Predictive models for the removal efficiency of emerging organic contaminants based on conventional water quality...

This study examines the prospects of predicting the removal efficiency of emerging organic contaminants (EOCs) (pharmaceuticals-PhCs and personal care products-PCPs) based on conventional water quality parameters (CWQPs) including total suspended solids (TSS), chemical oxygen demand (COD), ammonium‑nitrogen (NH₄⁺-N) and total phosphorus (TP) in horizontal subsurface flow constructed wetlands (HFCWs). Although previous research demonstrates a correlation between CWQPs and EOCs removal in CWs, research is lacking in evaluating the possibilities of developing predictive models. To bridge this gap, compound specific predictive models of six PhCs (acetaminophen, diclofenac, erythromycin, ibuprofen, ketoprofen and naproxen) and two PCPs (tonalide and triclosan) are developed underpinned by the principal component, correlation and multiple linear regression analyses conducted using a dataset compiled from peer-reviewed publications. The results showed that the CWQPs are reasonably good predictors of EOCs removal in HFCWs. For instance, the best predictive model for diclofenac incorporated COD and NH₄⁺-N (coefficient of determination (R²): 0.826; probability value (p): 3.434E^-07; root mean squared error (RMSE): training set: 7%; validation set: 14%). The best predictive model for tonalide incorporated COD and NH₄⁺-N (R²: 0.882; p: 1.910E^-04; RMSE: training set: 6%; validation set: 20%). The same combination of CWQPs formed credible models for other PhCs and PCPs. These novel models can serve as a screening tool to support assessment for making predictions for a number of EOCs. The new approach and resulting model developed in this study offer a robust framework for preliminary assessment of EOC removal under data limited situations alongside guiding research and optimizing HFCW design and operation.