Erratum: Probabilistic application of a fugacity model to predict triclosan fate during wastewater treatment



This article corrects:

  1. Probabilistic application of a fugacity model to predict triclosan fate during wastewater treatment Volume 6, Issue 3, 393–404, Article first published online: 1 February 2010

  • Owing to a publisher's error, the authors' proof corrections failed to appear in the first printing; they are shown here as they should have appeared. Original pagination has been retained.


The fate and partitioning of the antimicrobial compound, triclosan, in wastewater treatment plants (WWTPs) is evaluated using a probabilistic fugacity model to predict the range of triclosan concentrations in effluent and secondary biosolids. The WWTP model predicts 84% to 92% triclosan removal, which is within the range of measured removal efficiencies (typically 70% to 98%). Triclosan is predominantly removed by sorption and subsequent settling of organic particulates during primary treatment and by aerobic biodegradation during secondary treatment. Median modeled removal efficiency due to sorption is 40% for all treatment phases and 31% in the primary treatment phase. Median modeled removal efficiency due to biodegradation is 48% for all treatment phases and 44% in the secondary treatment phase. Important factors contributing to variation in predicted triclosan concentrations in effluent and biosolids include influent concentrations, solids concentrations in settling tanks, and factors related to solids retention time. Measured triclosan concentrations in biosolids and non-United States (US) effluent are consistent with model predictions. However, median concentrations in US effluent are over-predicted with this model, suggesting that differences in some aspect of treatment practices not incorporated in the model (e.g., disinfection methods) may affect triclosan removal from effluent. Model applications include predicting changes in environmental loadings associated with new triclosan applications and supporting risk analyses for biosolids-amended land and effluent receiving waters. Integr Environ Assess Manag 2010;6:393–404. © 2010 SETAC


Triclosan (5-chloro-2-(2,4-dichlorophenoxy)-phenol) is an antimicrobial compound that has been widely used in personal health care products for more than 35 y (Waltman et al. 2006). Many consumer products include triclosan, such as soap, deodorant, skin cream, toothpaste, detergents, and cosmetics. In consumer products, triclosan is typically used at concentrations ranging from 0.1% to 0.3% active ingredient by weight (Waltman et al. 2006). Additionally, triclosan is used in many household products such as clothing, countertops, carpets, and trash cans. Approximately 96% of the consumer products containing triclosan are disposed through residential drains (Reiss et al. 2002).

The primary pathway for triclosan to enter the environment is through municipal or industrial wastewater treatment plants (WWTPs). Approximately 74% of the United States (US) population is served by municipal WWTPs (USEPA 2000a), and 84% of the European Union population has access to urban wastewater systems with varying levels of treatment (EC 2001). WWTPs are designed primarily to remove settleable solids, biological oxygen demand, P, and N, but are not designed to remove specific chemicals such as triclosan. Any triclosan remaining in WWTP effluent enters receiving waters, and any triclosan remaining in the activated sludge component may be processed into biosolids that are subsequently applied to land (e.g., fields or forests). The aquatic and terrestrial fate and effects of triclosan are addressed in companion publications (Fuchsman et al. 2010 [this issue]; Lyndall et al. 2010 [this issue]). This study applies a probabilistic fugacity-based model to evaluate the fate and partitioning of triclosan within modern WWTPs (i.e., activated sludge WWTPs with primary and secondary treatment components). Additional applications of the model include predicting changes in environmental loadings associated with new triclosan applications and supporting risk analyses for biosolids-amended land and effluent receiving waters.

Previous formulations of wastewater treatment models rarely address system variability. Thus, they have provided little information on the impacts of differences in WWTP design and operation on the fate of personal care products. Our analysis is intended to account for system variability in the estimation of triclosan concentrations in effluent and biosolids. Additionally, this analysis allows an assessment of what processes and parameters have the greatest impact on the fate of triclosan in a WWTP.


A fugacity model, based on the Sewage Treatment Plant (STP) model of Clark et al. (1995), is used to evaluate triclosan partitioning within various WWTP compartments (Figure 1). The original STP model predicts the removal of a chemical in a WWTP based on default point estimate values that the authors determined were representative values for Canadian WWTPs. Trent University updated the original STP equations to increase clarity. These updated equations and descriptions can be found in the help files of the current version of the STP model (

Figure 1.

