Controlling effluent suspended solids in the aerobic granular sludge process

Accepted Manuscript Controlling effluent suspended solids in the aerobic granular sludge process Edward van Dijk, Mario Pronk, Mark van Loosdrecht PII:

S0043-1354(18)30767-X

DOI:

10.1016/j.watres.2018.09.052

Reference:

WR 14102

To appear in:

Water Research

Received Date: 9 July 2018 Revised Date:

19 September 2018

Accepted Date: 26 September 2018

Please cite this article as: van Dijk, E., Pronk, M., van Loosdrecht, M., Controlling effluent suspended solids in the aerobic granular sludge process, Water Research (2018), doi: https://doi.org/10.1016/ j.watres.2018.09.052. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

Effluent

AC C

EP

TE D

M AN U

SC

Stripping phase + Scum baffles

RI PT

Effluent

Scum baffle Scum/fatlike particles Biomass

Influent

Floating biomass with attached N2 bubble

Influent

ACCEPTED MANUSCRIPT

2

Controlling effluent suspended solids in the aerobic granular sludge process

3

Edward van Dijka,b,1,∗, Mario Pronka,b,1 , Mark van Loosdrechta

5 6

7

a

Department of Biotechnology, Delft University of Technology, Van der Maasweg 9, Delft, 2629 HZ, The Netherlands b Royal HaskoningDHV, Laan1914 35, Amersfoort, 3800 AL, The Netherlands

RI PT

4

Abstract

SC

1

M AN U

The main processes contributing to elevated effluent suspended solids in the full-scale aerobic granular sludge process were studied. The two processes found to be most important were (1) rising of sludge due to degasification of nitrogen gas (produced by denitrification) and (2) wash-out of particles that intrinsically do not settle such as certain fats and foams.

D

A mathematical model was made to describe the process of degasification of nitrogen gas

TE

during the feeding phase in an AGS reactor. The process of rising sludge due to degasification could be limited by stripping out the nitrogen gas before starting the settling phase in the

EP

process cycle. The wash-out of scum particles could be reduced by introducing a vertical scum baffle in front of the effluent weir, similar to weirs in traditional clarifiers. A full-

AC C

scale Nereda® reactor in the municipality of Utrecht, The Netherlands, was operated with a nitrogen stripping phase and scum baffles for 9 months at an average biomass concentration of 10 g L−1 and an average granulation grade of 84 %. In this period the influent suspended solids concentration was 230 ± 118 mg L−1 and the concentration of effluent suspended solids was 7.8 ± 3.8 mg L−1 . 8

Keywords: aerobic granular sludge, denitrification, degasification, nitrogen gas, scum 1

ACCEPTED MANUSCRIPT

10

layer, rising sludge, effluent suspended solids

1. Introduction

RI PT

9

Aerobic Granular Sludge (AGS) technology is a relatively new technology for the treat-

12

ment of both domestic and industrial wastewater (Pronk et al., 2015). AGS technology

13

was developed at laboratory scale in the late nineties (Heijnen and van Loosdrecht, 1998;

14

Morgenroth et al., 1997), a full-scale demonstration installation in Gansbaai was started

15

up in 2008 (Giesen et al., 2013) and in 2011 the first full-scale application was reported

16

(van der Roest et al., 2011). Effluent requirements for COD, total nitrogen en total phos-

17

phorus were easily met, with average suspended solids effluent concentrations reported for

18

full-scale applications in the range of 5 mg L−1 to 20 mg L−1 (see table 1).

M AN U

SC

11

TE

Concentration Reference mg L−1 Epe 10 - 20 (van der Roest et al., 2011) Dinxperlo 15 - 20 (van der Roest et al., 2011) Gansbaai <5 (Giesen et al., 2013) Garmerwolde 20 (Pronk et al., 2015) Ryki 13 (Giesen et al., 2016) Vroomshoop 10 (Giesen et al., 2016)

AC C

EP

Plant

D

Table 1: Reported effluent suspended solids concentrations

19

Some reports indicate an increase in effluent suspended solids concentrations for AGS

20

reactors with increasing granulation (Li et al., 2008; Rockt¨aschel et al., 2015). In most

21

laboratory studies the focus is on the granulation process and wasting of sludge for ease of ∗

Corresponding author Email address: [email protected] (Edward van Dijk) 1 Both authors contributed equally to this manuscript

Preprint submitted to Water Research

September 29, 2018

ACCEPTED MANUSCRIPT

operation is combined with decanting of the effluent (de Kreuk et al., 2005). As a result,

23

biomass is present in the effluent of laboratory reactors. This poses no problem under

24

laboratory conditions, but for the full-scale treatment of sewage the removal of suspended

25

solids is an integral part of the purification process and is usually limited by law. Since

26

hydraulics of laboratory and even pilot plant installations are not representative for full-scale

27

conditions, effluent suspended solids can only effectively be studied at full-scale installations.

28

In full-scale applications the effluent and sludge withdrawal are separated processes. This

29

leads to relatively low suspended solids effluent concentrations, certainly when compared

30

to influent values (Pronk et al., 2015). Still, the reported average values of 5 mg L−1 to

31

20 mg L−1 are higher than usually reported for conventional activated sludge processes, which

32

can be considerably lower than 10 mg L−1 (Parker et al., 2001).

M AN U

SC

RI PT

22

Elevated levels of effluent suspended solids have been intensively studied over the last

34

decades in conventional activated sludge systems. Commonly reported issues in these sys-

35

tems are bulking sludge and foam formation (Jenkins, 1992). Another reported issue in

36

secondary clarifiers is the rising of sludge due to denitrification (Henze et al., 1993). Denitri-

37

fication can lead to nitrogen bubble formation (degasification) if over-saturation of nitrogen

38

gas in the liquid is reached. The bubbles attach to the sludge flocs which then rise towards

39

the top of the clarifier and wash out with the effluent.

AC C

EP

TE

D

33

40

Sludge settling too fast, indicated by a low sludge volume index (SVI), is also reported

41

to cause elevated levels of suspended solids in the effluent of conventional activated sludge

42

systems, although this effect is doubted by others reporting a strong relation with the

43

flocculator design (Parker et al., 2001, 1996; Vitasovic et al., 1997). Since reported SVI 3

ACCEPTED MANUSCRIPT

values for full-scale AGS reactors are generally low (35 mL g−1 to 70 mL g−1 ) (Pronk et al.,

45

2015; van der Roest et al., 2011) this might also be a potential cause of elevated levels of

46

effluent suspended solids.

