬与 O3 的反应性较低，因此它们在臭氧化过程中的去除主要依赖于•OH。然而，预计布洛芬的降解程度将高于阿特拉津，因为它的 kOH 值大约高出三倍。咖啡因与 O3 的反应性适中，因此 O3 和•OH 的贡献可能都很重要。卡马西平与 O3 的反应性很高，直接与 O3 的反应可能是其在臭氧化过程中去除的主要途径。
选择了四种微污染物（即，阿特拉津、咖啡因、卡马西平和布洛芬），考虑到它们与氧化剂反应的速率常数。支持信息的表 S9 展示了所选微污染物的 kO3 和 kOH 值。阿特拉津和布洛芬与 O3 的反应性较低，因此它们在臭氧化过程中的去除主要依赖于•OH。然而，预计布洛芬的降解程度将高于阿特拉津，因为它的 kOH 值大约高出三倍。咖啡因与 O3 的反应性适中，因此 O3 和•OH 的贡献可能都很重要。卡马西平与 O3 的反应性很高，直接与 O3 的反应可能是其在臭氧化过程中去除的主要途径。
The selected MPs were spiked in 15 natural water and wastewater effluent samples (i.e., total 30 samples). The percent removals of MPs in the ozonation process after the depletion of O3 were measured (refer to Figure S18 of the Supporting Information). For natural water samples (refer to Figure S18a of the Supporting Information), atrazine was degraded by 30–90% and ibuprofen by 60–99%, and caffeine was completely degraded except for one sample. For wastewater effluent samples (refer to Figure S18b of the Supporting Information), atrazine was degraded by 20–60%, ibuprofen by 40–90%, and caffeine by 60–99%. For all cases, carbamazepine was almost completely degraded.
所选的微量污染物（MPs）在 15 个自然水和废水排放样本中进行了检测（即总共 30 个样本）。在臭氧消耗后，测量了臭氧化过程中的 MPs 去除百分比（参见支持信息的图 S18）。对于自然水样本（参见支持信息的图 S18a），阿特拉津的降解率为 30-90%，布洛芬为 60-99%，咖啡因几乎完全降解，只有一个样本例外。对于废水排放样本（参见支持信息的图 S18b），阿特拉津的降解率为 20-60%，布洛芬为 40-90%，咖啡因为 60-99%。在所有情况下，卡马西平几乎完全降解。
所选的微量污染物（MPs）在 15 个自然水和废水排放样本中进行了检测（即总共 30 个样本）。在臭氧消耗后，测量了臭氧化过程中的 MPs 去除百分比（参见支持信息的图 S18）。对于自然水样本（参见支持信息的图 S18a），阿特拉津的降解率为 30-90%，布洛芬为 60-99%，咖啡因几乎完全降解，只有一个样本例外。对于废水排放样本（参见支持信息的图 S18b），阿特拉津的降解率为 20-60%，布洛芬为 40-90%，咖啡因为 60-99%。在所有情况下，卡马西平几乎完全降解。
Meanwhile, the abatement of the selected MPs was calculated from eq 2 using 
and 
values predicted by the four RF models (calculated for the water samples in which the MP abatement experiments were conducted), and the outcomes were compared to the values measured by experiments (Figure 6). Overall, the model predictions fit reasonably with the experimental measurements. The FEEM-FullRes model showed the highest R2 (0.904) and the lowest RMSE (6.60%) values, followed by FEEM-HighRes > FEEM-LowRes = FEEM-Free. The difference in prediction accuracy among the models was less significant, compared to the cases of 
and 
prediction (Figures 4 and 5), indicating that the predicted values of 
and 
somewhat offset each other.
与此同时，所选微塑料的减少量是根据方程 2 使用四个随机森林模型预测的 和 值计算得出的（该计算是针对进行微塑料减少实验的水样），并将结果与实验测得的值进行了比较（图 6）。总体而言，模型预测与实验测量结果相当吻合。FEEM-FullRes 模型显示出最高的 R²值（0.904）和最低的均方根误差（RMSE）值（6.60%），其次是 FEEM-HighRes > FEEM-LowRes = FEEM-Free。与 和 预测的情况相比，各模型之间的预测准确性差异不太显著（图 4 和图 5），这表明 和 的预测值在某种程度上相互抵消。
and values predicted by the four RF models (calculated for the water samples in which the MP abatement experiments were conducted), and the outcomes were compared to the values measured by experiments (Figure 6). Overall, the model predictions fit reasonably with the experimental measurements. The FEEM-FullRes model showed the highest R2 (0.904) and the lowest RMSE (6.60%) values, followed by FEEM-HighRes > FEEM-LowRes = FEEM-Free. The difference in prediction accuracy among the models was less significant, compared to the cases of and prediction (Figures 4 and 5), indicating that the predicted values of and somewhat offset each other.与此同时，所选微塑料的减少量是根据方程 2 使用四个随机森林模型预测的
和 值计算得出的（该计算是针对进行微塑料减少实验的水样），并将结果与实验测得的值进行了比较（图 6）。总体而言，模型预测与实验测量结果相当吻合。FEEM-FullRes 模型显示出最高的 R²值（0.904）和最低的均方根误差（RMSE）值（6.60%），其次是 FEEM-HighRes > FEEM-LowRes = FEEM-Free。与 和 预测的情况相比，各模型之间的预测准确性差异不太显著（图 4 和图 5），这表明 和 的预测值在某种程度上相互抵消。Figure 6 图 6
Figure 6. Plots of the measured vs predicted MP abatement for four RF models, (a) FEEM-Free, (b) FEEM-LowRes, (c) FFEM-HighRes, and (d) FEEM-FullRes (Raw data for measured values are presented in Figure S18 of the Supporting Information) ([MP]0 = 0.1 μM, [O3]0 = 2.5 mg/L).
图 6. 四个 RF 模型的测量与预测 MP 减排的图表，(a) FEEM-Free，(b) FEEM-LowRes，(c) FFEM-HighRes，(d) FEEM-FullRes（测量值的原始数据见支持信息的图 S18）([MP]0 = 0.1 μM, [O3]0 = 2.5 mg/L)。
Implications and Research Imperatives
影响及研究迫切性
In this study, comprehensive models for the prediction of MP abatement by ozonation were developed. It was demonstrated that using a machine learning technique with FEEM data (input variables to characterize organic substances) could successfully be used to predict oxidant exposures, and in turn, MP abatement, during the ozonation of different water samples. The use of higher-resolution FEEM data generally offers more accurate prediction. However, the increased resolution requires an increased number of input variables for the prediction model, causing longer computation time and possible overfitting. We believe that FEEM-based models can be applied in predicting the abatement of target MPs in drinking water and wastewater ozonation processes and in determining the optimal O3 dose for intended purposes. For the practical application, the prediction models need to be more advanced; in fact, the enhancement in prediction accuracy by using the FEEM data was not significant for the prediction of MP abatement. In order to improve the reliability and comprehensibility of the prediction models, the machine learning models need to be trained with a greater number of water samples, with a wider spectrum of organic characteristics. In addition, the method of processing the FEEM data into the input variables needs to be further advanced (e.g., optimizing the division of the FEEM contour plot for FRI and taking account of time-dependent FEEM data during ozonation).
在本研究中，开发了用于预测臭氧化过程中微塑料（MP）去除的综合模型。研究表明，使用机器学习技术结合荧光增强光谱（FEEM）数据（用于表征有机物的输入变量）可以成功预测氧化剂暴露，从而预测不同水样的微塑料去除。使用更高分辨率的 FEEM 数据通常能提供更准确的预测。然而，分辨率的提高需要更多的输入变量，这导致计算时间延长和可能的过拟合。我们认为基于 FEEM 的模型可以应用于预测饮用水和废水臭氧化过程中的目标微塑料去除，并确定用于特定目的的最佳臭氧剂量。对于实际应用，预测模型需要更先进；实际上，使用 FEEM 数据提高预测准确性的效果在微塑料去除的预测中并不显著。 为了提高预测模型的可靠性和可理解性，机器学习模型需要用更多的水样本进行训练，并且这些样本应具有更广泛的有机特征。此外，处理 FEEM 数据以生成输入变量的方法需要进一步改进（例如，优化 FRI 的 FEEM 等高线图的划分，并考虑臭氧化过程中时间依赖的 FEEM 数据）。
在本研究中，开发了用于预测臭氧化过程中微塑料（MP）去除的综合模型。研究表明，使用机器学习技术结合荧光增强光谱（FEEM）数据（用于表征有机物的输入变量）可以成功预测氧化剂暴露，从而预测不同水样的微塑料去除。使用更高分辨率的 FEEM 数据通常能提供更准确的预测。然而，分辨率的提高需要更多的输入变量，这导致计算时间延长和可能的过拟合。我们认为基于 FEEM 的模型可以应用于预测饮用水和废水臭氧化过程中的目标微塑料去除，并确定用于特定目的的最佳臭氧剂量。对于实际应用，预测模型需要更先进；实际上，使用 FEEM 数据提高预测准确性的效果在微塑料去除的预测中并不显著。 为了提高预测模型的可靠性和可理解性，机器学习模型需要用更多的水样本进行训练，并且这些样本应具有更广泛的有机特征。此外，处理 FEEM 数据以生成输入变量的方法需要进一步改进（例如，优化 FRI 的 FEEM 等高线图的划分，并考虑臭氧化过程中时间依赖的 FEEM 数据）。
Supporting Information 支持信息
Click to copy section link
点击复制部分链接Section link copied!
点击复制部分链接Section link copied!
The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.est.0c05836.
支持信息可免费获取，网址为 https://pubs.acs.org/doi/10.1021/acs.est.0c05836。
Additional details, hyperparameters, and development procedures of the RF model, comparison of RF and ridge regression models, variable importance for FEEM-Free and FEEM-HighRes models, HPLC operating conditions, oxidant exposure values, water characteristic values, percent fluorescence response values, regression performance and AIC values, second-order rate constant values, river water sampling sites, time-dependent concentration profiles of ozone, time-dependent concentration profiles of pCBA, time-dependent exposure profiles of ozone, time-dependent exposure profiles of •OH, oxidant exposures as functions of pH and alkalinity, FEEM contour plots, division of FRI regions, comparison of oxidant exposures for RF and ridge regression models, variable importance for the prediction of oxidant exposures, and MP abatement experiment results (PDF)
RF 模型的附加细节、超参数和开发程序，RF 与岭回归模型的比较，FEEM-Free 和 FEEM-HighRes 模型的变量重要性，HPLC 操作条件，氧化剂暴露值，水体特征值，荧光响应百分比值，回归性能和 AIC 值，二级反应速率常数值，河水采样点，臭氧的时间依赖浓度曲线，pCBA 的时间依赖浓度曲线，臭氧的时间依赖暴露曲线，•OH 的时间依赖暴露曲线，氧化剂暴露与 pH 和碱度的关系，FEEM 等高线图，FRI 区域的划分，RF 与岭回归模型的氧化剂暴露比较，氧化剂暴露预测的变量重要性，以及 MP 减排实验结果（PDF）
Terms & Conditions 条款与条件
Most electronic Supporting Information files are available without a subscription to ACS Web Editions. Such files may be downloaded by article for research use (if there is a public use license linked to the relevant article, that license may permit other uses). Permission may be obtained from ACS for other uses through requests via the RightsLink permission system: http://pubs.acs.org/page/copyright/permissions.html.
Author Information
Click to copy section linkSection link copied!
- Changha Lee - School of Chemical and Biological Engineering, Institute of Chemical Process (ICP), Seoul National University, 1 Gwanak-ro, Gwanak-gu, Seoul 08826, Republic of Korea;
http://orcid.org/0000-0002-0404-9405;
- Seok Won Hong - Water Cycle Research Center, Korea Institute of Science and Technology (KIST), 5 Hwarang-ro 14-gil, Seongbuk-gu, Seoul 02792, Republic of Korea;http://orcid.org/0000-0003-0961-3842
Acknowledgments
Click to copy section linkSection link copied!
This work was supported by the Korea Environment Industry & Technology Institute (KEITI) through the Developing Innovative Drinking Water and Wastewater Technologies Project, funded by the Korea Ministry of Environment (MOE) (2019002710003), as well as Open Innovation R&D (Together PRO) Project supported by K-Water.
References
Click to copy section linkSection link copied!
This article references 37 other publications.
- 1Hoigné, J. Chemistry of Aqueous Ozone and Transformation of Pollutants by Ozonation and Advanced Oxidation Processes. In Quality and Treatment of Drinking Water II, Part of The Handbook of Environmental Chemistry (Part C: Water Pollution); Hrubec, J., Ed.; Springer: Berlin, 1998; Vol 5, pp 83– 141.Google ScholarThere is no corresponding record for this reference.
- 2Lee, Y.; von Gunten, U. Advances in predicting organic contaminant abatement during ozonation of municipal wastewater effluent: reaction kinetics, transformation products, and changes of biological effects. Environ. Sci.: Water Res. Technol. 2016, 2, 421– 442, DOI: 10.1039/c6ew00025hGoogle Scholar2https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28Xjs1Crtrc%253D&md5=6a736f485739952184f79df8a3e7b0ccAdvances in predicting organic contaminant abatement during ozonation of municipal wastewater effluent: reaction kinetics, transformation products, and changes of biological effectsLee, Yunho; von Gunten, UrsEnvironmental Science: Water Research & Technology (2016), 2 (3), 421-442CODEN: ESWRAR; ISSN:2053-1419. (Royal Society of Chemistry)Ozonation of municipal wastewater effluent has been considered in recent years as an enhanced wastewater treatment technol. to abate trace org. contaminants (micropollutants). The efficiency of ozonation for micropollutant abatement depends on (1) the reactivity of ozone and OH radical (̇ OH) with the target micropollutant, (2) the dosage of ozone and the stability of ozone anḋ OH in a given water matrix, (3) the removal of undesirable effects (e.g., biol. activities) of a micropollutant after structural transformation, and (4) the biodegradability of transformation products in biol. post-treatment. In this article, recent advances in predicting org. micropollutant abatement during ozonation of municipal wastewater effluents are reviewed with a focus on (i) principle-based approaches for describing and modeling the reaction kinetics of ozone anḋ OH, (ii) transformation products and pathways, (iii) changes of biol. activities, and (iv) biodegrdn. of transformation products in biol. post-treatment. Using the chem. kinetics based on ozone anḋ OH rate consts. (i.e., compd.-specific information) and exposures (i.e., water matrix-specific information), a generalized prediction of the abatement efficiency of various micropollutants in varying water quality appears to be possible. QSAR-type correlations based on Hammett coeffs. or quantum chem. energy calcns. or (semi)empirical models have been developed for predicting the ozone anḋ OH rate consts. and exposures, resp. Models based on the ozone anḋ OH reaction rules can be used to predict the transformation products of micropollutants by ozone anḋ OH. Reaction rule-based models in combination with the chem. kinetics information will enable the prediction of transformation product evolution during ozonation. The biol. activities of transformation products have been assessed by an effect-driven approach using in vitro bioassays. Biol. activities with specific modes of action (e.g., receptor-binding activities) were found to be quite efficiently removed, upon slight structural modifications by ozone oṙ OH. The formation of new biol. activities has also been obsd., which warrants identification of the responsible toxicophore(s) and quant. exposure-based risk assessment. Finally, there is only limited exptl. information on the biodegradability of transformation products; however, biodegradability probability models can be used to make first ests. In future research, the discussed principle-based approaches can be more actively applied to det. and predict not only the abatement levels of the parent micropollutants but also the formation of transformation products and the consequent changes of biol. activities and biodegradability, which dets. the overall treatment efficiency.