Modeled wastewater treatment plant process.

The model described herein is based on the same processes and equations as the updated STP model but incorporates system variability by using a probabilistic approach. Thus, the model is designed to predict the range of conditions and removal rates associated with treatment in a WWTP. In each iteration of the STP model, triclosan enters the WWTP with influent at a constant loading rate and goes to steady state through biodegradation, sorption, and/or other loss mechanisms (in effluent or through volatilization) from each WWTP compartment. Model outputs include total removal and effects of the 3 contributing processes (biodegradation, sorption to sludge, and volatilization and/or effluent removal) based on the range of typical operating conditions in a WWTP (Clark et al. 1995; Qasim 1999; Spellman 2003).

For the purposes of this model, sorption and biodegradation processes are limited by the residence time within the given WWTP compartments. Thus, while biodegradation of triclosan may continue in wasted sludge, only the portion that had biodegraded during the compartment residence time is considered in the model. The WWTP model only simulates processes through the secondary settling tank; subsequent treatment of wasted sludge (e.g., aerobic or anaerobic digestion, composting) is not considered. Similarly, advanced effluent treatments such as disinfection are not simulated. Triclosan is ionizable with a pKa of approximately 8.0 (Budavari 1989; Jäkel 1990); however, only a small fraction is ionized within the normal pH range in WWTPs. Therefore, this evaluation only addresses the neutral species of triclosan.

Model formulation

The model is written in fugacity format as described by Clark et al. (1995) and Mackay (2001). Fugacity is the equivalent of a partial pressure of a substance within a given phase and can be used to predict chemical partitioning among various phases. A full accounting of fugacity is beyond the scope of this document; readers are referred to the original STP paper (Clark et al. 1995) or Mackay (2001) and the materials at The following is a brief summary of the STP model.

There are 3 sets of intermediate parameters (fugacity capacity [Z], partitioning coefficients [K], and fugacity rate parameters [D]) that must be calculated before calculating fugacity (f). In the STP model, the Z values for each compartment (air, water, and biomass) are calculated according to the following equations:

equation image(1)
equation image(2)
equation image(3)

where R = gas constant (Pa m3/mol K), T = temperature (K), H = Henry's Law constant (Pa m3/mol), and biomass is the equivalent of both the sludge and microbial community compartments. The partitioning coefficients (K) between each compartment can be calculated using the following formulae:

equation image(4)
equation image(5)

where Kaw = the partition coefficient between air and water, and Kbw = the partition coefficient between biomass and water. Volatilization rate is expressed as an overall mass transfer coefficient (Kv), calculated as:

equation image(6)

where Kv = the overall partitioning coefficient, Ka = the air side mass transfer coefficient (Table 1), and Kw = the water side mass transfer coefficient (Table 1).

Table 1. Physicochemical properties of triclosan
PropertyValueSourceInternally Consistent Valuea
  • a

    Calculated according to Beyer et al. (2002). Average measured values for vapor pressure and water solubility were given twice the weight of values for Henry's Law constant and log Kow, based on data quantity and quality.

Henry's Law constant4.99 × 10−9 atm·m3/mole @ 25 °CMeylan and Howard 19932.27 × 10−8 atm·m3/mole @ 25 °C
Log n-octanol/water partition coefficient (log KOW)4.78 L/kgWyler and Jäkel 19904.76 L/kg
Vapor pressure4 × 10−6 mm Hg @ 20 °CBudavari 19891.8 × 10−4 Pa @ 25 °C
 6.45 × 10−7 mm Hg @ 25 °CNeely and Blau 1985 
 7.00 × 10−4 Pa @ 25 °CCiba Specialty Chemical 2004 
Solubility in water10 mg/L @ 25 °CYalkowsky and Dannenfelser 199223.4 mg/L
 12 mg/L @ 20 °CPointurier and Volz 1990 
Dissociation constant (pKa)8.14 @ 20 °CJäkel 1990Not calculated
 7.9Budavari 1989 

