DESIGN CRITERIA FOR THE TREATMENT OF MILKING FACILITY WASTEWATER IN
A COLD WEATHER VERTICAL FLOW WETLAND
By
Emily Loraine Campbell

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Biosystems Engineering—Master of Science
2014

ABSTRACT
DESIGN CRITERIA FOR THE TREATMENT OF MILKING FACILITY WASTEWATER IN
A COLD WEATHER VERTICAL FLOW WETLAND
By
Emily Loraine Campbell

The wastewater produced by Michigan milking facilities is harmful to the environment
and costly to treat. A field demonstration of a subsurface treatment wetland in mid-Michigan has
had continual success and has proven the capacity of constructed wetlands to effectively and
affordably treat milking facility wastewater in this climate. However, the wetland’s design was
inefficient and over-sized. Data from two lab-scale wetlands were collected and analyzed in
order to create sizing and design criteria for a cost-effective design.
As a result, design and sizing criteria were created for two types of wash water: manure
contaminated and manure free. Several types of nutrients were monitored, including phosphorus,
organic matter, total nitrogen, ammonia, nitrate, and pH. Since nitrogen levels in the manure
contaminated wash water was three times larger than that in the manure free wash water,
different designs were needed, though both designs were successful at treatment of COD and
nitrogen. Phosphorus treatment was adequate, but temporary, as the wetland media lost capacity
with time. Clogging potential was also taken into consideration. Although the change in
hydraulic conductivity was indiscernible in the lab wetlands, the field site, which has been in
operation over five years, was examined and no sign of clogging was found in sampling.

ACKNOWLEDGEMENTS

There is a number of people without whom this thesis would not be possible. First and
foremost, my advisor Dr. Steven Safferman for his help and guidance in every aspect of the
project from conception to the final result. In addition to being a technical advisor, he was an
excellent mentor and emotional support.
Next, I’d like to thank my statistician, Wenjuan Ma, who was extremely proactive and
patient to make sure I completed the statistics portion of the project successfully.
I’d also like to thank my committee, the undergraduate research team, and the staff in the
Biosystems department. Without these people, I wouldn’t have passed my dissertation or built
my lab scale wetlands.
A special thanks to Chuck and Linda Grettenberger as well as Aron Whitaker for
allowing us to take wash water from their farms, and, in the case of Grettenberger farm, allowing
us to play with their cats and dig up the field wetland.
Finally, I’d like to thank my friends and family for being so supportive and rooting for
me all along.

iii

TABLE OF CONTENTS

LIST OF TABLES……..………………………………………………………………………....vi
LIST OF FIGURES.……………………………………………………………………………...ix
INTRODUCTION………………………………………………………………………………...1
HYPOTHESIS…………………………………………………………………………………….2
OBJECTIVES……………………………………………………………………………………..3
LITERATURE REVIEW………………………………..…………….………………………….4
CONSTRUCTED WETLANDS…………………………….……………..……………………..5
Types of Constructed Wetlands…………………………………………………………...6
Constructed Wetland Components………………………………...……………………...8
Clogging………………………………………………………………………………….11
Effect of Cold Weather on Constructed Wetlands……………………………………….11
Milking Facility Wash Water…………………………………………………………….13
Example 1: Italy HSSF for Milking Parlor Wash Water and Domestic Sewage
(Mantovi et al. 2003)……………………………………………………………..15
Example 2: Canada VFCW for Milking Parlor Wash Water
(McKague 2012)…………………………………………………………………16
Example 3: Local Michigan VFCW for Milking Parlor Wash Water…………………...18
TREATMENT MECHANISMS FOR SELECT NUTRIENTS…………………………………21
Phosphorus……………………………………………………………………………….21
Organic Material…………………………………………………………………………22
Nitrogen………………………………………………………………………………….22
Ammonia Volatilization………………………………………………………….23
Matrix Adsorption………………………………………………………………..23
Ammonification………………………………………………………………….23
Nitrification………………………………………………………………………23
Denitrification……………………………………………………………………24
Anammox………………………………………………………………………...25
Plant Uptake……………………………………………………………………...25
DESIGN OF CONSTRUCTED TREATMENT WETLANDS…………………………………26
METHODS AND MATERIALS………………………………………………………………...28
Structure………………………………………………………………………………….28
Wetland Cells…………………………………………………………………………….30
Media…………………………………………………………………………………….32
System Hydraulics……………………………………………………………………….33
Wash Water...…………………………………………………………………………….38
Simulating Cold Temperature……………………………………………………………39
Monitoring Procedures…………………………………………………………………...41
Maintenance……………………………………………………………………………...43
iv

Sampling Procedure……………………………………………………………………...44
Analytical Testing………………………………………………………………………..47
Hydraulic Conductivity…………………………………………………………………..48
Statistical Methods……………………………………………………………………….50
RESULTS AND DISCUSSION…………………………………………………………………58
Starting Conditions………………………………………………………………………58
Phases…………………………………………………………………………………….59
Phosphorus……………………………………………………………………………….61
COD……………………………………………………………………………………...67
Nitrogen………………………………………………………………………………….74
Clogging………………………………………………………………………………….96
Sizing and Design………………………………………………………………………..97
CONCLUSIONS……………………………………………………………………………….100
APPENDICES………………………………………………………………………………….104
Appendix A: Detailed Results of Michigan Field Site………………………………...105
Appendix B: Monitoring Sheet…………………………………………………………122
Appendix C: Temperature Monitoring Data……………………………………………125
Appendix D: Analytical Recording Sheet………………………………………………129
Appendix E: Hydraulic Conductivity Calculation……………………………………...131
Appendix F: Wetland Field Site Clogging Examination……………………………….136
Appendix G: Analytical Results………………………………………………………..137
Appendix H: QAQC……………………………………………………………………142
Appendix I: Statistics Results…………………………………………………………..144
REFERENCES…………………………………………………………………………………184

v

LIST OF TABLES

Table 1: Loading values for milking facility wastewater………………………………………..15
Table 2: Nutrient loadings at Canadian milking facilities……………………………………….17
Table 3: Treatment at Michigan three-celled VFCW……………………………………………20
Table 4: Pump Detail…………………………………………………………………………….38
Table 5: Method Numbers……………………………………………………………………….47
Table 6: Operational Conditions…………………………………………………………………60
Table 7: Percent Difference between Phases for Phosphorus in MF Wetland…………………..62
Table 8: Percent Removal per Cell for Phosphorus in MF Wetland…………………………….63
Table 9: Percent Difference between Phases for Phosphorus in MC Wetland………….……….65
Table 10: Percent Removal per Cell for Phosphorus in MC Wetland…………………………...65
Table 11: Differences in Phosphorus Levels between MF and MC Wetlands…………………..66
Table 12: Percent Difference between Phases for COD in MF Wetland………………………..68
Table 13: Percent Removal per Cell for COD in MF Wetland…………………………………..69
Table 14: Percent Difference between Phases for COD in MC Wetland……………….……….71
Table 15: Percent Removal per Cell for COD in MC Wetland………………………………….72
Table 16: Differences in COD Levels between MF and MC Wetlands…………………………72
Table 17: Percent Difference between Phases for TN in MF Wetland…….…………………….75
Table 18: Percent Removal per Cell for TN in MF Wetland…………………………………….76
Table 19: Percent Difference between Phases for Ammonia in MF Wetland…….……………..77
Table 20: Percent Removal per Cell for Ammonia in MF Wetland……………………………..77
Table 21: Percent Difference between Phases for Nitrate in MF Wetland……………….……...78
Table 22: Percent Removal per Cell for Nitrate in MF Wetland………………………………...79
vi

Table 23: Percent Difference between Phases for pH in MF Wetland…………….…………….81
Table 24: Percent Removal per Cell for pH in MF Wetland…………………………………….81
Table 25: Percent Difference between Phases for TN in MC Wetland………………….………84
Table 26: Percent Removal per Cell for TN in MC Wetland……………………………………85
Table 27: Percent Difference between Phases for Ammonia in MC Wetland…………………..87
Table 28: Percent Removal per Cell for Ammonia in MC Wetland……………………………..88
Table 29: Percent Difference between Phases for Nitrate in MC Wetland……………………...89
Table 30: Percent Removal per Cell for Nitrate in MC Wetland…………………………...…...90
Table 31: Percent Difference between Phases for pH in MC Wetland………………….……….92
Table 32: Percent Removal per Cell for pH in MC Wetland…………………………………….93
Table 33: Differences in Nitrogen Speciation between MF and MC Wetlands…………………94
Table 34: Effluent Concentration Comparison across Different Technologies………………….98
Table 35: Loading Values………………………………………………………………………..98
Table 36: Field Site Alkalinity Removal……………………………………………………….106
Table 37: Field Site Ammonia Removal……………………………………………………….108
Table 38: Field Site BOD Measurements……………………………………………...……….110
Table 39: Field Site COD Removal…………………………………………………………….111
Table 40: Field Site Nitrate Removal…………………………………………………….…….113
Table 41: Field Site pH Measurements…………………………………………………..…….115
Table 42: Field Site Phosphorus Removal……………………………………………….…….117
Table 43: Field Site TS Removal……………………………………………………………….119
Table 44: Field Site VS Removal………………………………………………………...…….119
Table 45: Temperature Monitoring Data……………………………………………………….126
vii

