The activated sludge used in this study was procured from the aeration basin of a pilot scale aerobic-anoxic-oxic system of 50 m³·d⁻¹ operated in a local advanced WWTP. Once it was collected, the activated sludge was transported to the laboratory. Then, moisturized air was supplied to the activated sludge for more than 3 h in order to oxidize any remaining organics and ammonia. After the pretreatment, the activated sludge was transferred to two 4 L batch reactors. A DO probe was installed on the one of the reactors and a pH probe was installed on the other. The temperature of the reactors was carefully maintained at 25 ± 1 °C since the activity of nitrifying bacteria is affected by changes in temperature.

The effluent from a primary sedimentation basin of the pilot plant was used as test wastewater, to which a toxicant was added. The general characteristics of the wastewater (i.e., COD, TKN, NH₄⁺, NO₃⁻, and alkalinity) and activated sludge (i.e., mixed liquor volatile suspended solids: MLVSS) are summarized in Table 1. All the parameters were analyzed according to the Standard Methods [25].

To evaluate the feasibility of the proposed strategy, three toxic chemicals were arbitrarily selected. They are Hg²⁺ (as HgCl₂), ATU (C₄H₈N₂S), and chloroform (CHCl₃). The selected test toxicants of analytical grade were purchased from Sigma-Aldrich (St. Louis, MO, USA). Firstly, a stock solution of each selected chemical was made and stored under −20 °C until used. Whenever needed, the stock solution was diluted to an appropriate concentration with deionized water. Table 2 shows the concentration levels of each toxicant applied in this study along with published concentrations causing inhibition to activated sludge.

Changes in the OUR and in the pH profiles of the reactors with nitrifying activated sludge were monitored using a DO probe (DO-350L, i-stek, Incheon, Korea) and a pH probe (HA 405-DXK-S8, Mettler Toledo, Zurich, Switzerland). Both sensors were connected to a data acquisition and control (DAC) module (NI cDAQ-9172, National Instrument, Austin, TX, USA). The visualization of signals from the sensors and the DAC module, and the nitrification inhibition detection were made using a graphic user interface development environment (LabVIEW 8.2; National Instruments).

Normally, the signals from the DO and the pH probes contain concomitant noises. Therefore, it is hard to draw their true profiles. Thereafter, a moving-average digital filter was implemented for smoothing data from sensors (Equation (1)). The pH and DO of reactors were collected 10 times a second, and they were averaged every 20 s; 200 data (N*) were averaged to produce one data point. Then, the derivatives of the pH and DO (i.e., dpH/dt and dDO/dt) were calculated once every 1 min [22], since the biological reaction did not change spontaneously. (1) y F ( k ) = 1 N * ∑ i = k − N * + 1 k y iwhere N* is the number of past data points that are being averaged.

In the beginning, 3 L of nitrifying activated sludge was filled into each of two 5 L reactors (working volume: 4 L). Then, 1 L wastewater was added to each reactor. Once wastewater was added, the aerator for each reactor was turned on to provide air. In the case of the reactor with the DO probe installed, the reactor DO was initially increased up to 7 mg·L⁻¹. Then, the aerator was turned off to monitor the DO profile and calculate dDO/dt (or OUR); during the air-off period, the liquid of the reactor was slowly mixed using a magnetic stirrer for keeping the sludge in suspension. If the reactor DO was lower than 3 mg·L⁻¹, then the aerator was turned on again to increase the DO to 7 mg·L⁻¹. While the reactor DO was raised to 7.0 mg·L⁻¹, the OUR was not calculated. In the case of the reactor with the pH probe installed, the reactor pH was monitored to calculate dpH/dt and to identify the inflection point (i.e., AV or dpH/dt = 0) on the pH profile during the aeration.

During each batch assay experiment, a 30 mL mixed liquor sample was collected once every 20 min directly from the reactor with the pH probe installed. Then, NH₄⁺ concentration of the samples was analyzed to demonstrate the relation between the pH profile and NH₄⁺ oxidation by nitrifiers during the air-on time [22].

After the batch bioassay experiments, a toxicity detector was built in order to continuously screen wastewater for the presence of a potential toxicant (Figure 2). The toxicity detector consisted of two reactors with a working volume of 3 L each, peristaltic pumps, air diffusers, magnetic stirrers, a pH probe, a DO probe, a DAC module, and an IBM-compatible PC. Feeding and draining test wastewater, turning air on/off, turning a mixer on/off, collecting signals from the DO and pH sensors and calculating the derivatives of the DO and pH profiles for detecting the presence of toxicants were realized with the DAC module and a program coded with LabVIEW 8.2.