The partitioning coefficients are then used to calculate the fugacity rate parameters (D1 to D9) for transfer, degradation, and vaporization among compartments (Figure 1). D values for transport, degradation, and volatilization are used. For transport from one compartment to another:

equation image(7)

where G = the flow rate (m3/h) of the phase (i.e., water, sludge, air), defined through mass balance calculations based on the influent flow rate and solids concentration (Clark et al. 1995). A transfer D value is calculated for multiple flow streams: influent (D1), primary tank to aeration tank (D2), primary tank to primary sludge (D3), inflow from air to aeration tank (D4), aeration tank to air (D5), aeration tank to settling tank (D6), settling tank to effluent (D7), settling tank to aeration tank by return sludge (D8), and settling tank to waste sludge (D9). For degradation in a compartment:

equation image(8)

where V = the phase volume (m3), and k is a first-order rate constant (h−1) calculated from the triclosan half-life.

Finally, the fugacity for each WWTP tank is calculated using the following equations:

equation image(9)
equation image(10)
equation image(11)

where fp = fugacity in the primary tank, E = influx of chemical into the plant (mol/h), D2…D9 = fugacity rate parameter, as described above, Dpv = the volatilization D value for primary tank, Dpb = the biodegradation D value for the primary tank, fa = fugacity in the aeration tank, Dab = the biodegradation D value for the aeration tank, Dsv = the volatilization D value for settling tank, Dsb = the biodegradation D value for the settling tank, and fs = fugacity in the settling tank. Knowing f and Z, the concentration (C) of a substance (mol/m3) in each compartment can be calculated by the formula:

equation image(12)

Model parameterization

The STP model was modified to accommodate probabilistic input parameters simulating the range of WWTP operating conditions and was run as a Monte Carlo simulation with 100 000 iterations. The model assumptions and equations are unchanged from those found in Clark et al. 1995 and the materials at The modified model incorporates a combination of deterministic and probabilistic parameters (Table 2). Influent flow rate is held constant for modeling purposes, because treatment processes are scaled to influent volume, such that concentrations in effluent and biosolids are not affected. Point estimate inputs are also used for chemical constants and certain tank dimensions. The chemical properties of triclosan were determined based on structure activity relationships, direct measurements, and the consistency of the parameters to each other. It has been observed that measurement errors in chemical properties can be significant (Beyer et al. 2002), resulting in physiochemical parameters that are not thermodynamically consistent with each other and, therefore, cannot be correct. Beyer et al.'s (2002) methods were used to mathematically determine internally consistent physicochemical properties for triclosan (Table 1). In so doing, vapor pressure and water solubility were weighted twice as heavily as the other parameters, because multiple high quality data sets were available for these properties.

Table 2. Modified STP model input parameters
 ParameterDistribution ValuesUnitsSource
  1. MLSS mean live suspended solids.