- 3Glaze, W. H. Drinking-water treatment with ozone. Environ. Sci. Technol. 1987, 21, 224– 230, DOI: 10.1021/es00157a001Google Scholar3https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADyaL2sXht1GisLc%253D&md5=95b59f88ef96d58d8186fe6b4504c7c9Drinking-water treatment with ozoneGlaze, William H.Environmental Science and Technology (1987), 21 (3), 224-30CODEN: ESTHAG; ISSN:0013-936X.A review, with 39 refs., on the use of O3 in drinking-water treatment, including O3 generation, the transfer of O3 into water, O3 as a disinfectant and as an oxidant, the chem. of O3, the reactions of O3 with natural and synthetic org. substances, the combination of ozonization with catalysts and with other processes, O3 in biol. treatment, and economic aspects.
- 4Huber, M. M.; Canonica, S.; Park, G.-Y.; von Gunten, U. Oxidation of pharmaceuticals during ozonation and advanced oxidation processes. Environ. Sci. Technol. 2003, 37, 1016– 1024, DOI: 10.1021/es025896hGoogle Scholar4https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD3sXnslylsQ%253D%253D&md5=215a3c7ce8bf9c09525067770d26392fOxidation of Pharmaceuticals during Ozonation and Advanced Oxidation ProcessesHuber, Marc M.; Canonica, Silvio; Park, Gun-Young; von Gunten, UrsEnvironmental Science and Technology (2003), 37 (5), 1016-1024CODEN: ESTHAG; ISSN:0013-936X. (American Chemical Society)Oxidn. of pharmaceuticals during conventional ozonation and advanced oxidn. processes (AOP) in drinking water purifn. was studied. In a first step, second-order rate consts. for reactions of selected pharmaceuticals with O3 (kO3) and OH- (kOH) were detd. in bench-scale expts. (apparent kO3 at pH 7 and 20°): bezafibrate (590 ± 50/M-s), carbamazepine (∼3 × 105/M-s), diazepam (0.75 ± 0.15/M-s), diclofenac (∼1 × 106/M-s), 17α-ethinylestradiol (∼3 × 106/M-s), ibuprofen (9.6 ± 1.0/M-s), iopromide (<0.8/M-s), sulfamethoxazole (∼2.5 × 106/M-s), and roxithromycin (∼7 × 104/M-s). For 5 of the pharmaceuticals, apparent kO3 at pH 7 was >5 × 104/M-s, indicating these compds. are completely transformed during ozonation. KOH values were from 3.3 to 9.8 × 109/M-s. Compared to other important micro-pollutants, e.g., Me tert-Bu ether (MTBE) and atrazine, the selected pharmaceuticals reacted about 2-3 times faster with OH-. In the second part of the study, oxidn. kinetics of selected pharmaceuticals were examd. in ozonation expts. performed in different natural water. Second-order rate consts. detd. in pure aq. soln. could be applied to predict behavior of pharmaceuticals dissolved in natural water. Overall it was concluded that ozonation and AOP are promising processes to efficiently remove pharmaceuticals in drinking water.
- 5von Gunten, U. Ozonation of drinking water: Part I. Oxidation kinetics and product formation. Water Res. 2003, 37, 1443– 1467, DOI: 10.1016/S0043-1354(02)00457-8Google Scholar5https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD3sXhtlGqsr4%253D&md5=a71315929c55e466b0f0f0a8908eebb3Ozonation of drinking water: Part I. Oxidation kinetics and product formationvon Gunten, UrsWater Research (2003), 37 (7), 1443-1467CODEN: WATRAG; ISSN:0043-1354. (Elsevier Science Ltd.)A review is given. The oxidn. of org. and inorg. compds. during ozonization can occur via ozone or OH radicals or a combination thereof. The oxidn. pathway is detd. by the ratio of ozone and OH radical concns. and the corresponding kinetics. A huge database with several hundred rate consts. for ozone and a few thousand rate consts. for OH radicals is available. Ozone is an electrophile with a high selectivity. The 2nd-order rate consts. for oxidn. by ozone vary over 10 orders of magnitude, <0.1-7 × 109/M-s. The reactions of ozone with drinking-water relevant inorg. compds. are typically fast and occur by an oxygen atom transfer reaction. Org. micropollutants are oxidized with ozone selectively. Ozone reacts mainly with double bonds, activated arom. systems and non-protonated amines. In general, electron-donating groups enhance the oxidn. by ozone whereas electron-withdrawing groups reduce the reaction rates. The kinetics of direct ozone reactions depend strongly on the speciation (acid-base, metal complexation). The reaction of OH radicals with the majority of inorg. and org. compds. is nearly diffusion-controlled. The degree of oxidn. by ozone and OH radicals is given by the corresponding kinetics. Product formation from the ozonation of org. micropollutants in aq. systems has only been established for a few compds.
- 6Schmidt, C. K.; Brauch, H.-J. N,N-dimethylsulfamide as precursor for N-nitrosodimethylamine (NDMA) formation upon ozonation and its fate during drinking water treatment. Environ. Sci. Technol. 2008, 42, 6340– 6346, DOI: 10.1021/es7030467Google Scholar6https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD1cXksl2nsrw%253D&md5=3d7616b055a8d7ab157479e76ae0aa1eN,N-Dimethylsulfamide as Precursor for N-Nitrosodimethylamine (NDMA) Formation upon Ozonation and its Fate During Drinking Water TreatmentSchmidt, Carsten K.; Brauch, Heinz-JuergenEnvironmental Science & Technology (2008), 42 (17), 6340-6346CODEN: ESTHAG; ISSN:0013-936X. (American Chemical Society)Application and microbial degrdn. of the fungicide tolylfluanide gives rise to a new decompn. product named N,N-dimethylsulfamide (DMS). In Germany, DMS was found in groundwaters and surface waters with typical concns. in the range 100-1000 and 50-90 ng/L, resp. Lab.-scale and field studies concerning its fate during drinking water treatment showed that DMS cannot be removed via riverbank filtration, activated C filtration, flocculation, and oxidn. or disinfection procedures based on H2O2, K permanganate, ClO2, or UV irradn. Even nanofiltration does not provide a sufficient removal efficiency. During ozonization about 30-50% of DMS are converted to the carcinogenic N-nitrosodimethylamine (NDMA). The NDMA being formed is biodegradable and can at least partially be removed by subsequent biol. active drinking water treatment steps including sand or activated C filtration. Disinfection with hypochlorous acid converts DMS to so far unknown degrdn. products but not to NDMA or 1,1-dimethylhydrazine (UDMH).
- 7von Sonntag, C.; von Gunten, U. Chemistry of Ozone in Water and Wastewater Treatment: From Basic Principles to Applications; IWA Publishing: London, 2012.Google ScholarThere is no corresponding record for this reference.
- 8Buxton, G. V.; Greenstock, C. L.; Helman, W. P.; Ross, A. B. Critical review of rate constants for reactions of hydrated electrons, hydrogen atoms and hydroxyl radicals (·OH/·O-) in aqueous solution. J. Phys. Chem. Ref. Data 1988, 17, 513, DOI: 10.1063/1.555805Google Scholar8https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADyaL1cXlvFyisLc%253D&md5=dae961496d7cfc2e28c3fcced28370f6Critical review of rate constants for reactions of hydrated electrons, hydrogen atoms and hydroxyl radicals (·OH/·O-) in aqueous solutionBuxton, George V.; Greenstock, Clive L.; Helman, W. Phillip; Ross, Alberta B.Journal of Physical and Chemical Reference Data (1988), 17 (2), 513-886CODEN: JPCRBU; ISSN:0047-2689.Kinetic data for the radicals H and OH in aq. soln., and the corresponding radical anions, O- and eaq-, are critically reviewed with many refs. Reactions of the radicals in aq. soln. have been studied by pulse radiolysis, flash photolysis, and other methods. Rate consts. for >3,500 reactions are tabulated, including reactions with mols., ions, and other radicals derived from inorg. and org. solutes.
- 9von Gunten, U.; Hoigné, J. Bromate formation during ozonation of bromide-containing waters: Interaction of ozone and hydroxyl radical reactions. Environ. Sci. Technol. 1994, 28, 1234– 1242, DOI: 10.1021/es00056a009Google Scholar9https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADyaK2cXktFOnu7s%253D&md5=079e49017a46ac0a97a4251b30e797c5Bromate Formation during Ozonization of Bromide-Containing Waters: Interaction of Ozone and Hydroxyl Radical Reactionsvon Gunten, Urs; Hoigne, JuergEnvironmental Science and Technology (1994), 28 (7), 1234-42CODEN: ESTHAG; ISSN:0013-936X.Kinetic simulations have been tested by lab. expts. to evaluate the major factors controlling bromate formation during ozonization of waters contg. Br-e. In the presence of an org. scavenger for OH radicals, bromate formation can be accurately predicted by the mol. O3 mechanism using published reaction rate data, even for waters contg. ammonium. In the absence of scavengers, OH radical reactions contribute significantly to bromate formation. CO32-, produced by the oxidn. of HCO3- with OH radicals, oxidize the intermediate hypobromite to bromite, which is further oxidized by O3 to bromate. During drinking water ozonization, mol. O3 controls both the initial oxidn. of bromide and the final oxidn. of bromite. OH radical reactions contribute to the oxidn. of the intermediate oxybromine species. Bromate formation in advanced oxidn. processes can be explained by a synergism of O3 and OH radicals.
- 10Elovitz, M. S.; von Gunten, U. Hydroxyl radical/ozone ratios during ozonation processes. I. The Rct concept. Ozone: Sci. Eng. 1999, 21, 239– 260, DOI: 10.1080/01919519908547239Google Scholar10https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADyaK1MXksF2rsb0%253D&md5=0bbbec53b7aedf164e7f11b716742299Hydroxyl radical/ozone ratios during ozonation processes. I. The Rct conceptElovitz, Michael S.; Von Gunten, UrsOzone: Science & Engineering (1999), 21 (3), 239-260CODEN: OZSEDS; ISSN:0191-9512. (Lewis Publishers)The ozonization of model systems and several natural waters was examd. in bench-scale batch expts. In addn. to measuring the concn. of O3, the rate of depletion of an in situ hydroxyl radical probe compd. was monitored, thus providing information on the transient steady-state concn. of hydroxyl radicals (•OH). A new parameter, Rct, representing the ratio of the •OH-exposure to the O3-exposure was calcd. as a function of reaction time. For most waters tested, including pH-buffered model systems and natural waters, Rct was a const. value for the majority of the reaction. Therefore, Rct corresponds to the ratio of the •OH concn. to the O3 concn. in a given water. For a given water source, the degrdn. of a micropollutant (e.g. atrazine) via O3 and •OH reaction pathways can be predicted by the O3 reaction kinetics and Rct.
- 11Kwon, M.; Kye, H.; Jung, Y.; Yoon, Y.; Kang, J.-W. Performance characterization and kinetic modeling of ozonation using a new method: ROH,O3 concept. Water Res. 2017, 122, 172– 182, DOI: 10.1016/j.watres.2017.05.062Google Scholar11https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXps1Knt7o%253D&md5=c042a3a40fc51d8ede841b4f45d57c29Performance characterization and kinetic modeling of ozonation using a new method: ROH,O3 conceptKwon, Minhwan; Kye, Homin; Jung, Youmi; Yoon, Yeojoon; Kang, Joon-WunWater Research (2017), 122 (), 172-182CODEN: WATRAG; ISSN:0043-1354. (Elsevier Ltd.)Ozonation is an effective treatment for removing various org. pollutants from aquatic systems. The Rct concept, which is defined as the ratio of ·OH exposure to O3 exposure, has been widely used to predict the removal efficiency of target compds., but it has significant variations by water temp. and initial O3 dose which are crucial parameters in drinking water plant. The ROH,O3 concept, which is defined as the ·OH exposure by O3 consumption, was proposed as a kinetic parameter for characterization and kinetic modeling for ozonation. The ROH,O3 concept is independent of temp. and initial O3 dose. A higher ROH,O3 value indicates a higher ·OH formation when the same amt. of O3 is consumed in different water samples; therefore, the ·OH yield from O3 decompn. of the water samples can be compared using the ROH,O3 values. The ROH,O3 concept can also be used to characterize and model the initial ozone demand phase, and it is more convenient method compared to Rct concept. Using the ROH,O3 concept, the dynamic O3 and ·OH kinetics and the removal efficiencies of iopromide and ibuprofen were well predicted (R2 = 0.98) over a wide range of exptl. conditions (n = 124).
- 12Kim, M. S.; Cha, D.; Lee, K.-M.; Lee, H.-J.; Kim, T.; Lee, C. Modeling of ozone decomposition, oxidant exposures, and the abatement of micropollutants during ozonation processes. Water Res. 2020, 169, 115230, DOI: 10.1016/j.watres.2019.115230Google Scholar12https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1MXitVygt7rE&md5=f8dd7b0530203b03085f3cd59fde44f0Modeling of ozone decomposition, oxidant exposures, and the abatement of micropollutants during ozonation processesKim, Min Sik; Cha, Dongwon; Lee, Ki-Myeong; Lee, Hye-Jin; Kim, Taewan; Lee, ChanghaWater Research (2020), 169 (), 115230CODEN: WATRAG; ISSN:0043-1354. (Elsevier Ltd.)This study demonstrates new empirical models to predict the decompn. of ozone (O3) and the exposures of oxidants (i.e., O3 and hydroxyl radical, ·OH) during the ozonation of natural waters. Four models were developed for the instantaneous O3 demand, first-order rate const. for the secondary O3 decay, O3 exposure (∫[O3]dt), and ·OH exposure ((∫[·OH]dt)), as functions of five independent variables, namely the O3 dose, concn. of dissolved org. carbon (DOC), pH, alky., and temp. The models were derived by polynomial regression anal. of exptl. data obtained by controlling variables in natural water samples from a single source water (Maegok water in Korea), and they exhibited high accuracies for regression (R2 = 0.99 for the three O3 models, and R2 = 0.96 for the ·OH exposure model). The three O3 models exhibited excellent internal validity for Maegok water samples of different conditions (that were not used for the model development). They also showed acceptable external validity for seven natural water samples collected from different sources (not Maegok water); the IOD model showed somewhat poor external validity. The models for oxidant exposures were successfully used to predict the abatement of micropollutants by ozonation; the model predictions showed high accuracy for Maegok water, but not for the other natural waters.