Table 46: Trial 1 for Hydraulic Conductivity Measurement…………………………………...132
Table 47: Trial 2 for Hydraulic Conductivity Measurement…………………………………...133
Table 48: MF Wetland Analytical Data ………………………………………………...……...138
Table 49: MC Wetland Analytical Data………………………………………………...……...140
Table 50: Relative Percent Difference between Replicates…………..………………………...143
Table 51: Relative Percent Difference for Standards……………………...…………………...143
Table 52: Statistical Data for Overall Models………………………………………..………...145
Table 53: Statistical Data for Manure Free Models…………………………..………………...148
Table 54: Statistical Data for Manure Contaminated Models………………………..………...151
Table 55: Statistical Data for Influent Models……………………..…………………………...154
Table 56: Statistical Data for Cell 1 Models………………………….………………………...157
Table 57: Statistical Data for Cell 2 Models…………………………..………………………..160
Table 58: Statistical Data for Cell 3 Models………………………………..…………………..163
Table 59: Statistical Data for Phase 1 Models……………………………………………….....166
Table 60: Statistical Data for Phase 2 Models…………………………………………..……...169
Table 61: Statistical Data for MF Phase 3 Models………………………………………...…...172
Table 62: Statistical Data for MC Phase 3 Models………………...…………………………...175
Table 63: Statistical Data for MF Phase 4 Models…………………………...………………...178
Table 64: Statistical Data for MC Phase 4 Models…………………………...………………...181

viii

LIST OF FIGURES

Figure 1: Cross-section of a HFCW……………………………………………………………….7
Figure 2: Cross-section of a VFCW……………………………………………………………….8
Figure 3: Cross-section of three-celled VFCW design…………………………………………..17
Figure 4: Lab-scale Wetland Structure…………………………………………………………..29
Figure 5: Inlet Schematic (Cells 1 and 3)………………………………………………………..31
Figure 6: Outlet Schematic (Cell 2)…….………………………………………………………..31
Figure 7: Hydraulic Conductivity Outlet…………………………………………………….…..32
Figure 8: Flow Schematic…………………………………………………….………………….34
Figure 9: Reservoir 1…………………………………………………….………………………35
Figure 10: Reservoir 2 and 3 (Side View)…………………………………………………….…36
Figure 11: Recycle System (Top View)…………………………………………………….……37
Figure 12: Cooling System (Side View)…………………………………………………….…...40
Figure 13: Sampling Locations…………………………………………………….…………….45
Figure 14: Sampling Set-Up…………………………………………………….……………….46
Figure 15: Hydraulic Conductivity Set-Up………………………………………………………49
Figure 16: Nested Data Diagram…………………………………………………….…………..52
Figure 17: Random Effect (1|Date)…………………………………………………….………...54
Figure 18: Random Effect (Stage|Date)…………………………………………………….……55
Figure 19: Random Effect (1+Stage|Date)………………………………………………………55
Figure 20: Graph of Analytically Measured Phosphorus Levels for MF Wetland………………61
Figure 21: Graph of Analytically Measured Phosphorus Levels for MC Wetland……………...64
Figure 22: Graph of Analytically Measured COD Levels for MF Wetland……………………..67
ix

Figure 23: Graph of Analytically Measured COD Levels for MC Wetland……………………..69
Figure 24: Cell 1 COD Removal…………………………………………………….…………..73
Figure 25: Graph of Analytically Measured TN Levels for MF Wetland……………………….74
Figure 26: Graph of Analytically Measured Ammonia Levels for MF Wetland………………..76
Figure 27: Graph of Analytically Measured Nitrate Levels for MF Wetland…………………...78
Figure 28: C:N Ratio for MF Wetland…………………………………………………….……..79
Figure 29: Graph of Analytically Measured pH for MF Wetland……………………………….80
Figure 30: Graph of Analytically Measured TN Levels for MC Wetland……………………….83
Figure 31: Graph of Analytically Measured Ammonia Levels for MC Wetland………………..86
Figure 32: Graph of Analytically Measured Nitrate Levels for MC Wetland…………………...88
Figure 33: C:N Ratio for MC Wetland…………………………………………………….…….91
Figure 34: Graph of Analytically Measured pH for MC Wetland……………………………….92
Figure 35: Nitrogen Speciation Comparison across Cells……………………………………….95
Figure 36: Percent TN Removal in Cell 1 with Increasing C:N Ratio…………………………..96
Figure 37: Field Site Alkalinity Graph…………………………………………………………107
Figure 38: Field Site Ammonia Graph…………………………………………………………109
Figure 39: Field Site COD Graph………………………………………………………………112
Figure 40: Field Site Nitrate Graph………………………………………………….…………114
Figure 41: Field Site pH Graph…………………………………………………………………116
Figure 42: Field Site Phosphorus Graph…………………………………..……………………118
Figure 43: Field Site TS Graph…………………………………………………………………120
Figure 44: Field Site VS Graph……………………………………………………………...…121
Figure 45: Monitoring Sheet (Front) …………………………………………………..………123
x

Figure 46: Monitoring Sheet (Back) ……………………………………………………...……124
Figure 47: Analytical Recording Sheet…………………………………………………………130
Figure 48: Location of Darcy’s Law Measurements…………………………………………...132
Figure 49: Picture of Field Site Media …………………………………………………………135
Figure 50: Picture of Sampling Hole…………………………………………...………………136

xi

INTRODUCTION
The wastewater produced by Michigan dairy milking facilities is harmful to the
environment and costly to treat. This industry produces an estimated 409 million to 1.28 billion
gallons of wastewater a year. Milking facility wash water is produced from rinsing milking
equipment for cleaning before and after the milking process. The waste is characterized by high
nutrient loads as well as total suspended solids (TSS), biological oxygen demand (BOD), and
cleaning chemicals. Options for handling this wastewater are limited in Michigan. Most
commonly, it is disposed of in liquid manure storage, where it can consume 25-50% of available
space. In addition, the costs associated with this wastewater are estimated at 2-5% of total
operations. (Safferman 2008)
More options for this waste are needed to give dairy farmers flexibility in treatment. An
effective treatment system would reduce carbon and nutrient loads in ground application and
minimize cost.

1

HYPOTHESIS
Cost-share programs for dairy farmers exist through the National Resource Conservation
Service (NRCS), but there are few options for milking facility wastewater. In order to cost-share
wetlands for treatment, NRCS standards with design parameters need to be developed. Research
is needed to determine appropriate hydraulic and organic loading and, therefore, sizing.
The hypothesis for this research follows: cost-effective wetland design is feasible and
will adequately treat milking facility wastewater to meet or exceed environmental standards.
Cost-effective means that life cycle costs will be as low if not lower than other options while still
meeting performance standards. The system will be considered to have adequately treated
milking facility wash water if effluent characteristics are comparable to the allowable levels for
septic systems in Michigan. Efficiency of the design will be determined by comparing the lab
treatment to field treatment.

2

OBJECTIVES
The primary objectives of this study are as follows:
1. Collect the necessary data to develop an interim design standard, which is a
mechanism that allows the NRCS to field test new technology (NRCS 2009).
2. Create a treatment design that is more cost efficient than the demonstration field site.
3. Design for operational control flexibility for the technology to adapt for field specific
conditions or further optimization.
4. Determine if the characteristics of wash water (manure-free/manure-contaminated)
are significant enough to be considered separately for design purposes.

3

LITERATURE REVIEW
A review of previously published papers and other literature concerning key parts of this
project was completed in order to inform the experiment and data interpretation.
Firstly, constructed wetlands were examined, beginning generally and then differentiating
by type, with particular attention to vertical flow constructed wetlands. The various components
in a constructed wetland were explored, after which the effect of temperature on wetland
performance was discussed.
Next, general characteristics of the project’s wash water, milking facility wastewater,
were presented. In addition, examples of milking facility wastewater treatment with constructed
wetlands were shown, including a very similar design from a local Michigan dairy farm. The
mechanisms for removal of each nutrient of interest were also documented in order to understand
how treatment may be occurring within the experimental wetland. Finally, parameters related to
design are discussed, including loading, sizing, and modelling.

