Evaluating the performance of a wastewater treatment plant (WWTP) requires a good characterization of the biodegradable substrate entering the plant. As is generally acknowledged, the wastewater characteristics under rain or storm weather conditions vary significantly from dry weather conditions but this is rarely accounted for in modelling exercises. In order to address this defect, a measurement campaign was set up and samples were analysed using respirometric assays. In this paper, some hurdles regarding measurement campaigns under dilute wastewater conditions and a better exploration of the impact of using respirometric assays are described and discussed. The dependence of the heterotrophic yield on different substrates severely hampers the evaluation of the respirograms under dilute wastewater conditions. In addition, the low load conditions limit the application of the assays due to insufficient sensitivity of the experiment and uncontrolled oxygen inputs. The results clearly demonstrate the need for further research in order to allow for a proper evaluation of WWTP performance under rain and storm weather conditions.

INTRODUCTION

Any (model-based) design or optimization study of biological wastewater treatment plants (WWTPs) requires a determination of the biological degradability of the carbonaceous substrate present in the wastewater. The biological degradable carbonaceous substrate (bCOD) determines largely the oxygen demand and its variation clearly also affects denitrification, enhanced biological phosphorus removal and high rate activated sludge systems' performance. In addition, the available oxygen supply at a WWTP is regularly limiting for nitrification during storm events and, as such, an accurate estimation of the oxygen demand related to the bCOD allows for a better appraisal of the available capacity for nitrification. Furthermore, the unbiodegradable particulate chemical oxygen demand (COD) affects the sludge production, whereas the unbiodegradable soluble COD is a major fraction of the effluent COD (Choubert et al. 2013).

Temporal (daily, seasonal and weather dependent) variation in flow rates and loading is an acknowledged phenomenon at WWTPs (Tchobanoglous et al. 2004; Henze 2008) and heavily influences the performance of the WWTP. However, the corresponding temporal variation in biological degradable (carbonaceous) substrate (bCOD) and its ratio over total COD, has received far less attention. Choubert et al. (2013) showed examples where, during dry weather, the readily biodegradable COD (SB) fraction varies between 5% and 25% of the total COD over a period of a few hours. Lagarde et al. (2005) sampled two different types of sewers during several rain weather days. The analysis revealed variations in total COD concentrations from 220 to 515 mg/L. The readily biodegradable fraction ranged between 24% and 32% of the total COD, while the slowly biodegradable fraction ranged from 40% to 49%.

Despite the increased attention, still no measurement campaign has demonstrated the (fast) dynamics occurring under rain and storm water conditions with regard to the biodegradable COD fraction. Understanding these dynamics requires extensive measurement campaigns. In a first attempt, Lagarde et al. (2005) set up a measurement campaign in order to investigate the use of respirometry for influent fractionation for samples collected under various rain conditions. However, the total COD concentration for their samples lies between 220 and 515 mg COD/L, which is in the high range, considering that Bixio et al. (2000) reported an average total COD value of about 300 mg COD/L for the WWTP of Ghent, Belgium.

Quantifying the biodegradable substrate entering a WWTP is essential for its proper evaluation, both under dry and rain (or storm) weather conditions. In order to quantify the variations, which are acknowledged to be significant, separate samples were collected at the WWTP of Roeselare (Belgium) and a measurement campaign was set up in rain weather conditions at the WWTP of Eindhoven (The Netherlands). The samples collected were consecutively analysed using both chemical-physical analyses and respirometric assays allowing the determination of the wastewater fractions.

This paper highlights some of the hurdles encountered due to the impact of low COD and bCOD concentrations, typical for rain or storm weather conditions, on respirometric assays. Possible improvements are proposed and the improved experiments are assessed for their applicability in high frequency (e.g. hourly composite samples) measurement campaigns.

MATERIAL AND METHODS

The respirometric analysis applied in this work comprises of measuring the dissolved oxygen concentration and subsequently deriving the oxygen uptake rate (OUR) in order to determine the biodegradable COD fraction in the wastewater.

The respirometric unit consists of a 2 L double-glass vessel, kept at a constant temperature of 20°C by means of a cooling system (Lauda Alpha RA8; VWR), which pumps water through the heat-jacked reactors. The sludge is constantly mixed at a speed of 100 rpm and aerated with the aid of an aeration stone at a constant airflow rate.

LabView (National Instruments, USA) is used for monitoring, control and data acquisition of the respirometer. Dissolved oxygen (DO) and pH are measured using an LED dissolved oxygen (LDO) probe (LnPro68701/12/220; Mettler Toledo, Elscolab) and pH probe (GA405-DXK-S8/120 PN: 104054287; Mettler Toledo, Elscolab). The probes are connected respectively to an oxygen transmitter (M400 Type 2) and a pH transmitter (Knick Stratos 2401). Communication between the transmitters and LabView is based on 4–20 mA signals. Before starting a respirometric assay, the pH probe and LDO probe and their communication to the transmitters are calibrated. The LDO probe has an accuracy of 1% (with a minimum of 8 ppb) and a response time after which 98% of the signal has been reached (t98) of less than 20 seconds.

Respirometric assays

Several types of respirometric assays exist and can be classified according to the behaviour of both the liquid and gas phase (Spanjers et al. 1998; Young & Cowan 2004). In this research, the flowing gas–static liquid (GF-LS) method and the static gas–static liquid (GS-LS) method are applied.

At the start of both assays, the reactor is filled with 1.9 L of mixed liquor, which was aerated overnight until endogenous respiration was reached, collected from the biological tank. Nitrification is inhibited by adding allylthiourea (ATU) to the reactor at a concentration of 10 mg/L. Finally, the reactor is operated depending on the selected method and the OUR is calculated by making a general mass balance for dissolved oxygen over the liquid phase (Spanjers et al. 1998).

