A novel phenomenological model for streamlined prediction and online monitoring of organic micropollutant removal in full-scale wastewater ozonation

Ozonation is a leading option for advanced treatment in full-scale WWTPs, yet rapid prediction and online monitoring of organic micropollutant (MP) removal remain challenging. We propose an integrated, streamlined phenomenological framework that addresses both needs through two complementary models. The formulation rests on conventional ozone–radical kinetics, using the Rct concept ( OH-exposure/O 3 -exposure) with an effective first-order ozone decay. After algebraic simplification, the model reduces to two identifiable parameters, and these assumptions were verified a posteriori by goodness-of-fit. Calibrated with full-scale observations, the two-parameter removal model accurately reproduced the behaviour of 12 MPs across 103 data points from three campaigns (R 2 = 0.92; MAPE = 5.4%), with compound-wise errors of 0.4–10%. A key contribution is the experimental determination of k OH for four directive-listed indicators, amisulpride, citalopram, candesartan, and 4/5-methyl-benzotriazole, experimentally determined by back-calculating k OH from measured removals with the calibrated removal model. To the best of our knowledge, these constants have not previously been measured experimentally, and the new values closely match recent predictions for amisulpride, citalopram, and candesartan and provide the first reported k OH for 4/5-methyl-benzotriazole. For online MP removal monitoring, we developed a surrogate-driven forward model built on the phenomenological formulation. A436 delivered the best balance of parsimony and accuracy, marginally outperforming A254: a joint two-parameter model, pooled across the same campaigns, achieved R 2 = 0.92 and MAPE = 5.2%. Overall, the framework links a minimal kinetic structure to easily measured signals, enabling rapid prediction and proxy-based monitoring under realistic conditions while providing new kinetic data that reduce uncertainty in indirect oxidation pathways. • Two-parameter phenomenological model calibrated on full-scale ozonation • Accurate 12 micropollutant removal predictions: R 2 = 0.92; n = 103 • Phenomenology-driven A436 surrogate model enables proxy-based online monitoring. • First experimental k OH for 4 EU-directive MP indicators; 3 match recent predictions