- 13Mobed, J. J.; Hemmingsen, S. L.; Autry, J. L.; McGown, L. B. Fluorescence characterization of IHSS humic substances: Total luminescence spectra with absorbance correction. Environ. Sci. Technol. 1996, 30, 3061– 3065, DOI: 10.1021/es960132lGoogle Scholar13https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADyaK28Xlt1eksrg%253D&md5=050cc17a151f5611131eadbbc100001dFluorescence Characterization of IHSS Humic Substances: Total Luminescence Spectra with Absorbance CorrectionMobed, Jarafshan J.; Hemmingsen, Sherry L.; Autry, Jennifer L.; McGown, Linda B.Environmental Science and Technology (1996), 30 (10), 3061-3065CODEN: ESTHAG; ISSN:0013-936X. (American Chemical Society)Total luminescence spectroscopy was applied to the fluorescence characterization of humic substances obtained from the International Humic Substances Society (IHSS). Total luminescence spectra, represented as excitation-emission matrixes (EEMs), may be used to discriminate between soil-derived and aquatic-derived IHSS humic substances and between humic and fulvic acids derived from the same source (soil or aquatic). Ionic strength at 0-1 M KCl and humic substance concn. at 5-100 mg/L had little effect on the fluorescence spectral characteristics of the humic substances, while pH had significant effects as expected. Absorbance correction is essential for accurate representation and comparison of the EEMs of the humic substances at high concns.
- 14Baker, A. Fluorescence excitation-emission matrix characterization of some sewage-impacted rivers. Environ. Sci. Technol. 2001, 35, 948– 953, DOI: 10.1021/es000177tGoogle Scholar14https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD3MXmtFyrtw%253D%253D&md5=743056946850c6b9d3a8173a4e9b4d60Fluorescence Excitation-Emission Matrix Characterization of Some Sewage-Impacted RiversBaker, AndyEnvironmental Science and Technology (2001), 35 (5), 948-953CODEN: ESTHAG; ISSN:0013-936X. (American Chemical Society)Fluorescence excitation-emission matrix (EEM) spectrophotometry was applied to 10 sample sites in 6 rivers in northeastern England, some of which were polluted by sewage treatment works (STW) discharges, to study whether STW discharges had a significantly distinct fluorescence signature. Upstream, downstream, and STW discharge samples for 2 STW demonstrated that treated sewage had a distinct fluorescence EEM, with high tryptophan and fulvic-like fluorescence intensities of approx. equal ratio. This signature was obsd. downstream samples. When all 10 sample sites were compared, 2 trend lines were apparent where STW impacted rivers plotted sep. from the other sample sites. Fluorescence EEM signatures were compared to absorption at 254 nm and demonstrated to provide a better fingerprint of sewage-impacted water. It is suggested that fluorescence EEM spectrophotometry can provide a useful tool to analyze grab samples collected for routine and investigative monitoring and has the potential for online monitoring of STW impacts on river systems.
- 15Yu, H.; Qu, F.; Zhang, X.; Shao, S.; Rong, H.; Liang, H.; Bai, L.; Ma, J. Development of correlation spectroscopy (COS) method for analyzing fluorescence excitation emission matrix (EEM): A case study of effluent organic matter (EfOM) ozonation. Chemosphere 2019, 228, 35– 43, DOI: 10.1016/j.chemosphere.2019.04.119Google Scholar15https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1MXnvFGrtLo%253D&md5=f53e82581db1c0735d9edef10c3ea211Development of correlation spectroscopy (COS) method for analyzing fluorescence excitation emission matrix (EEM): A case study of effluent organic matter (EfOM) ozonationYu, Huarong; Qu, Fangshu; Zhang, Xiaolei; Shao, Senlin; Rong, Hongwei; Liang, Heng; Bai, Langming; Ma, JunChemosphere (2019), 228 (), 35-43CODEN: CMSHAF; ISSN:0045-6535. (Elsevier Ltd.)Two-dimensional correlation spectroscopy has been used as a powerful tool for analyzing spectral features, but it has never been applied to fluorescence excitation-emission matrix (EEM) data due to the incompatible dimensions. This study first investigated EEM-COS by reducing the dimensions of the EEM (using parallel factor anal., PARAFAC) for fitting to 2DCOS (EEM-PARAFAC-COS). The fluorescence changes of effluent org. matter during ozonation were studied using EEM-COS and synchronous fluorescence (SF)-2DCOS. The conventionally used SF-2DCOS proved to be biased due to the intrinsic drawback of SF, while the EEM-PARAFAC-COS gave accurate and trustworthy results. Homo-EEM-PARAFAC-COS indicated that the fluorescence protein-like and fulvic-like substances in EfOM were preferentially ozonated compared to humic-like substances. Hetero-EEM-PARAFAC-COS analyses on the EEM, FTIR, UV-vis absorbance, and size-exclusion chromatog. showed that the fluorescence protein-like and fulvic-like substances in EfOM were assocd. with lower mol. wt. (MW, ∼0.95 kDa), UV absorbance at ∼280 nm, and more electron-enriched aroms. (with amide and phenolic groups), which explained their ozonation preference, while humic-like substances were related to carboxylic groups, UV absorbance at ∼255 nm, and orgs. at MW of ∼4.50 kDa. This work demonstrated the great potential of EEM-PARAFAC-COS in studying fluorescence change and correlating fluorescence with other spectra.
- 16Cui, Y.; Yu, J.; Su, M.; Jia, Z.; Liu, T.; Oinuma, G.; Yamauchi, T. Humic acid removal by gas-liquid interface discharge plasma: performance, mechanism and comparison to ozonation. Environ. Sci.: Water Res. Technol. 2019, 5, 152– 160, DOI: 10.1039/c8ew00520fGoogle Scholar16https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXit1eitr7P&md5=c43f29a4d57e93d83bc501d3ab478531Humic acid removal by gas-liquid interface discharge plasma: performance, mechanism and comparison to ozonationCui, Yanyan; Yu, Jianwei; Su, Ming; Jia, Zeyu; Liu, Tingting; Oinuma, Gaku; Yamauchi, TokikoEnvironmental Science: Water Research & Technology (2019), 5 (1), 152-160CODEN: ESWRAR; ISSN:2053-1419. (Royal Society of Chemistry)The removal of natural org. matter (NOM) is a major objective for drinking water treatment. In this study, a novel advanced oxidn. process (AOP) based on plasma in gas-liq. interface discharge was investigated for removing humic acid (HA), and ozonation treatment effect was evaluated as a comparison. Several indexes including UV254, dissolved org. carbon (DOC), specific UV absorbance (SUVA), and excitation-emission matrix (EEM) fluorescence spectra were employed to characterize HA transformation. The results showed that typical species including ozone, hydrogen peroxide and HȮ were produced in the plasma process, which made better removal efficiency than ozonation. Esp., UV254 was removed effectively by 96.5% and 90.3% within 15 min by plasma and ozonation, while DOC was 49.2% and 41.1%, resp. HȮ radicals were majorly responsible for the degrdn. of HA in the plasma process, which were mainly generated via reaction of in situ produced ozone and H2O2 during the discharge process. UV/vis absorbance and EEM results further showed that HA with the large mol. wt. fraction was decompd. to small mol. wt. fraction. This study indicates that plasma technol. has a significant application prospect as a replacement for ozone in drinking water industry.
- 17Li, L.; Rong, S.; Wang, R.; Yu, S. Recent advances in artificial intelligence and machine learning for nonlinear relationship analysis and process control in drinking water treatment: A review. Chem. Eng. J. 2021, 405, 126673, DOI: 10.1016/j.cej.2020.126673Google Scholar17https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXhs1Cqu7bM&md5=7b0f1ca10517b5a26a4aed1e2213bf6dRecent advances in artificial intelligence and machine learning for nonlinear relationship analysis and process control in drinking water treatment: A reviewLi, Lei; Rong, Shuming; Wang, Rui; Yu, ShuiliChemical Engineering Journal (Amsterdam, Netherlands) (2021), 405 (), 126673CODEN: CMEJAJ; ISSN:1385-8947. (Elsevier B.V.)A review. Because of its robust autonomous learning and ability to address complex problems, artificial intelligence (AI) has increasingly demonstrated its potential to solve the challenges faced in drinking water treatment (DWT). AI technol. provides tech. support for the management and operation of DWT processes, which is more efficient than relying solely on human operations. AI-based data anal. and evolutionary learning mechanisms are capable of realizing water quality diagnosis, autonomous decision making and operation process optimization and have the potential to establish a universal process anal. and predictive model platform. This review briefly introduces AI technologies that are widely used in DWT. Moreover, this paper reviews in detail the mature applications and latest discoveries of AI and machine learning technologies in the fields of source water quality, coagulation/flocculation, disinfection and membrane filtration, including source water contaminant monitoring and identification, accurate and efficient prediction of coagulation dosage, anal. of the formation of disinfection byproducts and advanced control of membrane fouling. Finally, the challenges facing AI technologies and the issues that need further study are discussed; these challenges can be briefly summarized as (a) obtaining more effective characterization data to screen and identify targeted contaminants in the complex background with the assistance of AI technologies and (b) establishing a macro intelligence model and decision scheme for entire drinking water treatment plants (DWTPs) to support the management of the water supply system.
- 18American Public Health Association (APHA), Standard Methods for the Examination of Water and Wastewater; Greenberg, A. E., Clesceri, L. S., Eaton, A. D., Eds.; APHA: Washington, DC, 1992.Google ScholarThere is no corresponding record for this reference.
- 19Chen, W.; Westerhoff, P.; Leenheer, J. A.; Booksh, K. Fluorescence excitation-emission matrix regional integration to quantify spectra for dissolved organic matter. Environ. Sci. Technol. 2003, 37, 5701– 5710, DOI: 10.1021/es034354cGoogle Scholar19https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD3sXovFeisrc%253D&md5=6fc1ca4c33117f277bf6d01dd21c9f5fFluorescence Excitation-Emission Matrix Regional Integration to Quantify Spectra for Dissolved Organic MatterChen, Wen; Westerhoff, Paul; Leenheer, Jerry A.; Booksh, KarlEnvironmental Science and Technology (2003), 37 (24), 5701-5710CODEN: ESTHAG; ISSN:0013-936X. (American Chemical Society)Excitation-emission matrix (EEM) fluorescence spectroscopy has been widely used to characterize dissolved org. matter (DOM) in water and soil. However, interpreting the >10,000 wavelength-dependent fluorescence intensity data points represented in EEMs has posed a significant challenge. Fluorescence regional integration, a quant. technique that integrates the vol. beneath an EEM, was developed to analyze EEMs. EEMs were delineated into five excitation-emission regions based on the fluorescence of model compds., DOM fractions, and marine waters or fresh waters. Volumetric integration under the EEM within each region, normalized to the projected excitation-emission area within that region and dissolved org. carbon concn., resulted in a normalized region-specific EEM vol. (Φi,n). Solid-state carbon NMR (13C NMR), Fourier transform IR (FTIR) anal., UV-visible absorption spectra, and EEMs were obtained for std. Suwannee River fulvic acid and 15 hydrophobic or hydrophilic acid, neutral, and base DOM fractions plus nonfractionated DOM from wastewater effluents and rivers in the southwestern USA. DOM fractions fluoresced in one or more EEM regions. The highest cumulative EEM vol. (ΦT,n = ΣΦi,n) was obsd. for hydrophobic neutral DOM fractions, followed by lower ΦT,n values for hydrophobic acid, base, and hydrophilic acid DOM fractions, resp. An extd. wastewater biomass DOM sample contained arom. protein- and humic-like material and was characteristic of bacterial sol. microbial products. Arom. carbon and the presence of specific arom. compds. (as indicated by solid-state 13C NMR and FTIR data) resulted in EEMs that aided in differentiating wastewater effluent DOM from drinking water DOM.
- 20Bader, H.; Hoigné, J. Determination of ozone in water by the indigo method. Water Res. 1981, 15, 449– 456, DOI: 10.1016/0043-1354(81)90054-3Google Scholar20https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADyaL3MXks1Olu78%253D&md5=72c0f23814ddfe196794a5ede4b84532Determination of ozone in water by the indigo methodBader, H.; Hoigne, J.Water Research (1981), 15 (4), 449-56CODEN: WATRAG; ISSN:0043-1354.O3 was detd. in natural water by the decolorization of K indigotrisulfonate [67627-18-3] at 600 nm (pH <4). The method is stoichiometric and fast. The change of absorbance vs. O3 added is -2.0 × 104/mol-cm and is independent of the concn. of aq. O3 at 0.005-30 mg O3/L. The precision is 2% for low concns. Visual detection can be used to det. 0.01 mg O3/L. The reagent soln. is stable for 3 mo. For example, for a sample of ozonized lake water, 0.60 mg O3/L was detd. by this method, compared with 0.61 mg/L by a direct UV method.
- 21Breiman, L.; Friedman, J. H.; Olshen, R. A.; Stone, C. J. Classification and Regression Trees; Chapman and Hall/CRC: Boca Raton, FL, 1984.Google ScholarThere is no corresponding record for this reference.
- 22Ließ, M.; Glaser, B.; Huwe, B. Uncertainty in the spatial prediction of soil texture. Geoderma 2012, 170, 70– 79, DOI: 10.1016/j.geoderma.2011.10.010Google ScholarThere is no corresponding record for this reference.
- 23Breiman, L. Random Forests. Mach. Learn. 2001, 45, 5– 32, DOI: 10.1023/A:1010933404324Google ScholarThere is no corresponding record for this reference.
