Evaluation of titration methods for volatile fatty acids measurement: effect of the bicarbonate interference and feasibility for the monitoring of anaerobic reactors

Volatile fatty acids (VFAs) are intermediate products of anaerobic digestion and have pKa typically between 4 and 5, providing buffering capacity in this pH range. High VFA concentrations lead to lowering of pH and inhibition of micro-organisms, resulting in instability and even system collapse. VFA build up is more likely to occur when there are shock loads due to the thermodynamic and kinetic limitations of the syntrophic acetogens and methanogens. Thus, VFA monitoring is essential to assess the operating conditions of anaerobic reactors, particularly when treating high-strength wastewaters.

VFAs can be analyzed by various methods: distillation, colorimetry, conductimetry, and especially gas and liquid chromatography, and titrimetry. The titration methods present several advantages for routine control, especially simplicity, speed and low cost. The main disadvantages of titrimetry are the impossibility of identifying the acid species and their imprecision due the presence of interfering compounds – e.g., alkalizing substances. The methods of Ripley (Ripley et al. 1986), Kapp (Kapp 1984 described by Buchauer 1998) and DiLallo (DiLallo & Albertson 1961) are widely applied for monitoring anaerobic reactors. In the last method, changes in calculation procedures – the modified DiLallo method (Cavalcanti & van Haandel 2000) – and using an ultrasonic bath for 10 minutes instead of boiling for 3 to remove CO₂ (Ribas et al. 2007) are reported. The abbreviations used in this paper for each method and the formulae used to calculate VFA concentrations are set out in Table 1.

DiLallo & Albertson (1961) developed the DiLallo method as simpler and more accurate for determining volatile acids compared to the distillation described in Standard Methods for the Examination of Water and Wastewater (APHA 1955, APHA et al. 2005). It is considered that bicarbonate and volatile acids account for most of the total alkalinity, measured by titration to pH 4 but bicarbonate must be accounted for to quantify the volatile acid alkalinity (VAA). The procedure is based on destruction of the bicarbonate ion, reducing the pH to 3.3, and its removal as carbon dioxide by boiling for 3 minutes. Subsequent titration to pH 4 and then to pH 7 measures mainly the alkalinity imparted by organic acids. The standard deviation of the results was found to be higher when VFA concentrations were below 250 mg/l, and the method is more accurate if the acid concentrations are above this.

The Kapp method (Kapp 1984 described by Buchauer 1998) was initially developed to monitor sludge digesters. It considers the interaction between alkalinity imparted by the volatile acids, at pHs between 5 and 4, and the carbonate buffer system /CO₂ over the measurement of VFA and bicarbonate alkalinity levels. Using an empirical and theoretical approach, the method is applicable to the quantification of VFAs and the determination of total, volatile acid and bicarbonate alkalinities.

The method developed by Ripley et al. (1986) arose in an attempt to simplify the relationship between volatile acids and alkalinity in anaerobic reactors through a dimensionless relationship between the intermediate and partial alkalinities, imparted by the relative buffering contributions of VFA and bicarbonate, respectively. It was initially applied to poultry manure digester samples. The authors found that the method was suitable for monitoring anaerobic reactors as the intermediate alkalinity/partial alkalinity ratio is analogous to the volatile acids/total alkalinity ratio, but easier to evaluate as it does not require volatile acids measurement.

The experimental procedures of the DiLallo and modified DiLallo methods (Cavalvanti & van Haandel 2000) are identical. They differ only in the conversion factor used to calculate VFA from the measured VAA. A factor of 1.2 is adopted in the modified DiLallo method, corresponding to conversion of the alkalinity in mg-CaCO₃/l into VFA concentration in mg-HAc/l (an alkalinity of 1.0 mg-CaCO₃/l corresponds to 1.2 mg-CH₃COOH/l). It also incorporates the assumption that the measured VAA corresponds to 85% of the VFA, so that the conversion factor adopted is equal to 1.41 (1.2 × 1/0.85). It is noted that McCarty (1964) originally proposed a factor of 0.85 for the analysis of total and bicarbonate alkalinity in anaerobic reactors, indicating that titration of the original pH to pH 4.0 only measures 85% of the VAA.

