A simplified method for estimating chemical oxygen demand (COD) fractions

Modeling has provided valuable insight into wastewater treatment plant (WWTP) optimization. When applied to activated sludge (AS) processes, modeling outcomes not only reduce operating costs but can also improve effluent quality (Vanrolleghem et al. 2003). In this paper, focus is placed on Activated Sludge Models (ASMs), which are based on chemical oxygen demand (COD) balances.

According to Henze (2000), Henze & Gujer (1987), Makinia (2010), Choubert et al. (2013) and Rieger et al. (2013) the COD fractions presented in Figure 1 have the following characteristics:

• Soluble Unbiodegradable (S_U): cannot be removed by any biological treatment process or waste AS, so it reaches the effluent unaltered. Its value limits the minimum soluble effluent COD concentration.

• Particulate Unbiodegradable (X_U): accumulates in the sludge impacting sludge production, mixed liquor suspended solids concentration and clarifier performance.

• Soluble Biodegradable (S_B): can be metabolized readily by microorganisms, i.e. transported to cells, and oxidized or converted into storage products or biomass. S_B affects denitrification, anoxic tank size and P removal.

• Particulate Biodegradable (X_B): is often not metabolized immediately by microorganisms. Before being taken up by heterotrophic biomass, X_B requires hydrolysis. In wastewater treatment modeling, ‘hydrolysis’ is understood as the conversion of slowly biodegradable substrate into a more readily biodegradable form. Like S_B, X_B is relevant for carbonaceous biological oxygen demand.

Because colloidal particles (0.001 to 1 μm) cannot be removed by sedimentation, chemical methods must be used (Metcalf & Eddy, Inc. (2013)). The ‘Water Environment Research Foundation’ (WERF) and the ‘Foundation for Applied Water Research’ (STOWA) influent fractionation methods contain a specific flocculation step to discriminate between colloidal and soluble fractions (Roeleveld & Van Loosdrecht 2002; Melcer 2003). Colloidal COD quantification is important in relation to plant-wide energy recovery and carbon footprint. Chemically enhanced primary treatment (CEPT) acts to improve removal of particulate COD (pCOD) and convert colloidal COD to particulate form allowing its removal. Primary settling increases the solid fraction of COD processed in anaerobic digestion. In consequence, biogas production increases when CEPT is used, with an associated positive effect in energy recovery. If, as consequence of coagulant addition, the soluble substrate, necessary for denitrification, is reduced, carbon must be added. Because of this, the energy cost savings by enhancing primary sedimentation would be reduced by the costs of both the coagulants and the supplemental carbon (Gori et al. 2013).

Ekama et al. (1986) proposed meticulous methods for estimating X_U, S_B and S_U. They estimated X_U first by calibrating the parameter of a steady state model as done in Marais & Ekama (1976). Second, S_B is estimated by one of three different methods depending on reactor operating conditions (flow-through AS, aerobic batch and anoxic batch). Third, S_U is calculated using filtered effluent COD for aerobic reactors operated with sludge age exceeding 5 days. Finally, the X_B could be estimated by calculating the difference between the total COD and the previously estimated fractions. Wentzel et al. (1995) calculated the S_B fraction and heterotrophic active biomass using bioassays to determine the oxygen utilization rate (OUR) through the application of respirometry tests. In this case OUR should be registered over about 20 hours.

Another common approach features respirometry tests under varying conditions, high and low ratios between wastewater COD and biomass (measured as volatile suspended solids, VSS). Spérandio et al. (2001) discuss this approach and the calibration of a mathematical model. The model they proposed represented a simplified form of ASM focused on the aerobic growth and death of heterotrophs, hydrolysis and adsorption. Their method established bCOD, X_B, and active heterotrophic biomass.

Hulsbeek et al. (2002) developed a procedure combining the biochemical oxygen demand (BOD₁₀) test and physico-chemical tests. Using their procedure, the following fractions can be determined: (I) bCOD is the sum of S_B and X_B. As BOD₅ does not represent bCOD, bCOD is based on daily BOD values observed during the BOD₁₀ test. (II) The method to determine S_B is based on a physical separation, which involves pre-flocculation of the sample with Zn(OH)₂ or other flocculants, followed by filtration (0.1 μm). As the results obtained are almost the same, using filtration or flocculants, either is acceptable. Ultimately, since both S_B and S_U pass through the filter, the concentration of S_U must be determined independently and then subtracted from the soluble COD to determine S_B. (III) S_U is calculated based on the concentration of inert COD in the WWTP's effluent. As for S_B, the soluble fraction is separated by filtration or flocculation. Unbiodegradable COD and X_U are calculated as the difference between the total COD and other COD fractions.