- 24Baek, S.-S.; Choi, Y.; Jeon, J.; Pyo, J.; Park, J.; Cho, K. H. Replacing the internal standard to estimate micropollutants using deep and machine learning. Water Res. 2021, 188, 116535, DOI: 10.1016/j.watres.2020.116535Google Scholar24https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXit1OmsrzI&md5=80bfa049559c39de2de6a17691be9833Replacing the internal standard to estimate micropollutants using deep and machine learningBaek, Sang-Soo; Choi, Younghun; Jeon, Junho; Pyo, JongCheol; Park, Jongkwan; Cho, Kyung HwaWater Research (2021), 188 (), 116535CODEN: WATRAG; ISSN:0043-1354. (Elsevier Ltd.)Similar to the worldwide proliferation of urbanization, micropollutants have been involved in aquatic and ecol. environmental systems. These pollutants have the propensity to wreak havoc on human health and the ecol. system; hence, it is important to persistently monitor micropollutants in the environment. Micropollutants are commonly quantified via target anal. using high resoln. mass spectrometry and the stable isotope labeled (SIL) std. However, the cost-intensiveness of this std. presents a major obstacle in measuring micropollutants. This study resolved this problem by developing data-driven models, including deep learning (DL) and machine learning (ML), to est. the concn. of micropollutants without resorting to the SIL std. Our study hypothesized that natural org. matter (NOM) could replace internal stds. if there was a specific mass spectrum (MS) subset, including NOM information, which correlated with an SIL std. peak. Therefore, we analyzed the MS to find the specific MS subsets for replacing the SIL std. peak. Thirty-five alternative MS subsets were detd. for applying DL and ML as input data. Thereafter, we trained four different DL models, namely, ResNet101, GoogLeNet, VGG16, and Inception v3, as well as three different ML models, i.e., random forest (RF), support vector machine (SVM), and artificial neural network (ANN). A total of 680 MS data were used for the model training to est. five different micropollutants, namely Sulpiride, Metformin, and Benzotriazole. Among the DL models, ResNet 101 exhibited the highest model performance, showing that the av. validation R2 and MSE were 0.84 and 0.26 ng/L, resp., while RF was the best in the ML models, manifesting R2 and MSE values of 0.69 and 0.58 ng/L. The trained models showed accurate training and validation results for the estn. of the five micropollutant concns. Therefore, this study demonstrates that the suggested anal. has a potential for alternative micropollutant measurement that has rapid and economic vantages.
- 25Sevgen, E.; Kocaman, S.; Nefeslioglu, H. A.; Gokceoglu, C. A novel performance assessment approach using photogrammetric techniques for landslide susceptibility mapping with logistic regression, ANN and Random Forest. Sensors 2019, 19, 3940, DOI: 10.3390/s19183940Google ScholarThere is no corresponding record for this reference.
- 26Singh, B.; Sihag, P.; Singh, K. Modelling of impact of water quality on infiltration rate of soil by random forest regression. Model. Earth Syst. Environ. 2017, 3, 999– 1004, DOI: 10.1007/s40808-017-0347-3Google ScholarThere is no corresponding record for this reference.
- 27Akaike, H. Information Theory and an Extension of the Maximum Likelihood Principle. In Selected Papers of Hirotugu Akaike; Parzen, E., Tanabe, K., Kitagawa, G., Eds.; Springer: New York, NY, 1998.Google ScholarThere is no corresponding record for this reference.
- 28Snipes, M.; Taylor, D. C. Model selection and Akaike information criteria: An example from wine ratings and prices. Wine Econ. Policy 2014, 3, 3– 9, DOI: 10.1016/j.wep.2014.03.001Google ScholarThere is no corresponding record for this reference.
- 29Buffle, M.-O.; von Gunten, U. Phenols and amine induced HO· generation during the initial phase of natural water ozonation. Environ. Sci. Technol. 2006, 40, 3057– 3063, DOI: 10.1021/es052020cGoogle Scholar29https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD28XivF2qsbk%253D&md5=3e9a3cc595c6b54acba2901b7aa621bbPhenols and Amine Induced HO• Generation During the Initial Phase of Natural Water OzonationBuffle, Marc-Olivier; von Gunten, UrsEnvironmental Science & Technology (2006), 40 (9), 3057-3063CODEN: ESTHAG; ISSN:0013-936X. (American Chemical Society)The initial phase of ozone decompn. in natural water (t < 20 s) is poorly understood. It has recently been shown to result in very high transient HO• concns. and, thereby, plays an essential role during processes such as bromate formation or contaminants oxidn. Phenols and amines are ubiquitous moieties of natural org. matter. Naturally occurring concns. of primary, secondary, and tertiary amines, amino acids, and phenol were added to surface water, and ozone decompn. as well as HO• generation were measured starting 350 ms after ozone addn. Six seconds into the process, 5 μM of dimethylamine and phenol had generated ∫HO•dt = 1 × 10-10 M·s and 1.8 × 10-10 M·s, resp. With 10 μM dimethylamine and 1.5 mgO3/L, Rct, (∫HO•dt/∫O3dt) reached 10-6, which is larger than in advanced oxidn. processes (AOP) such as O3/H2O2. Expts. in the presence of HO•-scavengers indicated that a significant fraction of phenol-induced ozone decompn. and HO• generation results from a direct electron transfer to ozone. For dimethylamine, the main mechanism of HO• generation is direct formation of O2•- which reacts selectively with O3 to form O3•-. Pretreatment of phenol-contg. water with HOCl or HOBr did not decrease HO• generation, while the same treatment of dimethylamine-contg. water considerably reduced HO• generation.
- 30Neta, P.; Dorfman, L. M. Pulse Radiolysis Studies. XIII. Rate Constants for the Reaction of Hydroxyl Radicals with Aromatic Compounds in Aqueous Solutions. In Radiation Chemistry Volume I. Aqueous Media, Biology, Dosimetry; Hart, E. J., Ed.; American Chemical Society: Washington, DC, 1968.Google ScholarThere is no corresponding record for this reference.
- 31Determann, S.; Reuter, R.; Wagner, P.; Willkomm, R. Fluorescent matter in the eastern Atlantic Ocean. Part 1: Method of measurement and near-surface distribution. Deep Sea Res., Part I 1994, 41, 659– 675, DOI: 10.1016/0967-0637(94)90048-5Google Scholar31https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADyaK2MXhslWgtbg%253D&md5=cd37e273cd7acc7c086fc7e52d11c71eFluorescent matter in the eastern Atlantic Ocean. Part 1: method of measurement and near-surface distributionDetermann, S.; Reuter, R.; Wagner, P.; Willkomm, R.Deep-Sea Research, Part I: Oceanographic Research Papers (1994), 41 (4), 659-75CODEN: DRORE7; ISSN:0967-0637.Fluorescence spectra of org. matter in seawater were measured during a cruise through the South and North Atlantic from Capetown, South Africa, to Bremerhaven, Germany. The data are calibrated by normalization to the water Raman scatter band, which allows their quantification without the need of fluorescence stds. Spectral structures are found which can be related to tryptophan- and tyrosine-like substances and to gelbstoff. Their distribution in the eastern Atlantic is discussed and compared with other hydrog. parameters.
- 32Ahmad, S. R.; Reynolds, D. M. Monitoring of water quality using fluorescence technique: Prospect of on-line process control. Water Res. 1999, 33, 2069– 2074, DOI: 10.1016/S0043-1354(98)00435-7Google Scholar32https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADyaK1MXjslGjs7w%253D&md5=445b62eb42cc06068aa81ae33a336a6dMonitoring of water quality using fluorescence technique: prospect of on-line process controlAhmad, S. R.; Reynolds, D. M.Water Research (1999), 33 (9), 2069-2074CODEN: WATRAG; ISSN:0043-1354. (Elsevier Science Ltd.)The wastewater from sewage processing plants shows characteristic fluorescence signatures when excited by UV light in the 240-300 nm wavelength band. A typical signature is a spectrum having a broad band centered at about 350 nm and two relatively less intense bands centered at about 390 and 430 nm. Samples of settled sewage, treated in an aerobic digester, show a substantial redn. of the intensity of the 350 nm band and comparatively much smaller redn. of the strength of the other two bands. The biodegradable chromophoric constituent species are, therefore, considered to be the major contributors to the overall fluorescence within this band. The intensity of this band has been found to have a good correlation with the BOD parameter. This parameter is universally used for assessing the sewage strength and the suitability of the treated effluent for discharge into rivers or reservoirs. Therefore, the fluorescence technique is considered to have the potential for use in noninvasive continuous water quality monitoring thereby, enabling online process control in sewage treatment plants. However, fluorescence strength is affected by the pH of the sample, particularly at higher values. This has to be taken into account for the practical utilization of this technique.
- 33Mounier, S.; Patel, N.; Quilici, L.; Benaim, J. Y.; Benamou, C. Fluorescence 3D de la matière organique dissoute du fleuve amazone. Water Res. 1999, 33, 1523– 1533, DOI: 10.1016/S0043-1354(98)00347-9Google Scholar33https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADyaK1MXitlehtLw%253D&md5=9ba28abc2bb9eb3272db13638b1879afThree-dimensional fluorescence of the dissolved organic carbon in the Amazon RiverMounier, S.; Patel, N.; Quilici, L.; Benaim, J. Y.; Benamou, C.Water Research (1999), 33 (6), 1523-1533CODEN: WATRAG; ISSN:0043-1354. (Elsevier Science Ltd.)Natural org. matter is an important pool that is not yet totally described. Two types of compds. are found: some chem. well characterized mols. (biopolymers) and uncharacterized humic substances (geopolymers). The spectroscopic properties of this pool of org. matter have recently been advanced by the excitation emission fluorescence matrix (EEFM) characterization of marine and terrestrial DOM in seawater using excitation-emission matrix spectroscopy. Three types of fluorophores are described by their excitation/emission wavelength at max. intensity (λex/λem)max: the tyrosine and tryptophan like structures, not examd. here, and the humic like fluorescent structures of type A (λ260/λ445) and type C (λ330/λ445). The EEFM applied to sequential tangential ultrafiltered (UFTS) Amazonian fresh waters give spectroscopic information on the fluorescent properties of particulate ( > 0.22 μm), colloidal and dissolved ( < 5 kDa) org. matter. Chromophores A and C are present in all sized fraction samples. Their (λex/λem)max are in the same domains as those of terrestrial humic substances. Differentiation of the type A and type C peaks in 3D diagrams are based on their (λex/λem)max position and the Ia/Ic ratio. Differences are obsd. between humic material extd. by hydrophobic resins or concd. sample from tangential sequential ultrafiltration (UFTS). Spectroscopic properties of the humic material are not modified by the ultrafiltration process. A particular attention is given on the differentiation between black water (Rio Negro River) and white water (Rio Solimoes and Rio Madeira River). Black waters are generally known as humic rich and low mineral content waters. Humic like fluorescent compds. of type C are preferentially retained by membranes with 5 kDa cut off.; these compds. have the larger mol. wt. In the presence of Cu cation, the type C compds. are divided in 2 groups according to their (λex/λem)max: the 1st one is invariant, the other one expands (20 nm) its efficient excitation wavelength domain. Photochem. reactions induced by UV irradn. also lead to a 20-nm expansion of the efficient excitation wavelength domain. From these spectroscopic and mol. wt. complementary data, it is proposed that A type fluorophores are close to fulvic acids while C fluorophores seems to be more related to humic acid.
- 34Coble, P. G. Characterization of marine and terrestrial DOM in seawater using excitation-emission matrix spectroscopy. Mar. Chem. 1996, 51, 325– 346, DOI: 10.1016/0304-4203(95)00062-3Google Scholar34https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADyaK28XnslWltg%253D%253D&md5=e4652be4c0e7edce63657af49497b93cCharacterization of marine and terrestrial DOM in seawater using excitation-emission matrix spectroscopyCoble, Paula G.Marine Chemistry (1996), 51 (4), 325-46CODEN: MRCHBD; ISSN:0304-4203. (Elsevier)High-resoln. fluorescence spectroscopy was used to characterize dissolved org. matter (DOM) in concd. and unconcd. water samples from a wide variety of freshwater, coastal and marine environments. Several types of fluorescent signals were obsd., including humic-like, tyrosine-like, and tryptophan-like. Humic-like fluorescence consisted of two peaks, one stimulated by UV excitation (peak A) and one by visible excitation (peak C). For all samples, the positions of both excitation and emission maxima for peak C were dependent upon wavelength of observation, with a shift towards longer wavelength emission max. at longer excitation wavelength and longer wavelength excitation max. at longer emission wavelength. A trend was obsd. in the position of wavelength-independent max. fluorescence (Exmax/Emmax) for peak C, with max. at shorter excitation and emission wavelengths for marine samples than for freshwater samples. Differences suggest that the humic material in marine surface waters is chem. different from humic material in the other environments sampled. These results explain previous conflicting reports regarding fluorescence properties of DOM from natural waters and also provide a means of distinguishing between water mass sources in the ocean.
- 35Singh, R. Membrane Technology and Engineering for Water Purification; Butterworth-Heinemann: Oxford, U.K., 2015.Google ScholarThere is no corresponding record for this reference.
- 36Chun, K. C.; Chang, R. W.; Williams, G. P.; Chang, Y. S.; Tomasko, D.; LaGory, K.; Ditmars, J.; Chun, H. D.; Lee, B.-K. Water quality issues in the Nakdong River basin in the Republic of Korea. Environ. Eng. Policy 1999, 2, 131– 143, DOI: 10.1007/s100220000024Google ScholarThere is no corresponding record for this reference.
- 37Nam, S.-N.; Amy, G. Differentiation of wastewater effluent organic matter (EfOM) from natural organic matter (NOM) using multiple analytical techniques. Water Sci. Technol. 2008, 57, 1009– 1015, DOI: 10.2166/wst.2008.165Google Scholar37https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD1cXntV2hur4%253D&md5=c2a8cd6adfeba15b5f966ab8615ee576Differentiation of wastewater effluent organic matter (EfOM) from natural organic matter (NOM) using multiple analytical techniquesNam, Seong-Nam; Amy, GaryWater Science and Technology (2008), 57 (7), 1009-1015CODEN: WSTED4; ISSN:0273-1223. (IWA Publishing)Using three anal. techniques of size exclusion chromatog. (SEC), fluorescence excitation-emission matrix (EEM), and dissolved org. nitrogen (DON) measurement, differentiating characteristics of effluent org. matter (EfOM) from natural org. matter (NOM) have been investigated. SEC reveals a wide range of mol. wt. (MW) for EfOM and high amt. of high MW polysaccharides, and low MW org. acids compared to NOM. Clear protein-like peaks using fluorescence EEM were a major feature of EfOM distinguishing it from NOM. Fluorescence index (FI), an indicator to distinguish autochthonous origin from allochthonous origin, differentiated EfOM from NOM by exhibiting higher values, indicating a microbial origin. In EfOM samples, DON present in higher amts. than NOM.
Cited By
Click to copy section linkSection link copied!
Citation Statements
beta
Smart citations by scite.ai include citation statements extracted from the full text of the citing article. The number of the statements may be higher than the number of citations provided by ACS Publications if one paper cites another multiple times or lower if scite has not yet processed some of the citing articles.
- SupportingSupporting0
- MentioningMentioning20
- ContrastingContrasting0
Explore this article's citation statements on scite.ai
This article is cited by 30 publications.