Owing to the constraints associated with the original DiLallo method, especially the possibility of volatile acid volatilization during boiling and the time required for the sample to cool, Ribas et al. (2007) suggested the use of 10 minutes of sonication as an alternative to remove CO₂ from the liquid, instead of boiling. They reported good results when they applied this procedure to the measurement of VFA in solutions containing equivalent concentrations of VFA and bicarbonate, between 50 and 1,000 mg/l.

The aim of this study was to evaluate the titration methods used for VFA measurement taking into consideration their sensitivity to bicarbonate, as bicarbonate is usually present in relatively high concentrations in anaerobic reactors and affects the potentiometric titration. Solutions containing various combinations of acetic acid and bicarbonate were analyzed to evaluate the effect of increasing concentrations of bicarbonate on the quantification of acids. Statistical analyses were performed to determine the significance of the interference of bicarbonate concentrations on each method, the differences in VFA measurements between the methods, and the accuracy of the methods compared to the theoretical VFA values. Finally, the titration methods were employed in determining anaerobically treated stillage samples from the work described by Mota et al. (2013), in which VFA monitoring played a crucial role in enabling process start-up and the maintenance of stability in a methanogenic reactor. The feasibility of the methods was evaluated considering the VFA levels measured in the effluent in all operating conditions and the convergence of the results.

To evaluate the accuracy of the titration methods and the bicarbonate interference on VFA measurement, nine solutions containing different concentration combinations of acetic acid and sodium bicarbonate were analyzed. The solutions were named according to the concentrations of acetic acid (A) and bicarbonate (B), in which the numbers 1, 2 and 3 correspond to concentrations of 100 mg/l, 1,000 mg/l and 4,000 mg/l, respectively: A1B1, A1B2, A1B3, A2B1, A2B2, A2B3. A3B1, A3B2, A3B3. These were prepared with acetic acid Synth^® (99.7%) and sodium bicarbonate Sigma^®. The sample volume was 50 ml and the analyses were performed in triplicate.

For the boiling procedure used in DiLallo's method, the beakers containing the samples were heated on magnetic stirrers (stirring was not applied). For DiLallo's method with sonication, the beakers were placed in a water bath in the ultrasonic equipment used for glassware cleaning. The pH of the solutions with initial pH below 5.75 (A1B1, A2B1, A2B2, A3B1, A3B2 and A3B3) was raised with concentrated NaOH solution to values of between 6.5 and 7.0 for the application of both Ripley's and Kapp's methods. The samples were titrated with sulfuric acid to pH 5.75, 5.0, 4.3, 4.0 and 3.3–3.5. After boiling or sonication, the pH of the samples was raised first to 4.0 and then 7.0. The normality of the sulfuric acid and sodium hydroxide used was between 0.02 N and 0.20 N, and 0.01 N and 0.40 N, respectively. The solutions of sulfuric acid were standardized with primary standard sodium carbonate (Na₂CO₃), while sodium hydroxide solutions were standardized with previously standardized solutions of sulfuric acid. The calculations for each method were performed as described in Table 1.

Treated sugarcane stillage from a two-stage anaerobic bioreactor (acidogenic-methanogenic) was analyzed for VFA levels from start-up to stabilization over 6 months' of operation, using the methods of Ripley, Kapp, DiLallo and modified DiLallo. The experimental apparatus and operating conditions are described by Mota et al. (2013).

Initially, the means of triplicate determinations, and their standard deviations and coefficients of variation were calculated for each method. Where the coefficient of variation exceeded 0.01, one of the three measurements was excluded. Finally, of the 216 results (9 solutions × 8 methods × 3) a total of 12 was excluded (Table 2).