  2. TSS total suspended solids.

  3. VSS volatile suspended solids.

Physicochemical PropertiesKowPoint 5.75 × 104L/kgInternally consistent value based on Beyer et al. 2002
 R (gas constant)Point 8.314Pa m3/mol K 
 Henry's Law ConstantPoint 2.27 × 10−3Pa m3/mol KInternally consistent value based on Beyer et al. 2002
 Molecular WeightPoint 289.54g/molCalculated
 TemperaturePoint 25°CClark et al. 1995
 Density of BiosolidsPoint 1000kg/m3Clark et al. 1995
 KwPoint 0.05m/hClark et al. 1995
 KaPoint 5m/hClark et al. 1995
Half-Life (2000 mg/L MLSS)Aeration and Settling TanksNormalMin10hBased on 28-d treatability of 50–70% (Hanstveit and Hamwijk 2003; Stasinakis et al. 2008) and corresponding USEPA (2000b) recommended values +/- 50%
 Primary TankNormal CalculatedhAeration tank half-life × 10 (Clark et al. 1995)
InfluentInfluent Flow ratePoint 1000m3/hClark et al. 1995
 TSSNormalMin45g/m3Range of values from Qasim 1999 (90–280, mean = 175), min = 1/2 lower range, SD calculated assuming upper range value is the 95th percentile
 Triclosan (U.S.)LognormalGeo. Mean4.79µg/LFitted from measured concentrations
   Geo. SD1.91µg/L 
 Triclosan (non-U.S.)LognormalGeo. Mean1.03µg/LFitted from measured concentrations
   Geo. SD4.15µg/L 
Primary TreatmentPrimary Tank AreaPoint 266.7m2Clark et al. 1995
 Primary Tank DepthPoint 3.8mClark et al. 1995
 Fraction of Influent TSS RemovedUniformMin0.4proportionQasim 1999; Spellman 2003
 Fraction of Tank Volume as VSSNormalMin0.0005proportionQasim 1999; Spellman 2003
 Sludge VSSPoint 50 000g/m3Clark et al. 1995; Qasim 1999; Spellman 2003
Aeration TankAeration Tank AreaUniformMin500m2Varied to result in retention time from 4 to 8 h (Qasim 1999; Spellman 2003)
 Aeration Tank DepthPoint 8mQasim 1999, Spellman 2003
 Fraction of Tank Volume as VSSNormal Calculatedproportion0.0025 x multiplier (Clark et al. 1995; Spellman 2003)
 Outflow VSSNormal Calculatedg/m32500 x multiplier (Clark et al. 1995; Spellman 2003)
 Multiplier (for VSS calculations)NormalMin0.1proportionBased on range of VSS values (Qasim 1999; Spellman 2003)
Secondary Settling TankSettling Tank AreaPoint 727.3m2Clark et al. 1995
 Settling Tank DepthPoint 3.8mClark et al. 1995
 Fraction of Tank Volume as VSSNormal Calculatedproportion0.00055 x multiplier (Clark et al. 1995; Spellman 2003)
 Sludge VSSNormal Calculatedg/m35500 x multiplier (Clark et al. 1995; Spellman 2003)
 Outflow VSSNormal Calculatedg/m310 x multiplier (Spellman 2003)
 Fraction of Flow RecycledUniformMin0.5proportionClark et al. 1995; Qasim 1999; Spellman 2003
 Fraction of Flow RemovedUniformMin0.005proportionClark et al. 1995; Qasim 1999; Spellman 2003

Probabilistic input parameters describe the range of solids loading and processing parameters and tank dimensions in activated sludge WWTPs in the United States, based on industry standard design and operating criteria (Clark et al. 1995; Qasim 1999; Spellman 2003). Probabilistic parameters include the influent solids load, the fraction of influent suspended solids removed in the primary tank, aeration tank volume, concentrations of volatile suspended solids (VSS) (i.e., the organic component of suspended solids) in the tanks and tank return, outflow of VSS, sludge removal from the secondary settling tank, and flow recycle from the secondary settling tank to the aeration tank. The variability in these parameters primarily impacts the mechanical removal of solids and the solids retention time. Biodegradation is expected to be directly related to solids retention time (Ternes et al. 2004).

The activated sludge half-life for triclosan is estimated from published laboratory biodegradability test results, extrapolated to in-plant conditions following recommendations of USEPA (2000b). As reviewed by NICNAS (2009), triclosan degradation in 28-d laboratory tests with relevant initial concentrations (i.e., excluding unrealistically high concentrations potentially toxic to activated sludge microbes) ranges from 50–70% (Hanstveit and Hamwijk 2003; Stasinakis et al. 2008). According to USEPA (2000b), 28-d degradation of 20–70% corresponds to an in-plant activated sludge half-life of 30 h, whereas ≥70% degradation corresponds to an in-plant half-life of 10 h. We represent the uncertainty of this derivation by assuming a half-life of 30 h plus or minus 50%.

For modeling purposes, triclosan loading is estimated using 2 approaches. To quantify removal efficiencies, influent loading is set to 1 gram of triclosan per h to simplify calculations. To estimate triclosan concentrations in biosolids and effluent, a distribution of triclosan concentrations in influent is fitted from published measurements collected internationally (Table 3; Figure 2), using the best-fit function of Crystal Ball®. Separate influent distributions are identified for US and non-US WWTPs. This facilitates comparison to measured triclosan concentrations in biosolids, because nearly all (97%) of the biosolids data compiled for this purpose are from US sources (Table 3).