- Rabbi Sikder, Tianyu Zhang, Tao Ye. Predicting THM Formation and Revealing Its Contributors in Drinking Water Treatment Using Machine Learning. ACS ES&T Water 2024, 4
(3)
, 899-912. https://doi.org/10.1021/acsestwater.3c00020
- Yangyang Wu, Yingze Li, Ze Jiang, Ziyang Xu, Mingbao Yang, Jiahui Ding, Changyong Zhang. Machine Learning Prediction of Phosphate Adsorption on Six Different Metal-Containing Adsorbents. ACS ES&T Engineering 2023, 3
(8)
, 1135-1146. https://doi.org/10.1021/acsestengg.3c00001
- Jose D. Hernandez-Betancur, Gerardo J. Ruiz-Mercado, Mariano Martin. Predicting Chemical End-of-Life Scenarios Using Structure-Based Classification Models. ACS Sustainable Chemistry & Engineering 2023, 11
(9)
, 3594-3602. https://doi.org/10.1021/acssuschemeng.2c05662
- Xin Yang, Fernando L. Rosario-Ortiz, Yu Lei, Yanheng Pan, Xin Lei, Paul Westerhoff. Multiple Roles of Dissolved Organic Matter in Advanced Oxidation Processes. Environmental Science & Technology 2022, 56
(16)
, 11111-11131. https://doi.org/10.1021/acs.est.2c01017
- Guangfei Yu, Yiqiu Wu, Hongbin Cao, Qingfeng Ge, Qin Dai, Sihan Sun, Yongbing Xie. Insights into the Mechanism of Ozone Activation and Singlet Oxygen Generation on N-Doped Defective Nanocarbons: A DFT and Machine Learning Study. Environmental Science & Technology 2022, 56
(12)
, 7853-7863. https://doi.org/10.1021/acs.est.1c08666
- Ning Wei, Zhenyu Jia, Zhengyu Men, Chunzhe Ren, Yanjie Zhang, Jianfei Peng, Lin Wu, Ting Wang, Qijun Zhang, Hongjun Mao. Machine Learning Predicts Emissions of Brake Wear PM2.5: Model Construction and Interpretation. Environmental Science & Technology Letters 2022, 9
(5)
, 352-358. https://doi.org/10.1021/acs.estlett.2c00117
- Xiangzhou Yuan, Manu Suvarna, Sean Low, Pavani Dulanja Dissanayake, Ki Bong Lee, Jie Li, Xiaonan Wang, Yong Sik Ok. Applied Machine Learning for Prediction of CO2 Adsorption on Biomass Waste-Derived Porous Carbons. Environmental Science & Technology 2021, 55
(17)
, 11925-11936. https://doi.org/10.1021/acs.est.1c01849
- Nohyeong Jeong, Tai-heng Chung, Tiezheng Tong. Predicting Micropollutant Removal by Reverse Osmosis and Nanofiltration Membranes: Is Machine Learning Viable?. Environmental Science & Technology 2021, 55
(16)
, 11348-11359. https://doi.org/10.1021/acs.est.1c04041
- Benjie Zhu, Chenxing Li, Jialiang Liu, Fei Gao, Fang Liu, Xiaoguang Duan. Adsorption promoted heterogeneous catalytic ozonation for removing emerging contaminations: The intrinsic correlations with biochar properties and complementary effect of adsorption and ozonation processes. Separation and Purification Technology 2025, 354 , 128726. https://doi.org/10.1016/j.seppur.2024.128726
- Wei Zhuang, Xiao Zhao, Qianqian Luo, Xinyuan Lv, Zhilin Zhang, Lihua Zhang, Minghao Sui. Task decomposition strategy based on machine learning for boosting performance and identifying mechanisms in heterogeneous activation of peracetic acid process. Water Research 2024, 267 , 122521. https://doi.org/10.1016/j.watres.2024.122521
- Homin Kye, Sook-Hyun Nam, Eunju Kim, Jaewuk Koo, Yonghyun Shin, Juwon Lee, Tae-Mun Hwang. Application of tryptophan-like fluorescence index to quantify the trace organic compounds removal in wastewater ozonation. Chemosphere 2024, 363 , 142862. https://doi.org/10.1016/j.chemosphere.2024.142862
- Dongwon Cha, Sanghun Park, Min Sik Kim, Jaesang Lee, Yunho Lee, Kyung Hwa Cho, Changha Lee. Prediction of hydroxyl radical exposure during ozonation using different machine learning methods with ozone decay kinetic parameters. Water Research 2024, 261 , 122067. https://doi.org/10.1016/j.watres.2024.122067
- Wei Zhuang, Xiao Zhao, Yiying Zhang, Qianqian Luo, Lihua Zhang, Minghao Sui. Autoencoded chemical feature interaction machine learning method boosting performance of piezoelectric catalytic process. Nano Energy 2024, 126 , 109670. https://doi.org/10.1016/j.nanoen.2024.109670
- Rakesh Kumar, Deepa Kundu, Tapos Kormoker, Sarita Joshi, Pawan Kumar Rose, Sundhanshu Kumar, Prafulla Kumar Sahoo, Prabhakar Sharma, Jasmeet Lamba. Phycoremediation of potentially toxic elements for agricultural and industrial wastewater treatment: Recent advances, challenges, and future prospects. Desalination and Water Treatment 2024, 319 , 100505. https://doi.org/10.1016/j.dwt.2024.100505
- Chandrasekhar Nettem, Gopalan Rajaraman. How do quantum chemical descriptors shape hydrogen atom abstraction reactivity in cupric-superoxo species? A combined DFT and machine learning perspective. Inorganic Chemistry Frontiers 2024, 11
(13)
, 3830-3846. https://doi.org/10.1039/D4QI00701H
- Abolhassan Banisheikholeslami, Farhad Qaderi. Applied machine learning to the determination of biochar hydrogen sulfide adsorption capacity. Machine Learning 2024, 113
(6)
, 3419-3441. https://doi.org/10.1007/s10994-023-06446-2
- Libo Xia, Xiaoxuan Hao, Yun Zhou. Machine learning-based global trends and the development prospects of wastewater treatment: A bibliometric analysis. Journal of Environmental Chemical Engineering 2024, 12
(3)
, 112732. https://doi.org/10.1016/j.jece.2024.112732
- Benjie Zhu, Yuting Wang, Chenxing Li, Fei Gao, Fang Liu, Guofei Jiang, Huaqing Zhang, Xiaoguang Duan. Construction of catalytic ozonation synergistic photo-self-Fenton system and analysis of synergistic catalysis and reaction mechanism activated by modified carbon nitride. Applied Catalysis B: Environmental 2024, 342 , 123408. https://doi.org/10.1016/j.apcatb.2023.123408
- Yongkai Huang, Yiling Chen. Prediction of Total Phosphorus Based on Distance Correlation and Machine Learning Methods—a Case Study of Dongjiang River, China. Water, Air, & Soil Pollution 2024, 235
(2)
https://doi.org/10.1007/s11270-024-06913-z
- Ahmed E. Alprol, Abdallah Tageldein Mansour, Marwa Ezz El-Din Ibrahim, Mohamed Ashour. Artificial Intelligence Technologies Revolutionizing Wastewater Treatment: Current Trends and Future Prospective. Water 2024, 16
(2)
, 314. https://doi.org/10.3390/w16020314
- Zow Afshan, Sherin Zafar, Safdar Tanvir. Prediction of Water Quality Through Machine Learning: A Review. 2024, 527-550. https://doi.org/10.1007/978-981-97-4228-8_35
- Wei Zhang, Waqar Muhammad Ashraf, Sachini Supunsala Senadheera, Daniel S. Alessi, Filip M.G. Tack, Yong Sik Ok. Machine learning based prediction and experimental validation of arsenite and arsenate sorption on biochars. Science of The Total Environment 2023, 904 , 166678. https://doi.org/10.1016/j.scitotenv.2023.166678
- Rongkui Su, Zishi Li, Xiancheng Ma, Yuying Li, Zheng Zeng, Liqing Li, Peng Sheng, Hanqing Wang, Shaobin Wang. Machine learning, experiments and molecular simulation demonstrate the adsorption mechanism of acetone on porous carbon at different pressures. Separation and Purification Technology 2023, 323 , 124480. https://doi.org/10.1016/j.seppur.2023.124480
- Caihong Qin, Rongrong Liu, Chaochao Jiang, Mengke Guo, Rui Yu, Jiayu Huang, Wenwen Bai, Xiaoqing Dang. Non-thermal plasma coupled liquid-phase catalysis /Fe2+ for VOCs removal: Enhanced mechanism of protocatechuic acid. Journal of Industrial and Engineering Chemistry 2023, 122 , 334-340. https://doi.org/10.1016/j.jiec.2023.02.033
- Xiancheng Ma, Wenjun Xu, Rongkui Su, Lishu Shao, Zheng Zeng, Liqing Li, Hanqing Wang. Insights into CO2 capture in porous carbons from machine learning, experiments and molecular simulation. Separation and Purification Technology 2023, 306 , 122521. https://doi.org/10.1016/j.seppur.2022.122521
- Sanghun Park, Jaegyu Shim, Nakyung Yoon, Sungman Lee, Donggeun Kwak, Seungyong Lee, Young Mo Kim, Moon Son, Kyung Hwa Cho. Deep reinforcement learning in an ultrafiltration system: Optimizing operating pressure and chemical cleaning conditions. Chemosphere 2022, 308 , 136364. https://doi.org/10.1016/j.chemosphere.2022.136364
- Mengyuan Zhu, Jiawei Wang, Xiao Yang, Yu Zhang, Linyu Zhang, Hongqiang Ren, Bing Wu, Lin Ye. A review of the application of machine learning in water quality evaluation. Eco-Environment & Health 2022, 1
(2)
, 107-116. https://doi.org/10.1016/j.eehl.2022.06.001
- Hanan Ahmed Hosni Mahmoud, Nada Ali Hakami, Alaaeldin M. Hafez, . An Intelligent Deep Learning Model for CO2 Adsorption Prediction. Adsorption Science & Technology 2022, 2022 https://doi.org/10.1155/2022/8136302
- Yuquan Zhao, Jian Shen, Jimeng Feng, Zhitong Sun, Tianyang Sun, Decai Liu, Mansong Xi, Rui Li, Xinze Wang. The Estimation of Chemical Oxygen Demand of Erhai Lake Basin and Its Links with DOM Fluorescent Components Using Machine Learning. Water 2021, 13
(24)
, 3629. https://doi.org/10.3390/w13243629
- Ruixing Huang, Chengxue Ma, Jun Ma, Xiaoliu Huangfu, Qiang He. Machine learning in natural and engineered water systems. Water Research 2021, 205 , 117666. https://doi.org/10.1016/j.watres.2021.117666
Learn about these metrics
Article Views are the COUNTER-compliant sum of full text article downloads since November 2008 (both PDF and HTML) across all institutions and individuals. These metrics are regularly updated to reflect usage leading up to the last few days.
Citations are the number of other articles citing this article, calculated by Crossref and updated daily. Find more information about Crossref citation counts.
The Altmetric Attention Score is a quantitative measure of the attention that a research article has received online. Clicking on the donut icon will load a page at altmetric.com with additional details about the score and the social media presence for the given article. Find more information on the Altmetric Attention Score and how the score is calculated.
Recommended Articles
Machine Learning: New Ideas and Tools in Environmental Science and Engineering
Machine Learning Models for Inverse Design of the Electrochemical Oxidation Process for Water Purification
Machine Learning in Environmental Research: Common Pitfalls and Best Practices
Data-Driven Machine Learning in Environmental Pollution: Gains and Problems
Prediction of Soil Heavy Metal Immobilization by Biochar Using Machine Learning
Sponsored Content
Learn all you need to know about polymer chemistry through a novel combination of lectures and labs.
Abstract
Figure 1
Figure 1. (a) Time–concentration profiles of O3. Inset: time-dependent exposures of O3 during ozonation. (b) Time–concentration profiles of pCBA. Inset: time-dependent exposures of •OH during ozonation (Data shown for two selected water samples, no. 2 in Table S2 of the Supporting Information for natural water and no. 5 in Table S3 of the Supporting Information for wastewater effluent) ([O3]0 = 2.5 mg/L).
Figure 2
Figure 2. (a) O3 exposures and (b) •OH exposures (for the 60 water samples) as functions of DOC concentration ([O3]0 = 2.5 mg/L).
Figure 3
Figure 3. FEEM contour plots of (a) natural water and (b) wastewater effluent (the data shown for two selected water samples, no. 4 in Table S2 of the Supporting Information for natural water and no. 5 in Table S3 of the Supporting Information for wastewater effluent).
Figure 4
Figure 4. Plots of measured vs predicted O3 exposures for four RF models, (a) FEEM-Free, (b) FEEM-LowRes, (c) FEEM-HighRes, and (d) FEEM-FullRes ([O3]0 = 2.5 mg/L).
Figure 5
Figure 5. Plots of measured vs predicted •OH exposures for four RF models, (a) FEEM-Free, (b) FEEM-LowRes, (c) FEEM-HighRes, and (d) FEEM-FullRes ([O3]0 = 2.5 mg/L).
Figure 6
Figure 6. Plots of the measured vs predicted MP abatement for four RF models, (a) FEEM-Free, (b) FEEM-LowRes, (c) FFEM-HighRes, and (d) FEEM-FullRes (Raw data for measured values are presented in Figure S18 of the Supporting Information) ([MP]0 = 0.1 μM, [O3]0 = 2.5 mg/L).
References
This article references 37 other publications.
- 1Hoigné, J. Chemistry of Aqueous Ozone and Transformation of Pollutants by Ozonation and Advanced Oxidation Processes. In Quality and Treatment of Drinking Water II, Part of The Handbook of Environmental Chemistry (Part C: Water Pollution); Hrubec, J., Ed.; Springer: Berlin, 1998; Vol 5, pp 83– 141.There is no corresponding record for this reference.
- 2Lee, Y.; von Gunten, U. Advances in predicting organic contaminant abatement during ozonation of municipal wastewater effluent: reaction kinetics, transformation products, and changes of biological effects. Environ. Sci.: Water Res. Technol. 2016, 2, 421– 442, DOI: 10.1039/c6ew00025h2https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28Xjs1Crtrc%253D&md5=6a736f485739952184f79df8a3e7b0ccAdvances in predicting organic contaminant abatement during ozonation of municipal wastewater effluent: reaction kinetics, transformation products, and changes of biological effectsLee, Yunho; von Gunten, UrsEnvironmental Science: Water Research & Technology (2016), 2 (3), 421-442CODEN: ESWRAR; ISSN:2053-1419. (Royal Society of Chemistry)Ozonation of municipal wastewater effluent has been considered in recent years as an enhanced wastewater treatment technol. to abate trace org. contaminants (micropollutants). The efficiency of ozonation for micropollutant abatement depends on (1) the reactivity of ozone and OH radical (̇ OH) with the target micropollutant, (2) the dosage of ozone and the stability of ozone anḋ OH in a given water matrix, (3) the removal of undesirable effects (e.g., biol. activities) of a micropollutant after structural transformation, and (4) the biodegradability of transformation products in biol. post-treatment. In this article, recent advances in predicting org. micropollutant abatement during ozonation of municipal wastewater effluents are reviewed with a focus on (i) principle-based approaches for describing and modeling the reaction kinetics of ozone anḋ OH, (ii) transformation products and pathways, (iii) changes of biol. activities, and (iv) biodegrdn. of transformation products in biol. post-treatment. Using the chem. kinetics based on ozone anḋ OH rate consts. (i.e., compd.-specific information) and exposures (i.e., water matrix-specific information), a generalized prediction of the abatement efficiency of various micropollutants in varying water quality appears to be possible. QSAR-type correlations based on Hammett coeffs. or quantum chem. energy calcns. or (semi)empirical models have been developed for predicting the ozone anḋ OH rate consts. and exposures, resp. Models based on the ozone anḋ OH reaction rules can be used to predict the transformation products of micropollutants by ozone anḋ OH. Reaction rule-based models in combination with the chem. kinetics information will enable the prediction of transformation product evolution during ozonation. The biol. activities of transformation products have been assessed by an effect-driven approach using in vitro bioassays. Biol. activities with specific modes of action (e.g., receptor-binding activities) were found to be quite efficiently removed, upon slight structural modifications by ozone oṙ OH. The formation of new biol. activities has also been obsd., which warrants identification of the responsible toxicophore(s) and quant. exposure-based risk assessment. Finally, there is only limited exptl. information on the biodegradability of transformation products; however, biodegradability probability models can be used to make first ests. In future research, the discussed principle-based approaches can be more actively applied to det. and predict not only the abatement levels of the parent micropollutants but also the formation of transformation products and the consequent changes of biol. activities and biodegradability, which dets. the overall treatment efficiency.