Table 3 and Figure 1 show the mean concentrations of the volatile acids measured in the samples containing, theoretically, 100 (A1), 1,000 (A2) and 4,000 (A3) mg-HAc/l against bicarbonate concentrations of 100 (B1), 1,000 (B2), and 4,000 (B3) mg-HCO₃/l. In Figure 1, the profiles of the results obtained by DBM and DUM methods were, of course, the same as those from DB and DU methods, as the values for the modified DiLallo methods are calculated applying conversion factors of 1.41 and 0.94, when the VFA concentrations calculated by the original DiLallo method are <180 mg/l and >180 mg/l, respectively.

The methods of Ripley and DiLallo with sonication were the most affected by the bicarbonate interference, wherein the higher the bicarbonate concentration, the higher the VFA concentration reported. In Ripley's method, this can be attributed to the fact that bicarbonate contributes to alkalinity in the pH range 4.3–5.75, and therefore on the measured intermediate alkalinity and the calculated value of VFA. This also may explain the highest value of VFA measured by Kapp's method in solution A1B3, in relation to solutions A1B2 and A1B1.

The highest VFA, with increased bicarbonate concentration measured by DiLallo's method with sonication, indicates that sonication was not effective for removing CO₂ from the liquid. In previous experiments (data not shown), apart from boiling and 10 minutes of sonication, a control version was used that included no procedure for CO₂. Sonication for periods of 20 and 40 minutes were also tested. These experiments showed that, compared to the control, sonication contributes to the removal of CO₂, and works better at longer exposure times, although, in all cases, the effect was unsatisfactory and the VFA concentrations reported were strongly affected by the bicarbonate concentration. Better results using DiLallo's method with sonication were obtained by Ribas et al. (2007), probably because the method was applied to solutions containing equal concentrations of VFA and bicarbonate, and therefore the relative bicarbonate concentrations were lower.

Although at much lower intensity, bicarbonate interfered with VFA measurements by DiLallo's original and modified methods, when the acetic acid concentration was 100 mg/l. This indicates that a small proportion of the bicarbonate was not transferred to the atmosphere as carbon dioxide. Indeed, the DiLallo method is more accurate quantifying volatile acids when their concentrations exceed 250 mg/l.

The statistical significance of the bicarbonate concentration interference on the VFA measurements was verified by applying Spearman's non-parametric correlation test, since most results were not distributed normally. Correlations were considered significant when p < 0.05. From the results in Table 3, it can be seen that at lower concentrations of acetic acid, corresponding to 100 mg-HAc/l, the bicarbonate concentration interfered significantly in all the methods tested. For the samples containing acetic acid concentrations between 1,000 and 4,000 mg-HAc/l, the bicarbonate interfered significantly only on VFA measurements by the methods of Ripley and DiLallo with sonication.

Thus, DiLallo's and Kapp's methods were little affected by bicarbonate interference in the determination of VFA, unlike the methods of Ripley and DiLallo with sonication, wherein relatively high concentrations of bicarbonate lead to overestimated VFA values.

Analyses of variance (ANOVA) complemented with the Tukey test were applied, to identify the statistical differences between the methods. Despite being a parametric test, ANOVA is robust and applies even for samples with a degree of deviation from normal distribution.

Table 4 shows the statistical differences between the methods for each solution (α = 0.05). Excluding the comparisons between the methods of DiLallo and modified DiLallo, and between DiLallo with sonication and modified DiLallo with sonication – DB × DBM, DU × DUM – the greatest similarity was found between the methods of Kapp and DiLallo – Kapp × DB, Kapp × DBM. On the other hand, the biggest differences were between Kapp and DiLallo with sonication, modified DiLallo and DiLallo with sonication, and Ripley and DiLallo methods – Kapp × DU, DBM × DU, Ripley × DB.

Thus, the smaller the difference and dispersion between the observed and expected values, the lower the X² value. From this analysis, which considered all the measurements performed in triplicate (n = 27 for each method), DiLallo and modified DiLallo methods showed greater approximation and lower dispersion in relation to the theoretical values of VFA, followed by Kapp's method. The calculated X² values were: DB = 446.6, DBM = 470.9, Kapp = 1,113.6, DUM = 14,912.4, DU = 18,302.1 and Ripley = 32,183.9.