4

CONSTRUCTED WETLANDS
Constructed wetlands are manufactured wastewater treatment systems that are based off
of the natural ecosystem of a wetland (Zhang 2011; Fountoulakis 2009; Langergraber 2008).
Although constructed wetlands have been used for wastewater treatment since the 1950s
(Stefanakis 2012), the complicated processes involved are not well understood (Kouki 2009).
Despite this, there is great interest in constructed wetlands due to a variety of benefits including
low cost, low maintenance, high efficiency, aesthetic value, wildlife habitat, usable plant
biomass, and relative efficiency in removing pollutants (Zhang 2011; Ying 2010; Chang 2012).
Constructed wetlands treat many types of wastewater for a variety of pollutants, but most
often from domestic, agricultural, and mining sources. Treatment of leachate, contaminated
groundwater, and industrial effluents with constructed wetlands is becoming increasingly
common (Stottmeister 2003). Organic matter, suspended solids, excess nutrients, metals, and
fecal coliforms are pollutants that wetlands are credited with removing adequately (Kouki 2009).
There are several types of constructed treatment wetlands, which are defined by
characteristics of flow. The first differentiation is the flow of water through the system: surface
or subsurface flow. Subsurface flow is further divided by the direction of water flowing through
the system: horizontal or vertical. The broad components of the wetland system consist of the
soil (or substrate), water, plants, and microorganisms; however it is believed that the microbial
activity is the most important factor in treatment efficiency (Ying 2010).

5

Types of Constructed Wetlands
Wetland nomenclature differs depending on specialty. Scientists define natural wetlands
by characteristic plant communities, but engineering definitions of wetlands focus primarily on
the way water moves through a system. As such, it is commonly said that the broad categories of
constructed wetland are either surface flow or subsurface flow.
Surface flow wetlands (also known as free water surface wetlands) are often
characterized by dense vegetation and horizontally directed flow, as opposed to vertically
(Langergraber 2008). Although subsurface flow is generally more popular in Europe, free water
surface wetlands are becoming increasingly popular for the associated lower cost and increased
wildlife benefits (Fountoulakis 2009). However there is a greater loss of treatment efficiency in
winter months.
Subsurface flow wetlands require wastewater to flow underneath a layer of substrate or
soil, which minimizes energy loss (Ying 2010). This top layer insulates, prevents freezing and
evaporation, and maintains higher temperatures, which delays plant hibernation (Ying 2010). A
concern with subsurface flow wetlands is the potential for clogging. There are two types of
subsurface flow wetlands: horizontal and vertical. Horizontal flow wetlands rely on water to flow
horizontally, while vertical wetlands allow water to move vertically through the soil column.
In horizontal flow constructed wetlands (HFCW), water is fed at the surface of a wetland
and trickles diagonally down through media to the other side (Figure 1). HFCW are often used
as secondary treatment and are efficient at removing organics, suspended solids, microbial
pollution, and heavy metals. Conditions in the treatment bed allow for denitrification, but limited
nitrification, due to a lack of oxygen. Removal of phosphorus is low unless a special material
with high sorptive capacity is used. The lifetime of a HFCW is limited by the selection of the

6

substrate media, which is prone to clogging. The area of a HFCW is usually around 5 m2/PE
(where PE is population equivalent). The main advantage of HFCW is the low associated
energy (no electricity needed once the water enters the wetland), but if nitrogen or phosphorus is
the target pollutant, HFCW may not be the right choice. Aeration may improve nitrogen
pollutant removal, but is too costly to justify in most cases. (Vymazal 2010)

Figure 1: Cross-section of a HFCW

Vertical flow constructed wetlands (VFCW) have many applications for treatment,
including on-site domestic wastewaters or sewage from small communities, industrial
wastewaters, or stormwater runoff (Vymazal 2010). Usually VFCWs distribute water evenly
over the top of the system, allow it to trickle through, and collect water at the bottom (Figure 2),
but there are also upward flow VFCWs that fill from the bottom (Vymazal 2010).
VFCWs gained popularity in the 1990s due to their increased capacity for oxidizing
ammonium (Fountoulakis 2009). Vertical wetlands also are preferred for their smaller size (1-3

7

m2/PE) and greater oxygen transfer capacity (Cui 2012; Vymazal 2010). A settling tank is often
utilized as a pretreatment for VFCWs to remove suspended solids (Langergraber 2008). A major
difference between HFCWs and VFCWs is that VFCW are fed in batches; the system is allowed
to flood and drain intermittently (Vymazal 2010). By doing so, oxygen availability is
periodically higher, which means that treatment is improved and clogging is less likely
(Stefanakis 2012). However, clogging is one of the major problems associated with VFCWs and
depends on substrate selection, even wastewater distribution, and hydraulic loading rate
(Vyzamal 2010).

Figure 2: Cross-section of a VFCW

Constructed Wetland Components
A constructed wetland consists of wetland plants and substrate, and is fed with
wastewater. These components foster the growth of a microbial community which is responsible
for most of the treatment. Wetland function is also affected by seasonal change and start-up time.

8

This review will dedicate a subsection to each physical component to further elaborate their
impact on constructed wetland treatment.
Wetland plants are an important physical component of treatment wetlands. Plants help
influence microbial communities, hold together the substrate, and may even uptake some
pollutants. Studies on plants in constructed wetlands have focused on the effect of their presence
and plant selection between species.
In most studies, the presence of plants was found to have a positive effect on pollutant
removal (Brisson 2009). The benefits are attributed to providing oxygen, carbon, and a surface
for microbial growth, as well as slowing the velocity of wastewater through the system,
insulation in winter, and stabilizing the substrate (Brisson 2009). There is some nutrient uptake
by plants, but it is not a major mechanism for removal (Stefanakis 2012). Studies have found that
as a carbon supplier, plants can have a big influence on microbial communities which synthesize
enzymes in the sediments (Choi 2009).
In terms of quantifying the effect of pollutant removal by plants, Stefanakis (2009) found
statistically relevant differences in pollutant removal in vegetated and non-vegetated wetlands.
The VFCWs were found to improve removal of organic matter by 6%, nitrogen by 10%, and
phosphorus by 11%. Ammonia removal is primarily through nitrification. In a study by
Stefanakis (2012), there was a lack of statistical support for phosphorus removal, indicating the
main mode of phosphorus removal is through sorption by the substrate, which in this case was
limited due to short contact time. (Stefanakis 2012)
Plant selection is usually based on growth rate, large biomass with a significant root
structure, and tolerance to wastewater treatment conditions (Brisson 2009). A study comparing
cattails and phragmites did not yield a statistical difference, but most studies found a difference

9

in removal efficiency (Stefanakis 2012). Occasionally, these differences were statistically
significant, contributing to the belief that plant selection does matter (Brisson 2009). Differences
in treatment by plant selection was primarily found for nitrogen removal: evidence for TSS,
organic matter, and phosphorus is less convincing (Brisson 2009).
The substrate provides structure for the plants and surface for the microbial
communities. It is usually made up of sand and gravel, but synthetic materials have also been
experimented with (Kouki 2009; Zhang 2012). The depth of substrate and its aeration capacity
are factors in nutrient removal (Stefanakis 2009). Substrate can clog and cause treatment
problems with clogging.
There is conflicting data on the use of artificial media to supplement treatment for
constructed wetlands. A study on horizontal subsurface flow showed a great need for a deviation
from the standard gravel due to excess clogging (Kouki 2009). Two alternatives explored
included light-expanded clay aggregates (LECA) and Filtralite. LECA was chosen due to higher
porosity and specific surface area in order to lead to better biofilm adhesion and hydraulic
conductivity. Filtralite proved to be beneficial for improved treatment in COD, ammonia, and
TSS, compared to the standard gravel. (Kouki 2009)
Stefanakis (2009) examined a variety of media compositions for VFCWs and found that a
sand (or other similar material) layer is integral to wetland performance because it decelerates
water flow, leading to longer contact time (i.e. treatment) with the substrates, wastewater, and
plant roots. This is important for the removal of ammonia as well as organic material. The study
also looked at special materials (zeolite and bauxite), but did not find any statistical
improvements in treatment to justify their use, though it did recommend special materials for use
in surface flow treatment systems. (Stefanakis 2009)

10

Clogging
Clogging is a big concern with substrate media in subsurface flow wetlands. Clogging
occurs when organic and inorganic solids form a mat that blocks soil pores. Wastewater from
milking facilities has been found to clog leach fields with solids from milk and manure (Payer
and Holmes, University of Wisconsin-Extension). A clogged wetland results in a number of
problems for treatment, including short circuiting and odor issues (Zhang 2011). Preferential
flow patterns resulting from clogging represent a failure of wetland operation. In order to prevent
clogging, mechanical pretreatment is often used in subsurface flow wetlands to settle out some of
the suspended solids (Langergraber 2008). A drying time is also necessary to avoid clogging so
that organic material can be oxidized (Stefanakis 2009). Many constructed wetlands also
implemented a porous substrate, such as gravel, in order to minimize clogging (Mantovi et al.
2003; McKague 2012).
Some studies (Beach et al 2005; Beal et al 2005 and 2006) have related the growth of a
biofilm mat with hydraulic conductivity. Since a major concern with constructed wetlands is
their lifetime due to clogging, hydraulic conductivity may give insight to the progression of
clogging in the cell.