- 3Glaze, W. H. Drinking-water treatment with ozone. Environ. Sci. Technol. 1987, 21, 224– 230, DOI: 10.1021/es00157a0013https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADyaL2sXht1GisLc%253D&md5=95b59f88ef96d58d8186fe6b4504c7c9Drinking-water treatment with ozoneGlaze, William H.Environmental Science and Technology (1987), 21 (3), 224-30CODEN: ESTHAG; ISSN:0013-936X.A review, with 39 refs., on the use of O3 in drinking-water treatment, including O3 generation, the transfer of O3 into water, O3 as a disinfectant and as an oxidant, the chem. of O3, the reactions of O3 with natural and synthetic org. substances, the combination of ozonization with catalysts and with other processes, O3 in biol. treatment, and economic aspects.
- 4Huber, M. M.; Canonica, S.; Park, G.-Y.; von Gunten, U. Oxidation of pharmaceuticals during ozonation and advanced oxidation processes. Environ. Sci. Technol. 2003, 37, 1016– 1024, DOI: 10.1021/es025896h4https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD3sXnslylsQ%253D%253D&md5=215a3c7ce8bf9c09525067770d26392fOxidation of Pharmaceuticals during Ozonation and Advanced Oxidation ProcessesHuber, Marc M.; Canonica, Silvio; Park, Gun-Young; von Gunten, UrsEnvironmental Science and Technology (2003), 37 (5), 1016-1024CODEN: ESTHAG; ISSN:0013-936X. (American Chemical Society)Oxidn. of pharmaceuticals during conventional ozonation and advanced oxidn. processes (AOP) in drinking water purifn. was studied. In a first step, second-order rate consts. for reactions of selected pharmaceuticals with O3 (kO3) and OH- (kOH) were detd. in bench-scale expts. (apparent kO3 at pH 7 and 20°): bezafibrate (590 ± 50/M-s), carbamazepine (∼3 × 105/M-s), diazepam (0.75 ± 0.15/M-s), diclofenac (∼1 × 106/M-s), 17α-ethinylestradiol (∼3 × 106/M-s), ibuprofen (9.6 ± 1.0/M-s), iopromide (<0.8/M-s), sulfamethoxazole (∼2.5 × 106/M-s), and roxithromycin (∼7 × 104/M-s). For 5 of the pharmaceuticals, apparent kO3 at pH 7 was >5 × 104/M-s, indicating these compds. are completely transformed during ozonation. KOH values were from 3.3 to 9.8 × 109/M-s. Compared to other important micro-pollutants, e.g., Me tert-Bu ether (MTBE) and atrazine, the selected pharmaceuticals reacted about 2-3 times faster with OH-. In the second part of the study, oxidn. kinetics of selected pharmaceuticals were examd. in ozonation expts. performed in different natural water. Second-order rate consts. detd. in pure aq. soln. could be applied to predict behavior of pharmaceuticals dissolved in natural water. Overall it was concluded that ozonation and AOP are promising processes to efficiently remove pharmaceuticals in drinking water.
- 5von Gunten, U. Ozonation of drinking water: Part I. Oxidation kinetics and product formation. Water Res. 2003, 37, 1443– 1467, DOI: 10.1016/S0043-1354(02)00457-85https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD3sXhtlGqsr4%253D&md5=a71315929c55e466b0f0f0a8908eebb3Ozonation of drinking water: Part I. Oxidation kinetics and product formationvon Gunten, UrsWater Research (2003), 37 (7), 1443-1467CODEN: WATRAG; ISSN:0043-1354. (Elsevier Science Ltd.)A review is given. The oxidn. of org. and inorg. compds. during ozonization can occur via ozone or OH radicals or a combination thereof. The oxidn. pathway is detd. by the ratio of ozone and OH radical concns. and the corresponding kinetics. A huge database with several hundred rate consts. for ozone and a few thousand rate consts. for OH radicals is available. Ozone is an electrophile with a high selectivity. The 2nd-order rate consts. for oxidn. by ozone vary over 10 orders of magnitude, <0.1-7 × 109/M-s. The reactions of ozone with drinking-water relevant inorg. compds. are typically fast and occur by an oxygen atom transfer reaction. Org. micropollutants are oxidized with ozone selectively. Ozone reacts mainly with double bonds, activated arom. systems and non-protonated amines. In general, electron-donating groups enhance the oxidn. by ozone whereas electron-withdrawing groups reduce the reaction rates. The kinetics of direct ozone reactions depend strongly on the speciation (acid-base, metal complexation). The reaction of OH radicals with the majority of inorg. and org. compds. is nearly diffusion-controlled. The degree of oxidn. by ozone and OH radicals is given by the corresponding kinetics. Product formation from the ozonation of org. micropollutants in aq. systems has only been established for a few compds.
- 6Schmidt, C. K.; Brauch, H.-J. N,N-dimethylsulfamide as precursor for N-nitrosodimethylamine (NDMA) formation upon ozonation and its fate during drinking water treatment. Environ. Sci. Technol. 2008, 42, 6340– 6346, DOI: 10.1021/es70304676https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD1cXksl2nsrw%253D&md5=3d7616b055a8d7ab157479e76ae0aa1eN,N-Dimethylsulfamide as Precursor for N-Nitrosodimethylamine (NDMA) Formation upon Ozonation and its Fate During Drinking Water TreatmentSchmidt, Carsten K.; Brauch, Heinz-JuergenEnvironmental Science & Technology (2008), 42 (17), 6340-6346CODEN: ESTHAG; ISSN:0013-936X. (American Chemical Society)Application and microbial degrdn. of the fungicide tolylfluanide gives rise to a new decompn. product named N,N-dimethylsulfamide (DMS). In Germany, DMS was found in groundwaters and surface waters with typical concns. in the range 100-1000 and 50-90 ng/L, resp. Lab.-scale and field studies concerning its fate during drinking water treatment showed that DMS cannot be removed via riverbank filtration, activated C filtration, flocculation, and oxidn. or disinfection procedures based on H2O2, K permanganate, ClO2, or UV irradn. Even nanofiltration does not provide a sufficient removal efficiency. During ozonization about 30-50% of DMS are converted to the carcinogenic N-nitrosodimethylamine (NDMA). The NDMA being formed is biodegradable and can at least partially be removed by subsequent biol. active drinking water treatment steps including sand or activated C filtration. Disinfection with hypochlorous acid converts DMS to so far unknown degrdn. products but not to NDMA or 1,1-dimethylhydrazine (UDMH).
- 7von Sonntag, C.; von Gunten, U. Chemistry of Ozone in Water and Wastewater Treatment: From Basic Principles to Applications; IWA Publishing: London, 2012.There is no corresponding record for this reference.
- 8Buxton, G. V.; Greenstock, C. L.; Helman, W. P.; Ross, A. B. Critical review of rate constants for reactions of hydrated electrons, hydrogen atoms and hydroxyl radicals (·OH/·O-) in aqueous solution. J. Phys. Chem. Ref. Data 1988, 17, 513, DOI: 10.1063/1.5558058https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADyaL1cXlvFyisLc%253D&md5=dae961496d7cfc2e28c3fcced28370f6Critical review of rate constants for reactions of hydrated electrons, hydrogen atoms and hydroxyl radicals (·OH/·O-) in aqueous solutionBuxton, George V.; Greenstock, Clive L.; Helman, W. Phillip; Ross, Alberta B.Journal of Physical and Chemical Reference Data (1988), 17 (2), 513-886CODEN: JPCRBU; ISSN:0047-2689.Kinetic data for the radicals H and OH in aq. soln., and the corresponding radical anions, O- and eaq-, are critically reviewed with many refs. Reactions of the radicals in aq. soln. have been studied by pulse radiolysis, flash photolysis, and other methods. Rate consts. for >3,500 reactions are tabulated, including reactions with mols., ions, and other radicals derived from inorg. and org. solutes.
- 9von Gunten, U.; Hoigné, J. Bromate formation during ozonation of bromide-containing waters: Interaction of ozone and hydroxyl radical reactions. Environ. Sci. Technol. 1994, 28, 1234– 1242, DOI: 10.1021/es00056a0099https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADyaK2cXktFOnu7s%253D&md5=079e49017a46ac0a97a4251b30e797c5Bromate Formation during Ozonization of Bromide-Containing Waters: Interaction of Ozone and Hydroxyl Radical Reactionsvon Gunten, Urs; Hoigne, JuergEnvironmental Science and Technology (1994), 28 (7), 1234-42CODEN: ESTHAG; ISSN:0013-936X.Kinetic simulations have been tested by lab. expts. to evaluate the major factors controlling bromate formation during ozonization of waters contg. Br-e. In the presence of an org. scavenger for OH radicals, bromate formation can be accurately predicted by the mol. O3 mechanism using published reaction rate data, even for waters contg. ammonium. In the absence of scavengers, OH radical reactions contribute significantly to bromate formation. CO32-, produced by the oxidn. of HCO3- with OH radicals, oxidize the intermediate hypobromite to bromite, which is further oxidized by O3 to bromate. During drinking water ozonization, mol. O3 controls both the initial oxidn. of bromide and the final oxidn. of bromite. OH radical reactions contribute to the oxidn. of the intermediate oxybromine species. Bromate formation in advanced oxidn. processes can be explained by a synergism of O3 and OH radicals.
- 10Elovitz, M. S.; von Gunten, U. Hydroxyl radical/ozone ratios during ozonation processes. I. The Rct concept. Ozone: Sci. Eng. 1999, 21, 239– 260, DOI: 10.1080/0191951990854723910https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADyaK1MXksF2rsb0%253D&md5=0bbbec53b7aedf164e7f11b716742299Hydroxyl radical/ozone ratios during ozonation processes. I. The Rct conceptElovitz, Michael S.; Von Gunten, UrsOzone: Science & Engineering (1999), 21 (3), 239-260CODEN: OZSEDS; ISSN:0191-9512. (Lewis Publishers)The ozonization of model systems and several natural waters was examd. in bench-scale batch expts. In addn. to measuring the concn. of O3, the rate of depletion of an in situ hydroxyl radical probe compd. was monitored, thus providing information on the transient steady-state concn. of hydroxyl radicals (•OH). A new parameter, Rct, representing the ratio of the •OH-exposure to the O3-exposure was calcd. as a function of reaction time. For most waters tested, including pH-buffered model systems and natural waters, Rct was a const. value for the majority of the reaction. Therefore, Rct corresponds to the ratio of the •OH concn. to the O3 concn. in a given water. For a given water source, the degrdn. of a micropollutant (e.g. atrazine) via O3 and •OH reaction pathways can be predicted by the O3 reaction kinetics and Rct.
- 11Kwon, M.; Kye, H.; Jung, Y.; Yoon, Y.; Kang, J.-W. Performance characterization and kinetic modeling of ozonation using a new method: ROH,O3 concept. Water Res. 2017, 122, 172– 182, DOI: 10.1016/j.watres.2017.05.06211https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXps1Knt7o%253D&md5=c042a3a40fc51d8ede841b4f45d57c29Performance characterization and kinetic modeling of ozonation using a new method: ROH,O3 conceptKwon, Minhwan; Kye, Homin; Jung, Youmi; Yoon, Yeojoon; Kang, Joon-WunWater Research (2017), 122 (), 172-182CODEN: WATRAG; ISSN:0043-1354. (Elsevier Ltd.)Ozonation is an effective treatment for removing various org. pollutants from aquatic systems. The Rct concept, which is defined as the ratio of ·OH exposure to O3 exposure, has been widely used to predict the removal efficiency of target compds., but it has significant variations by water temp. and initial O3 dose which are crucial parameters in drinking water plant. The ROH,O3 concept, which is defined as the ·OH exposure by O3 consumption, was proposed as a kinetic parameter for characterization and kinetic modeling for ozonation. The ROH,O3 concept is independent of temp. and initial O3 dose. A higher ROH,O3 value indicates a higher ·OH formation when the same amt. of O3 is consumed in different water samples; therefore, the ·OH yield from O3 decompn. of the water samples can be compared using the ROH,O3 values. The ROH,O3 concept can also be used to characterize and model the initial ozone demand phase, and it is more convenient method compared to Rct concept. Using the ROH,O3 concept, the dynamic O3 and ·OH kinetics and the removal efficiencies of iopromide and ibuprofen were well predicted (R2 = 0.98) over a wide range of exptl. conditions (n = 124).
- 12Kim, M. S.; Cha, D.; Lee, K.-M.; Lee, H.-J.; Kim, T.; Lee, C. Modeling of ozone decomposition, oxidant exposures, and the abatement of micropollutants during ozonation processes. Water Res. 2020, 169, 115230, DOI: 10.1016/j.watres.2019.11523012https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1MXitVygt7rE&md5=f8dd7b0530203b03085f3cd59fde44f0Modeling of ozone decomposition, oxidant exposures, and the abatement of micropollutants during ozonation processesKim, Min Sik; Cha, Dongwon; Lee, Ki-Myeong; Lee, Hye-Jin; Kim, Taewan; Lee, ChanghaWater Research (2020), 169 (), 115230CODEN: WATRAG; ISSN:0043-1354. (Elsevier Ltd.)This study demonstrates new empirical models to predict the decompn. of ozone (O3) and the exposures of oxidants (i.e., O3 and hydroxyl radical, ·OH) during the ozonation of natural waters. Four models were developed for the instantaneous O3 demand, first-order rate const. for the secondary O3 decay, O3 exposure (∫[O3]dt), and ·OH exposure ((∫[·OH]dt)), as functions of five independent variables, namely the O3 dose, concn. of dissolved org. carbon (DOC), pH, alky., and temp. The models were derived by polynomial regression anal. of exptl. data obtained by controlling variables in natural water samples from a single source water (Maegok water in Korea), and they exhibited high accuracies for regression (R2 = 0.99 for the three O3 models, and R2 = 0.96 for the ·OH exposure model). The three O3 models exhibited excellent internal validity for Maegok water samples of different conditions (that were not used for the model development). They also showed acceptable external validity for seven natural water samples collected from different sources (not Maegok water); the IOD model showed somewhat poor external validity. The models for oxidant exposures were successfully used to predict the abatement of micropollutants by ozonation; the model predictions showed high accuracy for Maegok water, but not for the other natural waters.
