Analysis of Selected Pharmaceuticals and Endocrine Disrupting Compounds and their Removal by Granular Activated Carbon in Drinking Water Treatment

Abstract

Over the last decade, endocrine disrupting compounds (EDCs) and pharmaceutically active compounds (PhACs) have been detected in drinking water at very low levels, mostly ng/L concentrations, suggesting that these compounds resisted removal through water treatment processes. Concerns have been raised regarding the effectiveness of common drinking water treatment technologies to remove these emerging contaminants. Adsorption processes were suggested to play an important role in the removal of PhACs and EDCs, based on the assumption that these compounds are similar to other conventional micropollutants such as pesticides in both physicochemical properties and concentration levels present in water. However, this remains to be demonstrated since the availability of adsorption data for PhACs and EDCs is extremely limited and their environmental concentrations are typically much lower than the ones for pesticides. The primary objective of this research was to evaluate in detail the removal of representative EDCs and PhACs at environmentally relevant concentrations by granular activated carbon (GAC) adsorption. In the first stage of this study, EDCs (15) were screened separately from the PhACs (86) with two different sets of assessment criteria due to the different nature and the availability of information for these two groups of compounds. As a result, 6 EDCs and 12 PhACs were selected for further evaluation. Subsequently, a multi-residue analytical method based on gas chromatography/mass spectrometry (GC/MS) was developed for the simultaneous determination of the selected PhACs and EDCs. Two key analytical steps - solid phase extraction and derivatization - were systematically optimized using full factorial design and a central composite design, respectively. The statistical experimental design in combination with the concept of the total desirability was demonstrated to be an effective tool for developing a multi-residue analytical method. The application of the developed method to Grand River water, a local raw water source, and finished drinking water from this source indicated that PhACs such as naproxen, carbamazepine, salicylic acid, ibuprofen, and gemfibrozil, and EDCs such as estrone (E1) and nonylphenol mono-ethoxy carboxylate (NP1EC) were the most common contaminants. Based on these results, the quality of the analytical data, and the physicochemical properties relevant to the adsorption on activated carbon, two PhACs (naproxen, carbamazepine) and one EDC (nonylphenol (NP)) were finally chosen for the adsorption studies. Adsorptions of the selected target compounds were evaluated on two types of activated carbon (coal-based Calgon Filtrasorb® 400 (F400) and coconut shell-based PICACTIF TE (PICA) by first investigating their isotherms at environmentally relevant concentrations (equilibrium liquid phase concentration ranging from 10 to 1000 ng/L). The single-solute isotherm data determined for both carbons showed that the relative adsorbabilities of the three target compounds were not in agreement with expectations based on their log Kow values. Overall, in this low concentration range, carbamazepine was most easily removed, and NP was least adsorbable. The adsorption of naproxen was negatively influenced by its dissociation in water. Comparison of single-solute isotherms on F400 carbon for the target compounds to those for other selected conventional micropollutants showed that naproxen and carbamazepine have generally comparable isotherms to 2-methylisoborneol (MIB) and geosmin. The isotherm tests in a post-sedimentation (PS) water from a full-scale plant demonstrated that the presence of background natural organic matter (NOM) significantly reduced the adsorption of all three target compounds, among which.NP was the least impacted compound. Based on the quantification of the direct competition using the ideal adsorbed solution theory (IAST) in combination with the equivalent background compound (EBC) approach, the minimum carbon usage rates (CURs) for removing 90% of the target compounds in PS water were calculated at two environmentally relevant concentrations (50 and 500 ng/L). This work confirmed that the percentage removal of the trace level target compound at a given carbon dosage was independent of the initial target compound concentration. Isotherm experiments were conducted for the target compound on GACs preloaded with PS water for various time intervals (up to 16 weeks) at the Mannheim Water Treatment Plant (Region of Waterloo, ON, Canada). The results indicated that the adsorption of all target compounds were subject to significant negative impacts from preloading of NOM, albeit to different extents. Among the three target compounds, reduction in adsorption capacity for naproxen was most severe, followed by carbamazepine and then NP. The three target compounds followed quite different patterns of decrease in adsorption capacity with increasing preloading time, thus revealing different competitive mechanisms at work for the different compounds. For naproxen, the change in heterogeneity of the carbons due to preloading suggests that some pre-adsorbed NOM could not be replaced by naproxen. However, both direct competitive and pore blockage mechanisms could successfully explain the adsorption performance of naproxen and carbamazepine. The removal of NP even at prolonged preloading times could be explained by absorption or partitioning in the NOM matrix on the surface of, or inside the carbons. The kinetic parameters for each target compound-virgin carbon pair were determined using the short fixed bed (SFB) approach based on the pore and surface diffusion model (PSDM). The SFB results and sensitivity analyses indicated that, under the very low influent concentration conditions, film diffusion (indexed as βL) exerts a much greater effect on breakthrough profiles than internal diffusion. The SFB tests on preloaded GACs showed that mass transport of all the target compounds decreased with increasing preloading time. Similar to the impact of preloading on adsorption capacity, naproxen was subject to the most deteriorative effect, followed by carbamazepine and then NP. In addition, potential mechanisms for the decay of the film diffusion coefficient with increased preloading time were discussed based on scanning electron microscope (SEM) images of virgin and preloaded GAC. Electrostatic interactions between the NOM/bio film formed on the preloaded carbon and dissociated naproxen may have contributed to the enhanced reduction in its film diffusion. Sensitivity analyses and subsequent calculations of the Biot numbers confirmed that film diffusion was also the predominant mechanism controlling the mass transport on preloaded carbon, in particular for naproxen. This suggests that the early breakthrough prediction of the target compounds at their environmentally relevant concentrations could be further simplified by only considering film diffusion and adsorptive capacity. Kinetic and isotherm parameters were used as input for modeling using time-variable PSDM. It was found that the varying trends for Freundlich KF and 1/n, and βL could be generally depicted by a corresponding empirical model. Pilot scale treatability tests were performed for the target compounds which subsequently validated the time-variable PSDM results thus demonstrating its effectiveness and robustness to model GAC adsorber performance for PhAC and EDC removal at environmentally relevant concentrations. The time-variable approach was further improved by adjusting for NOM surface loading differences between the preloading and the pilot columns, which successfully compensated for the prediction errors at the early phase. The validated NOM surface loading associated time variable PSDM was used to predict performances of hypothetical F400 and PICA full-scale adsorbers. Both adsorbers were expected to provide satisfactory performance in achieving 90% removals for the neutral target compounds (carbamazepine and NP). Naproxen was predicted to break through fast since both, capacity and kinetic parameters decay quickly due to carbon fouling by NOM and the physicochemical properties of this compound. Initial recommendations on the choice of adsorption process (GAC vs. PAC) for removing EDCs and PhACs can be made based on the comparison of carbon usage rates (CUR) which were calculated for a GAC adsorber using the validated improved PSDM and for PAC using the minimum applied dosages predicted by the IAST-EBC model.