Effect of Cold Weather on Constructed Wetlands
Temperature does have an effect on pollutant removal. In order to design a wetland for
the “worst-case” scenario, the overall treatment capability in summer and winter conditions are
compared.

11

Organic phosphorus removal was found to increase statistically significant with an
increase in temperature (Stefanakis 2009; Stefanakis in 2012). The same study found a similar
trend for total phosphorus, though only one configuration was found to be significant.
In a study by Stefanakis (2012), it was found that organic material (OM) removal was
positively affected with increased temperatures. However, while BOD was statistically
significant for all configurations of VFCWs tested, the same was not true for COD. COD
removal was not statistically significant in an unplanted unit. Increasing temperature had a
significant positive effect on OM removal, despite the negative effect temperature has on
dissolved oxygen (DO), which indicates that VFCW have adequate oxygen throughout year.
(Stefanakis 2012)
Nitrogen pollutant removal is positively affected with increased temperatures, though less
of an increase in treatment compared to COD (Stefanakis 2012). Vymazal (2007) found that the
microbial activity for organic matter decomposition, nitrification, and denitrification is favored
at higher temperatures.
Because the most significant difference for temperature variation was for organic matter
removal, which increases with temperature (Stefanakis 2009), and microbial reactions, plant
nutrient uptake, and phosphorus sorption are also favored at higher temperatures (Stefanakis
2009; Ying 2010), the limiting rate for a treatment wetland in Michigan is in winter. Although
dissolved oxygen is higher during this time, plants are dormant and microbial activity (the main
treatment mechanism) is slower.

12

Milking Facility Wash Water
This subsection begins by introducing milking facility wash water, in general, and
concludes with several case studies on treating milking facility wastewater with constructed
wetlands.
Wastewater produced by Michigan milking facilities is harmful to the environment and
costly to treat. The Michigan dairy industry produces an estimated 409 million to 1.28 billion
gallons of wastewater a year. Milking facility wash water is produced from cleaning milking
equipment before and after the milking process. The waste is characterized by high nutrient loads
as well as TSS, BOD, and cleaning chemicals. Options for handling wash water are limited in
Michigan. Most commonly, it is disposed of in liquid manure storage, where it can consume 2550% of available space. In addition, the costs associated with wash water are estimated at 2-5%
of total operations. (Safferman 2008).
Daily wastewater production in Ontario, Canada was found to range from 6 to 28
L/cow/day with an average of 14 L/cow/day (Hawkins 2011). Researchers in Italy determined
an estimate of 25-40 L/cow/day for an efficiently designed dairy including wastewater from the
holding area, milking system, milk room, milking parlor, and milk tank (Mantovi et al. 2003).
Additional wastewaters may be included from cooling milk, softening water, and washing hands
or boots (Payer and Holmes, University of Wisconsin-Extension). It should be noted that the
amount of wastewater produced does not increase linearly with the amount of cows (USDA,
1992). Small to medium dairy farms are typically expected to produce 200-400 gallons/day of
milking facility wastewater, and large farms may produce 1000 or more gallons/day (Payer and
Holmes, University of Wisconsin-Extension).

13

After milking cows, the associated equipment is thoroughly cleaned with a procedure that
includes a rinse, acid rinse, and sanitization (Hawkins 2011). The chemicals used in this process
include high concentrations of chlorinated alkaline solutions, acidified waters, and detergents
(Hawkins 2011). Wash water includes phosphorus from phosphate detergent and phosphoric
acid, which is used to remove oils and grease and sanitize (Hawkins 2011). Phosphate-free
detergent is not recommended because it usually utilizes nitrates instead, which are particularly
prone to contaminate groundwater (Hawkins 2011). Bacteria and oxygen consumed during
decomposition make milk itself a contaminant as well (Hawkins 2011). Additional substances
that may contaminate milking facility wash water include manure, urine, dirt, feed, and bedding
(Payer and Holmes, University of Wisconsin-Extension).
The components of the wash water are considered pollutants in several ways. Nutrients,
particularly phosphorus, are hazardous to the aquatic environment through the creation of
eutrophic waters. Ammonia is poisonous to fish, and nitrates can contaminate groundwater and
cause methemoglobinemia, also known as “blue baby syndrome,” which is fatal in human
infants. (Payer and Holmes, University of Wisconsin-Extension)
A variety of treatment techniques for milking facility wastewater have been tested. In
Ontario, Canada, the options available to dairy farmers include sediment tank and treatment
trench systems, vegetated filter strips, and engineered wetlands (Hawkins 2011). Unfortunately,
in Michigan, constructed wetlands are not yet an approved option, which this study hopes to
remedy. In Michigan, aerobic treatment units (ATU) and filter mounds were shown to
successfully treat milking facility wash water (Larson and Safferman 2009, Rathbun et al. 2012).
The NRCS in Wisconsin recommends the use of ridge and furrow, wetland, subsurface
absorption, and vegetative buffer treatment systems for milking facility waste (Wisconsin

14

NRCS). Note that a common septic field by itself is not recommended for this waste due to the
concern with solids clogging the septic field prematurely. Three examples of milking facility
wastewater treatment with wetlands are discussed below.

Example 1: Italy HSSF for Milking Parlor Wash Water and Domestic Sewage
(Mantovi et al. 2003)
A two-cell, horizontal subsurface flow constructed treatment wetland was used to treat
milking facility wash water and some domestic sewage from a staff bathroom and nearby
farmhouses. An average of 80 cows were at the facility during the experiment duration. The
majority of the milking parlor wastewater came from the holding area containing manure, with
an estimated input of 20-40 l/day.
Loading values for the site included a hydraulic loading of 6.3 m3/day, 4.4 m3/day of
which was from the milking parlor. The study measured influent streams of wash water
separately and differentiated the wash water caused by milking from that produced by washing.
Table 1 below shows the loading values for milking and washing waste.

Loading
Hydraulic

Milking Waste
1.6 m3/day

Washing Waste
2.8 m3/day

Organic

612 mg/l O2

618 mg/l O2
1.965 mg/l

COD

1.165 mg/l O2
TKN
83.78 mg/l
76.20 mg/l
Ammonical N 34.48 mg/l
22.05 mg/l
Nitrates
>0.5 mg/l
>0.5 mg/l
TP
11.78 mg/l
19.95 mg/l
TSS
0.48 g/l
1.22 g/l
Table 1: Loading values for milking facility wastewater

15

The milking parlor wash water traveled through a two-celled HSSF wetland. Waste was
pretreated with a septic system and filter to remove TSS. Recirculation was included in the
design for the last cell. The first cell was filled with medium gravel, and the second cell had fine
gravel. The cells were planted with Phragmites australis, which is common for constructed
wetlands in Europe, but considered an invasive species in Michigan.
Removal rates for COD, Total Kjedhal Nitrogen (TKN), TP and TSS were 91.9%,
48.5%, 60.6% and 90.8%, respectively. The first cell was responsible for most removal, and the
second cell served mostly as a polishing step. Nitrogen removal was lacking, and it was thought
that the partially anoxic conditions did not allow for efficient nitrification.