- 13Mobed, J. J.; Hemmingsen, S. L.; Autry, J. L.; McGown, L. B. Fluorescence characterization of IHSS humic substances: Total luminescence spectra with absorbance correction. Environ. Sci. Technol. 1996, 30, 3061– 3065, DOI: 10.1021/es960132l13https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADyaK28Xlt1eksrg%253D&md5=050cc17a151f5611131eadbbc100001dFluorescence Characterization of IHSS Humic Substances: Total Luminescence Spectra with Absorbance CorrectionMobed, Jarafshan J.; Hemmingsen, Sherry L.; Autry, Jennifer L.; McGown, Linda B.Environmental Science and Technology (1996), 30 (10), 3061-3065CODEN: ESTHAG; ISSN:0013-936X. (American Chemical Society)Total luminescence spectroscopy was applied to the fluorescence characterization of humic substances obtained from the International Humic Substances Society (IHSS). Total luminescence spectra, represented as excitation-emission matrixes (EEMs), may be used to discriminate between soil-derived and aquatic-derived IHSS humic substances and between humic and fulvic acids derived from the same source (soil or aquatic). Ionic strength at 0-1 M KCl and humic substance concn. at 5-100 mg/L had little effect on the fluorescence spectral characteristics of the humic substances, while pH had significant effects as expected. Absorbance correction is essential for accurate representation and comparison of the EEMs of the humic substances at high concns.
- 14Baker, A. Fluorescence excitation-emission matrix characterization of some sewage-impacted rivers. Environ. Sci. Technol. 2001, 35, 948– 953, DOI: 10.1021/es000177t14https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD3MXmtFyrtw%253D%253D&md5=743056946850c6b9d3a8173a4e9b4d60Fluorescence Excitation-Emission Matrix Characterization of Some Sewage-Impacted RiversBaker, AndyEnvironmental Science and Technology (2001), 35 (5), 948-953CODEN: ESTHAG; ISSN:0013-936X. (American Chemical Society)Fluorescence excitation-emission matrix (EEM) spectrophotometry was applied to 10 sample sites in 6 rivers in northeastern England, some of which were polluted by sewage treatment works (STW) discharges, to study whether STW discharges had a significantly distinct fluorescence signature. Upstream, downstream, and STW discharge samples for 2 STW demonstrated that treated sewage had a distinct fluorescence EEM, with high tryptophan and fulvic-like fluorescence intensities of approx. equal ratio. This signature was obsd. downstream samples. When all 10 sample sites were compared, 2 trend lines were apparent where STW impacted rivers plotted sep. from the other sample sites. Fluorescence EEM signatures were compared to absorption at 254 nm and demonstrated to provide a better fingerprint of sewage-impacted water. It is suggested that fluorescence EEM spectrophotometry can provide a useful tool to analyze grab samples collected for routine and investigative monitoring and has the potential for online monitoring of STW impacts on river systems.
- 15Yu, H.; Qu, F.; Zhang, X.; Shao, S.; Rong, H.; Liang, H.; Bai, L.; Ma, J. Development of correlation spectroscopy (COS) method for analyzing fluorescence excitation emission matrix (EEM): A case study of effluent organic matter (EfOM) ozonation. Chemosphere 2019, 228, 35– 43, DOI: 10.1016/j.chemosphere.2019.04.11915https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1MXnvFGrtLo%253D&md5=f53e82581db1c0735d9edef10c3ea211Development of correlation spectroscopy (COS) method for analyzing fluorescence excitation emission matrix (EEM): A case study of effluent organic matter (EfOM) ozonationYu, Huarong; Qu, Fangshu; Zhang, Xiaolei; Shao, Senlin; Rong, Hongwei; Liang, Heng; Bai, Langming; Ma, JunChemosphere (2019), 228 (), 35-43CODEN: CMSHAF; ISSN:0045-6535. (Elsevier Ltd.)Two-dimensional correlation spectroscopy has been used as a powerful tool for analyzing spectral features, but it has never been applied to fluorescence excitation-emission matrix (EEM) data due to the incompatible dimensions. This study first investigated EEM-COS by reducing the dimensions of the EEM (using parallel factor anal., PARAFAC) for fitting to 2DCOS (EEM-PARAFAC-COS). The fluorescence changes of effluent org. matter during ozonation were studied using EEM-COS and synchronous fluorescence (SF)-2DCOS. The conventionally used SF-2DCOS proved to be biased due to the intrinsic drawback of SF, while the EEM-PARAFAC-COS gave accurate and trustworthy results. Homo-EEM-PARAFAC-COS indicated that the fluorescence protein-like and fulvic-like substances in EfOM were preferentially ozonated compared to humic-like substances. Hetero-EEM-PARAFAC-COS analyses on the EEM, FTIR, UV-vis absorbance, and size-exclusion chromatog. showed that the fluorescence protein-like and fulvic-like substances in EfOM were assocd. with lower mol. wt. (MW, ∼0.95 kDa), UV absorbance at ∼280 nm, and more electron-enriched aroms. (with amide and phenolic groups), which explained their ozonation preference, while humic-like substances were related to carboxylic groups, UV absorbance at ∼255 nm, and orgs. at MW of ∼4.50 kDa. This work demonstrated the great potential of EEM-PARAFAC-COS in studying fluorescence change and correlating fluorescence with other spectra.
- 16Cui, Y.; Yu, J.; Su, M.; Jia, Z.; Liu, T.; Oinuma, G.; Yamauchi, T. Humic acid removal by gas-liquid interface discharge plasma: performance, mechanism and comparison to ozonation. Environ. Sci.: Water Res. Technol. 2019, 5, 152– 160, DOI: 10.1039/c8ew00520f16https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXit1eitr7P&md5=c43f29a4d57e93d83bc501d3ab478531Humic acid removal by gas-liquid interface discharge plasma: performance, mechanism and comparison to ozonationCui, Yanyan; Yu, Jianwei; Su, Ming; Jia, Zeyu; Liu, Tingting; Oinuma, Gaku; Yamauchi, TokikoEnvironmental Science: Water Research & Technology (2019), 5 (1), 152-160CODEN: ESWRAR; ISSN:2053-1419. (Royal Society of Chemistry)The removal of natural org. matter (NOM) is a major objective for drinking water treatment. In this study, a novel advanced oxidn. process (AOP) based on plasma in gas-liq. interface discharge was investigated for removing humic acid (HA), and ozonation treatment effect was evaluated as a comparison. Several indexes including UV254, dissolved org. carbon (DOC), specific UV absorbance (SUVA), and excitation-emission matrix (EEM) fluorescence spectra were employed to characterize HA transformation. The results showed that typical species including ozone, hydrogen peroxide and HȮ were produced in the plasma process, which made better removal efficiency than ozonation. Esp., UV254 was removed effectively by 96.5% and 90.3% within 15 min by plasma and ozonation, while DOC was 49.2% and 41.1%, resp. HȮ radicals were majorly responsible for the degrdn. of HA in the plasma process, which were mainly generated via reaction of in situ produced ozone and H2O2 during the discharge process. UV/vis absorbance and EEM results further showed that HA with the large mol. wt. fraction was decompd. to small mol. wt. fraction. This study indicates that plasma technol. has a significant application prospect as a replacement for ozone in drinking water industry.
- 17Li, L.; Rong, S.; Wang, R.; Yu, S. Recent advances in artificial intelligence and machine learning for nonlinear relationship analysis and process control in drinking water treatment: A review. Chem. Eng. J. 2021, 405, 126673, DOI: 10.1016/j.cej.2020.12667317https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXhs1Cqu7bM&md5=7b0f1ca10517b5a26a4aed1e2213bf6dRecent advances in artificial intelligence and machine learning for nonlinear relationship analysis and process control in drinking water treatment: A reviewLi, Lei; Rong, Shuming; Wang, Rui; Yu, ShuiliChemical Engineering Journal (Amsterdam, Netherlands) (2021), 405 (), 126673CODEN: CMEJAJ; ISSN:1385-8947. (Elsevier B.V.)A review. Because of its robust autonomous learning and ability to address complex problems, artificial intelligence (AI) has increasingly demonstrated its potential to solve the challenges faced in drinking water treatment (DWT). AI technol. provides tech. support for the management and operation of DWT processes, which is more efficient than relying solely on human operations. AI-based data anal. and evolutionary learning mechanisms are capable of realizing water quality diagnosis, autonomous decision making and operation process optimization and have the potential to establish a universal process anal. and predictive model platform. This review briefly introduces AI technologies that are widely used in DWT. Moreover, this paper reviews in detail the mature applications and latest discoveries of AI and machine learning technologies in the fields of source water quality, coagulation/flocculation, disinfection and membrane filtration, including source water contaminant monitoring and identification, accurate and efficient prediction of coagulation dosage, anal. of the formation of disinfection byproducts and advanced control of membrane fouling. Finally, the challenges facing AI technologies and the issues that need further study are discussed; these challenges can be briefly summarized as (a) obtaining more effective characterization data to screen and identify targeted contaminants in the complex background with the assistance of AI technologies and (b) establishing a macro intelligence model and decision scheme for entire drinking water treatment plants (DWTPs) to support the management of the water supply system.
- 18American Public Health Association (APHA), Standard Methods for the Examination of Water and Wastewater; Greenberg, A. E., Clesceri, L. S., Eaton, A. D., Eds.; APHA: Washington, DC, 1992.There is no corresponding record for this reference.
- 19Chen, W.; Westerhoff, P.; Leenheer, J. A.; Booksh, K. Fluorescence excitation-emission matrix regional integration to quantify spectra for dissolved organic matter. Environ. Sci. Technol. 2003, 37, 5701– 5710, DOI: 10.1021/es034354c19https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD3sXovFeisrc%253D&md5=6fc1ca4c33117f277bf6d01dd21c9f5fFluorescence Excitation-Emission Matrix Regional Integration to Quantify Spectra for Dissolved Organic MatterChen, Wen; Westerhoff, Paul; Leenheer, Jerry A.; Booksh, KarlEnvironmental Science and Technology (2003), 37 (24), 5701-5710CODEN: ESTHAG; ISSN:0013-936X. (American Chemical Society)Excitation-emission matrix (EEM) fluorescence spectroscopy has been widely used to characterize dissolved org. matter (DOM) in water and soil. However, interpreting the >10,000 wavelength-dependent fluorescence intensity data points represented in EEMs has posed a significant challenge. Fluorescence regional integration, a quant. technique that integrates the vol. beneath an EEM, was developed to analyze EEMs. EEMs were delineated into five excitation-emission regions based on the fluorescence of model compds., DOM fractions, and marine waters or fresh waters. Volumetric integration under the EEM within each region, normalized to the projected excitation-emission area within that region and dissolved org. carbon concn., resulted in a normalized region-specific EEM vol. (Φi,n). Solid-state carbon NMR (13C NMR), Fourier transform IR (FTIR) anal., UV-visible absorption spectra, and EEMs were obtained for std. Suwannee River fulvic acid and 15 hydrophobic or hydrophilic acid, neutral, and base DOM fractions plus nonfractionated DOM from wastewater effluents and rivers in the southwestern USA. DOM fractions fluoresced in one or more EEM regions. The highest cumulative EEM vol. (ΦT,n = ΣΦi,n) was obsd. for hydrophobic neutral DOM fractions, followed by lower ΦT,n values for hydrophobic acid, base, and hydrophilic acid DOM fractions, resp. An extd. wastewater biomass DOM sample contained arom. protein- and humic-like material and was characteristic of bacterial sol. microbial products. Arom. carbon and the presence of specific arom. compds. (as indicated by solid-state 13C NMR and FTIR data) resulted in EEMs that aided in differentiating wastewater effluent DOM from drinking water DOM.
- 20Bader, H.; Hoigné, J. Determination of ozone in water by the indigo method. Water Res. 1981, 15, 449– 456, DOI: 10.1016/0043-1354(81)90054-320https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADyaL3MXks1Olu78%253D&md5=72c0f23814ddfe196794a5ede4b84532Determination of ozone in water by the indigo methodBader, H.; Hoigne, J.Water Research (1981), 15 (4), 449-56CODEN: WATRAG; ISSN:0043-1354.O3 was detd. in natural water by the decolorization of K indigotrisulfonate [67627-18-3] at 600 nm (pH <4). The method is stoichiometric and fast. The change of absorbance vs. O3 added is -2.0 × 104/mol-cm and is independent of the concn. of aq. O3 at 0.005-30 mg O3/L. The precision is 2% for low concns. Visual detection can be used to det. 0.01 mg O3/L. The reagent soln. is stable for 3 mo. For example, for a sample of ozonized lake water, 0.60 mg O3/L was detd. by this method, compared with 0.61 mg/L by a direct UV method.
- 21Breiman, L.; Friedman, J. H.; Olshen, R. A.; Stone, C. J. Classification and Regression Trees; Chapman and Hall/CRC: Boca Raton, FL, 1984.There is no corresponding record for this reference.
- 22Ließ, M.; Glaser, B.; Huwe, B. Uncertainty in the spatial prediction of soil texture. Geoderma 2012, 170, 70– 79, DOI: 10.1016/j.geoderma.2011.10.010There is no corresponding record for this reference.
- 23Breiman, L. Random Forests. Mach. Learn. 2001, 45, 5– 32, DOI: 10.1023/A:1010933404324There is no corresponding record for this reference.