Integrating the surface under the OUR curve provides the total amount of oxygen consumed at the expense of all degraded substrate. From the amount of oxygen consumed, the bCOD is calculated (Equation (1)), for which knowledge of the heterotrophic biomass yield (YOHO) is indispensable (Barnett et al. 1998; Vanrolleghem et al. 1999; Petersen 2000). For municipal wastewater, a value of 0.67 gCOD/gCOD for YOHO is deemed appropriate and generally accepted (Kappeler & Gujer 1992; Fall et al. 2011). 
formula
1
where bCODR is the bCOD in the reactor (mg/L) and tfinal is the endpoint of the integration interval, i.e. the time instant for which the endogenous state is reached again (min). The concentration of biodegradable COD in the wastewater (bCODWW) is then determined by correcting for the dilution (Vanrolleghem et al. 1999; Orhon & Okutman 2003; Gatti et al. 2010).

The flowing gas–static liquid (GF–LS) respirometric assay

During the GF–LS method the batch reactor is continuously aerated (flowing gas) while the liquid is kept in the reactor (static liquid) (Spanjers et al. 1998). For this assay, the following consecutive steps are proposed: (i) the determination of the equilibrium dissolved oxygen concentration (SO2,Eq), (ii) the estimation of the oxygen transfer coefficient (kLa) and (iii) the measurement of the exogenous oxygen uptake rate (OURex).

Gas flow, bubble size, reactor dimensions, stirring of mixed liquor (turbulence and sludge flocculation state), temperature of mixed liquor, and air pressure have a major influence on kLa. Therefore, the following conditions must be ensured during the subsequent experiments (Ros et al. 1988): (i) a constant airflow through the whole experiment, (ii) a reactor with known volume and shape has to be used for all measurements, (iii) constant stirring must be provided and (iv) a constant temperature of mixed liquor during the measurements.

The third step, finalising the respirometric assay, is the determination of the exogenous OUR by adding a specific volume of substrate (in this work wastewater) to the mixed liquor and measuring the dynamic response of dissolved oxygen concentration. Based on this dynamic response, the OUR profile can be derived from Equation (2). 
formula
2

The static gas–static liquid (GS–LS) respirometric assay

During the GS–LS respirometric assays, aeration in the batch reactor is stopped (static gas) before dosing the sample while the liquid is kept in the reactor (static liquid) (Spanjers et al. 1998). For this assay, the following consecutive steps are performed: (i) aerating the mixed liquor in the reactor until a dissolved oxygen concentration of about 8 mg/L is reached, (ii) stopping the aeration and monitoring the (endogenous) respiration rate until the oxygen concentration has dropped about 1.5 mg/L and (iii) dosing the substrate and monitoring of the total OUR, which is the sum of both endogenous and exogenous respiration.

During this type of assay, the mass balance reduces to Equation (3) (Drtil et al. 1993; Gernaey et al. 2001). 
formula
3
During phase I, oxygen is utilized at a constant rate, which is determined by the endogenous activity (OURI) of the microorganisms (Figure 1). At a certain time, a known volume of substrate is added to the batch reactor resulting in a temporary increase in respiration rate (OURII) due to substrate degradation, i.e. exogenous activity (phase II). The increased respiration can be noticed by an increased slope in the DO curve. If the substrate is completely degraded and removed (phase III), the respiration returns to the endogenous activity (OURIII), with the same slope in the DO curve as before the substrate addition.
Figure 1

DO-profile obtained after addition of 13.6 mg of sodium acetate trihydrate to 1.9 l of mixed liquor of the WWTP of Roeselare.

Figure 1

DO-profile obtained after addition of 13.6 mg of sodium acetate trihydrate to 1.9 l of mixed liquor of the WWTP of Roeselare.

According to Vanrolleghem (2002) the differential term in Equation (3) can be approximated with a finite difference term resulting in the following derivation for bCOD (Equation (4)), which will be referred to as direct parameter abstraction method. 
formula
4
where bCODWW is the biodegradable COD in the wastewater [mg/L] and ΔtS is the time needed to degrade the biodegradable matter present in the wastewater (min).

For the direct parameter abstraction method, in order to calculate the confidence intervals on the calculations, the combined standard uncertainties are calculated using the summation in quadrature method (BIPM et al. 1993). Multiplying the combined standard uncertainty with a coverage factor of 2 (for a confidence level of 95%) yields the expanded uncertainty, which can be used as an estimate to the confidence intervals.

Automatic parameter estimation for dynamic process models

As an alternative to the direct parameter abstraction method, as described above, a model-based evaluation, using dynamic process models including activated sludge models (ASM) (Henze et al. 2000), is evaluated as well. The model was implemented in the WEST® modelling and simulation software (mikebydhi.com) taking into account the respirometer configuration. The configuration reflects the actual set-up and consists of a sample container, two timers (for controlling the timing and volume of the sample dosed and for controlling the aeration) and the respirometer. The biological reactions in the respirometer are dynamically modelled based on ASM. Both ASM1 or ASM3 (Henze et al. 2000) were selected in order to account for the impact of the assumptions made in the models with regard to the possible storage of substrate.

To fit the model to the experimental datasets, automatic parameter estimation experiments are performed. During parameter estimation, a unique set of model parameters, resulting in the best fit of the model prediction to the experimental data, is searched for. This is achieved by minimization of the root mean square error between simulated and measured oxygen profiles. Two different optimization algorithms are used, namely the simplex method (Nelder & Mead 1965) or the Praxis method (Brent 1973).

In first instance, for the GS–LS respirometric assays, the initial values describing the endogenous respiration behaviour of the respirometer are estimated on the oxygen profiles before the addition of the wastewater sample. In second instance, the concentration of biodegradable COD in the sample is estimated with the total curve. In the last instance, the uncertainty bounds on the estimated parameters are determined using the inverse of the Fisher Information Matrix (FIM) (Donckels 2009).