Example 2: Canada VFCW for Milking Parlor Wash Water
(McKague 2012)
The application of a three celled vertical flow constructed treatment wetland was tested at
three milking facilities in Ontario, Canada, consisting of two sheep farms and one cow farm.
Lloyd Rozema of Aqua Treatment Technologies was responsible for the design.
The three milking facility sites were named St. John, Kenilworth, and Elmira. St. John
treated wash water from a cow herd of 20-25 and did not include manure contamination. The
inflow was 254 L/day, including a rainwater input. Elmira and Kenilworth treated wash water
from sheep herds (170-180 and 340 head herds, respectively) and both had manure
contamination. The Elmira and Kenilworth wetlands had average inflows of 886 and 2895 L/day,
respectively. Therefore, the average hydraulic loading for all three sites ranged from 3.5 to 40
L/m2/day. A summary of the nutrient loadings from each site is in Table 2.

16

2

Loading (mg/day/ft )
Site
Phosphorus COD Ammonia +Ammonium (N) Nitrate (N)
St. Jacobs
36.13
250.57
5.02
0.06
Elmira
23.90
782.12
39.71
0.07
Kenilworth
57.47
723.91
5.65
0.62
Table 2: Nutrient loadings at Canadian milking facilities

The treatment system design included a septic tank for pretreatment, a collect sump to
receive overflow from the septic tank and pump into the wetland, a three-cell VFCW, and an
effluent land application to a treatment trench. The three-cell VFCW was responsible for most
of the treatment. Wash water moved sequentially through the cells, with recycling in the first and
third cells. The first cell was aerobic and responsible for reducing the majority of BOD and
nitrification. Next, the second cell remained saturated to maintain anaerobic conditions for
denitrification. Finally, the third cell was the polishing cell, to remove additional BOD. Figure 3
provides additional details.

Figure 3: Cross-section of three-celled VFCW design

17

Each wetland cell was 4 ft (1.2 m) deep with a surface area of either 8 x 8 ft (6 m2) or 16
x 16 ft (12.5 m2), depending on the characteristics of the wastewater. The cells were lined with
an impermeable membrane.
The media was pea gravel. There was no top soil, but plants grew adequately due to the
high nutrient loadings of wastewater. Pea gravel was chosen for its capacity to create large void
spaces within the soil matrix; this follows the logic that larger pores will not clog, maximizing
the life of the system and allowing substantial oxygen transfer.
Two PVC pipe manifolds were placed in each system to distribute the wastewater: on the
surface, and 1.5 ft (0.5 m) below the surface. The former was the “summer manifold” that
distributes wastewater for the warm term. The latter was the “winter manifold” that distributes
wastewater for the cold term. Switching between manifolds was done manually. In each
manifold, pipes were laid parallel to each other, 2 ft (0.6 m) apart.
On average, the wetland systems at each of the sites yielded approximately 99% removal
of COD, 78% removal of phosphorus, and 80% removal ammonia+ammonium. Nitrate, as
nitrogen, increased by a large factor: 1080% on average after the first cell. Effluent values of
nitrate averaged about 4.2 mg/L.

Example 3: Local Michigan VFCW for Milking Parlor Wash Water
A wetland based on the design of the Ontario wetlands described above was implemented
in Michigan to serve as a demonstration site that proved the design was applicable for Michigan.
This system was a prerequisite to the lab-optimization study described in this report. The field
wetland was constructed in 2009 and has been in use ever since. Intensive scientific monitoring
of the site began in June 2009 and ended in January 2010.
18

Like the Rozema wetlands described in the previous section, the Michigan wetland had
three cells. In this particular case, each cell had 8 x 8 ft (6 m2) surface area and was 4 ft (1.2 m)
deep. Pea gravel was the selected substrate. Wash water was applied to the wetland twice a day
from a small Michigan dairy farm consisting of about 25 cows. The wash water was free from
manure contamination.
After the monitoring period, results of the study indicated that the system successfully
treated Michigan dairy wash water. Table 3 summarizes the results and details are provided in
Appendix A.

19

Parameter

Location
Average of all Samples
308
Sediment Tank
Alkalinity
353
(mg/L of Wetland Cell 1
360
Wetland Cell 2
CaCO3)
340
Wetland Cell 3
6.38
Sediment Tank
3.99
Wetland Cell 1
pH
7.21
Wetland Cell 2
7.35
Wetland Cell 3
1880
Sediment Tank
149
Wetland Cell 1
COD
66
Wetland Cell 2
31
Wetland Cell 3
15
Sediment Tank
Ammonia6
Wetland Cell 1
Nitrogen
4
Wetland Cell 2
(mg/L)
1
Wetland Cell 3
25
Sediment Tank
Nitrate4
Wetland Cell 1
Nitrogen
5
Wetland Cell 2
(mg/L)
4
Wetland Cell 3
30
Sediment Tank
14
Phosphorus Wetland Cell 1
(mg/L)
10
Wetland Cell 2
5
Wetland Cell 3
Table 3: Treatment at Michigan three-celled VFCW

As a result of the successful implementation of the treatment wetland in Michigan, an
optimization study to determine a cost-effective design was conducted and is described in this
report. With the data from the field site, in combination with lab data, the design may be
standardized for use across the state and implemented as a USDA Michigan NRCS interim
practice standard.

20

TREATMENT MECHANISMS FOR SELECT NUTRIENTS
A large variety of pollutants can be removed in a wetland including organic matter, TSS,
nitrogen, phosphorus, fecal coliforms, and metals (Kouki 2009). For this project, the pollutants
of interest include phosphorus, organic material, and nitrogen. The mechanisms for removal for
each pollutant may be biotic or abiotic. Microorganisms or macrophytes can contribute to biotic
removal. Mechanisms of removal for each nutrient are described in the sequential subsections.

Phosphorus
In order to remove phosphorus, it is necessary to harvest plants and/or remove saturated
media (Zhang 2011). Past research has shown that with some specialty artificial material,
phosphorus removal can reach 98-100% removal in constructed wetlands (Zhang 2011). In most
cases for VFCWs, phosphorus removal is quite low (around 20-30%) (Stefanakis 2009).
Phosphorus removal occurs through plant uptake, microbial activity, substrate adsorption,
chemical precipitation, and organic matter incorporation (Stefanakis 2009). Removal by plant
uptake is low, but is most likely the most sustainable process (Stefanakis 2009).
Phosphorus removal mechanisms in wetlands are primarily abiotic. Abiotic removal
includes adsorption and precipitation. Biotic removal primarily relies on macrophytes activity,
including plant adsorption and uptake. Macrophyte involvement is most sustainable, but
mirobiotic (bacteria and microalgal) removal is also possible. (Vymazal 1998)
Wetlands function more as a “sink” for phosphorus rather than removing it from the
system completely. In addition, it is important to note that studies have found that wetlands are
less suited to phosphorus accumulation than terrestrial ecosystems. (Vymazal 1998)

21

Organic Material
Organic pollutant removal is measured by COD. Abiotic removal mechanisms include
deposition and filtration; biotic removal is characterized by macrophyte uptake and a variety of
microbial activity. Deposition and filtration are typically the removal method for settleable
solids. Soluble organics are treated primarily by microbes. (Vymazal 1998).
Microbial removal can be aerobic or anaerobic, and many different populations are
responsible. Aerobic removal is dominated by heterotrophic bacteria because of faster metabolic
rates, but autotrophic bacteria (including nitrifiers) also play a role. Anaerobic removal is a
multi-step process and is completed by facultative or obligate anaerobes. Macrophyte uptake in
constructed wetlands is considered negligible in comparison to microbial degradation (Vymazal
1998).