- 24Baek, S.-S.; Choi, Y.; Jeon, J.; Pyo, J.; Park, J.; Cho, K. H. Replacing the internal standard to estimate micropollutants using deep and machine learning. Water Res. 2021, 188, 116535, DOI: 10.1016/j.watres.2020.11653524https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXit1OmsrzI&md5=80bfa049559c39de2de6a17691be9833Replacing the internal standard to estimate micropollutants using deep and machine learningBaek, Sang-Soo; Choi, Younghun; Jeon, Junho; Pyo, JongCheol; Park, Jongkwan; Cho, Kyung HwaWater Research (2021), 188 (), 116535CODEN: WATRAG; ISSN:0043-1354. (Elsevier Ltd.)Similar to the worldwide proliferation of urbanization, micropollutants have been involved in aquatic and ecol. environmental systems. These pollutants have the propensity to wreak havoc on human health and the ecol. system; hence, it is important to persistently monitor micropollutants in the environment. Micropollutants are commonly quantified via target anal. using high resoln. mass spectrometry and the stable isotope labeled (SIL) std. However, the cost-intensiveness of this std. presents a major obstacle in measuring micropollutants. This study resolved this problem by developing data-driven models, including deep learning (DL) and machine learning (ML), to est. the concn. of micropollutants without resorting to the SIL std. Our study hypothesized that natural org. matter (NOM) could replace internal stds. if there was a specific mass spectrum (MS) subset, including NOM information, which correlated with an SIL std. peak. Therefore, we analyzed the MS to find the specific MS subsets for replacing the SIL std. peak. Thirty-five alternative MS subsets were detd. for applying DL and ML as input data. Thereafter, we trained four different DL models, namely, ResNet101, GoogLeNet, VGG16, and Inception v3, as well as three different ML models, i.e., random forest (RF), support vector machine (SVM), and artificial neural network (ANN). A total of 680 MS data were used for the model training to est. five different micropollutants, namely Sulpiride, Metformin, and Benzotriazole. Among the DL models, ResNet 101 exhibited the highest model performance, showing that the av. validation R2 and MSE were 0.84 and 0.26 ng/L, resp., while RF was the best in the ML models, manifesting R2 and MSE values of 0.69 and 0.58 ng/L. The trained models showed accurate training and validation results for the estn. of the five micropollutant concns. Therefore, this study demonstrates that the suggested anal. has a potential for alternative micropollutant measurement that has rapid and economic vantages.
- 25Sevgen, E.; Kocaman, S.; Nefeslioglu, H. A.; Gokceoglu, C. A novel performance assessment approach using photogrammetric techniques for landslide susceptibility mapping with logistic regression, ANN and Random Forest. Sensors 2019, 19, 3940, DOI: 10.3390/s19183940There is no corresponding record for this reference.
- 26Singh, B.; Sihag, P.; Singh, K. Modelling of impact of water quality on infiltration rate of soil by random forest regression. Model. Earth Syst. Environ. 2017, 3, 999– 1004, DOI: 10.1007/s40808-017-0347-3There is no corresponding record for this reference.
- 27Akaike, H. Information Theory and an Extension of the Maximum Likelihood Principle. In Selected Papers of Hirotugu Akaike; Parzen, E., Tanabe, K., Kitagawa, G., Eds.; Springer: New York, NY, 1998.There is no corresponding record for this reference.
- 28Snipes, M.; Taylor, D. C. Model selection and Akaike information criteria: An example from wine ratings and prices. Wine Econ. Policy 2014, 3, 3– 9, DOI: 10.1016/j.wep.2014.03.001There is no corresponding record for this reference.
- 29Buffle, M.-O.; von Gunten, U. Phenols and amine induced HO· generation during the initial phase of natural water ozonation. Environ. Sci. Technol. 2006, 40, 3057– 3063, DOI: 10.1021/es052020c29https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD28XivF2qsbk%253D&md5=3e9a3cc595c6b54acba2901b7aa621bbPhenols and Amine Induced HO• Generation During the Initial Phase of Natural Water OzonationBuffle, Marc-Olivier; von Gunten, UrsEnvironmental Science & Technology (2006), 40 (9), 3057-3063CODEN: ESTHAG; ISSN:0013-936X. (American Chemical Society)The initial phase of ozone decompn. in natural water (t < 20 s) is poorly understood. It has recently been shown to result in very high transient HO• concns. and, thereby, plays an essential role during processes such as bromate formation or contaminants oxidn. Phenols and amines are ubiquitous moieties of natural org. matter. Naturally occurring concns. of primary, secondary, and tertiary amines, amino acids, and phenol were added to surface water, and ozone decompn. as well as HO• generation were measured starting 350 ms after ozone addn. Six seconds into the process, 5 μM of dimethylamine and phenol had generated ∫HO•dt = 1 × 10-10 M·s and 1.8 × 10-10 M·s, resp. With 10 μM dimethylamine and 1.5 mgO3/L, Rct, (∫HO•dt/∫O3dt) reached 10-6, which is larger than in advanced oxidn. processes (AOP) such as O3/H2O2. Expts. in the presence of HO•-scavengers indicated that a significant fraction of phenol-induced ozone decompn. and HO• generation results from a direct electron transfer to ozone. For dimethylamine, the main mechanism of HO• generation is direct formation of O2•- which reacts selectively with O3 to form O3•-. Pretreatment of phenol-contg. water with HOCl or HOBr did not decrease HO• generation, while the same treatment of dimethylamine-contg. water considerably reduced HO• generation.
- 30Neta, P.; Dorfman, L. M. Pulse Radiolysis Studies. XIII. Rate Constants for the Reaction of Hydroxyl Radicals with Aromatic Compounds in Aqueous Solutions. In Radiation Chemistry Volume I. Aqueous Media, Biology, Dosimetry; Hart, E. J., Ed.; American Chemical Society: Washington, DC, 1968.There is no corresponding record for this reference.
- 31Determann, S.; Reuter, R.; Wagner, P.; Willkomm, R. Fluorescent matter in the eastern Atlantic Ocean. Part 1: Method of measurement and near-surface distribution. Deep Sea Res., Part I 1994, 41, 659– 675, DOI: 10.1016/0967-0637(94)90048-531https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADyaK2MXhslWgtbg%253D&md5=cd37e273cd7acc7c086fc7e52d11c71eFluorescent matter in the eastern Atlantic Ocean. Part 1: method of measurement and near-surface distributionDetermann, S.; Reuter, R.; Wagner, P.; Willkomm, R.Deep-Sea Research, Part I: Oceanographic Research Papers (1994), 41 (4), 659-75CODEN: DRORE7; ISSN:0967-0637.Fluorescence spectra of org. matter in seawater were measured during a cruise through the South and North Atlantic from Capetown, South Africa, to Bremerhaven, Germany. The data are calibrated by normalization to the water Raman scatter band, which allows their quantification without the need of fluorescence stds. Spectral structures are found which can be related to tryptophan- and tyrosine-like substances and to gelbstoff. Their distribution in the eastern Atlantic is discussed and compared with other hydrog. parameters.
- 32Ahmad, S. R.; Reynolds, D. M. Monitoring of water quality using fluorescence technique: Prospect of on-line process control. Water Res. 1999, 33, 2069– 2074, DOI: 10.1016/S0043-1354(98)00435-732https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADyaK1MXjslGjs7w%253D&md5=445b62eb42cc06068aa81ae33a336a6dMonitoring of water quality using fluorescence technique: prospect of on-line process controlAhmad, S. R.; Reynolds, D. M.Water Research (1999), 33 (9), 2069-2074CODEN: WATRAG; ISSN:0043-1354. (Elsevier Science Ltd.)The wastewater from sewage processing plants shows characteristic fluorescence signatures when excited by UV light in the 240-300 nm wavelength band. A typical signature is a spectrum having a broad band centered at about 350 nm and two relatively less intense bands centered at about 390 and 430 nm. Samples of settled sewage, treated in an aerobic digester, show a substantial redn. of the intensity of the 350 nm band and comparatively much smaller redn. of the strength of the other two bands. The biodegradable chromophoric constituent species are, therefore, considered to be the major contributors to the overall fluorescence within this band. The intensity of this band has been found to have a good correlation with the BOD parameter. This parameter is universally used for assessing the sewage strength and the suitability of the treated effluent for discharge into rivers or reservoirs. Therefore, the fluorescence technique is considered to have the potential for use in noninvasive continuous water quality monitoring thereby, enabling online process control in sewage treatment plants. However, fluorescence strength is affected by the pH of the sample, particularly at higher values. This has to be taken into account for the practical utilization of this technique.
- 33Mounier, S.; Patel, N.; Quilici, L.; Benaim, J. Y.; Benamou, C. Fluorescence 3D de la matière organique dissoute du fleuve amazone. Water Res. 1999, 33, 1523– 1533, DOI: 10.1016/S0043-1354(98)00347-933https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADyaK1MXitlehtLw%253D&md5=9ba28abc2bb9eb3272db13638b1879afThree-dimensional fluorescence of the dissolved organic carbon in the Amazon RiverMounier, S.; Patel, N.; Quilici, L.; Benaim, J. Y.; Benamou, C.Water Research (1999), 33 (6), 1523-1533CODEN: WATRAG; ISSN:0043-1354. (Elsevier Science Ltd.)Natural org. matter is an important pool that is not yet totally described. Two types of compds. are found: some chem. well characterized mols. (biopolymers) and uncharacterized humic substances (geopolymers). The spectroscopic properties of this pool of org. matter have recently been advanced by the excitation emission fluorescence matrix (EEFM) characterization of marine and terrestrial DOM in seawater using excitation-emission matrix spectroscopy. Three types of fluorophores are described by their excitation/emission wavelength at max. intensity (λex/λem)max: the tyrosine and tryptophan like structures, not examd. here, and the humic like fluorescent structures of type A (λ260/λ445) and type C (λ330/λ445). The EEFM applied to sequential tangential ultrafiltered (UFTS) Amazonian fresh waters give spectroscopic information on the fluorescent properties of particulate ( > 0.22 μm), colloidal and dissolved ( < 5 kDa) org. matter. Chromophores A and C are present in all sized fraction samples. Their (λex/λem)max are in the same domains as those of terrestrial humic substances. Differentiation of the type A and type C peaks in 3D diagrams are based on their (λex/λem)max position and the Ia/Ic ratio. Differences are obsd. between humic material extd. by hydrophobic resins or concd. sample from tangential sequential ultrafiltration (UFTS). Spectroscopic properties of the humic material are not modified by the ultrafiltration process. A particular attention is given on the differentiation between black water (Rio Negro River) and white water (Rio Solimoes and Rio Madeira River). Black waters are generally known as humic rich and low mineral content waters. Humic like fluorescent compds. of type C are preferentially retained by membranes with 5 kDa cut off.; these compds. have the larger mol. wt. In the presence of Cu cation, the type C compds. are divided in 2 groups according to their (λex/λem)max: the 1st one is invariant, the other one expands (20 nm) its efficient excitation wavelength domain. Photochem. reactions induced by UV irradn. also lead to a 20-nm expansion of the efficient excitation wavelength domain. From these spectroscopic and mol. wt. complementary data, it is proposed that A type fluorophores are close to fulvic acids while C fluorophores seems to be more related to humic acid.
- 34Coble, P. G. Characterization of marine and terrestrial DOM in seawater using excitation-emission matrix spectroscopy. Mar. Chem. 1996, 51, 325– 346, DOI: 10.1016/0304-4203(95)00062-334https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADyaK28XnslWltg%253D%253D&md5=e4652be4c0e7edce63657af49497b93cCharacterization of marine and terrestrial DOM in seawater using excitation-emission matrix spectroscopyCoble, Paula G.Marine Chemistry (1996), 51 (4), 325-46CODEN: MRCHBD; ISSN:0304-4203. (Elsevier)High-resoln. fluorescence spectroscopy was used to characterize dissolved org. matter (DOM) in concd. and unconcd. water samples from a wide variety of freshwater, coastal and marine environments. Several types of fluorescent signals were obsd., including humic-like, tyrosine-like, and tryptophan-like. Humic-like fluorescence consisted of two peaks, one stimulated by UV excitation (peak A) and one by visible excitation (peak C). For all samples, the positions of both excitation and emission maxima for peak C were dependent upon wavelength of observation, with a shift towards longer wavelength emission max. at longer excitation wavelength and longer wavelength excitation max. at longer emission wavelength. A trend was obsd. in the position of wavelength-independent max. fluorescence (Exmax/Emmax) for peak C, with max. at shorter excitation and emission wavelengths for marine samples than for freshwater samples. Differences suggest that the humic material in marine surface waters is chem. different from humic material in the other environments sampled. These results explain previous conflicting reports regarding fluorescence properties of DOM from natural waters and also provide a means of distinguishing between water mass sources in the ocean.
- 35Singh, R. Membrane Technology and Engineering for Water Purification; Butterworth-Heinemann: Oxford, U.K., 2015.There is no corresponding record for this reference.
- 36Chun, K. C.; Chang, R. W.; Williams, G. P.; Chang, Y. S.; Tomasko, D.; LaGory, K.; Ditmars, J.; Chun, H. D.; Lee, B.-K. Water quality issues in the Nakdong River basin in the Republic of Korea. Environ. Eng. Policy 1999, 2, 131– 143, DOI: 10.1007/s100220000024There is no corresponding record for this reference.
- 37Nam, S.-N.; Amy, G. Differentiation of wastewater effluent organic matter (EfOM) from natural organic matter (NOM) using multiple analytical techniques. Water Sci. Technol. 2008, 57, 1009– 1015, DOI: 10.2166/wst.2008.16537https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD1cXntV2hur4%253D&md5=c2a8cd6adfeba15b5f966ab8615ee576Differentiation of wastewater effluent organic matter (EfOM) from natural organic matter (NOM) using multiple analytical techniquesNam, Seong-Nam; Amy, GaryWater Science and Technology (2008), 57 (7), 1009-1015CODEN: WSTED4; ISSN:0273-1223. (IWA Publishing)Using three anal. techniques of size exclusion chromatog. (SEC), fluorescence excitation-emission matrix (EEM), and dissolved org. nitrogen (DON) measurement, differentiating characteristics of effluent org. matter (EfOM) from natural org. matter (NOM) have been investigated. SEC reveals a wide range of mol. wt. (MW) for EfOM and high amt. of high MW polysaccharides, and low MW org. acids compared to NOM. Clear protein-like peaks using fluorescence EEM were a major feature of EfOM distinguishing it from NOM. Fluorescence index (FI), an indicator to distinguish autochthonous origin from allochthonous origin, differentiated EfOM from NOM by exhibiting higher values, indicating a microbial origin. In EfOM samples, DON present in higher amts. than NOM.
Supporting Information
Supporting Information
The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.est.0c05836.
Additional details, hyperparameters, and development procedures of the RF model, comparison of RF and ridge regression models, variable importance for FEEM-Free and FEEM-HighRes models, HPLC operating conditions, oxidant exposure values, water characteristic values, percent fluorescence response values, regression performance and AIC values, second-order rate constant values, river water sampling sites, time-dependent concentration profiles of ozone, time-dependent concentration profiles of pCBA, time-dependent exposure profiles of ozone, time-dependent exposure profiles of •OH, oxidant exposures as functions of pH and alkalinity, FEEM contour plots, division of FRI regions, comparison of oxidant exposures for RF and ridge regression models, variable importance for the prediction of oxidant exposures, and MP abatement experiment results (PDF)
Terms & Conditions
Most electronic Supporting Information files are available without a subscription to ACS Web Editions. Such files may be downloaded by article for research use (if there is a public use license linked to the relevant article, that license may permit other uses). Permission may be obtained from ACS for other uses through requests via the RightsLink permission system: http://pubs.acs.org/page/copyright/permissions.html.
