Predicting the Occurrence of Disinfection by-products (DBPs) in Aquatic Centres: Novel Models for Pool-Level Assessments

Managing disinfection by-products (DBPs) in swimming pools is a significant concern due to their potential toxicity. Empirical models serve as an effective alternative to conventional, time- and cost-intensive DBP monitoring. However, the models developed to date for swimming pools have limitations in terms of inclusiveness and generalizability. This paper presents the first effort to develop robust and generalizable empirical models for trihalomethanes (THM-4) and haloacetic acids (HAA-5) species across various pool types (lap, leisure, and hot tubs) in indoor swimming pools (ISPs). A total of 133 samples were collected from four chlorinated ISPs and analyzed for target parameters. Of these, 93 samples from three ISPs were used to develop the models, while 40 samples from a different facility were used to validate them externally. Multiple linear regression (MLR) was used to model selected DBPs as a function of total organic carbon (TOC), pH, temperature, free chlorine, total chlorine (TC), chloramines, number of swimmers, and flow rate (Q). The finalized models showed good predictive performance, except for the hot tubs, with regression coefficients (R²) ranging from 0.69 to 0.98 and comparable adjusted and predicted R² values. External validation, assessed using R² and standard error metrics, indicated good generalizability of the pool-specific (lap and leisure pool) models, whereas facility-level models performed poorly on external data. Results showed that TOC primarily influenced HAA formation across all pools. In contrast, TC and Q were the key factors for THM formation and accumulation in the leisure and lap pools, respectively. A framework has been proposed to incorporate the empirical models into a multi-barrier, risk-based approach to DBP management.