Nitrogen
Nitrogen removal mechanisms are numerous and can be complicated. Abiotic techniques
include ammonia volatilization and matrix adsorption. Ammonification,
nitrification/denitrification, anammox, and plant uptake are biotic removal processes. The biotic
techniques can be further divided into catabolic processes (nitrification, nitrification,
denitrification, dissimilatory nitrate reduction, and anammox) and anabolic processes
(ammonium uptake, assimilatory nitrate reduction, and nitrate fixation). Ammonification, also a
biological removal process, is just a necessary part of the biological food chain. (Jetten et al
2009)

22

Ammonia Volatilization
Ammonia volatilization is a physiochemical process that occurs as hydroxyl and gaseous
forms of ammonia try to reach equilibrium with each other. Factors that influence rates of
ammonia volatilization include pH, temperature, concentration, wind velocity, solar radiation,
and characteristics of the plant community. (Vymazal 1998)

Matrix Adsorption
Matrix adsorption is another equilibrium-based removal mechanism. Reduced ammonia
can bind to soil matrix, but it is not a long term removal solution; the bound ammonia re-releases
rapidly as equilibrium shifts due to nitrification. (Vymazal 1998)

Ammonification
In ammonification, organic nitrogen is converted to inorganic nitrogen, particularly
ammonia. Ammonification is a biotic mechanism that relies on aerobic communities. As such,
the rates of ammonification are fastest in oxygenated zones and decrease as the community shifts
to facultative and obligate anaerobic conditions. Factors influencing ammonification include pH,
temperature, carbon to nitrogen ratio (C:N ratio), available nutrients, and soil (texture and
structure). The ideal pH is between 6.5 and 8.5, and the rate of ammonification doubles for
temperatures over 10˚C, within a set range. (Vymazal 1998)

Nitrification
Nitrification is a two-step biological process that converts ammonium to nitrate, with the
intermediate production of nitrite. The first step, transforming ammonium to nitrite, is done by

23

strictly chemolithotrophic bacteria, which require oxygen. The conversion of nitrite to nitrate is
completed by facultative chemolithotrophic bacteria, which may degrade organic matter instead
of carbon dioxide and do not require oxygen. (Vymazal 1998)
Nitrifiers are sensitive bacteria and many factors influence the rate of nitrification
including temperature, pH, alkalinity, amount of carbon dioxide, microbial population,
concentration of ammonium, and DO. The ideal ranges for temperature and pH (in soil) are 3040˚C and 7.5-8.6, respectively. During nitrification, a lot of alkalinity (approximately 8.64 mg
HCO3-) is consumed per milligram of ammoniacal nitrogen oxidized. The amount of ammonium
can actually inhibit bacteria at high concentrations. (Vymazal 1998)

Denitrification
Denitrification is the reduction of nitrate to nitrogen molecules or nitrogen gas. Anoxic
conditions are required for denitrification to progress, but most denitrifying bacteria are
chemoheterotrophs. That is, the bacteria consume organic material and do not require oxygen. It
is important to note that while the bacteria do not need oxygen, it is preferred, and if oxygen is
present, the bacteria will produce carbon dioxide and water instead of nitrogen gas. (Vymazal
1998)
Factors influencing denitrification include anoxic conditions, redox potential, soil
moisture, temperature, pH, presence of denitrifiers, soil type, organic matter, and the presence of
overlying water. The ideal redox potential and pH is approximately 300 mV and 7-8,
respectively. Denitrification is known to slow down at temperatures lower than 5˚C. (Vymazal
1998)

24

Anammox
Anaerobic ammonium-oxidizing bacteria, commonly known as anammox bacteria,
convert ammonium and nitrite into nitrogen gas in the absence of oxygen. It is suspected that
50% of dinitrogen gas released from the marine environment is due to anammox bacteria. The
bacteria are characterized by slow growth and strictly anaerobic conditions: anammox bacteria
cells double once every 11-20 days and are inhibited at 2 µM oxygen. The carbon source for
anammox bacteria is bicarbonate, which make the bacteria autotrophs. Research has shown that
anammox bacteria are capable of growth in a variety of temperatures, ranging from -2°C to 43°C
with some studies indicating anammox presence in deep sea hyperthermal vents at 60 and 85°C.
(Jetten et al 2009)

Plant Uptake
In ideal conditions, plants can take up to 10-16% of nitrogen in wastewater, but generally
the amount of nitrogen that can be removed by harvesting plants is insignificant compared to the
amount entering the system. The plants must be harvested to remove nitrogen from the system,
otherwise, nitrogen will return as the plant decomposes. Factors influencing plant uptake include
growth rate, concentration of nutrients in plant tissue, and biomass per unit area. Standing height
positively impacts biomass per unit area. (Vymazal 1998)

25

DESIGN OF CONSTRUCTED TREATMENT WETLANDS
Constructed wetlands have been historically designed by empirical approaches based on
specific surface area (Langergraber 2008). Modelling technology for wetlands has developed in
the past decades and is considered a valid design tool (Kadlec and Wallace 2009). Unfortunately,
design methods for vertical flow wetlands have not developed past empirical techniques (Kadlec
and Wallace 2009). Treatment performance for VFCW is “significantly affected” by operational
conditions (Kadlec and Wallace 2009). Vertical flow wetlands are most commonly designed
using scaling factors (Kadlec and Wallace 2009). Scaling factors are typically based on loading
rates.
Scaling rules for VFCW typically concentrate on BOD removal or nitrification to
calculate the wetland area (Kadlec and Wallace 2009). One of the largest concerns about the
feasibility of wetlands as treatment is the belief that wetlands require a low hydraulic loading
rate and a high hydraulic retention time, which leads to a large area (Fountoulakis 2009).
VFCWs were created in part to address this concern (Chang 2012). Not only do VFCWs require
less surface area, but studies, including one by Stefanakis in 2012, have shown that the systems
are able to treat strong wastewater with high concentrations of organic matter and nitrogen.
Organic matter in this study was sufficiently decomposed with a drying time of 4 days and an
organic loading rate of 125 g COD/m2d, but for similar nitrogen removal a longer drying period
was required to ensure the wetland reached an aerobic state (Stefanakis 2012). It was noted that
the drying stage is integral for the function of VFCWs because it allows for oxidation of
accumulated organic matter in the substrate, thus avoiding clogging (Stefanakis 2012). In another
study in France, it was found that European standards could be met while applying COD
overloads up to10 times the amount of the dry weather flow in VFCWs (Fountoulakis 2009).
26

Other types of constructed wetland have also been studied in terms of hydraulic loading.
The United States Environmental Protection Agency (EPA) suggested BOD loading limits for
free water surface and horizontal surface flow of 11.2 and 6 g BOD/m2d, respectively
(Fountoulakis 2009). Another study found that there was no significant difference in COD
removal in two horizontal surface flow wetlands operated at hydraulic loading rates of 6 and 23 g
COD/m2d (Fountoulakis 2009).
Vertical flow wetlands for cold climate applications in North America are typically
vegetated recirculating gravel filter beds (Kadlec and Wallace 2009). According to Kadlec and
Wallace (2009), the systems are mainly considered treatment wetlands due to the characteristic
vegetation, even though the plants do not significantly contribute to treatment. Essentially, such
wetlands are gravel filters and depend on mass loading, recycle ratio, and media type for
treatment (Kadlec and Wallace 2009). Current design methods typically ignore vegetation and
design the system empirically as recirculating gravel filters (Kadlec and Wallace). The wetland
area is typically derived from a mass loading scaling factor based on BOD (Kadlec and Wallace
2009).

27

METHODS AND MATERIALS
The design for treatment of milking facility wash water tested in this report is based on
the full-size field wetland in Michigan described previously. In fact, wash water from the same
dairy farm was used in the lab wetland. Because the field wetland was considered to be oversized, the lab began with a specific surface area loading that was slightly higher. Due to cost, the
research was limited to two lab-scale wetlands. Therefore, the study was primarily a longitudinal
study.
This section is organized in the following manner. First, the materials for the construction
of the system are described, starting with the basic structure, and moving on to the wetland cells,
media, hydraulic system, and ending with the electrical system. Next, the milking facility
wastewater selected for testing is described. A discussion of the cooling system and the
monitoring procedure and maintenance follow. The sampling and analytical testing procedures
are described, followed by the method for measuring hydraulic conductivity. Finally, the section
concludes with the statistical methods.

Structure
The basic structure for the wetland system was built in the spring of 2013. A wooden
support system was constructed as shown in Figure 4. Because the system requires flow-bygravity into reservoirs at several stages, the additional depth was incorporated. The conservative
height allowed 50 in (127 cm) of head space to add hanging lights if it was decided to include
plants in future studies. The structure allowed for two offset, back-to-back, lab-scale wetlands,
each with three cells. The bench allowed adequate space underneath the wetland cells to place
pumps and reservoirs on the base of the system (Figure 4). In addition, the structure included two

28

rails 8 in (20 cm) above the top of the wetland cells (Figure 4) for use in the hydraulic
conductivity study (See the “Hydraulic Conductivity” subsection).