Table 3. Measured triclosan concentrations in WWTP influent and removal efficiencies from WWTP systems
LocationInfluent ConcentrationPrimary EffluentSecondary EffluentFinal EffluentSource
Mean (Range) (µg/L)Removala (%)Removal (%)Removal (%)
  • a

    Removal percentages for each treatment phase are compiled from multiple WWTPs at multiple time points and therefore should not be considered additive.

  • b

    Primary effluent concentration exceeded influent concentration.

Activated Sludge
 Crofton, UK21.93995Sabaliunas et al. 2003
 Columbus, Ohio, USA5.23595McAvoy et al. 2002
 Loveland, Ohio, USA10.73696McAvoy et al. 2002
 Rhine-Ruhr Region, Germany7.396Bester 2005
 Rhine-Ruhr Region, Germany4.8316387Bester 2005
 Dortmund, Germany1.2 (1.1–1.3)96Bester 2003
 Denton, Texas7.3 (2.7–26.8)98Waltman et al. 2006
 Arlington, Virginia3108798Thomas and Foster 2005
 Alexandria, Virginia3.3455399Thomas and Foster 2005
 Fairfax Country, Virginia3.6178299Thomas and Foster 2005
 Almería, Spain1.8 (0.4–4.2)89Gómez et al. 2007
 Mid-Atlantic Region, USA4.7 (0.8–10.8)99Heidler and Halden 2007
 Ontario, Canada1.9 (0.01–4.0)94Lishman et al. 2006
 Tokyo, Japan0.545–93Nakada et al. 2006
 Lyon, France0.3855Paxeus 2004
 Patras, Greece2.1794Paxeus 2004
 Naples, Italy1.3773Paxeus 2004
 Lund, Sweden0.3858Paxeus 2004
 Copenhagen, Denmark0.0988Paxeus 2004
 Adelaide, Australia0.85414193Ying and Kookana 2007
 Perth, Australia0.5875971Ying and Kookana 2007
 Coslech, South Wales, UK0.23 (0.03–0.46)75Kasprzyk-Hordern et al. 2009
 Moorhead, Minnesota, USA2.21b8791Shelver et al. 2007
Trickling Filter
 Meltham, UK7.52195Sabaliunas et al. 2003
 Glendale, Ohio, USA3.8758McAvoy et al. 2002
 West Union, Ohio, USA16.68686McAvoy et al. 2002
 West Union, Ohio, USA15.4483483McAvoy et al. 2002
 United Kingdom (Biological filter)1.55 (1.0–2.1)39–4663–8889Thompson et al. 2005
 Cilfynydd, South Wales, UK0.087 (<0.097–0.24)71Kasprzyk-Hordern et al. 2009
 Fargo, North Dakota, USA2.56b7292Shelver et al. 2007
Other WWTPs
 United Kingdom (Rotating biological contactor)2.5 (1.3–3.7)40–8925–3075Thompson et al. 2005
 United Kingdom (Oxidation ditch)3.35 (1.6–5.1)9798Thompson et al. 2005
Adelaide, Australia (Lagoon)0.5985Ying and Kookana 2007
Adelaide, Australia (Bioreactor)0.818589Ying and Kookana 2007
Adelaide, Australia (Oxidation ditch)0.799192Ying and Kookana 2007
West Fargo, North Dakota, USA (Lagoon)4.1298Shelver et al. 2007
Figure 2.

Comparison of measured triclosan concentrations with fitted distributions (influent) and modeled concentrations (effluent and biosolids).

Model calibration

The model is calibrated to measured triclosan concentrations in effluent and biosolids. Effluent data were compiled from the sources listed in Table 3 and from Boyd et al. (2003), Lindström et al. (2002), and Sabaliunas et al. (2003). Biosolids data sources are summarized in Table 4; information on WWTP type and biosolids treatment processes (e.g., digestion, composting) is not available for most of the biosolids measurements. To weight each WWTP equally, distributions of measured triclosan concentrations in effluent and biosolids are based on mean concentrations from individual WWTPs.