An example of an alternative approach using parametric assumptions for estimating COD fractions has been proposed by Gori et al. (2011). In this procedure, COD fractions are calculated from commonly measured water quality parameters like BOD₅, COD, and VSS. Additional information is also needed regarding WWTP particulate removal in primary sedimentation. Two, more uncommon but easily measured, water quality parameters are also required: the soluble non-biodegradable COD and pCOD. The ratio pCOD/VSS is a relevant input in this method, a relationship between pCOD and VSS between 1.07 and 1.87 was adopted, and is the basis for estimating the particulate and soluble COD fractions. BOD₅ is the only input required to determine the biodegradable (bCOD = 1.6 * BOD₅). Other COD fractions as S_S, X_B and X_U are calculated by subtraction from values determined above (Gori et al. 2011).

The literature abounds in methods for estimating wastewater COD fractions. The examples presented attest to the diversity available, i.e. alternatives to combinations of physical and bioassay methods. Sometimes, the authors fail to provide clear directions for implementing their methods, which has occasionally led to subjective data interpretation. The latter is pertinent to interpreting changes in slope and in the definition of areas associated with respirometry tests. The lack of a standard set of conditions and the subjectivity of data interpretation hinder the implementation of the procedures discussed (Spérandio et al. 2001).

In this paper, a method is presented for estimating COD fractions. It is based on dedicated lab-scale experimental methods for wastewater characterization. It consists primarily of bioassays and physico-chemical methods intended to produce a simple, readily interpretable alternative to the most frequently used methods. Evaluation of the proposed method's effectiveness involved a case study of COD fractions in domestic wastewater in Bogotá, Colombia, at two different sites. The results obtained were compared to reference values reported in the scientific literature.

The Salitre facility is the only WWTP in the Colombian capital, Bogotá. The influent enters the CEPT through a pumping station. The plant's capacity is not adequate to treat all of the wastewater discharged to the city's sanitary sewer system, so some of the influent is treated and some dumped (untreated) into the Bogotá River. There are many theories about the sedimentation and biological degradation occurring in the channel, and their effect on influent wastewater characteristics. Because of the WWTP's inability to treat the city's wastewater, an upgrade of Salitre WWTP and a second WWTP dubbed Canoas are planned. The design includes an AS process but many have doubts about the ability of such processes to treat Bogotá's wastewater. This has spurred research, one research line entailed constructing an AS pilot plant, known as Gibraltar, using a feed with characteristics reflecting those of the influent for Canoas' projected WWTP. There are still questions surrounding the actual quality and variability of the Canoas influent.

Two sources of domestic wastewater, Salitre and Gibraltar, were chosen for COD fractionation. The first comprises raw sewage from Salitre WWTP, the second is the raw sewage flowing into the Gibraltar experimental plant. Six experiments were conducted: four using Gibraltar samples and two using Salitre samples. Samples were kept in refrigerators for an average of 12 hours at below 4 °C and analyzed within 24 hours. AS from the aerobic reactor of the AS pilot plant was collected for the respirometry tests.

Respirometry is used to measure and interpret the biological oxygen consumption rate under well-defined conditions. As oxygen consumption is directly associated with biomass growth and substrate removal, respirometry helps monitor, model and control the AS process (Vanrolleghem 2002), and enabled tracking of OUR.

The first step involves collecting sufficient volumes of untreated wastewater and AS to fill the bioreactor vessel and provide samples for laboratory analysis – see Table 1.

The AS used in the experiment must be aerated, as it facilitates the endogenous phase and ensures residual substrate removal. Köhler (2008) says that at least 6 hours should be allotted for aeration.

If wastewater comes from an AS process, sludge samples should be collected from the same system. Otherwise, if AS comes from a different system, allowance must be made for sludge acclimatization. The AS is required to be acclimatized to the wastewater. During this acclimatization phase, the microorganisms have to adjust to their new surrounding environment before the assessment of COD fractions.

Once the bioreactor vessel and related elements are sterilized, installation must be carried out with great care; proper calibration of the DO and pH sensors proves particularly important. Likewise, optimal kinetics cannot be achieved for the microorganisms unless the temperature and pH are constant. For the former, Köhler (2008) suggests 25 °C. For the latter, the pH of the mixed liquor must be approximately 7 during testing to ensure the correct environmental conditions for the microorganisms' metabolic processes. Thus, an acid or base (H₂SO₄ or NaOH, 1 M) should be added automatically when pH deviates from the target value.

Food/microorganism (F/M) ratio is based upon the ratio of food fed to the microorganisms each day to the mass of microorganisms held under aeration. It is a simple calculation, using the results from the influent COD test and the mixed liquor volatile suspended solids (MLVSS) test. The former (food) is the product of WW COD concentration and its volume, whereas the latter (microorganism) is the product of the MLVSS concentration and its volume in the bioreactor vessel. Ekama et al. (1986), and Melcer (2003) argue that a 0.6 F/M ratio is ideal for producing easily interpretable respirograms.

Nitrification during respirometry tests results in modifications to the respirogram, thus complicating it and the COD fraction interpretation. To reduce the rate at which ammonium is converted to nitrate, Ekama et al. (1986), and Melcer (2003) propose the use of a nitrification inhibitor – e.g., N-Allylthiourea in concentrations of 5 mg/l.

