Use of multiple regression models for predicting the formation of bromoform and dibromochloromethane during ballast water treatment based on an advanced oxidation process

Abstract

Disinfection byproducts (DBPs) generated by ballast water treatment have become a concern worldwide because of their potential threat to the marine environment. Predicting the relative DBP concentrations after disinfection could enable better control of DBP formation. However, there is no appropriate method of evaluating DBP formation in a full-scale ballast water treatment system (BWTS). In this study, multiple regression models were developed for predicting the dibromochloromethane (DBCM) and bromoform (TBM) concentrations produced by an emergency BWTS using field experimental data from ballast water treatments conducted at Dalian Port, China. Six combinations of independent variables [including several water parameters and/or the total residual oxidant (TRO) concentration] were evaluated to construct mathematical prediction formulas based on a polynomial linear model and logarithmic regression model. Further, statistical analyses were performed to verify and determine the appropriate mathematical models for DBCM and TBM formation, which were ultimately validated using additional field experimental data. The polynomial linear model with four variables (temperature, salinity, chlorophyll, and TRO) and the logarithmic regression model with seven variables (temperature, salinity, dissolved oxygen, pH, turbidity, chlorophyll, and TRO) exhibited good reproducibility and could be used to predict the DBCM and TBM concentrations, respectively. The validation results indicated that the developed models could accurately predict DBP concentrations, with no significant statistical difference from the measured values. The results of this work could provide a theoretical basis and data reference for ballast water treatment control in engineering applications of emergency BWTSs.