Enhancing the nitrogen removal from swine wastewater digested liquid in a trickling biofilter with a soil layer

Trickling biofilters (TFs) allow for a simultaneous nitrification and denitrification (SND) process, and offer a favorable solution for the treatment of swine-wastewater digested liquid due to their simple operation and low cost. In this study, a soil trickling biofilter (STF) was developed to enhance nitrogen removal. A gravel trickling filter (GTF) and a woodchip trickling filter (WTF) were also constructed and operated synchronously to demonstrate the advantage of micron-sized media. The results showed that the STF had a higher ammonium nitrogen (NH4+-N) removal capacity of 21.4%, 24.9%, and 18.3% in comparison to the GTF when the influent NH4+-N was 192.9 mg L−1, 500.2 mg L−1 and 802.1 mg L−1, respectively. The total nitrogen (TN) removal capacity of the STF was 104.6%, 89.4%, and 37.5% higher than that of the WTF. Thus, the addition of micron-sized soil to TF could increase the systemic nitrogen removal capacity.


Introduction
Swine wastewater from large-scale farms can have a high chemical oxygen demand (COD) and high ammonium nitrogen (NH 4 + -N) concentration, thus making it a major pollutant in vast rural areas. 1 Swine wastewater that is not properly disposed of can cause serious problems in the agricultural environment, for example, eutrophication and zoonosis. [2][3][4] With respect to the economy, environmental sustainability, and social acceptability, the technologies for swine wastewater treatment should be low-cost, energy-saving, low-maintenance, effective, and stable. 5,6 The key point of swine wastewater treatment is the removal of nitrogen from the digested liquid with a low carbon/nitrogen ratio (C/N). 7 Most technologies for nitrogen removal are based on the conventional nitrication and denitrication processes. Generally, the traditional biological nitrogen removal process includes plug ow reactors [e.g., anoxic oxic (A/O)] 8 and continuous stirred tank reactors [e.g., sequencing batch reactor (SBR)]. 9,10 Although these techniques have a favorable effect on nitrogen removal during swine-wastewater digested liquid treatment, they are disadvantageous in terms of their running costs.
The modern nitrogen removal processes include single high ammonia removal over nitrite (SHARON), anaerobic ammonium oxidation (ANAMMOX), complete autotrophic nitrogen removal over nitrite (CANON), and oxygen-limited autotrophic nitrication-denitrication (OLAND). These methods have several advantages in terms of their low running costs due to the money saved for aeration, the organic carbon source for deni-trication, and the reactor volume. [11][12][13] Unfortunately, these technologies require strict operational conditions to maintain the stability of the system, which increases the operation difficulties and restrict their widespread application.
Trickling biolter (TF) is a biolm wastewater treatment process that has been generally recognized as providing an effective simultaneous nitrication and denitrication (SND) process with a low-cost and convenient management. TF has been widely used to treat domestic wastewater, piggery wastewater, textile wastewater, and leachate. 14 The biochemical process is the main pollutant removal pathway, whereby the microbes growing on the surface of the lter material play the major role. 15 Hence, the characteristics of the lter material are key factors affecting the efficiency of TF.
An excellent lter material should have a large specic surface area and a good biolm adhesion. Various materials have been investigated as lter material for use in TF systems, for example, zeolite, sand, plastic, woodchip and complex material. 16 This study focused on the surface characteristics of the lter media to obtain favorable treatment efficiency. With the exception of the surface characteristics, the size of the lter material is the most direct factor affecting the specic surface area of the material. In principle, the smaller the size of a lter material, the larger specic surface area is for the enrichment of microbes. Soil, as the natively weathered gravel, generally has a small size between millimeters and microns. Hence, soil inltration technology has long since been used for water treatment. 17,18 However, the clogging and blocking of the soil layer limit the treatment capacity of soil inltration, and also prevent the utilization of soil in TF systems. 1 In this study, a novel soil TF (STF) is proposed to solve the clogging problem of a micron-sized soil lter material under a high hydraulic and organic load. This novel technique involves the micron-sized soil adhering to a woodchip framework. Thus, the soil layer that becomes xed to the framework can resist a certain hydraulic load and has a large specic surface area to adhere to the biolm. In order to compare the advantages of this STF process, a gravel TF (GTF) and a woodchip TF (WTF) were also constructed and synchronously operated. The aim of using GTF is to demonstrate the size advantage of the micron-size soil for the enrichment of microbes, whereas the aim of using the WTF is to show the effect of the woodchip framework in the STF.

