Evaluation of Mechanistic Models for Nitrate Removal in Woodchip

Apr 10, 2017 - (50) The algorithm finds the set of parameters within a trust-region specified by the ...... Schmidt , C. A.; Clark , M. W. Deciphering...
0 downloads 0 Views 773KB Size
Subscriber access provided by University of Newcastle, Australia

Article

Evaluation of mechanistic models for nitrate removal in woodchip bioreactors Brian James Halaburka, Gregory H. LeFevre, and Richard G Luthy Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.7b01025 • Publication Date (Web): 10 Apr 2017 Downloaded from http://pubs.acs.org on April 10, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Environmental Science & Technology is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 31

Environmental Science & Technology

1

Evaluation of mechanistic models for nitrate

2

removal in woodchip bioreactors

3

Brian J. Halaburka1,2, Gregory H. LeFevre1,3, Richard G. Luthy1,2,*

4

1. Re-inventing the Nation’s Urban Water Infrastructure (ReNUWIt), National Science

5

Foundation Engineering Research Center

6

2. Department of Civil & Environmental Engineering, Stanford University, Stanford, California,

7

94305-4020 USA

8

3. Department of Civil & Environmental Engineering, University of Iowa, Iowa City, Iowa,

9

52242, USA

10

ABSTRACT

11

Woodchip bioreactors (WBRs) are increasingly being applied to remove nitrate from runoff. In

12

this study, replicate columns with aged woodchips were subjected to a range of measured flow

13

rates and influent nitrate concentrations with an artificial stormwater matrix. Dissolved oxygen

14

(DO), nitrate, and dissolved organic carbon (DOC) were measured along the length of the

15

columns. A multi-species reactive transport model with Michaelis-Menten kinetics was

16

developed to explain the concentration profiles of DO, nitrate, and DOC. Four additional models

17

were developed based on simplifying assumptions, and all five models were tested for their

ACS Paragon Plus Environment

1

Environmental Science & Technology

Page 2 of 31

18

ability to predict nitrate concentrations in the experimental columns. Global sensitivity analysis

19

and constrained optimization determined the set of parameters that minimized the root-mean-

20

squared error (RMSE) between the model and the experimental data. A k-fold validation test

21

revealed no statistical difference in RMSE for predicting nitrate concentrations between a zero-

22

order model and the other multi-species reactive transport models tested. Additionally, the multi-

23

species reactive transport models demonstrated no significant differences in predicting DO and

24

DOC concentrations. These results suggest that denitrification in an aged woodchip bioreactor at

25

constant temperature can effectively be modeled using zero-order kinetics when nitrate

26

concentrations >2 mg-N L-1. A multi-species model may be used if predicting DOC or DO

27

concentrations is desired.

28

INTRODUCTION

29

The National Academy of Engineering has identified managing the nitrogen cycle as one of the

30

14 Great Challenges for Engineering in the 21st Century,1 and woodchip bioreactors (WBRs)

31

have emerged as a promising approach to reduce nitrate exports in agricultural runoff and urban

32

stormwater.2-4 Woodchips are inexpensive and renewable, and current forest management

33

practices generate a substantial volume of low quality/low value wood.5 Additionally, woodchips

34

support high permeability, have a high C:N ratio (ranging from 30:1 to 3000:1), and robust

35

durability.6,7 Long-term field experiments indicate wood-particle media can provide consistent

36

nitrate removal for up to 15 years.8-10

37

Despite the increasing application and perceived advantages, the mechanisms governing nitrate

38

removal rates in WBRs are still poorly understood. The literature reports a wide range of

39

denitrification rates,6,11,12 ranging from 0.7-22.0 g-N m-3 media d-1. Numerous factors are

40

suggested for the large range of denitrification rates measured in the field, such as woodchip

ACS Paragon Plus Environment

2

Page 3 of 31

Environmental Science & Technology

41

age,7,11 temperature,13,14 and carbon substrate.15 Nevertheless, there is no clear consensus on the

42

appropriate model to explain woodchip-based denitrification. The most popular model to predict

43

reactor performance is a simple zero-order model, where the denitrification rate is constant and

44

nitrate removal is linearly related to hydraulic residence time (HRT).6,13,16 However, other

45

models have been proposed. Leverenz et al.17 suggested that a first-order model provides a better

46

fit for reactors operating at low nitrate concentrations and reduced temperatures. Hoover et al.18

47

reported influent nitrate concentration influences nitrate reduction rates up to 30-50 mg-N L-1,

48

suggesting first-order or Michaelis-Menten reaction kinetics.

