Discussion of case study data

6.3.8.1. Plant feed characteristics

A constant reduction of community wastewater production over the entire period of investigation was observed through flow measurement campaigns. Wastewater flow to the plant was approximately 65% design flow at the start of operation in 2010 and 44% in 2013.

Apart from a slight decrease of connected people between 2010 and 2011 constant numbers of connected users can be assumed which are close to design estimations. It is therefore concluded that the organic load to the plant remained approximately constant over the entire period of investigation.

Well water measurements yielded a mean raw-water alkalinity of 468 mg CaCO3 l^-1 with a standard deviation of 59 mg CaCO3 l^-1.

6.3.8.2. Effect of flow surges on plant reactors

It is evident that at irregular intervals storm-water entered the sewer system increasing the hydraulic load to the plant. A significant flow increase was measured during such rainfall in 2012.

The ABR sludge appeared to be remarkably unaffected by these flow surges during Phase I: the sludge-levels in all chambers accumulated regularly, apparently without being influenced by strong sludge migration within the compartments.

The matter was entirely different during Phase II, where sludge accumulation soared during the wet season, most probably due to washed out sludge from the digester. There was also a strong shift of sludge towards the rear compartments (which was supported by TS and VS sludge concentration measurements) and obvious washout from the reactor. Whether SMA, as shown by the available data, truly increased over the wet season needs to be confirmed by future investigations. A possible reason for the observed activity increase could have been washout of active sludge from the digester to the ABR.

Field observations did not indicate obvious signs of strong water level fluctuations inside the ABR.

6.3.8.3. Estimated digester load and treatment

Figure 80 a to d present the loading and treatment parameters OLR, HRT, biogas production and effluent COD concentrations for BGD 2. They compare design estimations and the results based on the investigations carried out in Phase I and II.

The OLR is in this case based on the number of connected users (shown as cap m^-1 d^-1) since representative feed concentration data was not available. The load doubled in Phase II compared to Phase I and design. In terms of hydraulic load however, the adjustments made during Phase II established the load situation for which the digester had been initially designed.

Figure 80 a, b, c and d: Loading and treatment parameters of BGD 2 in Phase I and II: OLR, HRT, biogas production and digester effluent concentrations, error-bars indicate standard deviations

Increasing the load certainly had a strong impact on the treatment of the digester: the digester effluent alkalinity increased significantly indicating stronger anaerobic activity. The biogas production more than doubled from 4.8 m³ d^-1 to 12 m³ d^-1.

The slight digester effluent CODp concentration increase in Phase II was not found to be statistically significant. The data however certainly underestimates the real particulate washout in Phase II:

observations made on ABR sludge accumulation rates in Phase II before the wet season indicated an increase of more than 300% compared to Phase I. The digester also became much more sensitive to hydraulic surges in the wet season when more than fifteen times more sludge accumulated in the ABR than in Phase I. Possibly the digester had incidentally reached its maximum sludge capacity by the time the load change occurred. However, no noticeable sludge washout was recorded in the previous wet season. Also, communal biogas digesters operating under tropical climate are generally known to have excellent sludge stabilisation abilities. They certainly require desludging after much longer periods than the two years BGD 2 had been operating when the loading change occurred.

A mass balance calculation across the biogas digester was attempted with the available data using the methodology described in Section 3.8.1. The average flow measured in 2013 of 15.7 m³ d^-1 was used as the value for Q. Repeated biogas composition measurements by an external laboratory and the BORDA research team yielded approximately 80% CH4 content. 12 m³ d^-1 of biogas production indicate a daily COD reduction of about 22 kg COD d^-1. Added to the amount of COD leaving the reactor this implies an approximate average digester feed concentration of 1.900 mg COD l^-1 and a per capita COD production of about 52 g COD cap^-1 d^-1. Both these values however surely underestimate the reality since a certain fraction of the produced methane certainly escaped the reactor, dissolved in the effluent wastewater. The design per capita production of 30 g COD cap^-1 d^-1 therefore certainly underestimated the real value.

The above implies an average digester treatment efficiency of 73%. Again, this value most probably does not take all produced CH4 into consideration and therefore underestimates the real treatment efficiency.

The CODs increase presented in Figure 80d has been shown to have been mainly induced by seasonal variations. The CODs concentration was therefore not significantly different across the phases.

The design effluent concentration was much lower than what was measured during both phases.

Design Phase I Phase II

Design Phase I Phase I Phase I

Phase II Phase II Phase II

It is interesting to note that significantly more biogas was produced during April, the hottest month of the year. Also, turbidity measurements indicated a start-up period which lasted approximately six months for the digester effluent particle content to reach a constant value.

6.3.8.4. Estimated ABR load and treatment

A steady decline of wastewater feed flow was observed over the complete investigation period, apparently due to water scarcity.

Figure 81 a to d present the loading and treatment parameters OLR, HRT, effluent COD concentrations and average reduction rates.

The Q used for the OLR and HRT measurement calculations was the average of the flow measured in 2010 and 2011 divided by four streets and the flow measured in August divided by three streets. A variation of 20% was estimated.

In the absence of better data, the OLRs were calculated with the available COD concentration values although these were shown to not be necessarily comparable. Additional information needs to be considered in order to correctly interpret the graph.

The OLRs in both phases are comparable to the design assumption. The generally higher OLR in Phase II shown in Figure 81 a stems from an apparent CODs increase in the Phase II feed flow, which was however shown to be insignificant. Phase II CODp ABRin data however certainly strongly underestimates the real average value since digester sludge was repeatedly washed into the ABR during storm water events in 2013 which the measured CODp values do not account for. The actual OLR in Phase II was therefore probably much higher than design expectations.