RI PT

44

Besides a different sludge morphology and likely flocculation behavior, current granular

48

sludge plants according to the Nereda® concept deviate also in hydraulics from secondary

49

settling tanks. Secondary clarifiers have typical superficial upflow velocities of 0.5 m h−1 to

50

1.0 m h−1 , while the complex flow in the clarifier is mainly caused by density currents (Zhou

51

et al., 1992). For the Nereda® technology effluent results from pumping influent in the

52

bottom of the granular bed. This generates an upward plug flow in the reactor with typical

53

superficial velocities in the range of 2 m h−1 to 4 m h−1 through a granular bed with on top of

54

that a layer of flocculent sludge. The relatively higher upflow velocities in Nereda® reactors

55

might be responsible for the washout of floating sludge and fat-like particles. In contrast

56

to conventional activated sludge systems, the mechanisms behind the formation and control

57

of effluent suspended solids for full-scale aerobic granular sludge reactors are not yet fully

58

understood.

EP

TE

D

M AN U

SC

47

The aim of this study was to evaluate the main processes leading to effluent suspended

60

solids in full-scale AGS reactors, and the options to minimize effluent suspended solids.

61

Therefore, a full-scale Nereda® reactor, situated in the municipality of Utrecht, The Nether-

62

lands was used to study the mechanisms leading to elevated suspended solids concentrations

63

in the effluent. In this study, we focused on both removal of floating material and degasifi-

64

cation of nitrogen gas. Hereto, full-scale experiments were performed and a mathematical

65

model was developed to describe the degasification of nitrogen gas during the operation of

AC C

59

4

ACCEPTED MANUSCRIPT

the aerobic granular sludge reactor.

67

2. Materials and methods

68

2.1. Description of the plant

RI PT

66

In 2013 the Dutch water authority Hoogheemraadschap de Stichtse Rijnlanden took a

70

1050 m3 aerobic granular sludge reactor in operation. The reactor is located in the munic-

71

ipality of Utrecht, The Netherlands and was designed by Royal HaskoningDHV under the

72

trade name Nereda® .

M AN U

SC

69

The reactor has a footprint of 150 m2 and a water depth of 7.00 m. The influent pumps

74

have a total capacity of 900 m3 h−1 . The influent is pumped directly from a main sewer line

75

of the Overvecht district in Utrecht. The sewage consists mainly of domestic wastewater.

76

Wastewater characteristics are given in table 2. After screening by a 6 mm perforated plate

77

screen, the wastewater is directly pumped into the reactor. During the trial the reactor

78

was operated at a MLSS concentration of 8 kg m−3 to 12 kg m−3 , a SVI5 of 34 mL g−1 to

79

57 mL g−1 , a SVI30 of 33 mL g−1 to 43 mL g−1 , a volumetric loading rate of 0.6 m3 m−3 d−1

80

to 2.7 m3 m−3 d−1 and an average sludge loading rate of 0.08 kgCOD/kgMLSS/d. About

81

84 % of the biomass present in the reactor was granulated during the experimental period

82

(i.e. granule size larger than 0.2 mm). Average granule size distributions are shown in table

83

3. The reactor is equipped with two blowers each with a capacity of 400 m3 h−1 . Air is

84

supplied to the reactor using micro bubble diffusors evenly distributed on the floor of the

85

tank.

AC C

EP

TE

D

73

5

ACCEPTED MANUSCRIPT

Table 2: Influent of the aerobic granular sludge wastewater treatment plant in Utrecht, The Netherlands Min mg L−1 442.4 35.0 25.1 5.3 3.2 160.0

Max mg L−1 1007.5 81.7 58.6 11.8 8.2 380.0

RI PT

Average mg L−1 707.0 64.0 46.1 8.9 5.6 230.0

SC

Parameter COD NKj N NH4+ N Total Phosphorus PO43– P Suspended solids

The reactor is operated as a sequencing fed batch process with constant volume, con-

87

sisting of a combined feeding and effluent withdrawal step, a reaction phase and a settling

88

and sludge withdrawal step. Process conditions are set to target full biological nitrogen and

89

phosphorus removal. No possibilities for chemical dosing are present in the reactor. During

90

the trial nitrogen removal was established by nitrification and denitrification.

D

M AN U

86

The cycle time varied between 4 and 8 hours, with 1 hour of anaerobic feeding and

92

simultaneous effluent withdrawal, a variable reaction phase and a phase for mixing, setting

93

and decanting of sludge of 30 to 60 minutes.

EP

TE

91

The reaction phase was stopped when the set points, 2 mg L−1 and 3 mg L−1 for respec-

95

tively NH4+ N and NO3– N were reached. During the reaction phase the dissolved oxygen

96

concentration was controlled in the range of 1.0 mg L−1 to 2.0 mg L−1 .

AC C

94

97

The plant is in operation since the 1st of April 2013. The reactor was started up without

98

baffles in front of the overflow weir. At the 3rd of June 2015 vertical baffles were placed in

99

front of the effluent weirs, to investigate the effect of scum baffles on the concentration of

6

ACCEPTED MANUSCRIPT

100

effluent suspended solids.

101

2.2. Online measurements The reactor was equipped with probes for dissolved oxygen, redox potential, water level,

103

temperature, turbidity, and nitrate. Ammonium and phosphate are continuously measured

104

using a filter unit and auto sampling device (Hach Lange; Filtrax, Amtax and Phosphax).

105

The filter unit was situated 0.5 meter below the water surface. Sampling was done at an

106

interval of 5 minutes. A turbidity sensor (Hach Lange; Solitax) was present in the effluent

107

gutter to observe the presence of suspended solids.

108

2.3. Sampling

M AN U

SC

RI PT

102

Samples for analyses of influent and effluent were collected using refrigerated auto sam-

110

plers, collecting flow proportional samples for both influent and effluent. Suspended solids

111

concentrations of both influent and effluent were measured by filtering 1 L of sample with a

112

glass fiber filter, which was dried at 105 ◦C until no weight change was measured.

TE

Grab samples for MLSS analyses were taken from the top of the reactor after at least 15

EP

113

D

109

minutes of aeration at full capacity to ensure sufficiently mixed conditions.