49

In addition to simple zero- or first-order reaction kinetic models, several one-dimensional

50

transport models have been proposed. Jaynes et al.19 proposed a dual porosity model that

51

specifies a mobile and immobile fraction of water within the woodchip reactor based on the

52

assumption that denitrification occurs primarily inside the woodchips. Results were inconclusive

53

whether zero- or first-order reaction kinetics best described the data, and the fitted parameters for

54

the mobile and immobile fraction were significantly different from experimentally measured

55

values.19 Ghane et al.20 proposed a non-Darcy transport model with Michaelis-Menten kinetics to

56

describe denitrification rates. The Forcheimer hydraulic model closely matched the tracer test

57

data for horizontal flow in woodchip reactor beds,21 but the Michaelis-Menten reaction kinetic

58

parameters were estimated without replicate measurements and thus should not be deemed fully

59

robust. In addition, the denitrification rates in the reactor were abnormally high with a maximum

60

nitrate removal rate of 7.1 mg-N L-1 hr-1 at 23.5 °C, or 144.8 g-N m-3 media d-1, indicating the

61

woodchips were not fully aged.

62

A number of factors may explain the wide range of denitrification rates and models reported in

63

the literature. New woodchips have higher denitrification rates within the first year of operation

ACS Paragon Plus Environment

3

Environmental Science & Technology

Page 4 of 31

64

due to the leaching of excess organic material, and after approximately one year of operation the

65

rate stabilizes.22-24 Many studies use insufficiently aged woodchips, and as a result the reported

66

rates do not reflect long-term performance. In addition, many studies have poor spatial resolution

67

along the length of the reactor or do not take replicate measurements, making the assertion of

68

trends tenuous. Temperature variation and packing density of carbon source also can

69

substantially alter denitrification rates,14,23 but wood type and grain size do not.12,15

70

The critical parameters needed to model nitrate removal in woodchip reactors are still poorly

71

understood.2 Under conditions where heterotrophic denitrification is controlled by dissolved

72

oxygen (DO), nitrate, and dissolved organic carbon (DOC) concentrations,25 a complete

73

mechanistic model of denitrification in woodchip reactors would include all three of these

74

parameters. Although a complete mechanistic model may improve understanding of the

75

processes involved in denitrification in woodchip bioreactors (WBRs), the optimal model to use

76

in practice is the most parsimonious and several simplifying assumptions may be made.

77

The objective of this work was to quantitatively evaluate mechanistic reactive transport models

78

describing denitrification in laboratory WBR columns using aged woodchips. Five models were

79

evaluated in this study. The models were calibrated using experimental data collected from

80

laboratory woodchip columns that were aged for over a year, then evaluated using sensitivity

81

analysis and a k-fold validation test. The results of this study will increase understanding of the

82

underlying mechanisms of denitrification in WBRs, while providing justification for the use of

83

the simplest model to describe WBR performance.

84

METHODS

85

Column Design

ACS Paragon Plus Environment

4

Page 5 of 31

Environmental Science & Technology

86

Woodchips were obtained from an arborist woodchip waste pile in Portola Valley, CA. The

87

woodchips were composed of a mix of species, including California redwood (Sequoia

88

sempervirens), oak (Quercus sp.), and Douglas fir (Pseudotsuga menziesii). The woodchips were

89

dried at 50 °C for 48 hours in a drying oven then sieved to a diameter between 2-10 mm.

90

Woodchip type and particle size have been reported to have no significant effect on nitrate

91

removal rates,12,15 thus additional woodchip composition analysis was not performed.