After determining the suitability of the methods of DiLallo and Kapp for VFA quantification in samples containing several concentrations of volatile acids and bicarbonate, the methods were used to monitor VFA in an anaerobic reactor. Samples of sugarcane stillage treated sequentially in acidogenic and methanogenic reactors coupled to membranes, during the experimental period, were analyzed periodically (Mota et al. 2013). In DiLallo's method, the beakers containing the samples were covered with aluminum foil during heating and boiling, to avoid VFA loss by volatilization.

Figure 2(a) shows VFA concentrations in the methanogenic reactor effluent during the start-up period, when VFA concentrations exceeded 250 mg/l. Figure 2(b) refers to the final period of operation, when the reactor was stable and there was a sharp reduction of VFA concentration in the effluent. The bicarbonate alkalinity was between −275 and 4,523 mg-CaCO₃/l (by adding NaHCO₃) during the first period of operation, and 411 and 1,153 mg-CaCO₃/l in the second period (data not shown).

The results obtained by Ripley's method were relatively close to those obtained by the other methods only when the acid concentrations were high, perhaps because the method was developed analyzing samples with high concentrations of volatile acids, ranging from 2,000 to 10,000 mg-HAc/l (Ripley et al. 1986).

Although reported by Cavalcanti & van Haandel (2000) and Ribas et al. (2007), these trials did not show that 3 minutes of boiling led to VFA losses high enough to underestimate the quantification of VFA in the standard solutions or anaerobic reactor effluent by the DiLallo method, compared to Kapp's method. This is probably because of differences in the sample boiling procedure. In fact, a major shortcoming of the DiLallo method is the difficulty of standardizing the boiling procedure (Lahav & Morgan 2004).

To verify the statistical differences between the methods, one-way ANOVA followed by the Tukey test were applied. At a significance level of 95%, the differences are considered significant when p values were less than 0.05 (Table 5).

When VFA concentrations ranged from 500 to 5,500 mg-HAc/l (Figure 2(a)), the methods of Ripley, Kapp, DiLallo and modified DiLallo were statistically equal, with the results obtained using the methods of Kapp and original DiLallo slightly closer (Table 5). However, when the VFA concentrations were lower, between approximately 30 and 200 mg-HAc/L, the results obtained using Ripley's method were significantly higher, while there were no statistical differences between the methods of Kapp, DiLallo and modified DiLallo. It is noted in relation to Figure 2(b) and the p-values (Table 5) that the results from Kapp's were closer to the original DiLallo method (DB) than to its modified version (DBM). This was not seen when these methods were applied to the standard solutions and is probably explained by the fact that, when the VAA levels were lower (<180 mg-CaCO₃/l) in the anaerobic reactor, the relative contribution of the phosphate ions to alkalinity was higher, and therefore the lowest conversion factor of VAA into VFA originally proposed in these situations (1.0) was most suitable.

With the standard solutions, DiLallo and modified DiLallo methods showed less dispersion and closer results to the theoretical values of VFA, followed by Kapp's method. It was found that Ripley's method and DiLallo's with sonication were highly sensitive to bicarbonate interference, leading to overestimated VFA measurements in proportion to the relative concentrations of bicarbonate in the samples. Applying these methods (except DiLallo with sonication) to the analysis of anaerobically treated stillage, it was observed that Ripley's method led to higher VFA values compared to the methods of Kapp, DiLallo and modified DiLallo, and that this difference was significant when the VFA concentrations determined by the other methods were below 250 mg/l. In view of the results of this study, the most suitable titration methods for VFA monitoring in anaerobic reactors were Kapp's and DiLallo's methods, which presented statistically equal and closer results, regardless of the levels of VFA in the effluent samples, and were little affected by bicarbonate interference. Kapp's method stands out due to its greater simplicity, standardization and applicability to the measurement of total and bicarbonate alkalinity.

The authors would like to thank the Brazilian government for the support provided for this research from the National Council of Technological and Scientific Development (CNPq) and the Research Support Foundation of the State of Minas Gerais (Fapemig).