Each of the two reactors was filled with 2.7 L activated sludge, which had been procured from the same pilot plant mentioned above. As in the case of the batch assay, changes on the pH and DO profiles drawn during nitrification were monitored to detect the inhibition to the activity of nitrifiers. A program logic for screening wastewater to identify the presence of potential toxicants is presented in Figure 3.

In case of the reactor with a pH probe, initially 300 mL wastewater (10% of the total reactor working volume) was added. The amount of sample provided to the toxicity detector was minimized to reduce the NH₄⁺ concentration in the reactor and to more rapidly detect the AV on the pH profile. Once the feeding of wastewater was completed, air was supplied to maintain the reactor DO at 4.0 mg·L⁻¹. Then, the reactor pH was monitored to calculate dpH/dt for identifying the AV on the pH profile (i.e., the termination of nitrification). Once the AV was detected, the activated sludge was allowed to settle down and the supernatant was removed from the top of the reactor. The new wastewater was then added to the reactor for the next run (Figure 3). If the time elapsed to detect the AV in the current measurement was 10% longer than the one of the previous measurement, the PC issued a warning.

In case of the reactor with a DO probe, the oxygen uptake by the activated sludge was monitored to calculate the OUR. If the calculated OUR was 20% or more lower than the one determined in the previous measurement, the PC issued a warning for the inflow of potential toxicants.

Batch assay tests were performed and the effects of the toxicants (i.e., Hg²⁺, ATU, and CH₃Cl) on the nitrification process were evaluated by monitoring the pH and DO profiles. During the nitrification, in general, the system pH first decreases due to the H⁺ from nitrification process. However, if the nitrification is completed, the pH increases again via ammonification, producing a local minimum (i.e., AV) on the pH profile (Figure 4(a–c)) [22].

Under normal conditions, the AV on the pH profile could be identified easily within less than an hour. However, if the activity of nitrifiers was negatively affected, the time for the AV detection could be prolonged. Figure 4 demonstrates that the ending time of nitrification and the time elapsed to detect the AV on the pH profile are identical. In this specific experiment, the nitrification was complete within 40–50 min after the aeration was initiated; the completion of the nitrification is characterized by the AV on the pH profile. In fact, the time for the AV occurrence on the pH profile was inversely proportional to the nitrification rate, and was proportional to the amount of a toxic chemical added. In other words, the time would be proportional to the degree of the nitrification inhibition, which is calculated using Equation (2): (2) Percent   nitrification   inhibition ( % ) = AUR normal − AUR with   toxicant AUR normal × 100 %where, AUR_normal and AUR_{with toxicant} are the ammonia utilization rate (mg NH₄⁺ L⁻¹·min⁻¹) under the normal condition and in the presence of a toxicant, respectively.

When 5 mg·L⁻¹ Hg²⁺ was present in the test wastewater (Figure 4(a)), the AV could not be observed on the pH profile, indicating very slow (or no) nitrification had occurred. In fact, based on the calculation with data provided in Figure 4(a), more than 70% nitrification inhibition was observed. ATU and CHCl₃ caused significant inhibition to the activity of nitrifiers even at lower concentrations; ATU of 0.5 mg·L⁻¹ or more and CHCl₃ of 0.25 mg·L⁻¹ or more could induce almost complete nitrification inhibition (Figure 4(b,c)). In the case of CHCl₃, the nitrification was inhibited even at 0.25 mg·L⁻¹ CHCl₃. In fact, CHCl₃ concentration of as low as 0.1 mg L⁻¹ could inhibit the activity of nitrifying activated sludge (data not shown).

From Figure 4(b), when 2.5 mg·L⁻¹ Hg²⁺ or 0.25 mg·L⁻¹ ATU was present in the test wastewater, the AV occurrence on the pH profile was delayed for about 20 min. Since the activity of the activated sludge does not change within several hours under the normal operating condition, significantly delayed nitrification can be regarded abnormal. Therefore, the upper limit for the time delay of the AV occurrence was arbitrarily set at 10%. This means that wastewater flowing into the WWTP of interest may inhibit the activity of the activated sludge if the time for detecting the AV in the current measurement is delayed 10% or more than the previous measurement.