RESULTS AND DISCUSSION

Dependence on the yield of heterotrophs

Respirometric assays according to the GFLS method, with dosing of acetate as readily biodegradable substrate, were performed in triplicate in two identical respirometers. The respirometers were filled with activated sludge collected at the WWTP of Eindhoven and 250 ml of a solution of sodium acetate trihydrate with a concentration of 92 mg COD/l was dosed. The calculated SB fraction equals to a concentration of 52.9 mg/l (with a standard deviation of 8.6%) for the first respirometer and to 61.2 mg/l (with a standard deviation of 8.3%) for the second respirometer.

The underestimation of the SB fraction, well below the dosed COD concentration, might be attributed by a real heterotrophic yield that is different from the default heterotrophic yield of 0.67 g COD/g COD as used in the evaluation. To confirm the possible difference in the yield, it has been recalculated based on the experimental data, according to Equation (5) (Majone et al. 1999; Strotmann et al. 1999). The recalculated yield amounts to 0.795 g COD/g COD (with a standard deviation of 2.9%), which proves to be significantly higher than the default yield value. 
formula
5
where Yobs is the observed yield [mg cell COD/mg COD], ΔO2 is the change in oxygen concentration [mg/l] and CODdeg is the COD removed.

Respirometric assays, similar to the assays with acetate, were performed with glucose. For this respirometric assay, the activated sludge was collected at the WWTP of Roeselare and 250 ml of a solution of glucose with a concentration of 213.5 mg COD/l was dosed in quadruplicate. After calculation of the concentration of biodegradable substrate with the default value for YOHO of 0.67 g COD/g COD, a concentration of only 59.6 mg COD/l (with a standard deviation of 7.7%) was obtained. Recalculating the yield according to Equation (5) provides a value of 0.91 g COD/g COD (with a standard deviation of 0.8%).

High yield values for the heterotrophic bacteria have also been reported by Guisasola et al. (2005) (0.666, 0.726, 0.757 and 0.792 g COD/g COD). They attribute the high yield values to excessively available substrate and extensive storage. The storage of polymers (usually, polysaccharides and lipids) can be caused by a feast and famine regime, which is present at the WWTP of Eindhoven, the origin of the sludge. As such arguing for the dependence of YOHO on (the history of) the biomass. Dircks et al. (1999) reported higher values for the heterotrophic yield for different substances, 0.72 g COD/g COD for acetate and 0.91 g COD/g COD for glucose. Goel et al. (1999) also reported high YOHO values for the degradation of glucose (0.9 g COD/g COD). Moreover, McCarty (2007) clearly demonstrated the variability of YOHO, in a dataset on yields for different individual substrates, which they used for the calibration of their thermodynamic electron equivalents model for bacterial yield prediction. From the dataset, the yield was calculated to range from 0.37 for methanol up to 0.65 for formate (gCOD cell/gCOD substrate). In a global sensitivity analysis, Cosenza et al. (2014) used, based on a literature review, a variance of the yield ranging from 0.38 up to 0.75 (gCOD cell/gCOD substrate).

As can be clearly deduced from Equations (1) and (4), the yield of heterotrophs is indispensable to determine the biodegradable COD from respirograms. The value of YOHO is important as a variation of 10% leads to a change of 18% in estimation of the biodegradable fraction (Spérandio & Paul 2000). However, as explained above, the yield of heterotrophs has a considerable variability related to both the (history of the) biomass and to the substrate consumed.

Depending on the length of the previous dry weather period and the intensity and length of the rain event the content of organic particulate matter changes significantly during the course of the rain events (Stumwohrer et al. 2003). Boogaard et al. (2014) listed stormwater quality data retrieved from a database based on data monitoring projects and compared them with data from the USA and other European countries. The quality data are significantly different from dry weather data. Also, typical ratios differ considerably compared to dry weather. The dry weather COD over BOD5 ratio mean value (2.06) reported by Rieger et al. (2012) is smaller than the ratio (5.61 for The Netherlands, 6.23 for Europe and 6.11 for the USA) calculated from the storm weather data presented by Boogaard et al. (2014). This higher ratio may be attributed to the scouring of deposited sludge, which may have undergone sludge stabilization reactions resulting in less organic biodegradable matter.

As the wastewater matrix changes the yield factor is also likely to change. As a consequence, there is a need to determine this yield factor for the wastewater composition during storm events. This, however, is beyond the scope of this work but has a strong recommendation for future projects. However, this additional experiment lengthens the procedure for determining the influent fractionation and may become a bottleneck when a measurement campaign is set up requiring a high sampling frequency.

Low load conditions

At the WWTP of Roeselare, influent wastewater samples were taken Monday mornings around 9 am on 24 March, 7 April and 14 April 2014. All the samples were analysed the next 2 days in the laboratory (Table 1). All 3 days were dry weather days. However, the first day and, in a lesser extent, the second and third day flow rates surpassed the normal dry weather flow considerably. The increased flow rate is probably attributed to the high infiltration rate caused by the extreme rain events in February and March 2014. The possibility of high infiltration rates was confirmed by consulting the ground water levels at the Flemish subsoil database (https://dov.vlaanderen.be). The total COD (CODT) is within the same range for the 3 days although it shows a moderate increase with decreasing flow rates, which could strengthen the hypothesis of the dilution by infiltration. BOD10 proves to be weaker the first day, indicating a lower biodegradability, which is confirmed by the wastewater fractionation.

