It is widely assumed that non-aerated selectors are very efficient in nutrient removal, and especially anaerobic basins may largely contribute to good sludge settleability as well. However, based on results measured in full-scale, this paper draws attention to the fact that with decreasing availability of readily biodegradable carbon source (rbCOD) being experienced worldwide, oxygen penetration into non-aerated basins through the uncovered surface may no longer be considered negligible. When the oxygen mass transfer is significant compared to the available influent rbCOD, non-aerated selectors should be regarded as basins with low dissolved oxygen (low DO) concentrations that may underperform with respect to nutrient removal and favor the growth of filaments, especially during low-loaded conditions. In order to fully exclude oxygen penetration, floating seals have been developed and applied at the North-Budapest Wastewater Treatment Plant in Hungary. Comparative full-scale studies showed prevention of significant amounts of influent rbCOD loss (up to 60 mg/L) through the application of this new technology. This amount of saved, non-oxidised but fermented carbon source could be accordingly used for enhancing biological P-removal. Due to the elimination of microaerophilic conditions, the undesirable growth of filamentous bacteria could also be avoided, leading to significantly better activated sludge settling.
Non-aerated selectors are considered to be useful for both achieving low effluent nutrient (N and/or P) levels and assure well settling sludge by favoring proliferation of floc-forming bacteria. It has been shown that anaerobic selectors may especially contribute to maintaining a low sludge volume index (SVI) (Parker et al. 2004) through promoting growth of polyphosphate accumulating organisms (PAOs) in the absence of oxygen and nitrate (Schuler & Jenkins 2003).
Anoxic selectors used for denitrification are also regarded to provide a metabolic advantage to floc-forming bacteria. However, this environment may be more sensitive due to the nitrate recirculation likely carrying oxygen back from the aerated basin. Moreover, these reactors may easily be diluted to relatively low substrate concentrations, especially at high nitrate recirculation rates and in the case of low influent rbCOD loads (Wanner & Jobbágy 2014). Due to its well-known overall metabolic advantage in carbon source consumption, the necessity of excluding oxygen is widely accepted and practiced in non-aerated selectors of lab and pilot scale systems (Wanner & Jobbágy 2014). Methods to accomplish this include the Zero-Headspace Reactor developed by Jobbágy et al. (2000) and the introduction of nitrogen gas bubbled through the system (Martins et al. 2004). The full-scale impact of oxygen penetration through the exposed surface of a non-aerated selector is commonly not taken into consideration.
It has been observed, however, that the overall amount of readily biodegradable carbon source expressed as COD (rbCOD) is decreasing in the influent to several wastewater treatment plants (Tardy et al. 2012). Enlargement of sewered areas through connections to regional treatment facilities leads to generally long retention times in the sewer systems where rbCOD content may decrease significantly, especially at relatively high water temperatures. Due to increasing water prices, both industrial and municipal water users have an incentive to decrease water consumption, and thereby sewer retention times may increase further. With decreasing dilution the wastewater constituent concentrations will increase, but the more readily biodegradable substrates may well be used up through the interactions with the sewer biofilm. In this case, the more slowly degradable parts may be significantly concentrated, especially ammonia (which in extreme cases may even exceed 100 mg/L) while the concentrations of readily biodegradable carbon sources may remarkably decrease during the long retention time leading to marginal substrate availability, or even deficiency, for nutrient removal (Tardy et al. 2012). The intensified primary clarification for enhanced digester loading to increase biogas production may also worsen the situation. While external dosing is getting more common (Swinarski et al. 2008; Czerwionka et al. 2012), consumption of valuable carbon source by oxygen penetration through the exposed basin surface is especially undesirable.
It is well known that oxygen does not just provide a metabolic advantage but also inhibits desired biochemical transformations in non-aerated selectors, such as hindering the production of denitrifying enzymes (Oh & Silverstein 1999; Plósz et al. 2003). Plósz et al. (2003) found that dissolved oxygen (DO) concentrations as low as 0.2 mg/L may cause the rate of denitrification to fall by half. It has also been pointed out in an experiment with a clean water diffused aeration system (DeMoyer et al. 2003) that oxygen penetration through the open surface may not at all be negligible compared to bubble aeration (the separately determined mass transfer coefficient from surface oxygen penetration reached 59–85% of the bubble mass transfer coefficient). Turbulence of activated sludge reactors may differ from site to site as well as from time to time, and may be enhanced by several factors, including rapidly changing influent flow rates, strong winds and air entrainment from mixers. In order to decrease the oxygen penetration, Barnard et al. (2015) worked out a technique that stops mixing in the anaerobic selector for a considerable time of the operation.
