Scum yield and effectiveness of a hydrostatic removal device: case study of Laboreaux STP

UASB (upflow anaerobic sludge blanket) reactors used for the treatment of domestic sewage are a consolidated technology in some warm-climate countries, such as Brazil, where several full-scale plants have been in operation for more than 15 years. However, in spite of their great advantages and broad application, UASB reactors still present some operational limitations, as discussed in Chernicharo & Stuez (2008). A major operational limitation reported in most of the Brazilian full-scale plants is the removal of scum that develops and accumulates inside the gas-liquid-solid (GLS) separators.

The scum formed in UASB reactors treating domestic sewage is a mixture of almost everything that enters the reactors, therefore constituting a very heterogenic layer of floating material containing fats, oils, waxes, soaps, food leftovers, fruit and vegetable skins, hair, paper, cotton, cigarette butts, plastics, and similar materials (Metcalf & Eddy 2003). Non-periodic removal causes build up of scum inside the GLS, resulting in the increase in thickness and concentration and posing severe difficulties for its removal, even when proper inspection hatches exist. Most of the times it is a very time consuming activity, since the top solid scum layer must be firstly broken (disaggregated) to allow its removal by a suction device (cesspool cleaning truck). Inside a GLS, the thick layer of scum can cause severe problems, such as blockage of the natural passage of gas and structural damage of the GLS, especially on those manufactured with thin walls, such as fiberglass. Problems related to scum have also been addressed by other authors (Halalsheh et al. 2005; Souza et al. 2006; Van Lier et al. 2010). Besides a better understanding of scum characteristics were achieved (Pereira et al. 2009; Garcia et al. 2012), we have also been focusing our research on practical aspects of scum management in full-scale plants, particularly at the Laboreaux sewage treatment plant (STP) (PE = 70,000 inhabitants) and Onça STP (PE = 1 million inhabitants), respectively located in the cities of Itabira and Belo Horizonte, Brazil (Chernicharo et al. 2009; Rosa et al. 2012).

The present work aims to evaluate different scum removal strategies as well as determining the scum yield inside the GLS of a full-scale UASB reactor treating domestic sewage, based on the protocols proposed by Rosa et al. (2010).

Although the UASB reactors of the Laboreaux STP were designed with a hydrostatic system of scum removal, the routine operations were not effective in the removal of this material, which resulted in the retention of a thick layer of scum within the GLS. The effective removal of scum was only achieved with the opening of the inspection hatches and the suction of accumulated material using cesspool cleaning trucks (Rosa et al. 2010). The complexity and time spent on this operation are emphasized by the fact that it is necessary to undertake this task in 56 GLS (8 reactors × 7 GLS per reactor). Figure 2 illustrates a schematic representation of seven GLS and the scum line for a typical reactor.

The effectiveness of the hydrostatic scum removal device was evaluated in two out of the seven GLS (GLS 1 and 4), in one out of the eight UASB reactors of the Laboreaux STP. For these two GLS, the scum was collected, quantified and systematically removed in two phases that comprised four testing periods. Table 2 shows the operational characteristics of the UASB reactor used in the experiments as well as the scum removal frequencies practiced in each phase/period. Before the beginning of the experiments, the coarse fraction of the scum previously accumulated inside all seven GLS was removed by a cesspool cleaning truck in order to re-establish its fluid condition.

To determine scum efficiency removal for all tests during the four periods the following routine was established: (i) aperture of the hatches of GLS 1 and 4; (ii) isolation of the scum surface area below the inspection hatches (0.25 m²); (iii) removal of the scum layer present in the surface sampling area (a sieve was used for this purpose); (iv) quantification of the scum volume according to Rosa et al. (2012); (v) closure of the hatches and pressurization of gas line according to procedures reported by Rosa et al. (2012); (vi) establishment of desired pressure inside the water seal; (vii) opening of the external valves of GLS 1 and 4, one at a time, and discharge of the scum towards a static sieve and drying bed. This operation lasted around two minutes, until almost total removal of scum in the GLS; approximately half of the drying bed volume was filled), producing 0.20 m³ scum per m² of scum weir area; (viii) opening of hatches of separators 1 and 4 for visual verification and quantification of the remaining material under the area of influence of each hatch.

The determination of the scum yield was carried in two out of the seven GLS (GLS 5 and 7) of the same reactor (as above). For these two GLS, scum was not removed along the sampling period but only sampled and quantified. The sampling and characterization procedures comprised: (i) aperture of the hatches of GLS 5 and 7; (ii) isolation of the scum surface area below the inspection hatches; (iii) removal of the scum layer present in the surface sampling area (as above, a sieve was used); and (iv) quantification of the scum volume according to the procedures used by Rosa et al. (2012). The scum accumulation rates (L.m⁻².d⁻¹) inside GLS 5 and 7 were estimated by dividing the specific volumes of scum measured under the inspection hatches (L.m⁻²) by the accumulation periods (d). The total volume of scum accumulated in all seven GLS of one reactor (L.d⁻¹) considers a surface area of the scum/gas interface of each GLS equal to 0.9 m² and 6.3 m² for all seven GLS. The scum yield was then determined by dividing the volume of scum accumulated during the monitoring period by the average COD load applied in the reactor, in the same period, being expressed as mL.kgCOD⁻¹.