The percent nitrification inhibition was also assessed empirically using the changes in the OUR measurements as shown in Equation (3). In this specific experiment, the OUR value was calculated to be 0.095–0.1 mg O₂ min⁻¹ under normal conditions. At 2.5 mg·L⁻¹ Hg²⁺ or 0.25 mg·L⁻¹ ATU, however, the OUR of the activated sludge in the batch reactor was lowered more than 20% (Table 3); the calculated OUR was lower than 0.08 mg O₂ min⁻¹. Therefore, measuring the OUR of the total activated sludge also could play an auxiliary role in the detection of the nitrification inhibition. In fact, 20% respiration inhibition has been suggested as a guideline for deciding if test wastewater is “safe” or “toxic” to activated sludge in literature [15]: (3) Percent   respiration   inhibition ( % ) = OUR normal − OUR with   toxicant OUR with   toxicant × 100 %where, OUR_normal and OUR_{with toxicant} are the oxygen utilization rate (mg O₂ L⁻¹·min⁻¹) under the normal condition, and in the presence of a toxicant, respectively.

As discussed above, both dpH/dt and dDO/dt of the batch reactors were monitored along with pH and DO in the presence or absence of Hg²⁺, ATU, or CHCl₃ (Figures 5–7). When no toxic chemical was present in the test wastewater, the dpH/dt of the reactor changes from (−) to (+) at the AV, although some noises were observed. In order to reliably identify the AV on the pH profile, the strategy suggested in a previous study [31] was adopted. Namely, the AV (A in Figure 4(a)) on the pH profile was identified by detecting B (dpH/dt = −0.001) and C (dpH/dt = 0.001) on the dpH/dt profile in sequence.

The OUR (or dDO/dt) of the activated sludge was determined by performing a linear regression; the linear regression was carried out using the data collected when DO concentration was between 5 and 7 mg·L⁻¹. Therefore, if the slope of the linear regression curve for the DO consumption in the current run was 20% or more less steeper than the one measured in the previous run, it was considered that the test wastewater contained toxicity-causing substances.

The feasibility of the proposed strategy was evaluated by applying it to screen wastewater with different concentrations of Hg²⁺. As shown in Figure 5(a), the inflection point (=AV) could be detected on the pH profile at 25 min after the aeration was initiated for the wastewater with 0 mg·L⁻¹ Hg²⁺. However, it could be detected at time 35 min for the wastewater with 2.5 mg·L⁻¹ Hg²⁺. Interestingly, the proposed method could not distinguish the toxicant-free wastewater and the wastewater with 1.25 mg·L⁻¹ Hg²⁺. No distinct delay was observed in the AV detection time for the wastewater with Hg²⁺ (28 min), comparing the test with the Hg²⁺-free wastewater (26 min). The dDO/dt in the absence of Hg²⁺ was −0.443 mg·L⁻¹·min⁻¹, while that in the presence of Hg²⁺ was −0.428 mg·L⁻¹·min⁻¹ (Figure 5). It indicates that the activity of the nitrifiers was not significantly inhibited by 1.25 mg L⁻¹ Hg^2+. In fact, it has been reported that Gram negative bacteria such as nitrifying bacteria could reduce Hg²⁺ to Hg(0) to some extent with NADPH inside the cell and could thus detoxify it [29,30,32].

However, in the presence of 2.5 or 5 mg·L⁻¹ Hg²⁺ in the test wastewater, the AV on the pH profile was detected 5 min or more later than under normal conditions. As shown in the Figure 5(b), dDO/dt also indicated the nitrification being inhibited when the test wastewater contained Hg²⁺ of 2.5 mg·L⁻¹ or more, showing its value of −0.340 mg·L⁻¹·min⁻¹ or higher.

In the case where ATU or CH₃Cl was present in the test wastewater, similar trends could be detected in the pH and DO profiles (Figure 6). However, the result from the test with ATU clearly showed the detection of nitrification inhibition based on the dpH/dt would be superior than the one based on the OUR. Even at 0.25 mg·L⁻¹ ATU, the occurrence of the AV on the pH profile delayed for about 5 min, comparing to the one at 0 mg·L⁻¹ ATU; a warning was issued by the system. However, the DO profile drawn in the presence of 0.25 mg·L⁻¹ ATU was not much different from the one in the absence of ATU (Figure 6(b)). If the toxicity detector system was operated only depending on the DO profile, a warning would not be issued.

At higher concentrations (i.e., 0.5 and 0.75 mg·L⁻¹), both the dpH/dt and dDO/dt profiles indicated the activity of the nitrifiers was being inhibited; in fact the AV could not be observed in the presence of ATU of 0.5 mg·L⁻¹ or more. On the other hand, the dDO/dt for the test wastewater was also increased to reveal the nitrification inhibition.