Table 1

Influent fractionation and ratios in respect to total COD of influent at the WWTP of Roeselare for sampling during dry weather days (Monday mornings around 9 am on 24 March, 7 April and 14 April 2014)

   24/03/14
 
07/04/14
 
14/04/14
 
Date Method Unit Mean StdevaMean StdevaMean Stdeva
Qb Logged m3/h 974 N.A. 730 N.A. 680 N.A. 
CODT Chemical mg/l 196 0.8 205 221 
BOD5c Respirometry mg/l 40.2 0.0 59.2 58.5 
BOD10c Respirometry mg/l 50.7 0.0 69.8 70.4 
bCOD Respirometry mg/l 22.4 21 19.3 12 63.7 16 
bCOD STOWAd mg/l 59.2 0.3 82.8 0.7 86.4 
SB Respirometry mg/l 11.6 22 14.7 13 18.8 12 
SB STOWAd mg/l 34.9 29.1 17 33.6 
XCB Respirometry mg/l 10.8 20 4.6 44 44.9 17 
XCB STOWAd mg/l 24.3 53.7 52.8 12 
bCOD/CODTe Calculation 11.4 21 9.4 12 28.9 16 
SB/CODTe Calculation 5.9 22 7.1 13 8.5 13 
XCB/CODTe Calculation 5.5 20 2.2 44 20.3 17 
   24/03/14
 
07/04/14
 
14/04/14
 
Date Method Unit Mean StdevaMean StdevaMean Stdeva
Qb Logged m3/h 974 N.A. 730 N.A. 680 N.A. 
CODT Chemical mg/l 196 0.8 205 221 
BOD5c Respirometry mg/l 40.2 0.0 59.2 58.5 
BOD10c Respirometry mg/l 50.7 0.0 69.8 70.4 
bCOD Respirometry mg/l 22.4 21 19.3 12 63.7 16 
bCOD STOWAd mg/l 59.2 0.3 82.8 0.7 86.4 
SB Respirometry mg/l 11.6 22 14.7 13 18.8 12 
SB STOWAd mg/l 34.9 29.1 17 33.6 
XCB Respirometry mg/l 10.8 20 4.6 44 44.9 17 
XCB STOWAd mg/l 24.3 53.7 52.8 12 
bCOD/CODTe Calculation 11.4 21 9.4 12 28.9 16 
SB/CODTe Calculation 5.9 22 7.1 13 8.5 13 
XCB/CODTe Calculation 5.5 20 2.2 44 20.3 17 

aStandard deviation in %.

bInstantaneous value at the time of sampling.

cBOD is performed in duplicate, while the other analysis were done in triplicate.

dPhysical-chemical fractionation method according to the STOWA protocol (Hulsbeek et al. 2002).

eRatios calculated based on the results of the respirometry tests.

In the first 2 days, the biodegradable fraction (bCOD) determined by respirometry is substantially lower than the bCOD determined with the STOWA method (Hulsbeek et al. 2002). On the last day, the difference between the methods is reasonably small. This indicates that the physical–chemical fractionation methods tend to estimate higher fractions of biodegradable matter in case of low biodegradability, conditions which seem to occur in rain weather or under a large infiltration. This could indicate that some of the biodegradable matter is only degraded in the BOD10 tests, which are applied in the STOWA protocol, and not in the short term respirometry experiments. In addition, the ratio of readily biodegradable substrate determined by respirometry (SB) to the total wastewater COD ranges between 5.9 and 8.5%. For the slowly biodegradable substrate (XCB), this ratio ranges between 2.2 and 20.3%. These low ratios may also be attributed to the high infiltration rate caused by the extreme rain events in February and March 2014.

Wastewater samples obtained on 25 February and 3 March at the WWTP of Roeselare were taken during wet weather conditions and analysed using respirometric assays according to the GFLS method. After determination of the kLa, 250 mL of wastewater sample was added. As can be seen in Figure 2, the DO-profile (left) shows the expected drop of DO immediately after the addition the wastewater sample. However, instead of an enduring decrease, the DO unexpectedly starts increasing again up to a level higher than the previous equilibrium DO concentration.
Figure 2

DO-profile (left) and OUR-profile (right) after the addition of 250 mL of a dilute PST influent to a batch reactor containing 2.0 L activated sludge (with 10 mg/L ATU to block nitrification).

Figure 2

DO-profile (left) and OUR-profile (right) after the addition of 250 mL of a dilute PST influent to a batch reactor containing 2.0 L activated sludge (with 10 mg/L ATU to block nitrification).

To eliminate the possibility that the organisms in the activated sludge did not have enough essential nutrients, limiting the degradation of the biodegradable substrate, sodium phosphate and ammonia sulphate were added to the activated sludge. However, after adding these nutrients, the same respirometric response was observed upon addition of the wastewater samples. Another possible cause for the low bCOD values could be the presence of toxic components in either the sludge or the wastewater. To check for the possible presence of toxic components, acetate and diluted synthetic wastewater were dosed to the batch reactor. This resulted in a fast, normal respirometric response (results not shown) for both substrates, implying normal microbial activity and, hence, the absence of toxic components.

COD measurements of the wastewater samples taken during wet weather conditions yielded a COD concentration ranging between 31.2 mg/L and 179.0 mg/L. In addition, 5-day BOD measurements were performed, resulting in a BOD5 value ranging between 30.3 mg/L and 43.7 mg/L indicating a low presence of biodegradable substrate in these wastewater samples. In comparison, samples obtained during dry weather conditions had a BOD5 value ranging between 58.5 mg/L and 145 mg/L. Apparently, the dilute wastewater samples contain mainly COD fractions that cannot be degraded during the short-term respirometric experiments.

Due to these large unbiodegradable COD fractions, the respirometric test meets its limits in regard to its sensitivity. I.e. the rate of oxygen consumption of the microorganisms in the activated sludge during substrate degradation does not exceed the oxygen supply, resulting in DO and OUR-profiles as shown in Figure 1. Therefore, the aeration was lowered to the minimal aeration rate (0.5 L/min) to reduce the oxygen supply. In an attempt to further decrease the oxygen supply, the aeration stone was removed to create larger air bubble sizes, resulting in a lower specific area for mass transfer and less efficient oxygen transfer. However, the same respirometric profile was still observed indicating that the oxygen supply was not reduced sufficiently.

The technical specifications of the dissolved oxygen probes may also play a role in the observed behaviour. The main factor playing a role in the dissolved oxygen probes used here is the response time (t98), which amounts to about 20 seconds. The profile may display a time lag but will represent the same trend as the real dissolved oxygen profile.