Figure 4: Lab-scale Wetland Structure

29

Wetland Cells
To simulate the field wetland cells, only the surface area of the cells were scaled down.
Therefore, the wetland cells maintained a four foot depth (1.2 m), with winter application inlets
1.5 ft (0.5 m) below the top of the wetland cell (Figure 4). The “top” of the lab-scale wetland is
equivalent to the ground surface of the field wetland.
The bodies of the cells were made up of plastic, corrugated tubing with a 4 in (10.2 cm)
diameter. Corrugation was used to limit short circuiting along the inside of the wall and aided in
the stability and placement of the cooling system tubing (See “Cooling System”). The tubing was
cut to a length of 4 ft and placed vertically, perpendicular to the bench. Plastic straps at three
locations along the height of the cell secured it to support beams on the structure (Figure 4).
Two openings for fluid flow were included in each wetland cell. At the end of the
corrugated tubing, a quarter in (0.6 cm) hose barb was placed directly in the cap at the bottom of
the cell. This generally served as the effluent for Cells 1 and 3 and the influent for Cell 2. This
barb was accessible due to a 4 in (10.2 cm) diameter hole cut into the wetland bench on which
the wetland cell rested. At a distance 1.5 ft from the top of the wetland cell, the inlets for Cells 1
and 3 and outlet for Cell 2 were installed. Schematics of the inlet and outlet are shown below
(Figures 5 and 6).

30

Figure 5: Inlet Schematic (Cells 1 and 3)

Figure 6: Outlet Schematic (Cell 2)

As seen in the schematic, the inlet and outlet designs are similar, though the outlet is
slightly less complicated because the waste water flows naturally through an open PVC pipe

31

instead of drilled holes. Plumber’s putty was used at the PVC/corrugated pipe interface to
prevent leaks.
The final opening for fluid flow in all six cells was implemented for use in a hydraulic
conductivity study (see “Hydraulic Conductivity” subsection). In terms of design, the hydraulic
conductivity opening was exactly identical to the outlet mentioned in the previous paragraph,
except 3 in higher on the body of the wetland cell (see Figure 7).

Figure 7: Hydraulic Conductivity Outlet

Media
The primary media used to fill each cell was pea gravel, obtained directly from the
corresponding field cells (i.e. gravel from Cell 1 in the field wetland was transplanted to Cell 1 in
the lab-scale wetland). This was done in order to accelerate the system’s acclimation period, in

32

the hope that microbial communities that flourished in each cell condition would still be attached
to the media.
Since the gravel was taken from the field wetland, there was a lot of excess dirt and
debris in the transplanted media. Consequently, the gravel was sieved and rinsed with deionized
water (approximately 10x per large handful) before placement into lab-scale cells. It was
estimated that the majority of gravel ranged in diameter from 0.3 to 0.7 in (0.8 to 1.8 cm) with an
average diameter of approximately 0.4 in (1 cm). The porosity of the pea gravel was determined
in the lab as 0.42.
In order to protect the hose barbs at the bottom of the wetland cells from lodged pea
gravel, a layer of store bought “River Rock” was added to the cells first, approximately 2-3 in
(5.1-7.6 cm) in depth. Pea gravel was then poured on top of the “River Rock” to fill the
remainder of the wetland cell.

System Hydraulics
The hydraulics for the wetland system was fairly complicated and involved the use of
timers, level switches, pumps, and recycle ratios. A basic schematic of flow through the system
is shown in Figure 8.

33

Figure 8: Flow Schematic

The “Influent” shown on the schematic is a five gallon carboy filled with wash water
from a local Michigan dairy farm and is stored in a refrigerator. Tubing stretches from the
bottom of the carboy, through the wall of the refrigerator, and is pumped up to a reservoir
(Figure 9) at the top of the refrigerator using a peristaltic pump.

34

Figure 9: Reservoir 1

35

Wash water flowed from Reservoir 1 into Cell 1 via gravity. Vinyl tubing (1/4” ID) was
used to connect the hose barb at Reservoir 1 to the hose barb at the inlet of Cell 1 (see Figure 9).
After wash water entered Cell 1, it was distributed through holes in PVC pipe (see Figure 5) and
trickled down through the gravel media all the way to the bottom of the cell. Once again, vinyl
tubing (1/4” ID) was used to connect the hose barb at the bottom of Cell 1 to the influent hose
barb of Reservoir 2. A schematic of Reservoir 2 is shown in Figure 10.

Figure 10: Reservoir 2 and 3 (Side View)

The wash water drips from the hose barb into the reservoir and raises the water level. A
float switch rises with the water level and when it closes, it triggers two pumps to turn on (Figure

36

11). The aforementioned pumps are identical (Masterflex L/S Economy Drive) and are
connected to the reservoir through tubing and hose barbs, as shown in Figure 11. The T-barb is
kept level to limit preferential flow. One pump leads all the way back to the influent reservoir as
the recycle line; the other pumps into the bottom of Cell 2. The ratio of the speed of flow
between reservoirs was called the recycle ratio, and was adjustable.

Figure 11: Recycle System (Top View)

The wash water was then pumped into the bottom of Cell 2 and travelled vertically
upwards to exit at an outlet 1.5 ft (0.5 m) below the top of the cell (see Figure 6). In doing so, the
cell was always saturated, allowing for anoxic conditions. Once the wash water exited from Cell
2, it flowed by gravity into the third and final reservoir (Figure 10). This reservoir had a
float/level switch as well. Water triggered the level switch to turn on a final pump. The wash
37

water was then pumped to the influent of Cell 3, also 1.5 ft (0.5 m) below the surface into an
influent opening (Figure 5). Water flowed down through Cell 3 via gravity and exited into the
effluent container. Reservoir 3 and Cell 3 were structurally identical to Reservoir 2 and Cell 1,
respectively.
The hydraulic retention time, approximately 6 days, was controlled by the saturated cell
(Cell 2). Cells 1 and 3 have a short retention time; just the time it takes water to trickle down 2.5
ft (0.76 m) through gravel.
Although vinyl tubing was used for the majority of transport between wetland cells, it
was not acceptable for use in the peristaltic pumps. Except for the influent pump, which
transported wastewater directly from the fresh influent carboy into Reservoir 1, connector hose
barbs were used to attach specialized tubing for use by the pumps (as is shown in Figure 11). A
list of pumps and the corresponding tubing is summarized in Table 4.

Pump Use
Influent to
Reservoir 1
Reservoir 2 to Cell
2
Reservoir 3 to Cell
3
Table 4: Pump Detail

Manufacturer

Model
L/S Economy Variable
Speed

1/8” Norprene

Masterflex

L/S Compact Drive

1/4” C-flex

Control Company

Variable Flow Mini Pump

1/4” Silicone

Masterflex

Tubing

Wash Water
Wash water was collected from two Michigan dairy farms. The first had a physical
separation of the manure from the milking wastewater, which affected nutrient and solids
contents. The wash water from the second dairy farm was contaminated with manure, but was
pretreated through primary treatment to remove solids.
38

Wash water was collected once a week during early phases of the project and every other
week in later phases. A hand pump (Beckson High Volume Utility Pump) was used to convey
water from the outdoor holding areas into 5 gallon carboys, which were transferred via truck to
the research facility. Once at the research site, the wash water was stored in a refrigerator at
approximately 40°F to preserve quality throughout the testing period.
According to interviews with the farmers, wash water is discharged twice a day in
approximate hour-long intervals. Based on that information, the wash water was applied to the
wetland twice daily through the use of timers (Traceable Lab Controller). Wash water was
pumped from the carboy into the inlet reservoir every day at 9 am and 9 pm until 10 am and 10
pm, respectively.

Simulating Cold Temperature
To simulate winter conditions, the cells were cooled to winter temperatures. Three
separate measurement events at the field wetland site during two Michigan winters determined
that the temperature profile through the media was fairly consistent at approximately 40°F.
Therefore, the goal of winter simulation was maintaining a constant 40°F in all six wetland cells.
Because the effective root zone in wetlands is 20-30 cm (8-13 in) in depth (Kadlec and Wallace
2009), and the winter influent is 50 cm (1.5 ft) below the soil surface, plants were not included in
this study.
A clear flexible tubing (1 in OD) was tightly wound around the body of each wetland
cell. Duct tape was also wrapped around each cell in order to adequately secure the tubing,
otherwise the tubing became loose and lost contact with the cell body. Coolant was pumped

39

through the tightly wrapped tubing was meant to cool the wetland and maintain coolness
throughout the experiment.
Each column’s coolant tubing was attached, on either end, to an influent or effluent
manifold (Figure 12), which was connected to a chiller (Thermo Neslab ULT-80). A mix of
refrigerant and water were pumped through the chiller for distribution at the bottom of the cells
(Figure 12). At the top of the wetland cell, the tubing conveyed coolant to the effluent manifold
in order to return to the chiller. The coolant was recycled repeatedly throughout the experiment,
although occasionally water was added in order to maintain its volume.