92

Three PCV column reactor columns (10 cm ID x 50 cm) were constructed with sample ports

93

installed every 5 cm (Figure S1). For sample ports, 3.81 cm long luer-lock needles (gauge #16)

94

with ball valves were wrapped with PTFE tape and press-fit into holes drilled in the side of the

95

columns (Figure S2), allowing sampling from the center of the column. A total of 11 sample

96

ports were installed on each column. A stainless steel screen (mesh #10) was placed at the

97

bottom of each column to support the woodchips. 700 g of the dried and sieved woodchips were

98

added to each column and lightly compacted every 5 cm such that woodchips filled the column

99

to the top, corresponding to a packing density of 0.18 g cm-3. Upon completion of the

100

experiments, drainable porosity (specific yield) of the columns was determined by draining the

101

columns from the bottom over a 1-hour period, measuring the weight of the drained water, and

102

subtracting the volume of the bottom cap from the total volume drained. The woodchips were

103

then removed from the column and specific retention was determined by measuring the

104

difference between the wet and dry media after 48 hours in a drying oven at 50 °C. Total

105

porosity was determined by summing drainable porosity and specific retention.

106

Tracer Tests

107

Prior to running the experiments, linear pore-water velocity and dispersion coefficients were

108

estimated for each column with an interval-pulse bromide tracer test at a flow rate of 26 mL min-

ACS Paragon Plus Environment

5

Environmental Science & Technology

Page 6 of 31

(described in the SI). Theoretical HRT (τ) was calculated as  = V n ⁄60Q where Vr is volume

109

1

110

of the reactor (mL), ne is effective porosity (-), and Q is flow rate (mL min-1). Effective porosity

111

is the fraction of the total volume that contributes to fluid flow, and drainable porosity was used

112

as an estimate of effective porosity for the calculation of τ. Actual mean HRT ( ̅) was calculated

113

as ̅ = / where L is reactor length (cm) and ν is porewater velocity (cm h-1). Hydraulic

114

̅ . The ideal reactor would have an eV efficiency, eV, of the reactor was calculated26 as  = ⁄

115

value of 1, indicating plug flow conditions. An eV value of less than 1 indicates short-circuiting,

116

while an eV value greater than 1 may indicate drainable porosity is less than effective porosity or

117

that physical retardation is occurring such as fluid entering micropores (specific retention) within

118

the woodchips.27

119

Column Experiments

120

The columns were operated at room temperature (21 °C) in up-flow mode using variable speed

121

digital peristaltic pumps (Masterflex) to maintain saturated hydraulic conditions. Each column

122

was fed artificial stormwater at a different measured flow rate (1.5 mL min-1, 3.8 mL min-1, and

123

8.4 mL min-1). The artificial stormwater matrix was composed of 0.75 mM CaCl2, 0.075 mM

124

MgCl2, 0.33 mM Na2SO4, 1 mM NaHCO3, 0.0715 mM NH4Cl, and 0.016 mM Na2HPO4,

125

representing the average concentration of major ions in urban stormwater.28 NaNO3 was added to

126

the stormwater matrix to achieve an initial nitrate concentration of 10 mg-N L-1. Columns were

127

aged for 13 months with the flowing stormwater matrix prior to conducting tracer tests and

128

column experiments. For the column experiments, all columns were exposed to three influent

129

nitrate concentrations of 11 mg-N L-1, 5 mg-N L-1, and 2 mg-N L-1. Preliminary measurements

130

showed that concentration profiles of all the columns reached steady-steady within 2-3 days

ACS Paragon Plus Environment

6

Page 7 of 31

Environmental Science & Technology

131

following a change of influent nitrate concentration. For each concentration, the columns were

132

allowed to run for one week at the new nitrate concentration before sampling began.

133 134

Sampling and Analysis

135

The columns were sampled along the entire length for DO, nitrate, and DOC. Sampling was

136

repeated four times during a period of one week to obtain replicate measurements. Thus four

137

replicates were collected from the reactors at three different flow rates and three different

138

influent nitrate concentrations for a total of nine different conditions tested. For each sampling

139

event, samples were collected from all sample ports as well as the artificial stormwater matrix

140

tank to verify the stability of the solution. DOC and nitrate samples were collected starting at the

141

top-most sample port (at outlet) and moving downward (toward inlet) such that each sample was

142

representative of the porewater at or just above the sample port. Fifteen milliliters of sample was

143

collected in a 25 mL plastic syringe and filtered using a sterile 0.45 μm PVDF filter into a 24 mL

144

glass vial baked at 450 °C for four hours in a muffle furnace. All samples were analyzed within

145

four hours of sample collection and in random order using a random number generator. Nitrate

146

was measured using a WestCo SmartChem 200 Discrete Analyzer (detection limit: 0.05 mg-N L-

147

1

148

a Unisense dissolved oxygen needle probe (model DO-NP) and the Unisense SensorTrace

149

Software.