From Figure 4 it is possible to identify that the specific volumes of scum prior the removal procedures were quite disparate, both considering the different monitoring periods (1 to 4) and the different GLS (1 and 4). These results reflect the different scum accumulation dynamics within each GLS, that may be associated with the different operating conditions to which the reactors were exposed, in terms of HDT, upflow velocities, applied organic loading rates (see Table 1) and biogas production, and also to the previous scum removal procedures, which may have resulted in larger or smaller volumes of scum remaining in the interior of each GLS. In relation to the specific volumes of the scum after the removal procedures, it can be seen that these were more reduced, as expected, when the scum volumes prior the removal procedures were already low (GLS 1 and 4: periods 2 and 3). Additionally, the final volumes of scum were also lower when the removal frequencies were higher (alternate days and twice a week), which were practised in periods 2 to 4. However, regardless of the final specific volumes, the removal efficiencies were always very high, as discussed below.

From the results shown in Figure 5, it can be seen that scum discharges performed with the frequency twice per week and every other day (periods 2 and 3, respectively) resulted in the highest scum efficiencies, in the order of 90% (median results). The minor median removal efficiencies, in the order of 65%, were observed in period 1, when the weekly frequency of discharge was tested, and in period 4, in which the same frequency of discharge of period 2, twice per week, was repeated. The lower efficiencies found for period 4 are due possibly to the operating conditions prevailing in this period, in which there was a significant increase in the influent flowrates and concentrations of organic matter to the UASB reactors of Laboreaux STP, as shown in Table 1. These operating conditions resulted in higher upflow velocities and biogas production in the reactors, aspects that have a direct impact on scum accumulation and efficiency removal from the GLS. After the implementation of protocol for hydrostatic removal of scum in all monitoring periods, the scum was identified within the GLS in fluid condition (liquid phase), not presenting problems of blockage as already reported by Rosa et al. (2010).

Based on the results presented in Figure 6 and on the operational conditions of the reactor shown in Table 2 it was possible to calculate the scum accumulation volumes and scum yields for all tests, as depicted in Table 3.

It can be noticed from Table 3 that period 2 was characterized by higher HDT (34.3 h) and lower influent organic load (146.3 kgCOD.d⁻¹), as compared to period 4. Therefore, much higher volumes of scum were observed in period 4, due to the increased flowrate and transference of scum to the inner part of the GLS, in accordance with the results obtained by Souza (2006). However, the scum yields were slightly higher for the monitoring period 2 (13.3 mL.kgCOD⁻¹), against 7.3 mL.kgCOD⁻¹ for period 4. The scum yield is in accordance with Santos (2014) that observed for this system a production of 6.79–10.33 mL.kgCOD⁻¹. Pereira et al. (2009) reported a scum yield inside GLS of 3.97 mL.kgCOD⁻¹.

The lower scum yields observed in period 4 are possibly due to higher biogas release rates at the surface of the scum/gas interface, which favours the disaggregation and reduction of the solid fraction contained in the scum layer, as reported by Pereira et al. (2009).

The results presented in Table 3 provide relevant information for the design of UASB reactors projects and in decision-making regarding the adoption of preventive measures of scum control in STPs. However, one should take into account that the scum accumulation rate has a direct relationship with the characteristics of the influent raw sewage and with the hydrodynamic conditions of the UASB reactor, as already reported by Souza (2006).

The scum that accumulates in the inner part of GLS separators of UASB reactors can be adequately removed via an internal gutter, as long as this procedure is carried out on a regular basis and if the scum level inside the GLS is properly set by controlling the biogas pressure in the gas line. The effectiveness of the hydrostatic scum removal device was demonstrated for different frequencies of discharge and various reactors’ operating conditions. At lower removal frequencies (once a week) and/or higher hydraulic and organic loadings, the median efficiencies of scum withdrawal were around 65%. At higher removal frequencies (alternate days and twice a week) and lower hydraulic and organic loadings, the median extraction efficiencies were around 90%. High removal efficiencies are expected to be attained independently of the reactor operating condition as long as the scum is kept fluid and the frequency of extraction is compatible with the accumulation rate.

The results of the present study can represent an important contribution towards the solution of this severe problem o scum accumulation in UASB reactors treating domestic sewage.

The authors wish to acknowledge the support obtained from Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq), Fundação de Amparo à Pesquisa do Estado de Minas Gerais (FAPEMIG) and Sistema Autônomo de Água e Esgoto de Itabira (SAAE Itabira).