Uncontrolled oxygen input

Another phenomenon witnessed during the tests is the unexpected increase in DO concentration after the dosing of the sampled wastewater (Figure 3). This increase in oxygen concentration is probably caused by an increased oxygen input during the addition of the sample. Gas flow, bubble size, reactor dimensions, stirring of mixed liquor (turbulence), temperature of mixed liquor, and air pressure have a major influence on oxygen transfer and the kLa. Due to the relatively large sample volume that is added, temporary swirls are created in the reactor, creating air-bubbles and a temporary higher oxygen transfer. As such, the conditions, which must be ensured during the subsequent experiments to justify the assumption of a constant kLa (Ros et al. 1988), are not met.
Figure 3

Effect of the oxygen increase on the DO-profile (left) and the OUR-profile (right) obtained after addition of 250.0 mL of PST influent (266.0 mg COD/L) to a batch reactor containing 1.9 L activated sludge (with 10 mg/L ATU to block nitrification).

Figure 3

Effect of the oxygen increase on the DO-profile (left) and the OUR-profile (right) obtained after addition of 250.0 mL of PST influent (266.0 mg COD/L) to a batch reactor containing 1.9 L activated sludge (with 10 mg/L ATU to block nitrification).

In order to test this hypothesis of the higher oxygen transfer, 250 mL of distilled wastewater was added to the reactor, to check if the same profile would be obtained. Upon addition of the sample, an immediate drop followed by a rapid increase in DO concentration is observed (Figure 4). Because there is no biodegradable substrate present in the distilled water, the DO concentration goes slowly back to the saturated DO concentration, while a faster decrease in DO concentration is observed in Figure 3 due to the degradation of biodegradable matter present in the wastewater sample.
Figure 4

DO-profile obtained after addition of 250 mL of distilled water to a batch reactor containing 1.9 L activated sludge (with 10 mg/L ATU to block nitrification).

Figure 4

DO-profile obtained after addition of 250 mL of distilled water to a batch reactor containing 1.9 L activated sludge (with 10 mg/L ATU to block nitrification).

This phenomenon of uncontrolled increased oxygen input is only witnessed because of the low biodegradable COD fractions. With higher substrate concentrations, the phenomenon is completely outweighed by the more pronounced biodegradation reactions. Furthermore, the phenomenon complicates the evaluation of the respirogram using the direct parameter abstraction method.

Another explanation for the observed behaviour could be that the samples contain substantially higher dissolved oxygen concentrations than the reactor (i.e. significantly higher than 8 mg/L). In the case of the wastewater this is rather unlikely as the collected wastewater was stored in closed barrels (20 L) with little or no headspace. In addition, the wastewater was stored overnight (at 4°C), which may cause some of the oxygen to be depleted before the whole sample reaches 4°C.

Changing the initial substrate concentration to biomass concentration ratio

Since the respirometric protocol yielded extremely low results on bCOD for dilute wastewater samples, several attempts were made to improve the respirometric response upon addition of the water samples. First of all, the initial substrate concentration to initial biomass concentration (S0/X0) was adapted. A low S0/X0 results in a tall and narrow curve due the fast utilization of the biodegradable substrate, while a high S0/X0 gives a shallow and wide OUR-curve (Ekama et al. 1986). In order to evaluate the effect, several S0/X0 combinations were examined.

First, a batch test was performed with diluted sludge to create a higher S0/X0 ratio (0.011 g COD/g VSS). 1.0 L of activated sludge was diluted with 1.0 L of distilled water. To mimic the wastewater composition under rain weather condition, influent of the PST was diluted with effluent of the WWTP. Despite the higher S0/X0 ratio, the DO-profile and OUR (Figure 5) remain narrow and low, when dosing 250 mL of the diluted wastewater (125.3 mg COD/L) to the batch reactor containing the diluted sludge (flowing gas–static liquid method). A total biodegradable substrate concentration of 13.2 mg/L is obtained, consisting of 9.7 mg/L readily biodegradable substrate and 3.5 mg/L slowly biodegradable substrate, which is still a meagre fraction of the total dosed COD.
Figure 5

DO-profile (left) and OUR-profile (right) obtained after the addition of 250 mL of a 125.3 mg COD/L wastewater solution to a batch reactor containing 1.9 L diluted activated sludge (with 10 mg/L ATU to block nitrification).

Figure 5

DO-profile (left) and OUR-profile (right) obtained after the addition of 250 mL of a 125.3 mg COD/L wastewater solution to a batch reactor containing 1.9 L diluted activated sludge (with 10 mg/L ATU to block nitrification).

A drawback of performing experiments with diluted sludge is the long time necessary for determining the kLa value. Normally, it takes approximately 40 min to determine one kLa value, but with diluted sludge it takes 1 h 40 min (about 5 h for a triplicate determination). Due to this elongated procedure, this method hampers the application for high-measuring frequencies (e.g. for hourly composite samples).

Another test was performed creating an even higher S0/X0 ratio (0.038 g C0D/g VSS) through adding 0.40 L of dilute wastewater (103.7 mg COD/L) to the batch reactor containing 0.50 L of activated sludge. After addition of the wastewater sample, the exogenous respiration rate reaches a maximum of 40.4 mg/L/h and then decreases to a value lower than the endogenous respiration rate. Thereafter, OURex increases again to OURend (Figure 6). Due to the addition of 400 mL to 500 mL of sludge, the total reactor volume almost doubles upon addition of the sample. The air bubbles can stay longer in the mixed liquor and there is a strong dilution of the activated sludge. These factors possibly change the oxygen transfer (kLa), resulting in an error in the calculated OUR values. A better OUR-profile could be obtained, if the change in kLa value and oxygen transfer would be accounted for in the calculation of the OUR value, for which dynamic process models seem to be appropriate. However, these models suffer from identifiability issues (Guisasola et al. 2005) and may require additional measurements during the respirometric tests (Gernaey et al. 2002).
Figure 6

DO-profile (a) and OUR-profile (b) obtained after the addition of 400 mL of a 103.7 mg COD/L wastewater solution to a batch reactor containing 500.0 mL concentrated activated sludge (with 10 mg/L ATU to block nitrification).