Under low substrate concentrations that favor aerobic growth of filaments over floc-formers, the substrate consumption rate is relatively low, especially at low temperatures when oxygen solubility is higher. It can be assumed that the widely experienced increase of SVI in wintertime (Guo et al. 2010; Jobbágy et al. 2012) can at least partly be attributed to increased level of DO in the non-aerated selectors. Besides increasing oxygen consumption under a low DO condition favoring the growth of filaments, this may also cause increasing kinetic inhibition of anoxic and anaerobic processes that could produce floc-formers otherwise. This assumption is in agreement with the lab-scale findings of Martins et al. (2004) recommending to exclude microaerophilic conditions from biological nutrient removal (BNR) systems in order to maintain good performance.
The purpose of this paper is to draw attention to the fact that decreasing substrate availability in the influent of activated sludge wastewater treatment plants may lead to low S-low DO conditions in the non-aerated selectors, especially at relatively low temperature, with dilution by rain or melted snow, or by nitrate recirculation and/or under inappropriate operation of the pretreatment units (Jobbágy et al. 2015). The advantages of seal-covering these reactors in order to exclude surface oxygen penetration, and thereby avoid repressing anoxic and anaerobic processes, as well as discourage undesired growth of filaments, are illustrated.
RESEARCH CONCEPT OF LOW S-LOW DO CONDITIONS IN THE NON-AERATED SELECTORS
Figure 1 illustrates the bioreactor arrangement of Szeged Wastewater Treatment Plant (Szeged WWTP) in Hungary. The activated sludge system has an entirely anaerobic selector, followed by a two-stage anoxic selector, and an alternatively aerated or mixed nitrification/denitrification unit.
Despite the staged non-aerated selector system, Figure 2 clearly shows that significantly decreasing BOD5 concentration of the preclarified influent resulted in highly increased SVI values, even though aluminum salts were dosed by the operator to control filaments (Weinpel et al. 2014). Besides frequent rain events and infiltration from the Tisza river, both significant decrease of BOD5 in the grit chamber and occasional chemical dosing to the primary clarifier for increasing biogas production could be tracked down as reasons leading to drastically decreased concentrations of the influent carbon source.
The high SVI values in Figure 2 do not at all characterize an efficient anaerobic selector followed by anoxic selectors, operating characteristically with the metabolic advantage of floc-forming bacteria. It can be assumed, however, that a significant amount of rbCOD entering the bioreactor was consumed under low S–low DO conditions, as observed similarly in an anaerobic selector by Daigger & Nicholson (1990). To minimize the loss of high amounts of biodegradable substrates, the sand/grit removed during mechanical treatment was washed and the coagulant dose was decreased, resulting in an increase in the BOD5 concentration of the preclarified influent. Consequently, SVI values decreased, illustrating the dependence on temperature alone (see Guo et al. 2010; Jobbágy et al. 2012).
All the effects outlined above may push the operational substrate concentration in the non-aerated selectors towards filamentous advantage. In order to illustrate the growth conditions in the anaerobic stage with significant surface oxygen penetration, as well as the impact of the real low S-low DO conditions on the growth advantage of filaments over floc-formers, simulated curves of Figure 3 were calculated for aerobic growth conditions by values of μmax, filamentous: 3 d−1, μmax, floc-forming: 5 d−1, KS, filamentous: 1 mgCOD/L, KS, floc-forming: 10 mgCOD/L, KSO2,filamentous: 0.1 mgO2/L, KSO2, floc-forming: 0.25 mgO2/L according to Niekerk et al. (1988) and Kappeler & Gujer (1994).
In Figure 3(a) on the specific growth rate (μ): dissolved biodegradable COD concentration (S) curves typical ranges of dissolved biodegradable COD concentrations measured in the first non-aerated, intentionally anaerobic selector are illustrated. The results suggest that under significant oxygen penetration and without a pronounced excess of influent rbCOD, consumption advantage of floc-formers could not be assumed to be characteristic. Even under generally detected S conditions, occasional aerobic growth advantage of filaments could not have been excluded. However, with a heavy rain event potential aerobic growth advantage of filaments got unambiguously significant. Therefore, in the case of not overwhelming substrate consumption with metabolic advantage of PAOs as floc-formers, i.e. at a relatively low influent rbCOD load, oxygen consumption in the intentionally anaerobic selector may lead to significant growth advantage of filaments, as illustrated in Figure 3(a) and 3(b) (Weinpel & Jobbágy 2017).