Table 4. Summary of data sources for measured triclosan concentrations in biosolids
LocationNumber of WWTPsRange (dry mg/kg)Median (dry mg/kg)Reference
  • a

    The maximum reported triclosan concentration (133 mg/kg) is excluded as an outlier, based on a data qualifier indicating possible matrix interference, as well as lack of agreement with other measured concentrations.

United States (mid-Atlantic region)120.0–55.037.5Heidler and Halden 2007
United States (AZ, CO, IA, KS, TX, WA, WI)90.4–10.52.6Kinney et al. 2006
United States (VA, NY, MD, CA)63.6–7.45.4La Guardia et al. 2004
United States (OH)30.5–15.61.5McAvoy et al. 2002
New Zealand11.0–22.011.5Speir and Northcott 2006
United States (WA)123.0–29.026.0USEPA 2003; Reimer 2003
United States730.33–34.07.2USEPA 2009a
Australia20.2–5.62.9Ying and Kookana 2007


The model's output is summarized in Table 5 and Figures 2 and 3. Triclosan concentrations in WWTP influent are higher in the United States than in other developed countries (Figure 2), consistent with higher triclosan usage in the United States (MacAvoy et al. 2002; Singer et al. 2002). Triclosan concentrations in non-US effluent are consistent with model predictions based on non-US influent data in that the medians are within 15%. However, the median triclosan concentration in US effluent is over-predicted in this model by 80% based on median US influent data (Figure 2). This result might possibly reflect differences between the United States and other countries in effluent treatment processes such as disinfection. Triclosan is effectively removed through oxidation by disinfection agents including free Cl, chloramines, and ozone, whereas ultraviolet disinfection shows limited removal of microcontaminants (Snyder et al. 2008). However, documentation of different disinfection practices internationally is limited, and chlorination remains in widespread use (Jacangelo and Trussell 2002); thus, any such explanation is highly speculative. Artifacts of the effluent data set (e.g., biases associated with inclusion of multiple WWTP types) do not appear to explain the observed results. The predicted and modeled biosolids concentrations for the United States are within 5% of each other. For non-US biosolids concentrations, modeled and measured concentrations cannot be reliably compared due to small sample size (n = 3).

Table 5. Modeled triclosan concentrations and removal efficiencies
PercentileUS Influent SourceNon-US Influent SourceRemoval Efficiency (%)
Effluent (µg/L)Biosolids (dry mg/kg)Effluent (µg/L)Biosolids (dry mg/kg)PrimarySecondaryTotal
Figure 3.

Relative frequency distributions of model output: solids retention time (a), overall triclosan removal due to biodegradation and sorption (b), and overall removal efficiency (c).

Interestingly, although not predicted by the model, measured triclosan removal efficiency may be related to influent concentration, with low influent concentrations associated with a wide range of removal efficiencies and higher influent concentrations generally associated with high removal efficiencies (Figure 4). The Spearman's rank correlation coefficient for this relationship is 0.54 (p < 0.001). However, the correlation for US data is not significant (rho = −0.21, p > 0.05). For non-US data, the correlation coefficient is 0.71 (p < 0.001). This does not necessarily reflect a true concentration-dependent removal mechanism, however, and could be a spurious correlation or a relationship to some other factor that co-varies with triclosan influent concentration in non-US treatment facilities.

Figure 4.

Concentration dependence of measured triclosan removal efficiency.

The removal efficiency of the modeled WWTP (median of 89%) is similar to measured removal efficiencies (median of 93%). In both the modeled system and the measured WWTPs, approximately 1/3 of triclosan is removed during primary treatment, while secondary treatment removes the majority of triclosan (Tables 2 and 4). The 2 primary mechanisms for triclosan removal are sorption to solids and biodegradation (in aqueous and solid phases). The overall modeled triclosan removal efficiency is 84% to 92% (10th to 90th percentile) during WWTP passage. Volatilization is negligible in all treatment compartments because of the very low Kaw of 9.17 × 10−7. Overall, approximately 30% to 52% (10th to 90th percentile) of triclosan is sorbed to sludge and 36% to 53% (10th to 90th percentile) is biodegraded (Table 6). The modified STP model results are similar to those reported based on mass balance (Bester 2003; Thompson et al. 2005; Heidler and Halden 2007). However, the model estimates higher sorption and lower degradation than the KOC partitioning estimate by Singer et al. (2002). This discrepancy is likely because Singer et al. (2002) began measuring triclosan after the influent was mechanically clarified. As a result, sorption in the primary treatment phase is underrepresented.