). DOC was measured using a Shimadzu TOC-L Autoanalyzer. DO was measured in situ using

150

Model Development

151

Five models were quantitatively evaluated to describe denitrification in the experimental

152

woodchip columns. The first model evaluated was a system of one-dimensional advection

153

dispersion equations with coupled Michaelis-Menten reaction kinetics to describe the transport

ACS Paragon Plus Environment

7

Environmental Science & Technology

Page 8 of 31

154

of DO, nitrate, and DOC (Model 1). This system of 1-D advection-dispersion equations was

155

chosen because similar reactive transport models have been used with success to describe

156

microbial substrate, oxygen, and nitrate uptake in porous media,29,30 contaminant degradation in

157

porous media,31,32 and denitrification in hyporheic zone sediments.33,34 The generalized model

158

for each constituent takes the form      = −  −  (1)    

159

where Ci is the concentration of the ith species (mg L-1), t is time (h), D is the dispersion

160

coefficient (cm2 h-1), ν is the effective porewater velocity (cm h-1), x is distance along the column

161

(cm), and Ri is the biological reaction rate term for the ith species (mg L-1 h-1). The biological

162

reactions modeled in the woodchip columns are aerobic respiration, denitrification, and cellulose

163

hydrolysis. For aerobic respiration of DOC, both the availability of DO and DOC can limit the

164

overall reaction rate. Without knowing the limiting substrate a priori, coupled Michaelis-Menten

165

kinetics is an effective method to model the overall microbial kinetics.35 The aerobic reaction

166

rate can be expressed in the form  =

 ! "

 ' &" & (2) #$ +  #( + '

167

where RO is the rate of oxygen uptake (mg-O2 L-1 h-1), XO is the concentration of aerobic

168

heterotrophs (mg-biomass L-1), VO is the maximum uptake rate of DO (mg-O2 mg-biomass-1 h-1),

169

C is the concentration of DOC (mg-C L-1), Kc is the half-saturation constant for DOC (mg-C L-

170

1

171

O2 L-1).

), O is the concentration of DO (mg-O2 L-1), and Ko is the half-saturation constant for DO (mg-

172

Denitrification can similarly be expressed as a coupled Michaelis-Menten reaction, with the

173

addition of a non-competitive inhibition term representing the inhibiting effect of DO on

174

denitrification. This reaction rate takes the form

ACS Paragon Plus Environment

8

Page 9 of 31

Environmental Science & Technology

* =

 + #, &" &" & (3) #$ +  #* + + #, + '

* !* "

175

where RN is the rate of denitrification (mg-N L-1 h-1), XN is the concentration of heterotrophic

176

denitrifiers (mg-biomass L-1), VN is the maximum rate of denitrification (mg-N mg-biomass-1 h-

177

1

178

N L-1), and KI is the inhibition constant of DO (mg-O2 L-1).

), N is the concentration of nitrate (mg-N L-1), KN is the half-saturation constant for nitrate (mg-

179

DOC is consumed through both aerobic respiration and denitrification, and the DOC reaction

180

rate is modeled as a combination of the Michealis-Menten reaction equations for the two

181

processes. The DOC reaction term takes the form . = /

 ' &" & + /* #$ +  #( + '

 ! "

* !* "

 + #, &" &" & (4) #$ +  #* + + #, + '

182

where RC is the rate of DOC uptake (mg-C L-1 h-1), βO is the uptake coefficient for DO (mg-C

183

mg-O2-1), βN is the uptake coefficient for nitrate (mg-C mg-N-1). The uptake coefficients for DO

184

and nitrate are the ratios of the mass of DOC consumed per mass of DO or nitrate consumed,

185

respectively.