With decreasing substrate concentrations, the advantage of oxygen consumers over PAOs, as well as denitrifiers, may increase further due to increasing DO level coming from decreasing substrate consumption rate, especially at decreasing temperatures when oxygen solubility is increasing. Figure 3(b) illustrates the impact of both the concentration of rbCOD (as S) and DO on the advantage of filaments over floc-formers (Weinpel et al. 2014). It is important to note that under low DO conditions, meaning basically the range of 0.04–0.2 mg DO/L, the growth advantage of filaments over floc-formers may become significant even at relatively high S values (Jobbágy et al. 2017).
METHODS AND RESULTS OF SEAL-COVERING NON-AERATED BASINS
While in aerobic selectors increasing DO level may help to avoid filamentous overgrowth (see Figure 3(b)), this technique is obviously not applicable in anaerobic or anoxic basins. Therefore, going into the opposite way, i.e. fully excluding oxygen penetration by seal-covering these reactors has been applied as a novel solution. In order to estimate the hydraulic impact on the turbulence during applying this new technology, calculations were made for possible non-aerated experimental basins (Szabó 2014). The results showed that depending on the depth of the basins, placement and size of the holes introducing the influent and recycled sludge as well as nitrate recirculation into the reactor, type and power consumption of the mixers, roughness of the surface, etc. the mass of oxygen entering through the surface may differ from site to site, and reach remarkable values. It was also found that even a relatively deep (8 m) bioreactor may operate as a CSTR (completely stirred tank reactor) where practically all parts of the biomass may get in touch with the oxygen having penetrated into the basin.
The first full scale non-aerated activated sludge reactors having been covered by a floating seal, as published by Wanner & Jobbágy (2014), were the two anoxic/anaerobic basins of Train I of the New Line at the North-Budapest WWTP (see arrangement in Figure 4). As described earlier (Jobbágy et al. 2012), the plant has an Old Line and a New Line, consisting of four trains each. All of the trains have two subsequent anoxic/anaerobic basins followed by an intermittently aerated and mixed reactor used both for nitrification and denitrification. A major difference in the sludge handling is that recycled sludge of the New Line is mixed before feeding back to the first non-aerated basin, whereas in the Old Line, the four trains are operated separately. The treatment plant receives an average dry weather flow of 160,000 m3/d. The sludge recycling rate is 100%, and the rate of nitrate recirculation is ∼200%.
The first experiment was started on September 1st, 2013 in the New Line having free way for effluent from the non-aerated selectors without scum retaining. Since the biomass structure, as well as the wastewater quality, could be expected to be basically the same in both of the influents, this part of the research focused on the estimation of possible carbon source amount being saved and fermented through excluding oxygen penetration. Bioprocesses of the Test system with seal-covered non-aerated selectors (see Wanner & Jobbágy 2014), as well as those of the Reference system with open surface were parallelly investigated. Samples were taken at the points indicated in the schematic of the New Line in Figure 4. The most characteristic results are illustrated in Figure 5, where samples 1/1 and 1/3 were taken from the feeding chambers of the Test and Reference systems, respectively, and 3a stands for the feeding points of the nitrate recirculation. S0 is the calculated starting concentration of the mixture of preclarified wastewater and recycled sludge (coming mixed from the feeding chamber) and of the appropriate nitrate recirculation.
Comparing the shapes of the graphs in Figure 5(a) and 5(b), a clear difference can be observed between chemical P-precipitation and biological P-release. This verifies that tracking the orthophosphate concentration profiles in the system may unambiguously reveal the causes of P-removal deducted from the difference of influent and effluent concentrations of the non-aerated selectors. In Figure 5(b), the influent mixed nitrate concentration proved to be generally low, and slightly higher in the Reference system attributable to more intensive aeration. Practically no nitrate could be detected in the non-aerated reactors. Under this condition, however, P-release proved to be significantly higher in the Test system than in the Reference system. Calculations based on mass balances of the non-aerated reactors taking also into account the slightly higher amount of nitrate removed in the Reference system, resulted in a maximum of as much as 60 mg/L advantage as preclarified rbCOD savings for fermentable carbon source in the seal-covered non-aerated reactors.