Table 6. Reported relative contributions to triclosan removal from wastewater treatment plant systems
MethodType of SystemBiodegradation (%)Sorption (%)Source
  • a

    Measurements do not include primary treatment and therefore underestimate sorption to sludge.

  • b

    Range based on 10th and 90th percentiles from model.

Mass balanceActivated sludge6530Bester 2003
Mass balanceActivated sludge4850Heidler and Halden 2007
Mass balanceActivated sludge79a15aSinger et al. 2002
Mass balanceFixed film40–90Thompson et al. 2005
Fugacity modelActivated sludge36–53b30–52bThis study

Evaluation of WWTP processes

The STP model indicates that sorption and degradation are each dominant in different WWTP treatment compartments (Table 5). Of the overall triclosan removal due to sorption (median = 40%), the model confirms that the majority of the removal occurs in the primary treatment phase (median = 31%), when settleable solids are removed. Secondary treatment accounts for the remainder of sorption (median = 9%), as triclosan sorbs to activated sludge in this phase. In comparison, the overall triclosan removal due to biodegradation (median = 48%) is relatively limited during primary treatment (median = 4%) but is dominant during secondary treatment (median = 44%). Based on these compartmental trends, it can be concluded that sorption is the key removal mechanism during primary treatment, while biodegradation is dominant during secondary treatment.

The relative contribution of biodegradation during secondary treatment is influenced by multiple factors, including dissolved oxygen, solids retention time, biodegradation half-life, and solids loading. Thompson et al. (2005) reported that sorption appears to be the main mechanism of triclosan removal at lower dissolved oxygen levels (e.g., rotating biological contactor and trickling filter treatment methods), while biodegradation is dominant when a higher oxygen level (dissolved oxygen level >1.5 mg/L) is maintained. This is consistent with observations that triclosan degradation in anaerobic systems is much slower than in aerobic systems (Christensen 1994a, 1994b).

Solids retention time is considered a key parameter affecting the removal of a variety of pharmaceutical and personal care product chemicals from wastewater (e.g., Clara et al. 2005). Adam (2006) reported that the WWTP process entails high initial triclosan sorption to solids followed by degradation over time within the solid phase. Federle et al. (2002) reported that more than 94% of triclosan was ultimately biodegraded in a laboratory experiment designed to simulate secondary treatment in an activated sludge WWTP. Similarly, Stasinakis et al. (2007) found that 97% of triclosan was ultimately biodegraded in a 96-d activated sludge WWTP. Thus, the amount of biodegradation occurring in a WWTP is a function of solids retention time rather than the partitioning of triclosan into a recalcitrant fraction that is resistant to biodegradation.

In the model, solids retention time is a function of aerated sludge volume (set based on STP defaults and standard practices) and aerated sludge removal rates (determined by the STP equations). Therefore, solids retention time is a model output. The modeled solid retention time ranged from 1 to 66 d, with a median solids retention time of approximately 10 d (Figure 3). The model output demonstrates a correlation between the solids retention time and overall triclosan removal (Spearman's rank correlation, p < 0.001; n = 10 000; rho = 0.30) (Figure 5). This relationship is modeled using Michaelis-Menton kinetics:

equation image(13)

where Fdeg is the proportion of triclosan biodegraded, Tr is the solids retention time, a is the maximum proportion biodegraded, and b is a constant (a = 0.71; b = 0.31). The equation predicts that for increasing solids retention times, up to approximately 10–15 d, there is an increase in triclosan biodegradation. Indeed, upgrading WWTPs to achieve a solids retention time of 12–15 d has been recommended to improve removal of a variety of chemicals commonly found in wastewater (Ternes et al. 2004).

Figure 5.