115

2.4. Physical characteristics of the sludge

AC C

114

116

To determine the granule size distribution 1 L of sample was poured over a series of

117

sieves with different mesh sizes (200 µm, 400 µm, 630 µm, 1000 µm and 2000 µm). These

118

sieve fractions together with 100 mL of unsieved sludge were dried at 105 ◦C. The sludge

119

volume index (SVI) was measured by pouring 1000 mL of mixed, undiluted sludge into a

120

graduated measuring cylinder. The sludge volumes were measured after 5 and 30 minutes. 7

ACCEPTED MANUSCRIPT

Table 3: Average granule size distribution of the sludge in the aerobic granular sludge wastewater treatment plant in Utrecht, The Netherlands

121

RI PT

Percentage % 16 % 3% 2% 4% 31 % 44 % 100 %

SC

Concentration kg m−3 1.7 0.3 0.2 0.4 3.2 4.6 10.4

M AN U

Sieve fraction µm 0 - 200 200 - 400 400 - 600 600 - 1000 1000 - 2000 >2000 Total 2.5. Microscopic analysis

A Leica Microsystems Ltd stereo zoom microscope (M205 FA) in combination with Leica

123

Microsystems Qwin (V3.5.1) image analysis software was used to analyze the morphology

124

of the sludge.

125

2.6. Modelling of N2 stripping

TE

D

122

A mathematical model was developed to calculate the concentrations of dissolved nitro-

127

gen gas in the reactor. The solubility of nitrogen gas in water, cs , is dependent on the partial

128

pressure of nitrogen gas in the gas phase, p, and is calculated using Henrys law:

AC C

EP

126

cs = M · kH · p

(1)

129

Here kH is the Henrys constant for the solubility of nitrogen gas in water and M is the

130

molar mass of nitrogen gas. The Henrys constant is a decreasing function of temperature,

131

meaning that the solubility of nitrogen gas in water decreases with an increasing tempera8

ACCEPTED MANUSCRIPT

ture. The partial pressure of nitrogen in the gas phase is both dependent on the gas fraction,

133

f , and the pressure in the reactor, the latter being the sum of atmospheric pressure, Patm

134

and hydrostatic pressure:

p = f · (Patm + h · ρ · g)

RI PT

132

(2)

Here h is the water depth, ρ is the mixture density and g is the gravitational acceleration.

136

The transfer of nitrogen gas between the air bubbles and the liquid is calculated by:

M AN U

SC

135

dc = αF · (kL a)N · (cs − c) dt

(3)

Here c is the nitrogen gas concentration in the water phase, kL is the mass transfer

138

coefficient and a is the interfacial area of the gas bubbles. The subscripts O and N indicate

139

the values for oxygen gas and nitrogen gas. The mass transfer rate is corrected for effects

140

of wastewater contaminants, α, and fouling on the membrane aerators, F . The value of the

141

product of kL and a for oxygen transfer (kL a)O , is commonly measured during commissioning

142

in full-scale wastewater treatment plants and depends on the temperature. For this study,

143

a value of (kL a)O including the effects of contaminants and fouling was used. Hereto the

144

values for α and F were set to 1.

AC C

EP

TE

D

137

145

This empirical value of (kL a)O can be used to derive a value for nitrogen gas by using

146

Higbies penetration theory (Higbie, 1935). Based on this model the following relation be-

147

tween the kL a for oxygen gas and nitrogen gas and the diffusion coefficient for oxygen gas

148

(DO ) and nitrogen gas (DN ) for can be derived: 9

ACCEPTED MANUSCRIPT

(kL a)O = (kL a)N

r

DO DN

(4)

Both the diffusion coefficient and the Henry coefficient need to be corrected for temper-

150

ature. For the diffusion coefficient this is based on the Stokes-Einstein equation (Einstein,

151

1905). In this equation the diffusion coefficient D is related to the temperature T and the

152

dynamic viscosity µT :

M AN U

T µ0 T0 µ T

SC

RI PT

149

D(T ) = D0

(5)

The viscosity is also depending on temperature according to equation the Vogel-Fulcher-

154

Tammann equation (Dagdug, 2000), in which parameters A, B and C are empirical param-

155

eters:

TE

D

153

B

µ(T ) = eA+ C+T

EP

(6)

The Henry coefficient is corrected for temperature using the van ’t Hoff equation (van’t

157

Hoff, 1884), in which kH0 is the Henry coefficient at reference temperature, T0 is the reference

158

temperature and Hc is the temperature correction factor:

AC C

156

kH (T ) = kH0 e

−kHc ( T1 − T1 ) 0

(7)

159

The stripping of nitrogen gas is done by aeration of the reactor with fine bubble aera-

160

tion. This leads to transport of dissolved nitrogen gas through the reactor. A simple axial 10

ACCEPTED MANUSCRIPT

dispersion model can be used to describe the liquid backmixing of dissolved nitrogen gas in

162

the reactor (Degaleesan and Dudukovic, 1998), in which x is the reactor depth.

dc d2 c = Dax,1 2 dt dx

RI PT

161

(8)

The axial dispersion coefficient, Dax,1 , is a lumped coefficient, containing the effect of

164

both turbulent dispersion and global convective recirculation in the reactor. Molecular

165

diffusion in this case is neglected due to an expected limited impact. A tracer experiment

166

with NH4+ was used to determine the value of Dax,1 .

M AN U

SC

163

To describe the transfer of nitrogen gas between the air bubbles and the water phase,

168

equation 4 is substituted in equation 3. If this equation is combined with the transport

169

of nitrogen gas in the water phase according equation 8 a description for the stripping of

170

nitrogen in a 1D bubble column is constructed:

TE

D

167

172

DN · (cs (x) − c) DO

(9)

The saturation concentration for nitrogen gas according to equation 1 can be combined

AC C

171

r

EP

d2 c dc = Dax,1 2 + αF · (kL a)O dt dx

with the actual concentration to calculate the nitrogen gas deficit in the bulk liquid, cd .

cd = cs − c

(10)

173

When over-saturation occurs during aeration (i.e. the actual concentration c is bigger

174

than cs at ambient pressure and temperature), it is assumed that N2 gas bubbles are formed 11

ACCEPTED MANUSCRIPT

immediately and efficiently stripped out of the bulk liquid. This is modelled by artificially

176

increasing the local kL a with an order of magnitude.