Wastewater
Raw wastewater was collected from a local pig farm. The inuent used in this study was the digested liquid with a low C/N ratio from the UASB treatment of the raw wastewater. The characteristics of the digested liquid are listed in Table 1.

Reactors
Three TF reactors of the same size were constructed, each with a 0.3 m high plexiglass column of 0.1 m diameter (Fig. 1). The total volume of each reactor was 2.0 L, with an effective volume of 1.5 L. A circular cone with a volume of 0.2 L was attached to the bottom of the plexiglass column. There were two symmetrical vents with an area of 2.5 cm 2 between the lter bed and the circular cone to allow natural ventilation.

Filter material
In this study, pine wood was chosen to make the woodchips, which were 3-5 cm high, 2-3 cm wide, and 0.5-0.8 cm thick. Gravel with a diameter of 1 cm was used for the lter material in the GTF. The soil was ground to a powder and sieved through a screen of 160 meshes per square centimeter. The bulk density, porosity, and ash content of the soil were 0.855 g cm À3 , 50% (v/v), and 81.12% (w/w), respectively. Woodchips (of the same size as those used as the lter material in the WTF) provided the framework in the STF.
To create a soil layer within the framework of the STF, the woodchips were soaked in fresh water and then rolled over the prepared dry soil powder. As the woodchips were damp, the soil adhered onto the surface of the woodchips. The bulk volume ratio of the woodchips to the soil in the STF was approximately 2 : 1. To compare the efficiency of the three TFs, the void volume ratio for each of the three TFs was about 50%.

Operation
The three TFs were fed with the same swine wastewater digested liquid. The wastewater was evenly sprayed onto the top of the lter bed with a hydraulic loading of 0.17 m 3 m À3 d À1 during all operating periods. The effluent was collected by the cone and discharged from the bottom. There were no inoculums with microorganisms for all reactors before starting-up. There were four operating periods according to the different organic volume loading rate (OLR) and inuent quality. All three reactors used natural aeration and were operated at an ambient temperature of between 15 C and 25 C. The operational parameters are illustrated in Table 1.

Wastewater quality analysis
The COD, NH 4 + -N, nitrite nitrogen (NO 2 À -N), nitrate nitrogen (NO 3 À -N), and total nitrogen (TN) were analyzed according to the standard methods for the examination of water and wastewater. 7 The concentrations were determined using a spectrophotometer (Shimadzu, UV2500). The pH was monitored by a DELTA 320 pH meter. Normally, the inuent and effluent were sampled once every 5 days for analysis.

Observation of the lter media
Before operating the reactors, the woodchips, gravel, soil, and woodchips wrapped with soil were all sampled. The samples were soaked in a 2.5% glutaraldehyde solution (pH 7.2) for 2 h. Then, all samples were washed three times with a 0.1 mol L À1 phosphate buffer (pH 7.2). Alcohol at a concentration of 50%, 60%, 70%, 80%, 90%, and 100% was in turn used to dehydrate the samples, whereby tert-butyl alcohol gradually replaced the alcohol. Aerward, the samples were dried by a desiccator (HCP-2, HITACHI). Finally, the samples were plated with a gold lm of 1 nm by an ion sputtering coater (E-1010, HITACHI) and observed with a scanning electron microscope (SEM).

Analyses using PCR and DGGE
On the last day of operation, lter material was sampled from the three TFs. Genomic DNA of all samples was extracted using DNA isolation kit components (MOBIO), which was performed according to the manufacturer's operations manual. The purity and quantity of the extracted DNA were determined by ultraviolet spectrophotometry at 260 nm and 280 nm, and the DNA was subsequently stored at À20 C. A nested polymerase chain reaction (PCR) was used to amplify the bacterial specic 16S rRNA gene for denaturing gradient gel electrophoresis (DGGE). The bacterial specic primers CTO189f and CTO654r were used in the PCR to amplify the DNA fragments. The specic procedures of the PCR and DGGE analyses were conducted according to. 7