In order to further investigate the possibility of enhanced biological P-removal, and reveal the impact of seal-covering non-aerated basins on the sludge settling characteristics, the research at North-Budapest WWTP has been continued from December 29, 2014 in the Old Line (see Figure 4) where returned sludge coming from the different trains is not mixed. However, in this part of the system, severe scum formation could be observed attributable to both microaerophilic conditions and scum retaining ability of basin walls. During the experiment, non-aerated selectors of the system having originally considerably less scum, looking more like a system with open surface, had been seal-covered as Test system, and the operator took care of removing the scum from the non-aerated selectors of the Reference system weekly, when the surface started to be significantly covered with it. The floating seal had been developed in cooperation with the company of Karsai Holding Pte. in an advanced form especially tailored for this purpose (see Figure 6).
Since the availability of samples was restricted by the scum and seal-covers, the presence of PAOs in the activated sludge was investigated in complementary batch experiments. Mixed liquor samples were taken close to the outlet of the non-aerated reactors, and transferred to the laboratory. Having been aerated for 1 h, the mixtures were filled into Zero-Headspace Reactors (Jobbágy et al. 2000) and spiked with 200 mg/L acetate. Extent of P-release was compared after 180 min of anaerobic operation and divided by the MLVSS concentration in order to calculate the values of specific P-release, included in Figure 7(a). It can be observed in both of Figure 7(a) and 7(b) that specific values of P-release had proved to be higher in the Reference system before the Test system was seal-covered, attributable to the significantly larger surface area of scum. However, an unambiguous change in this behavior could be detected afterwards. The higher efficiency of biological P-removal in the seal-covered Test system can be attributed to ensuring favorable and relatively stable growth conditions for PAOs. The decreasing difference between the specific values of P-release referring also to the uptakes of the carbon source for biological P-removal in the systems can be attributed both to the random accumulation of scum on the surface of the Reference system and to chemical dosing.
Data presenting the SVI values in Figure 8 derive from extremely cautious sampling in the aerated phase of the alternatively operated Ae/Ax basins. Since both of the experimental trains consist of four sub-trains all of the SVI values are average data of four simultaneously executed sampling and measurement of both MLSS and V30 for assuring accuracy. While settling characteristics had proven to be practically identical before seal-covering the surface of the non-aerated Test basins, an obvious advantage of this system could be detected afterward, through the elimination of microaerophilic conditions and thereby the undesirable proliferation of filamentous bacteria. It is important to note that while the floating foam was reproduced from time to time and had to be removed from the surface of the non-aerated reactors of the Reference system, there was no scum removal from the seal-covered area of the Test system. This observation supported that the elimination of microaerophilic conditions hindered the scum formation. It was also detected that the scum originally present or sent to the surface of the seal-covered reactor was eliminated by fermentation. Drastic decrease of SVI values from the end of February can be attributed to an overflow of primary clarifiers carrying chemicals appropriate for precipitation. This could also have caused the decreased P-release experienced as shown in Figure 7(a) and 7(b).
The research supported that non-aerated activated sludge selectors should rather be regarded as low-DO basins when amount of oxygen penetrating into the reactor is not negligible compared to the influent rbCOD load. This may favor proliferation of filaments, and result in disappointing nutrient (N and/or P) removal efficiency, especially under low substrate concentrations. Under low-DO conditions, meaning basically the range of 0.04–0.2 mg DO/L, growth advantage of filaments over floc-formers may become significant even at relatively high substrate concentrations. Results suggest that an open non-aerated reactor surface may help the filaments to outcompete floc-formers under microaerophilic conditions, whereas covering the reactor by retained scum may enhance biological P-removal on a short term, but lengthen the retention time of filamentous scum-formers and cause escaping into the bulk of the mixed liquor, leading to poor sludge settling.
Excluding surface oxygen penetration through covering the non-aerated reactors by a floating seal could considerably contribute to enhanced nutrient removal observed in increased anaerobic P-release by saving and fermenting of rbCOD up to 60 mg/L expressed in preclarified wastewater. Improving SVI values up to 30% referred to significantly better performance as well.
The projects have been funded by Szeged Waterworks Ltd and Budapest Sewage Works Pte Ltd. The authors highly acknowledge the especially valuable contribution of CEO György Palkó and Magdolna Makó project manager, and express their thanks for assistance of Pál Román, plant manager, and Zsófia Kassai, deputy plant manager of Budapest Sewage Works Pte Ltd. Valuable contribution of Dezső Bodor, technical director, János Révész, plant manager, and Péter Pataki, deputy plant manager of Szeged Waterworks Ltd is also highly acknowledged. The authors express their special thanks to CEO Béla Karsai for personal cooperation in developing the floating seal and to József Simon for the excellent technical support. The contribution of Professor Timothy Ellis through advising appropriate formulation for better understanding differences is highly acknowledged.