177

2.7. Modelling of mixing during imperfect plug-flow feeding

RI PT

175

During feeding the bulk liquid in the reactor is pushed up towards the top of the reactor

179

where the overflow weirs are situated. To describe the transport of nitrogen gas during

180

feeding, the same axial dispersion model of equation 8 is used with a different value for

181

Dax . Also an additional term is added to describe the transport of nitrogen gas through

182

convection. Here, v is the upflow velocity of the liquid in the reactor. The value of Dax,2

183

was experimentally established by measuring a step response curve with NH4+ as a tracer.

M AN U

SC

178

(11)

D

dc d2 c dc = Dax,2 2 + v dt dx dx

The influent is fed from the bottom of the reactor. It is assumed that the influent con-

185

centration of dissolved nitrogen is equal to the equilibrium concentration of nitrogen gas at

186

atmospheric pressure. During the feeding phase denitrification can take place. The denitri-

187

fication rate depends on many parameters, but in this case a lumped specific denitrification

188

rate, r, is used. The specific denitrification rate was measured in the reactor in the days be-

189

fore the experiment, by measuring the decrease of the NO3– concentration during an anoxic

190

phase in the cycle. This rate is multiplied by the MLSS concentration, X, which is a function

191

of the reactor depth. The sludge bed is assumed to be in steady state and thus invariable

192

in time. In the simulations it was assumed that the sludge bed occupies the lower 4 meters

193

of the reactor and that the sludge is evenly distributed. We can then write equation 11 as

AC C

EP

TE

184

12

ACCEPTED MANUSCRIPT

194

follows:

dc d2 c dc = Dax,2 2 + v + rX(x) dt dx dx

RI PT

(12)

In this model the water in the reactor can get over-saturated with nitrogen gas. This

196

means that the local concentration of nitrogen gas, c, becomes larger than the local satu-

197

ration concentration according to equation 1. It is assumed that nitrogen gas bubbles are

198

formed immediately after over-saturation. Most likely supersaturation of nitrogen gas can

199

occur, but the effect is assumed to be limited (Henze et al., 1993).

M AN U

SC

195

4.

202

3. Results

203

3.1. Reactor operation

TE

201

D

To solve the model a Python code was written. All model parameters are given in table

200

The data were collected between June 2015 and May 2016. In this period the reactor

205

was run stable and without interruption with average effluent values of COD of 41 mg L−1 ,

206

Nkj N of 4.05 mg L−1 , NH4+ N of 1.29 mg L−1 , NO3– N of 3.72 mg L−1 , Ptot of 0.51 mg L−1

207

and PO43– P of 0.26 mg L−1 , showing an highly efficient N and P removal of respectively

208

94 % and 91 %. The average granule size distribution is reported in table 3. The sludge in

209

the reactor mainly consisted of smooth granules bigger than 1 mm, which is also shown by

210

the microscopic image of the sludge in figure 1A. The water temperature was in the range

211

of 9 ◦C to 23 ◦C.

AC C

EP

204

13

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

Figure 1: Biomass in the aerobic granular sludge plant in Utrecht, The Netherlands: (A)

D

granules in the reactor, (B) light microscopic picture of effluent suspended solids, (C) scum

3.2. Degasification of nitrogen gas

EP

212

TE

on top of the reactor after installation of the baffles, (D) biomass from the top of the reactor

Due to denitrification and upward transport of liquid (decreasing local pressure) over-

214

saturation of the water phase with nitrogen gas can occur, leading to gas bubble formation

215

in the feeding/decanting period. To confirm that over-saturation of nitrogen gas is present

216

during parts of the cycle an experiment was performed. Two comparable batches were fed

217

to the reactor. The batch sizes are given as volumetric exchange ratio (VER), which is

218

the ratio between the batch size (f.e. 400 m3 ) and the volume of the reactor (1000 m3 ).

219

Before the first batch (batch 1) the reactor was intensely aerated before feeding the reactor,

AC C

213

14

ACCEPTED MANUSCRIPT

while in the second batch (batch 2) this stripping phase was omitted. The gas stripping

221

period was 30 minutes long, with an air flow of 800 m3 h−1 . This high value was chosen

222

to strip a maximum amount of nitrogen gas. The process conditions in the two batches

223

were comparable (batch 1: volumetric exchange ratio 40 %, nitrogen removed in previous

224

batch 13.8 mg L−1 , average temperature 13.6 ◦C; batch 2: volumetric exchange ratio 40 %,

225

nitrogen removed in previous batch 13.9 mg L−1 , average temperature 13.5 ◦C). The effect

226

of a stripping phase is clearly illustrated in figure 2, showing the effluent turbidity for both

227

batches. Batch 1 showed an overall lower level effluent turbidity with a turbidity slightly

228

decreasing over the batch, with an average value of 4.9 FNU. The turbidity in batch 2 was

229

overall higher than in batch 1, with an average value of 21.0 FNU. At a volumetric exchange

230

ratio of 30 % the effluent turbidity started to increase up to a value of 48.5 FNU at the end

231

of batch 2. In figure 3 photographs recorded by a security camera are showing the gradual

232

formation of a scum layer due to degasification during batch 2.

20 10 0

AC C

Turbidity [FNU]

30

EP

batch 1 batch 2

40

TE

D

M AN U

SC

RI PT

220

0

5

10 15 20 25 30 Volumetric Exchange Ratio [%]

35

40

Figure 2: turbidity of the effluent in a batch with gas stripping (batch 1) and without gas stripping (batch 2)

15

SC

RI PT

ACCEPTED MANUSCRIPT

Figure 3: Degasification and the resulting floatation of biomass during feeding in a batch

233

M AN U

without stripping: (A) VER = 0 % (B) VER = 20 % (C) VER = 40 %

3.3. Stripping of nitrogen gas

Dissolved nitrogen gas can be stripped out by aerating right before the feeding phase in

235

order to prevent rising of biomass as is shown above. Equation 9, together with equation

236

1 and 2 for pressure correction, equation 5, equation 6 and equation 7 for temperature

237

correction and equation 4 for the conversion of the kL a can be used to evaluate the effect of

238

the stripping period on the dissolved N2 concentration in the reactor.