Observation of lter media before operation
Fig. 2A-D shows the surface of the woodchips, gravel, micronsized soil, and soil with woodchips in magnied SEM images (Â100). As can be seen in Fig. 2A and B, some cut stripes were clear on the woodchips and a crude surface was observed on the gravel. As Fig. 2C shows, the size of a soil granule ranged between 50 and 100 mm, which indicates that, aer grinding and screening through the 160 mesh, the natural soil could easily maintain a micron-size. The woodchips that were wrapped with soil are illustrated in Fig. 2D, which clearly displays a thin layer of micron-sized soil adhering to the woodchips. The cut stripes on the surface of the woodchips had been completely covered by the micron-sized soil. Moreover, Fig. 2E-G reveal the surface details of the woodchips, gravel, and micron-sized soil in magnied SEM images (Â3000). In particular, Fig. 2E illustrates the smooth wood surface and the side of the wood ber layer. Comparing Fig. 2F and G, the surfaces of the soil and gravel were similar and both were crude; hence, the main This journal is © The Royal Society of Chemistry 2020 RSC Adv., 2020, 10, 23782-23791 | 23785 difference between the soil and gravel was their size due to natural weathering and our operation. The pores on the woodchip surface were barely noticeable ( Fig. 2A and E), even when magnied Â3000. The pore size of the woodchips with soil adhered (inferred as being 8-10 mm from Fig. 2D) was signicantly smaller than that of the gravel and soil only, which contributed to its larger specic surface area for microorganisms living.

Removal of COD
The three TF reactors were started at a hydraulic loading volume of 0.17 m 3 m À3 d À1 with an organic volume loading rate (OLR) of 50.8 g m À3 d À1 COD. The removal performances for the COD in the three TFs are presented in Fig. 3A and B. During the rst 30 days, the COD of the effluent from each of the three TFs all reduced rapidly. Aer day 30, the COD stabilized. From day 30 to the end of start-up period, the average COD removal efficiencies of the WTF, GTF, and STF were 46.38%, 56.77%, and 38.28%, respectively. Moreover, the COD in the effluent in the STF was higher than that in the inuent during the rst 5 days because some intrinsic organic matter in the soil was removed by the effluent.
The inuent COD during period I, II, and III was 180.9 mg L À1 , 293.1 mg L À1 , and 508.9 mg L À1 with an organic load of 30.8 g m À3 d À1 , 49.8 g m À3 d À1 , and 86.5 g m À3 d À1 , respectively (Fig. 3A). The COD removal efficiencies of all three TFs recovered and were stable again within 5 days once the inuent changed. This indicates that the TF process in all of the reactors had a good capacity to resist the organic load impact, which agrees with previous studies. 19,20 The average COD removal rate of the WTF during the stable stages of periods I, II, and III (i.e., day 56 to day 155, day 156 to day 255, and day 256 to day 346, respectively) was 11.0%, 44.2%, and 60.6%, respectively. Meanwhile, the COD removal rate of the GTF during the stable stages of periods I, II, and III was 35.1%, 41.7%, and 60.1%, and that of the STF was 3.9%, 21.7%, and 49.6%, respectively (Fig. 3B).

Ammonium-nitrogen removal and nitrication
The NH 4 + -N concentration and removal rate of the three TFs are displayed in Fig. 3C  . The moisture content of the STF was higher than that of the WTF and GTF; therefore, the DO concentration in the solution of the STF was relatively high. As shown in Fig. 3D and E, the NH 4 + -N removal rate was highest in the STF, nevertheless, the NO 2 À -N concentration in the effluent of the STF was also the highest of three TFs. This indicates that nitrogen removal by AOB was higher than that by NOB. The NO 2 À -N and NO 3 À -N concentrations in the effluents of the three TFs are shown in Fig. 3E and F. All three TFs had a similar variation in the NO 2 À -N concentration, which rst increased and then decreased. When the effluent NO 2 -N concentration reduced, the effluent NO 3 À -N concentration increased, which agrees with the general rule of bio-nitrication. The highest effluent NO 2 -N concentration in the WTF, GTF, and STF occurred on day 21 (282.8 mg L À1 ), day 31 (268.6 mg L À1 ), and day 31 (222.4 mg L À1 ), respectively. Towards the end of the start-up period, the effluent NO 3 À -N concentration of the WTF, GTF, and STF was 52.6 mg L À1 , 45.6 mg L À1 , and 70.5 mg L À1 , whereas the effluent NO 2 À -N decreased to 200.9 mg L À1 , 200.5 mg L À1 , and 155.6 mg L À1 .
There were two main processes involved in the removal of NH 4 + -N: physiochemical and microbe catalysis. During the start-up period, the abundance of microbes was lower with a relatively low activity; thus, absorption to the lter media was the main process of NH 4 + -N removal. Although Buelna 21 considered that NH 4 + -N stripping was a main removal process at high pH, absorption may have been more important in the start-up period of TFs in this study. Different trends were evident when comparing the NH 4 + -N removal in the WTF, GTF, and STF, which could have been caused by adsorption to the lter material. Due to the high adsorption to soil, the STF had a relatively high NH 4 + -N removal during the rst 10 days.
During period I, the average inuent NH 4 + -N concentration was 192.9 mg L À1 with an average nitrogen loading ratio (NLR) of 32.79 g m À3 d À1 . The NH 4 + -N removal rate of the three TFs increased gradually (Fig. 3C) and the effluent NO 2 À -N concentration reduced sharply as the NO 3 À -N concentration increased ( Fig. 3E and F). All three reactors subsequently entered a stable stage from day 81 to day 151, during which, the NH 4 + -N removal rate of the WTF, GTF, and STF was 84.50%, 69.80%, and 84.86%, respectively. The average NO 2 -N concentration in the effluents of the WTF, GTF, and STF during this period was 2.1 mg L À1 , 11.7 mg L À1 , and 2.5 mg L À1 , respectively, whereas the NO 3 À -N concentration was 90.6 mg L À1 , 90.0 mg L À1 , and 88.4 mg L À1 , respectively.