186

In addition to the degradation term, the DOC transport equation includes a DOC production

187

term for cellulose hydrolysis. Cellulose hydrolysis is the enzymatic process by which microbes

188

cleave crystalline cellulose into smaller soluble oligosaccharides.36 Cellulose in woody material

189

is obstructed by lignin such that only certain surface binding sites are available for cellulase

190

adsorption. As more cellulose is hydrolyzed, fewer binding sites are available so the DOC

191

hydrolysis rate decreases over time following a power-law function.22,37 After an initial sharp

192

decrease in DOC release, the power law function reaches a quasi-steady state wherein the DOC

193

release rate remains relatively constant. This behavior is observed in field-scale woodchip

194

reactors where reaction rates drop rapidly within the first year of operation but then stay

195

relatively constant.6,11,24 The presence of fungi in oxic zones may increase the rate of DOC

ACS Paragon Plus Environment

9

Environmental Science & Technology

Page 10 of 31

196

release due to their ability to break down lignin, making available more cellulose to be

197

hydrolyzed.36 To account for different DOC release rates in oxic and anoxic zones, DOC release

198

kinetics are modeled using the equation #1 = !12 "

' #, & + !1 " & (5) # + ' #, + '

199

where Kh is the DOC production rate (mg-C L-1 h-1), Vh1 is the aerobic maximum DOC

200

production rate (mg-C L-1 h-1), and Vh2 is the anaerobic maximum DOC production rate (mg-C L-

201

1

202

for DO uptake and denitrification, respectively.

h-1). The values for KO and KI are assumed to be the same as those in the reaction rate equations

203

Three coupled equations (DO, nitrate, and DOC) comprise the model. The model assumes that

204

(1) the system is in steady-state and the transient term is zero, (2) the microbial biomass is

205

constant within each region and fixed to surfaces so that the microbial biomass terms XO and XN

206

can be combined with VO and VN, (3) substrate and electron acceptors (O2 and NO3-) are the only

207

limitations to sustaining growth, (4) all DOC is labile and bioavailable, and only dissolved

208

substrate is taken up by bacteria, (5) insoluble substrate mass remains relatively constant due to

209

the longevity of woodchips,16 and (6) sorption and intra-particle diffusion of DOC through

210

woodchips is not important because the equations are solved at steady-state. Thus the three

211

partial differential equations to model DO, nitrate, and DOC mass transport are

212

213

Dissolved Oxygen:

Nitrate:

(Model 1)  ' '  ' − ! " &" & 0 =   −    #$ +  #( + '

0= 214

 + +  + #, −  − ! " & " & " & *    #$ +  #* + + #, + '

Dissolved Organic Carbon:

ACS Paragon Plus Environment

10

Page 11 of 31

Environmental Science & Technology

    '  + #, 0 =   −  − / ! " &" & − /* !* " &" &" &   #$ +  #( + ' #$ +  #* + + #, + ' ' #, + !12 " & + !1 " & # + ' #, + '

215

The second model (Model 2) simplifies Model 1 by assuming DO does not significantly

216

impact the overall denitrification rate. It is well established that DO inhibits denitrification;35

217

however, denitrification has been observed in WBRs13,38 with DO concentrations between 0.5-

218

4.5 mg L-1. One explanation is that micropores within the woodchips create anaerobic

219

environments where denitrification can occur,25 suggesting that bulk solution DO concentrations

220

have little effect on the overall denitrification rate of WBRs. Alternatively, aerobic respiration

221

and denitrification may occur in the bulk solution, but aerobic respiration occurs at a much faster

222

rate and DO inhibition has a relatively minor effect on the over nitrate removal rate. The DO

223

terms are removed from the system of equations, and the model takes the form

224

Nitrate:

(Model 2) 0=

225

Dissolved Organic Carbon: 0=

 + +  + −  − !* " &" &    #$ +  #* + +

    + −  − / ! " & " & + !1 * *    #$ +  #* + +

226

The third model (Model 3) is an alternate simplification of Model 1 by assuming

227

denitrification is not dependent on DOC concentrations. DOC has been identified as the limiting

228

reactant in WBR denitrification.13,39 Nevertheless, a number of published denitrification models

229

ignore DOC concentrations, yet adequately fit experimental nitrate data.20,40 If denitrification in

230

WBRs is carbon limited, then C