EP

TE

D

234

In figure 4 the effect of stripping of an initially fully N2 saturated situation in the reactor

240

is shown at a temperature of 20 ◦C. The dissolved nitrogen gas concentration in the reactor

241

(top at 0 m, bottom at 7 m) is shown. The figure shows three colored zones: the green

242

zone marks the area where the nitrogen gas concentration is lower than the equilibrium

243

concentration for air (i.e. no degasification expected). The red zone marks the area where

244

the nitrogen gas concentration is higher than the equilibrium concentration for pure nitrogen

245

gas (i.e. degasification expected). The white zone marks the area where only a partial

AC C

239

16

ACCEPTED MANUSCRIPT

nitrogen gas deficit exists. The boundary between the green area and the white area is

247

calculated according to equation 1 and equation 2 with a gas fraction of nitrogen gas in

248

air at ambient water pressure and temperature (f = 78 %). The boundary between the

249

white and the red area is calculated in a similar manner, but with pure nitrogen gas bubbles

250

(f = 100 %).

RI PT

246

The mixing of the reactor has a large impact on the effect of the stripping phase. Dis-

252

solved nitrogen gas is convectively transported through the reactor, leading to an over-

253

saturated situation near the top of the reactor and an undersaturated situation near the

254

bottom of the reactor. When the aeration continues, the nitrogen gas is slowly stripped out

255

of the reactor and a N2 deficit is created at the bottom of the tank. After 15 minutes, due

256

to the mixing, the N2 concentration near the bottom of the reactor is lower than the air

257

equilibrium concentration at 7 meters water depth. However, near the top of the reactor

258

the N2 concentration is still over-saturated. Even after 60 minutes of aeration the nitrogen

259

concentration is still at saturation concentration near the top of the reactor due to upmixing

260

of dissolved N2 gas. Note that for operational conditions over-saturation at the top of the

261

reactor is not important since there is no sludge bed/blanket near the top of the reactor

262

during effluent withdrawal.

AC C

EP

TE

D

M AN U

SC

251

17

N2 concentration

0

1

2

over saturated

M AN U

max deficit

C

3 4 5 reactor depth [m]

max deficit

B

N2 concentration

over saturated

RI PT

max deficit

A

N2 concentration

over saturated

SC

over saturated

D

36 34 32 30 28 26 24 22 20 18 16 36 34 32 30 28 26 24 22 20 18 16

6

70

TE

concentration [mg L 1]

concentration [mg L 1]

ACCEPTED MANUSCRIPT

D 1

2

N2 concentration

max deficit

3 4 5 reactor depth [m]

6

7

EP

Figure 4: Stripping of N2 gas from fully saturated situation (A) and after 5 minutes (B),

AC C

15 minutes (C) and 60 minutes of stripping (D) at 20 ◦C. Black solid line: the actual N2 concentration, red area: N2 concentration above saturation, green area: N2 concentration below equilibrium with air, boundary between red area and white area: N2 concentration at equilibrium with pure nitrogen gas at ambient water pressure, boundary between green area and white area: N2 concentration at equilibrium with air at ambient water pressure.

18

ACCEPTED MANUSCRIPT

Table 4: Model parameters

3.4. Degasification during feeding

Unit mol m−3 Pa−1 K g mol−1 Pa kg m3 m s−2 h−1 m2 s−1 m2 s−1 m2 s−1 m2 s−1

SC

RI PT

Value 6.4 × 10−6 1300 28 0.78 101,325 1000 9.81 1 1 5.5 2.10 × 10−9 1.88 × 10−9 -10.6265 578.919 -137.546 1.0 × 10−4 1.0 × 10−2

D

263

Symbol kh khc M f Patm ρ g α F (kL a)O DO DN A B C Dax,1 Dax,2

M AN U

Parameter Henry’s constant for N2 (25 ◦C) Temperature coefficient N2 Molar mass N2 Gas fraction of N2 in air Atmospheric pressure Density of water Gravity acceleration Alpha factor Fouling factor Gas transfer rate for oxygen Diffusion coefficient O2 (25 ◦C) Diffusion coefficient N2 (25 ◦C) Empirical coefficient A Empirical coefficient B Empirical coefficient C Axial dispersion during feeding Axial dispersion during aeration

The effect of denitrification during feeding on degasification of nitrogen gas was also

265

evaluated by extending the model according to equation 12. In figure 5 an example is shown

266

for a batch comparable with the previously described batch 1 in which a stripping phase of

267

30 minutes is applied, with T = 13.5 ◦C, r = 0.25 mg g−1 d−1 .

AC C

EP

TE

264

268

The model shows limited over-saturation during the feeding period of 1 hour, but the

269

over-saturation only occurs above the sludge blanket so rising of sludge due to bubble for-

270

mation is not likely to happen.

19

25

N2 concentration MLSS concentration

over saturated

biomass concentration [g L 1]

N2 concentration MLSS concentration

over saturated

20

RI PT

15 10 5

B

over saturated

N2 concentration MLSS concentration

over saturated

0

1

2

M AN U

max deficit

C

3 4 5 reactor depth [m]

max deficit

6

70

D 1

2

0 25

N2 concentration MLSS concentration

biomass concentration [g L 1]

max deficit

SC

A

D

36 34 32 30 28 26 24 22 20 18 16 36 34 32 30 28 26 24 22 20 18 16

TE

concentration [mg L 1]

concentration [mg L 1]

ACCEPTED MANUSCRIPT

20 15 10 5

max deficit

3 4 5 reactor depth [m]

6

7

0

EP

Figure 5: effect of feeding after a stripping phase on the dissolved nitrogen gas concentration

AC C

from the start (A), after 15 minutes (B), after 30 minutes (C) and after 50 minutes (D). T = 13.5 ◦C, r = 0.25 mg g−1 d−1 . Black solid line: the actual N2 concentration, gray solid line: biomass concentration, red area: N2 concentration above saturation, green area: N2 concentration below equilibrium with air, boundary between red area and white area: N2 concentration at equilibrium with pure nitrogen gas at ambient water pressure, boundary between green area and white area: N2 concentration at equilibrium with air at ambient water pressure. 20

ACCEPTED MANUSCRIPT

If the stripping phase is reduced to 5 minutes, again with T = 13.5 ◦C, r = 0.25 mg g−1 d−1 ,

272

a situation comparable with the above described batch 2 experiment can be modelled. The

273

results are shown in figure 6. The water in the vicinity of the sludge blanket is almost

274

immediately slightly over-saturated with nitrogen gas. After 30 minutes the top of the

275

sludge blanket is fully over-saturated with nitrogen gas, which most likely will lead to rising

276

sludge as the escaping nitrogen bubbles get caught by small flocs near the supernatant

277

sludge interface.