Proportion of triclosan predicted to be biodegraded as a function of calculated solids retention time in wastewater treatment plant aeration and settling tanks during secondary treatment. Points represent individual model iterations. Dashed line represents the nonlinear regression: y = 0.71x / (x + 0.31).

Parameter contributions to model variability

Sensitivity analysis tools in Crystal Ball® quantify the influence of probabilistic model parameters on model predictions. When a distribution is used to describe the influent concentration of triclosan, it is the most important parameter by a wide margin, accounting for approximately 80% of the variability in the effluent and biosolids triclosan concentrations. Thus, the predictions of the model are very sensitive to the influent concentrations, and biases in this parameter potentially have a significant impact on the results. One source of uncertainty is the sensitivity of the fitted influent concentration to small changes in the influent dataset. The sensitivity of the geometric mean to individual observations was explored using piecewise elimination of single influent data points. The elimination of individual values resulted in changes in the geometric mean of 50% or less, meaning that the geometric mean of input distribution is likely to be within 50% of the true value, considerably less than the range of 3 orders of magnitude in the observed influent concentrations. Therefore, the magnitude of the uncertainty associated with the fitted influent data is minor in comparison to the variability associated with using a distribution of influent concentrations.

When the variability associated with triclosan loading in influent is controlled by using a single influent concentration for all model runs, the impacts of the variability in other probabilistic parameters can be assessed. Variations in effluent and biosolids concentrations are primarily linked to the factors controlling sorption to wasted sludge and biodegradation. The most important factor for triclosan sorption is the VSS concentration in the settling tanks. Variation in VSS concentration accounts for approximately 60% of the variation in effluent concentrations and 70% of the variation in the biosolids concentration of triclosan. Among the variables related to solids retention time, the aeration tank area is the most important contributor to predicted variation in triclosan concentrations, accounting for 20% of the variation in effluent and 15% of the variability in biosolids. Other important factors controlling biodegradation include the rate of solids wasting from the secondary tank and the rate of biosolids recycling back to the aeration tank, collectively accounting for 10% of the variability in the effluent and biosolids concentrations. Of these parameters, the aeration tank area represents a WWTP design consideration that could be adjusted to maximize the removal efficiency of a plant. The other parameters are generally adjusted to maximize the removal efficiency of biological oxygen demand, N, and phosphate (Qasim 1999; Spellman 2003), which typically govern the design and operation of a WWTP. The activated sludge half-life of triclosan, although uncertain, contributes negligibly to the variation in model output.

When all other parameters are held constant and the aeration tank volume is set to 4000, 6000, and 9000 m3, biosolids retention times are 5.2, 7.5, and 9.7 days; the biodegradation removal efficiency is 62%, 70%, and 75%, resulting in total removal efficiencies of 87%, 90%, and 92%. Clearly, increasing the size of the aeration tank can have a dramatic impact on biodegradation and, therefore, the concentration of triclosan in biosolids. Increasing the aeration tank volume from 4000 to 9000 m3 results in a 35% reduction in the triclosan concentration in biosolids.


Wastewater treatment processes remove triclosan to a high degree of efficiency by sorption to solids and biodegradation. Removal by sorption to solids is dominant in the WWTP primary treatment phase as solids settle out. Biodegradation is dominant in the WWTP secondary treatment phase, where high dissolved oxygen concentrations and longer solids retention times can further enhance biodegradation of both aqueous and sorbed triclosan. Our probabilistic modification of the STP model effectively predicts the distribution of triclosan concentrations in biosolids and non-US effluent, but concentrations in US effluent are over-predicted. Further investigation is warranted to identify treatment practices by which US WWTPs achieve low triclosan concentrations in effluent, despite higher influent concentrations than in other countries.


An expert panel consisting of Don Mackay, Lawrence Barnthouse, and Michael C. Newman provided thoughtful reviews of an earlier version of this study, resulting in substantial improvements to the analysis. We also thank two anonymous reviewers for their thoughtful comments. Kannan Vembu provided assistance in understanding WWTP design and processes. Thanks also to Michael Ferguson for model quality control. Colgate-Palmolive Company funded preparation of this manuscript.

Disclaimer—The peer-review process for this article was managed by the Editorial Board without involvement of Editor-in-Chief R. Wenning.