Figure 6

DO-profile (a) and OUR-profile (b) obtained after the addition of 400 mL of a 103.7 mg COD/L wastewater solution to a batch reactor containing 500.0 mL concentrated activated sludge (with 10 mg/L ATU to block nitrification).

Reducing oxygen input

In order to further reduce the oxygen input for the analysis of the diluted wastewater samples, the aeration was stopped before the addition of the sample, i.e. the GS–LS respirometric assays. The method is validated on the basis of an experiment dosing about 0.1 mol of sodium acetate trihydrate (CH3COONa.3H2O) and the theoretical COD for the sample was calculated (5.6 mg/L). The experimental results were analysed using the direct parameter abstraction method and also using a dynamic model based on either ASM1 or ASM3 (Table 2).

Table 2

Concentration of biodegradable substrate bCOD after addition of sodium acetate trihydrate determined using static gas-static liquid respirometry

Method Note Quantity Value 
Sample CH3COONa.3H2Dry weight 13.6 ± 0.1 mg/L 
ThCOD 5.6 ± 0.1 mg/L 
Direct parameter abstraction YOHO = 0.67 bCOD 4.7 ± 0.3 mg/L 
YOHO = 0.795 bCOD 7.6 ± 0.3 mg/L 
ASM1 YOHO = 0.67 bCOD 4.9 ± 0.2 mg/L 
YOHO = 0.795 bCOD 7.4 ± 0.3 mg/L 
ASM3 YOHO = 0.63 bCOD 7.7 ± 0.4 mg/L 
Method Note Quantity Value 
Sample CH3COONa.3H2Dry weight 13.6 ± 0.1 mg/L 
ThCOD 5.6 ± 0.1 mg/L 
Direct parameter abstraction YOHO = 0.67 bCOD 4.7 ± 0.3 mg/L 
YOHO = 0.795 bCOD 7.6 ± 0.3 mg/L 
ASM1 YOHO = 0.67 bCOD 4.9 ± 0.2 mg/L 
YOHO = 0.795 bCOD 7.4 ± 0.3 mg/L 
ASM3 YOHO = 0.63 bCOD 7.7 ± 0.4 mg/L 

The direct parameter abstraction method, using the default heterotrophic yield (0.67 g COD/g COD), provides a slightly too low estimate for the biodegradable COD (4.7 mg/L vs. a ThCOD of 5.6 mg/L). Applying the higher yield (0.795 g COD/g COD), as associated to the degradation of acetate before, provides a too high estimate (7.6 mg/L).

Estimating bCOD with the ASM1 model, using the default yield (0.67 g COD/g COD), also provides a reasonable estimate (4.9 mg/L vs. a ThCOD of 5.6 mg/L). Although the use of dynamic models is assumed to be hampered by identifiability issues (Gernaey et al. 2002; Orhon et al. 2007), the uncertainty bounds are relatively small, resulting in small confidence intervals (respectively 0.2 mg/L, 0.3 mg/L and 0.4 mg/L, Table 2). The offset in the values could be attributed to an incorrect YOHO value or the fact that not all acetate has been degraded. Applying the methods that could indicate the occurrence of storage, i.e. the higher YOHO value (0.795 g COD/g COD) determined before and the use of ASM3, overestimate the bCOD (respectively 7.4 and 7.7 mg/L vs. a ThCOD of 5.6 mg/L) similar as with the direct parameter abstraction method. This possibly indicates that storage of readily biodegradable substrate is not significant, which is contradictory to the flowing gas–static liquid experiment.

After the validation of the method, the experiment was repeated with 250 mL of dilute wastewater sample (52.7 mg COD/L). During this method, there is no aeration, so the oxygen decrease due to substrate degradation should be visible upon addition of the sample. However, no increase in respiration rate due to substrate degradation was observed (Figure 7, left). Moreover, an increase in DO concentration was observed after sample addition. The same increase in DO concentration upon addition of 250 mL of distilled water is observed (Figure 7, right). Similar to the flowing gas–static liquid method, this increase in DO concentration could be caused due to the sample addition, creating swirls in the reactor, inducing reaeration.
Figure 7

DO-profile obtained after addition of (left) 250 mL of PST influent (57.3 mg COD/L) and (right) 250 mL of distilled water to 1.9 L active sludge (arrow indicating the addition of the substrate).

Figure 7

DO-profile obtained after addition of (left) 250 mL of PST influent (57.3 mg COD/L) and (right) 250 mL of distilled water to 1.9 L active sludge (arrow indicating the addition of the substrate).

To evaluate the minimum biodegradable substrate concentration leading to an increased respiration rate after sample addition, dilute wastewater samples with different COD concentrations were made (Table 3). For the dilute wastewater sample with a COD concentration of 52.4 mg/L, no increased respiration rate upon addition of the sample was visible. For the other samples, with a higher COD concentration, a visible respirometric response was observed. After calculation of the biodegradable substrate concentration, a very low value is obtained. The calculated concentrations of the biodegradable substrate in the wastewater sample are probably an underestimation of the real biodegradable substrate concentrations, as was the case with acetate.

Table 3

Concentration of biodegradable substrate bCOD in dilute wastewater sample determined with static gas-static liquid respirometry

COD of sample (mg/L) bCOD (mg/l) Stdev (%) bCOD/CODT (%) Stdev (%) 
52.4 B.Da B.Da B.Da B.Da 
80.2 4.1 83.2 5.1 4.2 
110.5 9.1 25.9 8.2 2.1 
158.9 12.5 28.8 7.9 2.3 
COD of sample (mg/L) bCOD (mg/l) Stdev (%) bCOD/CODT (%) Stdev (%) 
52.4 B.Da B.Da B.Da B.Da 
80.2 4.1 83.2 5.1 4.2 
110.5 9.1 25.9 8.2 2.1 
158.9 12.5 28.8 7.9 2.3 

aBelow detection limit.