AC C

EP

TE

D

M AN U

SC

RI PT

271

21

25

N2 concentration MLSS concentration

over saturated

biomass concentration [g L 1]

N2 concentration MLSS concentration

over saturated

20

RI PT

15 10 5

B

over saturated

N2 concentration MLSS concentration

over saturated

0

1

2

M AN U

max deficit

C

3 4 5 reactor depth [m]

max deficit

6

70

D 1

2

0 25

N2 concentration MLSS concentration

biomass concentration [g L 1]

max deficit

SC

A

D

36 34 32 30 28 26 24 22 20 18 16 36 34 32 30 28 26 24 22 20 18 16

TE

concentration [mg L 1]

concentration [mg L 1]

ACCEPTED MANUSCRIPT

20 15 10 5

max deficit

3 4 5 reactor depth [m]

6

7

0

EP

Figure 6: effect of feeding without a prior stripping phase on the dissolved nitrogen gas

AC C

concentration from the start (A), after 15 minutes (B), after 30 minutes (C) and after 50 minutes (D). T = 13.5 ◦C, r = 0.25 mg g−1 d−1 . Black solid line: the actual N2 concentration, gray solid line: biomass concentration, red area: N2 concentration above saturation, green area: N2 concentration below equilibrium with air, boundary between red area and white area: N2 concentration at equilibrium with pure nitrogen gas at ambient water pressure, boundary between green area and white area: N2 concentration at equilibrium with air at ambient water pressure. 22

ACCEPTED MANUSCRIPT

278

3.5. Identification of suspended solids in the effluent The suspended solids present in the effluent were examined under a stereo-zoom micro-

280

scope to determine the characteristics and possible origin. Figure 1b and 1d show that the

281

suspended solids in the effluent are small dense light brown flocs. No filamentous organisms

282

like Thiothrix spp. or Microthrix spp. were observed that would normally be associated

283

with bad settling sludge (figure 1d). It was also observed that after light shaking of the sam-

284

ple flask the suspended solids quickly settled to the bottom indicating that the suspended

285

solids were able to settle well.

286

3.6. Effect of the vertical baffle

M AN U

SC

RI PT

279

During the first experimental period (between April 2013 and May 2015) the average sus-

288

pended solids effluent concentration was 30 mg L−1 with a 50-percentile value of 19.8 mg L−1 .

289

In the same period, many values below 10 mg L−1 were also recorded with a 25-percentile

290

value of 9.3 mg L−1 . On the 3rd of June 2015 vertical baffles were installed in front of the

291

overflow weirs. Composite samples of the effluent of the reactor were taken in the period

292

immediately before and after the baffles were installed; i.e. under similar operational and

293

sludge characteristics. Figure 7 shows the effluent suspended solids in this period. In the

294

period immediately before installation of the baffles the average effluent suspended solids

295

concentration was 23 mg L−1 , which is 10 % of the influent suspended solids concentration.

296

In the period directly after installation of the baffles only 1 sample was above 10 mg L−1

297

with a total average of 7.2 mg L−1 . In the period June 2015 to May 2016, 175 composite

298

influent and effluent samples were taken by the local water authority. The suspended solids

AC C

EP

TE

D

287

23

ACCEPTED MANUSCRIPT

300

pended solids was 7.8 ± 3.8 mg L−1 with a 95-percentile value of 13.6 mg L−1 . The average

301

removal efficiency of suspended solids was 97 %. After installation of the baffles a thin layer

302

of scum was present in most batches during feeding (figure 1c), but in this period no built

303

up of scum was observed. The thin layer of scum always disappeared in the first few minutes

304

of the reaction phase of the cycle.

SC

RI PT

concentration in the influent was 230 ± 118 mg L−1 and the concentration of effluent sus-

30 25 20 15 10 5

M AN U

before baffle after baffle

5

10

15 time [d]

20

25

TE

0

D

concentration [mgL 1]

299

EP

Figure 7: Suspended solids concentration in the effluent before and after installation of

AC C

vertical baffle in front of the overflow weirs

305

4. Discussion

306

4.1. Comparison of AGS reactors with activated sludge systems

307

In this study we examined the main processes for production of effluent suspended solids

308

in full-scale aerobic granular sludge reactors. The problem of rising sludge is known from

309

both continuously fed activated sludge systems (Henze et al., 1993) as from sequencing batch 24

ACCEPTED MANUSCRIPT

reactors (Tor`a et al., 2011). Comparison of the flow patterns in a secondary clarifier and the

311

AGS reactor reveals some important differences (figure 8). The outflow from an aeration

312

tank towards a secondary clarifier generally flows over a weir making the water pressure

313

almost atmospheric and generating a lot of turbulence. This is an effective way to remove

314

at least part of the dissolved nitrogen gas. Subsequently, the water from the aeration tank

315

flows into the secondary clarifier through the center well were the water pressure is close to

316

atmospheric pressure again. In the clarifier the flow splits into effluent and return sludge.

317

The return sludge is removed from the bottom of the clarifier. During dry weather conditions

318

a limited amount of sludge is present in the clarifier and this sludge will be near the bottom

319

of the clarifier at a water pressure well above atmospheric pressure. If the water pressure

320

increases at higher water depth, automatically the nitrogen gas deficit increases. Thus, if

321

nitrogen gas is formed in a clarifier due to denitrification this will mainly take place at

322

the bottom of the clarifier, where a relatively high nitrogen gas deficit is present. This is

323

probably why degasification in a clarifier is not a very common problem and is only known

324

from high loaded systems at high temperatures with high denitrification rates (Sarioglu and

325

Horan, 1996) or when the inflow to the clarifier is not degassed at an overflow weir. In the

326

AGS reactor the influent is fed from the bottom of the reactor, pushing the with nitrogen gas

327

saturated water in the reactor upwards to lower water pressure. This process is comparable

328

with the process occurring in a sequencing batch reactor, where degasification can occur

329

while decanting the effluent and thereby decreasing the water pressure. Thus, while feeding

330

the reactor, the upwards moving liquid gets a lower saturation concentration for nitrogen

331

gas, while in a clarifier it gets increased. In the AGS reactor during feeding, denitrification of

AC C

EP

TE

D

M AN U

SC

RI PT

310

25

ACCEPTED MANUSCRIPT

remaining NO3– on storage polymers by GAO and PAO like organisms, built up in previous

333

cycles, can occur (Kuba et al., 1996). The combination of a lower nitrogen gas deficit due

334

to pressure decrease and an higher denitrification potential make the AGS reactor more

335

susceptible for degasification of nitrogen gas in comparison to a clarifier. Therefore washout

336

of suspended solids due to degasification of nitrogen gas is less common in clarifiers than in

337

AGS reactors.