The result of the test with CHCl₃ as a toxicant is shown in Figure 7. Apparently, the activity of nitrifiers was seriously inhibited by the presence of CHCl₃. In the presence of 0.25 mg·L⁻¹ CHCl₃, the occurrence of the AV on the pH profile was delayed for more than 10 min, indicating that the activity of nitrifiers inhibited. However, the DO profiles did not show significant difference between with and without 0.25 mg·L⁻¹ CHCl₃. Increase of dDO/dt which was smaller than the alarming criterion for the OUR was observed; 18% of OUR change was observed.

At higher concentrations (i.e., 0.75 and 1.25 mg·L⁻¹), both the dpH/dt and dDO/dt profiles indicated the activity of the nitrifiers was being inhibited; in fact the AV could not be observed in the presence of CHCl₃ of 0.75 mg·L⁻¹ or more. On the other hand, the dDO/dt for the test wastewater with 0.75 mg·L⁻¹ or more was almost similar to that of wastewater with 0.25 mg·L⁻¹. This means that the activity of nitrifiers could be severely inhibited by CHCl₃ at a very low concentration. In addition, the oxygen consumption shown in Figure 6(b) might only represent the respiration by the heterotrophs in the detector system.

The result from the tests with three different toxicants at three levels is summarized in Table 4. In general, the strategy based both on the OUR and pH profiles could successfully screen test wastewater for potential inhibition to the nitrification process, except the wastewater containing 1.25 mg·L⁻¹ Hg²⁺.

Based on the result from the batch bioassay experiments, a nitrification inhibition detector was built to continuously screen incoming wastewater by real-time calculating dpH/dt and dDO/dt (i.e., OUR). Once new wastewater was fed into the two reactors of the detector, the detector evaluates if the wastewater contained toxic substances inhibiting the activity of nitrifiers both by checking the time taken for the AV to occur and by calculating the OUR.

If the time for the AV occurrence (i.e., dpH/dt = 0) on the pH profile in the current run was 10% or more longer than the previous run, the detector was programmed to issue a warning. For the first 5 min when new wastewater was fed into the pre-tester, the reactor dpH/dt was not monitored.

In addition, if the OUR for the current run was 20% or much smaller than in the previous run, the detector was programed to issue a warning. Therefore, a warning signal could be issued if new wastewater for the current run could not pass either the screening criteria based on the dpH/dt or the one based on the OUR.

First, the system was operated for the wastewater without any toxicant to obtain the DO, dDO/dt, pH, and dpH/dt profiles (Figure 8). In Figure 8(a,b), the DO and pH profiles show little change over time, except the pH decrease in the first four runs. Nonetheless, the dDO/dt and dpH/dt did not issue a warning, indicating the activated sludge in the detector system was not inhibited. Over more than 20 h, the AVs were successfully detected by the system; AVs occurred in the range between 6.95 and 7.25 with the mean of 7.03 and standard deviation of 0.06.

However, if test wastewater contained toxicity-causing substances, the occurrence of the AV on the pH profile was delayed. For example, when 2.5 mg·L⁻¹ Hg²⁺ was present in the test wastewater, the system could easily identify the inhibition of nitrifiers with both the delayed occurrence of the AV on the pH profile and the reduced OUR (Figure 9).

As shown in the last section, if a toxicant exerting low acute toxicity to nitrifying activated sludge is present in the influent wastewater to a WWTP, the system may not screen the wastewater. In fact, 1.25 mg·L⁻¹ Hg²⁺ or 0.25 mg·L⁻¹ ATU did not inhibit the activity of nitrifiers much; only little delay of the AV occurrence on the pH profile could be observed. However, if wastewater containing a low concentration of chemicals continuously flows into a WWTP, it could cause a chronic toxicity in its biological processes.

Therefore, the system was applied to screen the wastewater containing a low concentration of toxicant. After each test for the nitrification inhibition, the system discharged the test wastewater and was replenished with new wastewater for the next test. Figure 10 represents the system response to the continuous inflow of the wastewater containing 1.25 mg·L⁻¹ Hg²⁺. For the first three cycles, the system could not detect the presence of a toxicant in the influent. However, from the 4th cycle, the relative chronic effect of Hg²⁺ of 1.25 mg·L⁻¹ on the nitrifying bacteria delayed the occurrence of the AV on the pH profile significantly enough for the system to issue an warning. In fact, even with Hg²⁺ of such a low concentration, the activity of the nitrifiers was affected; the time of the AV occurrence was delayed in each run of the test (+0.5, +1.3, +7.7, and +10 min for 1st through 4th run, respectively). On the other hand, a significant change in dDO/dt could be observed in the 5th test.