Due to the above observations, a parameter estimation experiment was performed in WEST. The same configuration was used as described for acetate. For the determination of the endogenous state of the activated sludge in the batch reactor, an additional experiment was performed, during which the activated sludge was aerated for 3 h. The initial biomass concentration of the heterotrophs and the kLa were estimated, so that the simulated conditions of endogenous respiration were matching to the real experimental conditions of endogenous respiration. It was noticed that the longer the experiment continued the estimated biomass concentration in the respirometer increased from 0.98 g/L to 4.28 g/L. Indeed, the longer the experiment lasted, the lower (more negative) the slope of the curves were. At the beginning of the experiment the endogenous respiration rate is 0.0018 mg/L.s, while at the end of the experiment (approximately 4 h later) the endogenous respiration rate is 0.0079 mg/L.s. This increase in respiration rate cannot only be caused by the growth of biomass due to substrate addition. Another possible explanation is the presence of slowly biodegradable substrate present in the wastewater samples, which could not be degraded in the short time frame. So this means that the microorganisms in the batch reactor are not in the endogenous state because they are still degrading slowly biodegradable substrate. However, endogenous conditions of activated sludge in the beginning of the test are crucial for a correct determination of the biodegradable substrate present in a dosed sample. Waiting until the slowly biodegradable substrate is degraded before the aeration is turned on again is not an option because of oxygen limitations. Alternatively, oxygen limitation can be avoided by a regular reaeration of the batch reactor.

General discussion

The work described shows the need for further investigation of respirometric assays, in particular for the application for storm water characterization. The research should either be directed into more sensitive assays, taking special care to avoid uncontrolled (and undesired) oxygen input, or into the possibility of concentrating the wastewater, e.g. by means of membrane filtration. For the latter, special attention needs to be addressed to the effect of the pore size of the filters for the characterization.

In order to obtain more sensitive assays, special care should be given to the assumptions made. One of these assumptions is the kLa considered to be constant. Many factors play a role in the kLa, including the surfactants. These surfactants degrade during the assay and, as such, alter the kLa. Possible approaches to account for changes in the kLa are either the elimination of the dependence on the kLa (e.g. using alternating aeration or hybrid respirometers (Vanrolleghem & Spanjers 1998)) or a more rigorous evaluation of the respirograms (e.g. using dynamic ASM).

A second issue that needs to be addressed in future research is the determination of the yield factor. A question that arises is whether the fractionation method or the model should be adapted, i.e. the model could address more components groups of which the yield is more stable. Fractionation methods could then heavily rely on more chemical techniques such as gas or liquid chromatography combined with the detection methods mass spectrometry or UV-VIS spectrophotometry.

CONCLUSIONS

In order to evaluate the performance of a WWTP under rain or storm weather conditions it is crucial to have a clear view on the quantity of biodegradable substrate entering the plant under such conditions. In order to quantify the variations in biodegradable substrate, separate samples were collected and a measurement campaign was set up in rain weather conditions. Several factors hamper the application of the respirometric assays to quantify the biodegradable COD.

First, the dependence on the yield of heterotrophs is demonstrated in accordance with previously published values. An accurate estimation of the yield factor is important as a 10% variation leads to a change of 18% in the estimation of the biodegradable COD. Moreover, the changing yield factor for different substrates is important for the evaluation under rain and storm water conditions as these conditions alter the wastewater characteristics drastically. This observation of altered wastewater characteristics should be considered in further model developments.

Secondly, the low load conditions confront the respirometric assay with its limitations. Attempts changing the initial substrate concentration to biomass concentration were not successful. Moreover, diluting the sludge enlarged the necessary time to perform the assays. Another option to improve the sensitivity of the assays, but also not entirely successful, was to attempt eliminating the oxygen input by applying the static gas–static liquid respirometric assay.

The evaluation of the respirometric assays using the direct parameter abstraction method proves to be delicate under these extreme conditions. The evaluation using dynamic models based on ASM proves to be promising but needs to address the issues of identifiability. In addition, further studies could investigate the possibility of concentrating the wastewater samples by membrane filtration.

This work is an important step in the better exploration of the impact of dilute wastewater conditions on biodegradable substrate fraction in wastewater.

ACKNOWLEDGEMENTS

The authors thank the plant staff of the WWTP of Roeselare and the research department of Aquafin NV for the assistance. In addition, the authors would like to express their gratitude to Waterboard the Dommel for the great discussions and the financial support making this work possible.

