Monitoring results: reactor COD concentrations and COD removal rates

Figure 77 presents CODp data as measured at the ABR feed and in ABR 5. The data shows no clear correlation with the seasonal factors included in the figure. CODp is difficult to measure accurately and intrinsically prone to large methodological error since it requires a difficult filtering step. The dataset presented in Figure 77 for instance included 3 incorrect, since negative, values which had to be removed for analysis. Two extremely high CODp outliers were also excluded based on comparison with turbidity measurements.

CODp data was generally very variable throughout the whole investigation period, especially in the years 2010, 2011 and 2012.

Figure 77: ABRin and ABR 5 turbidity and CODp concentrations, the light red areas indicate the warmest period of the year, the light blue areas indicate the wettest period of the year

Since turbidity is known to be a robust indicator for particulate wastewater content it was used to check the plausibility of the available CODp data. As can be seen on Figure 77, the turbidity measurements confirm that the particulate content of the wastewater did not correlate with the seasons.

As opposed to the CODp data, turbidity was very high in the first half year of operation which appears plausible since the digesters had just been started up. The linear reduction of the turbidity data supports its credibility since this is exactly what would have been expected to happen during reactor start-up. The CODp data on the other hand was unexpectedly low until the end of 2010. It also showed a very high variability in 2010 and 2011 which is not consistent with the comparably constant turbidity values.

This raises questions concerning the accuracy of the COD values measured in the first two years. It was therefore decided not to include them for the further analyses of the dataset, also since the operational conditions in 2010 were obviously not comparable to the following years.

The remaining CODp dataset was then subjected to statistical investigations in order to assess whether there were significant differences between the reduction rates of the two phases. Data was normally distributed. Paired-sample t-tests failed to reject the null hypotheses that CODp concentrations were similar for ABRin and ABR 5 across phases (see Table 27 for details). Significant increase of CODp

reduction from one phase to the next therefore appears statistically improbable.

The t-tests were however repeated with the turbidity data after asserting their normal distribution and, as opposed to the CODp values, showed a significant increase of the ABRin values and decrease of the ABR 5 values across the phases (see Table 27). This implies that also the particulate reduction from ABRin to ABR 5 increased from Phase I to Phase II.

Hot season Wet season

Start of operation

Phase I Phase II

0 100 200 300

01.01.2010 01.01.2011 01.01.2012 31.12.2012 31.12.2013

NTU/ mg CODpl-1

Turbidity ABR in Turbidity ABR 5 CODp ABR in CODp ABR5

Table 27: Details of t-tests investigating the difference between CODp and NTU values across phases seen, seasonal wastewater temperature fluctuation and wet seasons most probably strongly influenced the CODs digester effluent concentration. Rain infiltrating the piping system could have lead to dilution and therefore reduction of measured wastewater concentrations. The high CODs values during and after the warmest season of the year could be explained by an increase of particulate organics solubilisation in the digester accompanied by rising of general SMA.

The average ABRin CODs concentrations were 368 mg CODs l^-1 and 434 mg CODs l^-1 for Phase I and Phase II respectively (see Figure 79). Figure 78 however shows that the apparent increase of average feed CODs was caused by a larger fraction of measurements taken during the warm and dry season in the Phase II dataset (57%) than in the Phase I dataset (33%). It is therefore not possible to compare the treatment of both phases only based on the available CODs (and therefore CODt) data.

The CODs concentrations measured in the supernatant of ABR 5 on the other hand were significantly lower during the warm dry season of Phase II than of Phase I. This implies a significantly higher CODs

reduction in Phase II. Influence of rainwater can be excluded in this case since the relevant data points only lay within the dry season.

Figure 78: ABRin and ABR 5 CODs concentrations and measured wastewater temperature, the light red areas indicate the warmest period of the year, the light blue areas indicate the wettest period of the year

Figure 79 a and b present the average COD concentration values measured in the different reactor chambers in Phase I and II. The represented values were computed only with measurement results from 2012 onwards and exclude certain outliers for reasons explained above. The COD values measured on June 11^th, 2013 were not considered since these were extremely high, leading to

non-Hot season Wet season

normality of the complete CODp and parts of the CODt and CODs dataset. Plausibility checks with turbidity measurements were made which supported the decision to remove these values. No rain occurred on the day of or prior to the sampling. The reason why these values differed from the rest of the dataset could not be identified.