The HRT indicates an approximately 50% design load under dry weather conditions in both phases. The average vup,max was 0.4 m h^-1 without rain water influence and therefore below the design value of 0.9 m h^-1.

Figure 81 a, b, c and d: Loading and treatment parameters of the first five ABR chambers in Phase I and II: OLR, HRT, effluent COD concentrations and COD reduction rates, OLR error-bars indicate combination of standard error of mean of CODt measurements and standard deviation of Q, all other error-bars indicate standard

Design Phase I Phase I Phase I

Phase II Phase II Phase II

Design Phase I Phase I Phase I

Phase II Phase II Phase II

6.3.8.4.1 Reactor load and performance

a) b) c)

a)

d)

Alkalinity and pH investigations indicated stable anaerobic treatment conditions throughout the reactors and operational phases.

The CODt treatment efficiency however was extremely low in Phase I with 35%. The comparably small CODs and CODp concentration reductions were found nevertheless to have been statistically significant.

A strong increase in treatment efficiency appears to have occurred from Phase I to Phase II during dry weather conditions although more measurements are needed to quantify this.

The available CODp data did not enable meaningful statistical testing because of its associated large measurement error. A significant increase in particulate reduction is however implied by the more reliable, since less error prone, turbidity measurements. But it is important to note that this only applies to dry-weather conditions since that data does not take rainfall into consideration. Rainfall however is known to have caused considerable sludge washout from the digester into the ABR. Also, the observed increase in turbidity reduction in Phase II correlates with a marked increase of sludge bed height which might have enhanced the filter effect of the reactor.

The CODs concentration values cannot be directly compared across phases for reasons explained above. The general trend of the data however strongly suggests improved treatment in Phase II.

A raise of CODs reduction implies an increase of sludge bio activity in Phase II. Two hypotheses are proposed to explain this phenomenon. Firstly, increase of ABR sludge activity was linked to the wash out of active sludge from the digester into the ABR in Phase II. Secondly, increase of ABR sludge activity was due to increased organic load in Phase II.

The CODp mass balance was calculated as detailed under Section 3.8.2:

The averages of measured values for Q, CODp concentrations of ABRin and ABR 5 and VS sludge concentration led to 4.2 (min = 1.9, max = 8.9) m³ y^-1 sludge increase in Phase I. It is assumed that the measured sludge VS concentration is representative for Phase I although it was measured during Phase II. Minimum and maximum values take into account a feed flow variation of 20%, the standard error of means of CODp concentrations and the standard deviation of sludge VS concentration.

Linear regression of sludge volumes measured in the first five ABR chambers led to a sludge build-up rate of 0.7 m³ y^-1. This is below the minimal rate calculated through mass balance. This discrepancy could be explained through unnoticed sludge washout on days on which no wastewater sampling took place. This however is improbable since no or very little sludge was found inside the chambers beyond ABR 5. The result therefore supports the hypothesis that anaerobic digestion did take place inside the ABR and significantly reduced the volume of retained biodegradable CODp. The further testing of this hypothesis with anaerobic digestion modelling is described in Chapter 7.

Comparing sludge build-up in Phase II to mass balance results was not attempted: it is obvious that strong sludge washout from the digester into the ABR and migration out of the first five ABR compartments occurred repeatedly during the period after mid June 2013 until the end of Phase II.

Since the Q and CODp measurements do not represent these washout events they cannot be compared 6.3.8.4.2 CODp massbalance

to sludge build-up. Attempting a comparison with the dataset gathered during Phase II before being affected by rain is unpromising since its size is very small (n = 4 for COD and sludge volume values).

Figure 82 a and b compare the average CODt measurement data of Phase I and II with predictions given by the ABR design calculation. The inputs for these calculations are the measured average flows (as described in Section 6.3.8.4.1) and average feed concentrations. The hydraulic load was very similar in both cases. The organic feed content during dry weather however was, as discussed above, similar in relation to soluble components but had a larger particulate fraction in Phase II. The organic feed load certainly increased strongly due to rainfall since large volumes of digester sludge were washed into the ABR. The “Initial design“ curve represents the treatment assumed at the design stage of the plant with a significantly lower feed concentration.

Figure 82 a and b: Measured average CODt concentration profiles in Phase I and II, initial design predictions („Initial design“) and design predictions with input variables adjusted to measured field values („Design prediction“)

The CODt reductions across single chambers (see Figure 82 a) were all shown to be statistically significant. They are however below design expectation for all chambers following ABR 1.

The CODt reductions across single chambers as presented in Figure 82 b were shown to be statistically significant for ABR 1, ABR 3 and ABR 5, with CODs reduction occurring primarily in the first compartments and CODp reduction throughout all the compartments. The concentration ranges are within design predictions.

Whether the increased treatment was due mainly to increased organic load or primarily to the accumulation of active digester sludge cannot be judged at this point but a combination of both factors is probable.

Most wastewater treatment clearly occurs in the first three chambers and more so in Phase II. This correlates with SMA measurements which show higher bioactivity in the front chambers.

0 200 400 600 800

ABR in ABR 1 ABR 2 ABR 3 ABR 4 ABR 5 mg CODtl-1

Phase I

0 200 400 600 800

ABR in ABR 1 ABR 2 ABR 3 ABR 4 ABR 5

Phase II Field measurements

Design prediction Initial design

6.3.8.4.3 Compartment performance ABR

a) b)