M AN U

SC

RI PT

332

Effluent

Scum/fatlike particles Biomass

Floating biomass with attached N2 bubble Effluent

TE

D

Overflow

Scum baffle

EP

Return sludge

AC C

Influent

AGS Reactor

Aeration tank + clarifier

Figure 8: Comparison of flow in an AGS reactor and flow in a conventional aeration tank and clarifier

26

ACCEPTED MANUSCRIPT

338

4.2. Effect of N2 stripping Rising of sludge due to degasification of dissolved N2 gas can cause elevated levels of

340

suspended solids in the effluent of activated sludge and AGS plants. Experimentally it was

341

shown that a nitrogen gas stripping phase just before the influent feeding phase was effective

342

to prevent rising of sludge in the AGS process. The developed mathematical model shows

343

that a limited amount of N2 gas deficit can be reached by applying a stripping phase. When

344

the water is pushed up during the feeding phase almost immediate over-saturation occurs

345

when no stripping phase is applied.

M AN U

SC

RI PT

339

The model presented by Henze et al. (Henze et al., 1993) only showed a steady state

347

N2 gas deficit in an unmixed situation. This model was extended to show the effect of non-

348

steady state aeration and feeding. A convection/dispersion term was added to the model

349

to describe both mixing due to aeration of the reactor and the plug flow behaviour of the

350

AGS reactor during feeding. The model shows that the local dissolved N2 gas concentration

351

changes over time and depth of the reactor. The maximum deficit for N2 gas depends highly

352

on the temperature, as do the denitrification rates. The model can easily be extended to

353

calculate the risk of rising sludge based on temperature, batch size and aeration history

354

of the reactor. Denitrification rates increase with temperature and the N2 gas deficit will

355

decrease with temperature. At temperatures above 20 ◦C the need for a stripping phase

356

increases. It should be realized that for the denitrification during the settling/feeding phase

357

mainly residual COD from the previous cycles is used. The COD containing influent is

358

replacing the nitrate containing liquid in the reactor. Therefore influent COD and nitrate

359

will only mix after the feeding phase when the aeration is switched on.

AC C

EP

TE

D

346

27

ACCEPTED MANUSCRIPT

360

4.3. Effect of the scum baffle Introduction of the vertical scum baffles in front of the effluent weir lowered the effluent

362

suspended solids concentration from 23 mg L−1 to 7 mg L−1 . Analyses of the suspended

363

solids in the effluent before installation of the baffles did not show any significant presence

364

of granules or filamentous bacteria. Microscopic analyses only revealed the presence of small

365

sludge flocs. The thin layer of sludge present on the surface of the reactor during feeding,

366

observed after installation of the baffles apparently was the result of the blocking of wash-out

367

of floating sludge by the baffles. The same effect is known from clarifiers in activated sludge

368

systems (Metcalf and Eddy, 2014).

M AN U

SC

RI PT

361

It was shown that at a MLSS concentration of 10.1 g L−1 and an influent suspended solids

370

of 230 mg L−1 an effluent suspended solids concentration below 10 mg L−1 were obtained over

371

long term operation. The flocs present in the layer on top of the reactor are assumed to

372

be grown in the reactor. Apparently not all biomass grown in the reactor is granular.

373

The flocculent mass likely consists of non biodegradable inert COD from the influent and

374

detached biomass from the granules. This is also shown by the fractionation of the sludge

375

showing an average 16 % of mass smaller than 200 µm.

AC C

EP

TE

D

369

376

The influent of the reactor contained an average of 230 mg L−1 of TSS after screening.

377

Since the only pre-treatment of the reactor was a 6 mm perforated plate screen, some floating

378

material from the sewerage will end up in the reactor. Although the influent consists merely

379

of domestic wastewater, the sewage of the Overvecht district is known to contain relatively

380

large fraction of fat. Some fatlike particles were observed in the reactor, but in the period of

381

10 months in which the experiment was run, no build up of scum or fat on the reactor was 28

ACCEPTED MANUSCRIPT

382

observed. Apparently the fat particles were converted or removed with the excess sludge. The reported values of effluent suspended solids of 20 mg L−1 for the Garmerwolde plant

384

(Pronk et al., 2015) are comparable with the levels measured in this study before installation

385

of the vertical baffles in front of the effluent weirs. Since the Garmerwolde plant mainly treats

386

domestic sewage and also the pre-treatment of the wastewater is comparable, these levels of

387

effluent suspended solids are to be expected. Also reported values for other plants (table 1)

388

were in the range of 10 mg L−1 to 20 mg L−1 and thus comparable with the values measured

389

in this study before installation of the vertical baffle. This value is to be expected from

390

a full-scale aerobic granular sludge reactor on domestic wastewater if no baffle is present.

391

The Gansbaai plant is the exception, with a value of 5 mg L−1 , but detailed information

392

is not available. Baffles have proven to be an effective and economic way to keep effluent

393

suspended solids below 10 mg L−1 , similar to well operated secondary clarifier effluents.

394

5. Conclusions

TE

D

M AN U

SC

RI PT

383

In this study, two main processes have been identified that can contribute to elevated

396

suspended solids concentration in the effluent of the aerobic granular sludge process in prac-

397

tice. These processes also play an important role in suspended solids control in secondary

398

clarifiers of activated sludge plants. Two solutions known from operation of secondary clar-

399

ifiers could successfully be incorporated into the AGS process leading to low concentration

400

of suspended solids in the effluent. The implementation of nitrogen stripping, before the

401

settling period, avoids gas bubble formation during feeding and thus the floatation of lighter

402

biomass. A mathematical model describing degasification was developed that explains the

AC C

EP

395

29

ACCEPTED MANUSCRIPT

observations of higher suspended solids concentrations in the effluent when treating sewage.

404

Introduction of a vertical baffle in front of the effluent weir showed to be an effective measure

405

to keep floating sludge and fat-like particle in the reactor, resulting in an effluent suspended

406

solids concentration lower than 10 mg L−1 with an average influent suspended solids concen-

407

tration of 230 mg L−1 . This was achieved with high biomass concentrations (10 g L−1 ) and a

408

high granulation grade (84 %).