Total nitrogen removal and denitrication
The three TFs commenced with a TN volume loading of $84.43 g m À3 d À1 . The TN concentrations of the inuents and effluents in each TF are shown in Fig. 3G Fig. 4, simultaneous nitrication and denitrication occurred in the three reactors, accounting for the nitrogen removal. The Ca 2+ and Na + adsorbed to the surface of the soil was replaced by NH 4 + , and the Ca 2+ and Na + precipitated aer combining with OH À . The cations associated with OH À , HCO À , and CO 3 2À anions form a dynamic balance of alternating ion changes. In the AOB process, oxygen is consumed for the  Paper RSC Advances oxidization of ammonia to nitrite. In addition, H + ions are generated from this reaction (eqn (1)). The nitrite and nitrate that are produced are reduced to N 2 during the denitrication process while H + ions are consumed (eqn (2)

Analysis of AOB by PCR-DGGE
When all operations were complete, biomass samples were obtained from the lter material of each TF. By using the CTO189f/ CTO654r primers, the AOB bacterial community was analyzed by PCR-DGGE. The gel is shown in Fig. 5. The 13 bands were identied and excised from the DGGE gel. Sequences of the bands were compared with the available sequences in GenBank, and the phylogenetic tree was constructed (Fig. 6).
The identied species belonged to Nitrosomonas (bands A1, A2, A4, A8, and A11), Nitrosospira (A9 and A10), Rhodanobacter (A6 and A12), Streptococcus australis (A13), Clostridiales (A3), Actinomycete (A7), and an unidentied bacteria (A5). The AOB were divided into the b-subclass and gsubclass. 23 There were two genera in the b-subclass: Nitrosomonas and Nitrosospira. The later included two subgenera: Nitrosolobus and Nitrosovibrio. 24 In this study, all identied AOB species belong to the b-subclass, although some iden-tied species were not AOBs, which was due to the specicity of the primers. The specicity of PCR is usually limited by only one AOB specic primer. The CTO189f/CTO654r primers held a higher specicity within the AOB specic primer. 25 Bacterium for bands A1, A2, A4, and A8 were identied (98%) as Nitrosomonas sp. Band A11 was identi-ed (99%) as Nitrosomonas eutropha, which has been reported to be a special AOB in aerobic environments and has a capacity for autotrophic denitrication. 26 Band A9 was 99% similar to an uncultured Nitrosospira sp. Band A10 was identied (100%) as Nitrosospira multiformis. Nitrosospira is a common AOB genus that can compose nitrite reductase and NO oxidoreductase. These enzymes can both catalyze nitrite to N 2 O. 27-29 Fig. 6 Neighbor-joining phylogenetic tree of sequences from the DGGE profile.