Three days are needed to complete this step, though respirograms (Figure 4) should be made before this to confirm that the endogenous phase has been reached. The latter is identified by the OUR series displaying an asymptote roughly horizontal to the origin and with two significant drops in the curve (e.g. the lower section of Figure 4). Once this phase has been confirmed, the experiment is finished. The whole process can take up to 7 days.

The magnitude of decreasing slopes in the DO concentration graph (Figure 4, upper) represents OUR. Graphing these slopes against time produces a respirogram (Figure 4, lower) whose values represent the organic matter degradation phases.

The curve's two horizontal asymptotes establish well-defined areas (Figure 4, lower). The first, between the black OUR curve and the green line, expresses the readily biodegradable chemical oxygen demand (S_B). The second, between the black OUR curve and the red line, indicates biodegradable particulate chemical oxygen demand (X_B) (Dold et al. 1986; Ekama et al. 1986; Melcer 2003). The method explained here requires only calculation of the S_B COD (first area).

Processing respirograms does not need specialized computer tools; any program that plots the negative slopes associated with DO series attained via respirometry tests can be used (e.g. Microsoft Excel). However, for the data obtained, the code in Appendix 1 is useful. It enables estimation of OUR based on the oxygen consumption slopes and is written in the programming language ‘Python’ (‘Respirogram.py’). A Python interpreter and respective console (e.g. PyCharm) must be used. In order to guarantee a unified interpretation of the results use of the proposed code is recommended.

The method presented also showed advantages over those currently in place, the paper provides detailed explanations of the steps required for the method and establishes a unified interpretation of the results obtained.

This step requires calculation of the X_B as outlined by Roeleveld & Van Loosdrecht (2002): S_B is subtracted from the bCOD, with bCOD estimated by analyzing the BOD curve constructed from the daily BOD readings collected over a 10-day period (Step 1). To calculate bCOD, the Python code ‘TotalCOD.py’ has been used successfully (see Appendix 2). The proposed code is recommended to guarantee a unified interpretation of the results.

Table 2 shows the results from the tests performed using the method described. In Table 3 those results are compared to typical values noted in the literature for the same fractions.

For influent wastewater in Gibraltar, the largest proportional COD fraction is S_B. At Salitre, the largest COD fraction is X_B. The results for the total biodegradable fraction (S_B + X_B) in the wastewaters analyzed show that, on average, this component comprises 70% of total COD at Gibraltar and 88% at Salitre. The main difference arises from the X_U fractions, which were 24% (Gibraltar) and 3% (Salitre) on average, respectively (Table 2).

The comparison demonstrates that, in terms of averages, the proposed method yielded values similar to those reported in the reference literature. The matches were not exact as some reflected reference maxima and minima. Some even fell outside the reference value range.

The atypical values may simply be related to particularities of the wastewaters studied. On the one hand, Gibraltar's S_U and X_B fractions were close to the average reference values. On the other, the proposed method's S_B values for Gibraltar were close to the maximum reference values while its X_U values were close to the minima reported in the literature. At Salitre, the S_B and S_U results are consistent with the reference literature averages, while its X_B exceeds the maximum reference values and X_U is below their minima.

In the AS process, the selector concept entails the selective growth of floc forming organisms at the initial stage of the biological process by providing a high F/M ratio at controlled DO levels. Selector design is based on biokinetic mechanisms that result in S_B uptake by the floc forming bacteria under aerobic or anoxic conditions. The wastewater at Gibraltar has characteristics favorable for selectors; with selectors the oxygen demand should be proportionally less for wastewaters with characteristics like Gibraltar's, especially as selectors favor the use of S_B without oxygen, in turn reducing the COD load on the aerobic reactor.

In the rational procedure the pCOD/VSS ratio has been defined as being in the range 1.07 to 1.87. The data available for Gibraltar yield a pCOD/VSS ratio of 2.3. The difference between pCOD and VSS explains the huge variability in X_U.

The method discussed here enables COD fractionation. Step-by-step directions for COD fractionation are provided. The method is oriented towards providing uniformly interpretable results. Some problems associated with other methods used for COD fractionation are obviated, in particular their limited applicability stemming from subjective data interpretation.

This method does have limitations, however. These include the time required for bioassays to determine COD fractions (up to 7 days) and the need for a 10-day BOD test. Furthermore, this method requires expensive materials and equipment (e.g. a bioreactor). In light of this there is a clear need to develop methods with shorter response times. Given that water quality variability often has diurnal patterns, the unambiguous explanation of testing procedures is critical, especially for ASMs. Among other things, the method presented here does not offer enough temporal resolution to capture COD fraction variability, which is an issue that needs to be dealt with.

This study was funded by the Pontificia Universidad Javeriana as part of Project 4541. Additionally, COLCIENCIAS (Colombian Administrative Department of Science, Technology and Innovation) has provided financial support for the Doctoral Studies of the corresponding author.