In order to take the above shown effect of seasonal variations on data into consideration, the following average reactor reduction rates were calculated as the average of all differences between corresponding ABRin and ABR 5 values measured on the same day:

In Phase I the average reduction from ABRin to ABR 5 implied by the available data is 35%, 49% and 26% for CODt, CODp and CODs respectively.

The calculated average reduction in Phase II is 58%, 73% and 50% for CODt, CODp and CODs

respectively.

The reactor reduction rates of all COD fractions within the same phase are obviously significant (see Figure 79 b so that further statistical testing was not deemed necessary in this case.

As to the significance of the treatment increase across phases, for reasons mentioned above, comparing the treatment efficiencies simply based on the available COD data could lead to wrong conclusions.

Available turbidity data for instance is believed to be less prone to analytical error than the CODp data.

It is therefore assumed to depict reality better in terms of particulate content. Thus, based on turbidity measurements a statistically significant increase of particulate, and therefore CODp, reduction across the phases is accepted as being the most credible scenario although CODp data itself indicates the opposite.

CODs concentrations on the other hand cannot be directly compared across phases because seasonal factors differently affected both datasets. Nevertheless a significant increase in reduction from Phase I to II is implied by the curve progression of the available data.

It is therefore concluded that COD reduction indeed increased significantly from Phase I to II. A quantification with the available data was however not possible.

Table 28 summarizes the outcomes of paired sample t-tests investigating the statistical significance of COD reductions measured across chambers during Phase I.

Figure 79 a and b: COD fraction concentration profiles as measured in reactor chambers, error-bars indicate standard deviations

The CODt reductions were statistically significant across all chambers. The only significant CODp

reduction however was measured between ABR 1 and 3, whereas the CODs concentration significantly decreased in ABR 1 and between ABR 3 and 5.

Table 28: Details of t-tests investigating the statistical significance of COD reductions measured across ABR chambers, Phase I

ABR in & 1 ABR 1 & 3 ABR 3 & 5 P Significance P Significance P Significance

CODt 0.001 yes 0.003 yes 0.02 yes

CODp 0.5 no 0.01 yes 0.5 no

CODs 0.0002 yes 0.6 no 0.03 yes

Table 29 summarizes the outcomes of paired sample t-tests investigating the statistical significance of COD reductions measured across chambers during Phase II.

Significant CODt reduction occurred until the rear chamber although not throughout all chambers.

Significant CODp reduction was only measured between ABR 2 and 3. ABRin CODp data could not be used in this test since it was not normally distributed. CODs concentrations only significantly declined in the first two ABR compartments.

Table 29: Details of t-tests investigating the statistical significance of COD reductions measured across ABR chambers, Phase II

ABR in & 1 ABR 1 & 2 ABR 2 & 3 ABR 3 & 4 ABR 4 & 5 P Significance P Significance P Significance P Significance P Significance

CODt 0.02 yes 0.09 no 0.005 yes 0.8 no 0.045 yes

CODp 0.9 no 0.02 yes 0.8 no 0.09 no

CODs 2*10^-6 yes 0.01 yes 0.06 no 0.9 no 0.1 no

No effluent BOD5 concentration was measured on this site.

Measured effluent concentrations were 336 (± 59) mg CODt l^-1 and 262 (± 64) mg CODt l^-1 in Phase I and II respectively. The effluent contained about 100 mg CODs l^-1 of non-biodegradable CODs. This result is based on two investigations performed during Phase II, both of which were done with quadruple measurements. Nonbiodegradable wastewater fractions strongly depend on user habits.

Since the population did not change significantly over the investigation period, the available value for Phase II is assumed to also be representative for Phase I. Consequently a large fraction of the COD leaving the reactor was still biodegradable.