Effect of ammonia removal and nitrication by the soil layer
A comparison of the performances of three TFs is presented in In a biochemical system, nitrication is carried out by AOBs, whereby their abundance has a primary impact on the NH 4 + -N removal capacity. The abundance of the AOBs was indicated by the brightness of bands A1, A2, and A4 in the lanes of the three TFs, which showed that the abundance of Nitrosomonas was similar in the three reactors. However, bands A8, A9, A10, and A11 in the STF somewhat brighter than those of the WTF and much brighter than those of the GTF. The abundance of Nitrosospira in the WTF and STF was greater than that in the GTF. This demonstrates that the soil layer in the STF was a more suitable micro-environment for the growth of AOBs, and thus more suitable for increasing their abundance and activity. There were two reasons for the higher NH 4 + -N removal and greater abundance of AOBs. The mixed microbes and cellulose formed a biolm as woodchips were decomposed in the WTF, and the soil layer further extended the appositional growth of the AOBs in the STF. Moreover, the soil layer could impede the transfer of oxygen in the STF, although this increased the ammonia concentration gradient to protect the AOBs in the soil layer against free-ammonia inhabitation, which may have led the better performance of the STF for NH 4 + -N removal.
The accumulation of nitrite in the effluents (Table 2) of the STF and WTF was similar in period I (2.5 mg L À1 and 2.1 mg L À1 ), whereas this was 31.0 mg L À1 and 58.6 mg L À1 in the STF during in period II and III, respectively, which were much higher in comparison to the WTF. This demonstrates that there was a much stronger "short-cut" nitrication in the STF. This may have been due to the DO gradient in the soil layer of the STF, which strengthened the anaerobic micro-environment on the surface of the lter material. Band A3 of anaerobic bacteria (Clostridiales) in the lane of the STF was brighter than that of the other two TFs, which suggests that the anaerobic micro-environment was relatively higher in the STF. It is known that NOB are more sensitive than AOB to low DO environments, 30,31 and such a lack of DO could decrease the activity of NOB.
Overall, the STF had a slightly higher ammonia oxidizing capacity in comparison to the WTF; however, the STF also had a much higher nitrite accumulation. This demonstrates that wrapping the soil on to the woodchips did not impact the ammonia oxidization, and "short-cut" nitrication was easily realized. Table 2 shows that the STF had a slightly higher TN removal efficiency (39.4%) in comparison to the WTF (38.1%) during period I. The TN removal rate of the STF was 41.8% and 35.9% during periods II and III, respectively, which were obviously higher than the TN removal rate of the WTF during the same periods (32.6% and 26.1%). This suggests that the addition of the soil layer in the STF improved denitrication. Bio-denitrication has been found to play a major role in TFs, whereas chemical nitrogen removal was determined to be very weak. 21 Bio-denitrication includes heterotrophic and autotrophic processes. 32,33 In terms of heterotrophic denitrication, the STF had better anaerobic conditions because the woodchips were wrapped by the soil layer. This could have advanced the decomposition of the woodchips, thus increasing the carbon resources for heterotrophic denitrication, which is supported by the high COD in the effluent of the STF. Moreover, the high accumulation of nitrite in the STF indicates that there should have been a strong "short-cut" nitrication, in which nitrite could directly converted to N 2 and the carbon resources used for nitrate denitrication could be decreased. This was, therefore, very benecial for the swine-wastewater digested liquid with a low COD/TN ratio.

Enhancement of denitrication by the soil layer
Some autotrophic nitrogen removal processes have been observed in wastewater treatment, for example, ANAMMOX. This requires the coexistence of nitrite accumulation and an anaerobic environment in a nitrication reactor. 6,[34][35][36] The STF in this study satised this requirement; hence, ANAMMOX may have occurred to a certain extent. Besides, some autotrophic nitrogen removal has been ascertained, for example, earlier research found that some AOBs could oxidize NH 4 + -N to N 2 O in a low DO environment. 26

Conclusion
Overall, the addition of a layer of micro-size soil in a TF could increase the ammonia oxidization, nitrite accumulation, and denitrication capacities. By calculating the NH 4 + -N and TN removal loads, the NH 4 + -N removal rate of the STF during periods I, II, and III was 21.4%, 24.9%, and 18.3% higher, respectively, than that of the GTF. In addition, the TN removal rate of the STF during periods I, II, and III was 104.6%, 89.4%, and 37.5% higher, respectively, than that of the WTF. The abundance of identied AOBs genera, belong to the b-subclass of Nitrosomonas and Nitrosospira, were enhanced by wrapping a soil layer around the woodchips in the STF. Moreover, the soil layer in the STF provided a better anaerobic micro-environment in comparison to the other TFs, which further strengthened the denitrication process.

Conflicts of interest
There are no conicts to declare.