409

6. Acknowledgements

M AN U

SC

RI PT

403

The authors wish to thank Mark Stevens, Jan Cloo and Tim van Erp for their end-

411

less dedication to this project. This work was supported by Royal HaskoningDHV, The

412

Netherlands.

413

Dagdug, L., 2000. A theoretical framework for the VogelFulcherTammann equation for covalent network

414

glasses derived by the stochastic matrix method. Journal of Physics: Condensed Matter 12 (00), 9573–

415

9589.

TE

D

410

de Kreuk, M., Pronk, M., van Loosdrecht, M., 2005. Formation of aerobic granules and conversion processes

417

in an aerobic granular sludge reactor at moderate and low temperatures. Water Research 39 (18), 4476–

418

4484.

420

421

422

AC C

419

EP

416

Degaleesan, S., Dudukovic, M. P., 1998. Liquid backmixing in bubble columns and the axial dispersion coefficient. AIChE Journal 44 (11), 2369–2378. ¨ Einstein, A., jan 1905. Uber die von der molekularkinetischen Theorie der W¨arme geforderte Bewegung von in ruhenden Fl¨ ussigkeiten suspendierten Teilchen. Annalen der Physik 322 (8), 549–560.

423

Giesen, A., De Bruin, L. M. M., Niermans, R. P., Van Der Roest, H. F., 2013. Advancements in the appli-

424

cation of aerobic granular biomass technology for sustainable treatment of wastewater. Water Practice &

425

Technology 8 (1), 47–54.

30

ACCEPTED MANUSCRIPT

426

Giesen, A., van Loosdrecht, M., Pronk, M., Robertson, S., Thompson, A., 2016. Aerobic Granular Biomass

427

Technology: recent performance data, lessons learnt and retrofitting conventional treatment infrastruc-

428

ture. Vol. 2016. Water Environment Federation.

434

435

436

437

438

439

440

441

442

443

444

RI PT

SC

433

cation. Water Research 27 (2), 231–236.

Higbie, R., 1935. The rate of absorption of a pure gas into still liquid during short periods of exposure. Institution of Chemical Engineers 35 (2), 36–60.

M AN U

432

Henze, M., Dupont, R., Grau, P., de la Sota, A., 1993. Rising sludge in secondary settlers due to denitrifi-

Jenkins, D., 1992. Towards a comprehensive model of activated sludge bulking and foaming. Water Science and Technology 25 (6), 215–230.

Kuba, T., Murnleitner, E., van Loosdrecht, M. C. M., Heijnen, J. J., 1996. A metabolic model for biological phosphorus removal by denitrifying organisms. Biotechnology and Bioengineering 52 (6), 685–695. Li, Z. H., Kuba, T., Kusuda, T., Wang, X. C., 2008. A Comparative Study on Aerobic Granular Sludge and

D

431

reactor. WO 9837027 (A1).

Effluent Suspended Solids in a Sequence Batch Reactor. Environmental Engineering Science 25 (4).

TE

430

Heijnen, J., van Loosdrecht, M., 1998. Method for acquiring grain-shaped growth of a microorganism in a

Metcalf, Eddy, 2014. Wastewater Engineering: Treatment and Resource Recovery, 5th Edition. McGrawHill.

Morgenroth, E., Sherden, T., van Loosdrecht, M., Heijnen, J. J., Wilderer, P. A., 1997. Aerobic granular

EP

429

sludge in a sequencing batch reactor. Water Research 31 (12), 3191–3194. Parker, D. S., Butler, R., Finger, R., Fisher, R., Fox, W., Kido, W., Merrill, S., Newman, G., Pope, R.,

446

Slapper, J., Wahlberg, E., 1996. Design and operations experience with flocculator-clarifiers in large

447

plants. Water Science and Technology 33 (12), 163–170.

448

449

450

451

AC C

445

Parker, D. S., Kinnear, D. J., Wahlberg, E. J., 2001. Review of Folklore in Design and Operation of Secondary Clarifiers. Journal of Environmental Engineering 127 (6), 476–484. Pronk, M., de Kreuk, M., de Bruin, B., Kamminga, P., Kleerebezem, R., van Loosdrecht, M., 2015. Full scale performance of the aerobic granular sludge process for sewage treatment. Water Research 84, 207–217.

31

ACCEPTED MANUSCRIPT

Rockt¨aschel, T., Klarmann, C., Ochoa, J., Boisson, P., Sørensen, K., Horn, H., 2015. Influence of the

453

granulation grade on the concentration of suspended solids in the effluent of a pilot scale sequencing

454

batch reactor operated with aerobic granular sludge. Separation and Purification Technology 142, 234–

455

241.

456

457

RI PT

452

Sarioglu, M., Horan, N., 1996. An equation for the empirical design of anoxic zones used to eliminate rising sludges at nitrifying activated sludge plants. Water Science and Technology 33 (3), 185–194.

Tor`a, J. A., Baeza, J. A., Carrera, J., Oleszkiewicz, J. A., 2011. Denitritation of a high-strength nitrite

459

wastewater in a sequencing batch reactor using different organic carbon sources. Chemical Engineering

460

Journal 172 (2-3), 994–998.

462

M AN U

461

SC

458

van der Roest, H. F., de Bruin, L., Gademan, G., Coelho, F., 2011. Towards sustainable waste water treatment with Dutch Nereda® technology. Water Practice & Technology 6 (3), 1–2. ´ van’t Hoff, J. H., 1884. Etudes de dynamique chimique. F. Muller & Co., Amsterdam.

464

Vitasovic, Z., Zhou, S., McCorquodale, J., Lingren, K., 1997. Secondary clarifier analysis using data from

465

the Clarifier Research Technical Committee protocol. Water Environment Research 69 (5), 999–1007.

466

Zhou, S., McCorquodale, J. A., Vitasovic, Z., 1992. Influences of Density on Circular Clarifiers with Baffles.

TE

EP

Journal of Environmental Engineering 118 (6), 829–847.

AC C

467

D

463

32

ACCEPTED MANUSCRIPT * Mechanisms for production of effluent suspended solids in AGS reactors unravelled * Stripping of nitrogen gas prevents rising sludge * Vertical baffle prevents washout of fat-like particles and scum

AC C

EP

TE D

M AN U

SC

RI PT

* Mathematical model of rising sludge describes stripping of nitrogen gas