REFERENCES

REFERENCES
Barnett
M. W.
Stenstrom
M. K.
Andrews
J. F.
1998
Dynamics and Control of Wastewater Systems
.
CRC Press
,
Lancaster, PA
,
USA
.
BIPM, I., I. IFCC, and I. IUPAC
1993
OIML, Guide to the Expression of Uncertainty in Measurement
.
International Organisation for Standardization
,
Geneva
,
Switzerland
.
Bixio
D.
van Hauwermeiren
P.
Thoeye
C.
2000
Impact of Primary Treatment Technologies on BNR: The Case Study of the STP of Ghent
.
The 5th IWA international symposium on systems analysis and computing in water quality management (Watermatex 2000),conference proceedings, Ghent, Belgium
.
Boogaard
F.
van de Ven
F.
Langeveld
J.
van de Giesen
N.
2014
Stormwater quality characteristics in (Dutch) urban areas and performance of settlement basins
.
Challenges
5
,
112
122
.
Brent
R. P.
1973
Algorithms for Minimization Without Derivatives
.
Prentice-Hall
,
New York
,
NY, USA
.
Choubert
J. M.
Rieger
L.
Shaw
A.
Copp
J.
Sperandio
M.
Sorensen
K.
Ronner-Holm
S.
Morgenroth
E.
Melcer
H.
Gillot
S.
2013
Rethinking wastewater characterisation methods for activated sludge systems – a position paper
.
Water Science & Technology
67
,
2363
2373
.
Cosenza
A.
Mannina
G.
Vanrolleghem
P. A.
Neumann
M. B.
2014
Variance-based sensitivity analysis for wastewater treatment plant modelling
.
Science of the Total Environment
470
,
1068
1077
.
Dircks
K.
Pind
P. F.
Mosbaek
H.
Henze
M.
1999
Yield determination by respirometry – the possible influence of storage under aerobic conditions in activated sludge
.
Water SA
25
,
69
74
.
Donckels
B.
2009
Optimal Experimental Design to Discriminate Among Rival Dynamic Mathematical Models
.
Ghent University, Ghent, Belgium
.
Drtil
M.
Németh
P.
Bodik
I.
1993
Kinetic constants of nitrification
.
Water Research
27
,
35
39
.
Ekama
G. A.
Dold
P. L.
Marais
G. v. R.
1986
Procedures for determining influent COD fractions and the maximum specific growth rate of heterotrophs in activated sludge systems
.
Water Science & Technology
18
,
91
114
.
Fall
C.
Flores
N. A.
Espinoza
M. A.
Vazquez
G.
Loaiza-Návia
J.
van Loosdrecht
M. C.
Hooijmans
C. M.
2011
Divergence between respirometry and physicochemical methods in the fractionation of the chemical oxygen demand in municipal wastewater
.
Water Environment Research
83
,
162
172
.
Gatti
M. N.
García-Usach
F.
Seco
A.
Ferrer
J.
2010
Wastewater COD characterization: analysis of respirometric and physical-chemical methods for determining biodegradable organic matter fractions
.
Journal of Chemical Technology & Biotechnology
85
,
536
544
.
Gernaey
A. K.
Petersen
B.
Ottoy
J.-P.
Vanrolleghem
P.
2001
Activated sludge monitoring with combined respirometric–titrimetric measurements
.
Water Research
35
,
1280
1294
.
Goel
R.
Mino
T.
Satoh
H.
Matsuo
T.
1999
Modeling hydrolysis processes considering intracellular storage
.
Water Science & Technology
39
,
97
105
.
Guisasola
A.
Sin
G.
Baeza
J. A.
Carrera
J.
Vanrolleghem
P. A.
2005
Limitations of ASM1 and ASM3: a comparison based on batch oxygen uptake rate profiles from different full-scale wastewater treatment plants
.
Water Science & Technology
52
,
69
77
.
Henze
M.
2008
Biological Wastewater Treatment: Principles, Modeling, and Design
,
IWA Publishing, London, UK
.
Henze
M.
Gujer
W.
Mino
T.
van Loosdrecht
M.
2000
Activated Sludge Models: ASM1, ASM2, ASM2d and ASM3: IWA Scientific and Technical Report
.
IWA Publishing
,
London, UK
.
Hulsbeek
J. J. W.
Kruit
J.
Roeleveld
P. J.
van Loosdrecht
M. C. M.
2002
A practical protocol for dynamic modelling of activated sludge systems
.
Water Science & Technology
45
,
127
136
.
Kappeler
J.
Gujer
W.
1992
Estimation of kinetic parameters of heterotrophic biomass under aerobic conditions and characterization of wastewater for activated sludge modelling
.
Water Science & Technology
25
,
125
139
.
Lagarde
F.
Tusseau-Vuillemin
M.-H.
Lessard
P.
Héduit
A.
Dutrop
F.
Mouchel
J.-M.
2005
Variability estimation of urban wastewater biodegradable fractions by respirometry
.
Water Research
39
,
4768
4778
.
Nelder
J. A.
Mead
R.
1965
A simplex method for function minization
.
Computer Journal
7
,
308
313
.
Orhon
D.
Okutman
D.
2003
Respirometric assessment of residual organic matter for domestic sewage
.
Enzyme and Microbial Technology
32
,
560
566
.
Petersen
B.
2000
Calibration, Identifiability and Optimal Experimental Design of Activated Sludge Models
.
PhD Thesis
,
Ghent University
,
Belgium
,
362
.
Rieger
L.
Gillot
S.
Langergraber
G.
Shaw
A.
2012
Good Modelling Practice: Guidelines for Use of Activated Sludge Models
.
IWA Publishing, London, UK
.
Ros
M.
Dular
M.
Farkas
P. A.
1988
Measurement of respiration of activated sludge
.
Water Research
22
,
1405
1411
.
Spanjers
H.
Vanrolleghem
P. A.
Olsson
G.
Dold
P. L.
1998
Respirometry in control of the activated sludge process: principles
.
Scientific and technical reports 7
.
IWA Publishing
,
London, UK
, p.
48
.
Stumwohrer
K.
Matsche
N.
Winkler
S.
2003
Influence of changes of the wastewater composition on the applicability of UV-absorption measurements at combined sewer overflows
.
Water Science & Technology
47
,
73
78
.
Tchobanoglous
G.
Burton
F. L.
Metcalf, Eddy
Stensel
H. D.
2004
Wastewater Engineering: Treatment and Reuse
.
McGraw-Hill, Columbus, OH, USA
.
Vanrolleghem
P.
2002
Principles of respirometry in activated sludge wastewater treatment
. In:
Proceedings International Workshop on Recent Development in Respirometry for Wastewater Treatment Plant Monitoring and Control, Taiwan, Taipei
.
Vanrolleghem
P.
Spanjers
H.
Petersen
B.
Ginestet
P.
Takacs
I.
1999
Estimating (combinations of) activated sludge model NO. 1 parameters and components by respirometry
.
Water Science & Technology
39
,
195
214
.
Young
J. C.
Cowan
R. M.
2004
Respirometry for Environmental Science and Engineering
.
SJ Enterprises, Springdale, AR, USA
.