Figure 12: Cooling System (Side View)

40

Two layers of insulation (Reflectix Double Reflective Double Bubble Insulation) covered
each wetland cell in order to limit heat transfer from the surrounding air. The insulation was
taped shut with insulation tape (Reflectix Foil Tape) after wrapping. In addition, tight horizontal
bands of duct tape around the insulation increased the quality of fit around the wetland cell.
In order to confirm the efficacy of the cooling system, temperature probes (Digi-Sense
Type K) were installed in each of the wetland cells and in the water bath of the chiller. Installing
the probes required that a small hole was drilled in each wetland cell approximately 15 in (38.1
cm) high. Each probe was inserted horizontally into the wetland cells to a depth of 2 in (5 cm)
so that the sensor would be approximately center in the wetland cell cross section. Plumber’s
putty created a water proof seal around the hole. The remainder of the probe, including a length
of wire with the connecting plug to the electronic temperature reader (Traceable Pocket-Size K
Thermometer °F), was left accessible for daily readings.

Monitoring Procedures
Each wetland underwent a daily series of checks (monitoring sheet in Appendix B),
included checking for clogs and leaks and recording the temperature readings. The temperature
readings were used to confirm and monitor the accuracy of the cooling system for simulating
winter conditions. The results of the temperature monitoring are included in the appendix
(Appendix C).
Weekly monitoring tasks were more involved. Two parameters were investigated: flow
rate and recycle ratio. The flow rate monitoring procedure was as follows.
1. Remove tubing from influent reservoirs and place over catch basin
2. Turn on pumps

41

3. Once a steady flow has been established, insert tubing into graduated cylinders and start
timer
4. At the end of a specified length of time, remove tubing and place over catch basin
5. Record the number of ml in the graduated cylinder on the monitoring sheet
6. Empty graduated cylinders in catch bucket
7. If necessary, make adjustments to pump speed and begin another timed measurement;
iterate until the amount of fluid in the cylinders is within at least 1 ml of the target 25 ml
8. Record final milliliter amount on monitoring sheet
9. Turn off pumps
10. Replace tubing in influent reservoirs, dispose of waste in catch basin

Adjusting the recycle ratio (referred to “flow ratio” on the monitoring sheet) was a
slightly more complicated procedure and is outlined below.
1. Record the starting cumulative flow from the flow meters on the monitoring sheet
2. Close hose clamp on the tubing under Cell 2 to prevent flow out of Cell 2
3. Undo the quick-disconnect beneath Cell 2, place catch bucket beneath open tubing
4. Pull the recycle line from the influent reservoir and place a catch bucket beneath it (eye
hooks were used to secure the tubing line to a convenient location for testing)
5. Remove cap/influent line from Reservoir 2
6. Pour approximately 500 mL of deionized water into Reservoir 2; pumps should be
triggered
7. After a steady flow has been established, place 1000 ml graduated cylinders under each
streaming flow at the same time

42

8. Wait for pumps to turn off and water to stop flowing
9. Record the amount of milliliters in each cylinder on monitoring sheet; compare ratio to
goal ratio
10. If necessary, make adjustments to pump speed: iterate until the amount of fluid in the
cylinders is within at least 5 ml away in either cylinder from the appropriate ratio
11. Record final volume on monitoring sheet
12. Replace tubing in influent reservoir, re-connect quick-disconnect, undo tube clamp, and
dispose of collected waste in the catch buckets
13. Repeat process for other wetland

Maintenance
Common maintenance for the system included unclogging influent reservoir/tubing,
cleaning out influent reservoirs, replacing worn out pump tubing, and repairing leaks. Quick
disconnects in the line between the effluent of Reservoir 1 (Influent Reservoir) and the influent
of Cell 1 were installed in order to give access for clog removal. Once the quick-disconnect was
opened, a long, malleable wire was inserted into the tubing to dislodge material and remove
clogs. Occasionally clogs were accessed from the inside of Reservoir 1 because clogging at the
hose barb was particularly vulnerable. Twice over the course of the experiment, standing water
in the influent reservoirs was strained with funnel constructed out of a paper towel to remove
thick, gelatinous biogrowth that caused clogging. It is believed these clogs were an artifact of the
small diameter tubing associated with the lab-scale wetlands; no clogging was observed in the
full-scale after over 5 years of operation (see Appendix F for more details).

43

Because peristaltic pumps were used, the associated tubing was constantly subject to
abrasion. Worn out tubing was replaced when leakage was detected.
Occasional leaking from the body of the wetland cells or reservoirs was a more involved
process. Vulnerable areas for leaking included connections for the hose barbs in the reservoirs,
the bottom of cells, the temperature probe insertions, and the inlets and outlets. In order to repair
these leaks, first the area was dried completely and roughened with coarse sand paper. Plumber’s
putty (William Harvey Plumber’s Epoxy Putty) was then applied to the area with particular care
to fill cracks and maintain a good seal on the surrounding area. It is important to use a putty that
dries via chemical reaction instead of air drying because at some junctures the putty may be laid
on thickly. After the putty was dry, a thin layer of caulk was spread over the putty and
surrounding area to be sure that the seal and small cracks were completely water tight.

Sampling Procedure
A total of 8 samples were collected for testing, four from each wetland. First, the influent
from each farm was collected. The carboys in the refrigerator were shaken to create a more
homogenous blend before pouring the wash water into plastic sample jars. Since the wetlands ran
periodically through use of a timer, the procedure to collect the other samples for testing had to
be aligned with the systems’ morning run. Figure 13 shows the sampling locations for the other 3
samples. A more detailed schematic of the sampling set up is shown in Figure 14. Reservoirs
were strapped to support beams with bungee cords to increase stability, and during sampling,
sample jars utilized the straps as well. Before the system started to run at 9 AM, the sample jars
were placed as shown in the schematic. As the system began to flow, effluent went directly
through Cell 1 and into the sampling bottle instead of into the reservoir. As a result, there was no

44

direct recycling input in the sample, however, testing was done specifically to look at changing
COD levels during the system’s run.

Figure 13: Sampling Locations

45

Figure 14: Sampling Set-Up

Smaller sample bottles were used for the effluent of Cells 1 and 2 because of the limited
amount of fluid flowing through the system. It was important to be present as the system ran in
order to transfer the cell’s effluent line from the sample bottle (once it was filled) back to the
reservoir so fluid flow could continue to the next sample bottle. Larger sample bottles were
utilized for the influent and Cell 3 effluent in order to perform tests that required a larger volume
of sample, including alkalinity and TSS.

46

Analytical Testing
Weekly tests were: COD, TP, TN, ammonium, nitrate, and pH. Alkalinity and TSS were
tested every other week. HACH standard procedures were used for measurement. Table 5
provides the corresponding HACH or USEPA method numbers of each test for reference. A
replicate, blank, and standard were included for each testing session (27% of samples per
session) for quality assurance/quality control.
The replicate procedure involved re-dilution as necessary from the grab sample. Samples
were chosen randomly for replication. Pipets were re-calibrated for each testing session, and
samples were always mixed before pipetting. Appendix D shows the recording sheet for
analytical testing.

Test
Alkalinity
Nitrate
Nitrogen, Ammonia
Nitrogen, Total
Oxygen Demand, Chemical
Phosphorus, Total
Solids, Total Filterable
Table 5: Method Numbers

HACH
Method 8203
Method 10206
Method 10205
Method 10208
Method 8000
Method 8190
Method 8163

USEPA
Standard Method 2320 B
40 CFR 141
EPA 350.1, EPA 351.1, EPA 351.2
N/A
Standard Method 5220 D
Standard Methods 4500 P-E
Standard Method 2540 C

To assure representative data, only data points under healthy or normal operating
conditions were collected and considered for analysis, where “healthy” implies a leak-free
column and “normal” discounts adjustment periods between phase changes. While a system was
shut down to perform maintenance on leaks, it was observed that the column’s microbial
community needed to adjust because of the disruption. While conservative sample collecting
was generally implemented to avoid collecting unrepresentative data, in some incidences,

47

researcher error miscalculated recovery time. For those incidences, data was thrown out.
Generally the system required one to two and a half weeks to recover from disruptions.

Hydraulic Conductivity
Studies by Beach et al. 2005, Beal et al. 2005, and Neal et al. 2006 have related the
growth of a biofilm mat with hydraulic conductivity. Since a major concern with constructed
wetlands is their lifetime due to clogging, a hydraulic conductivity test was done to give insight
to the progression of clogging in the cell. In addition, the field wetland media was examined to
determine if clogging issues are occurring.
In order to perform the hydraulic conductivity study, a constant head test was used on the
first cell, the rationale being that the first cell would be the first to clog if it were an issue. Figure
15 is a schematic of the test set up. The top outlet within the cell, a few inches (approximately
7.6 cm) above the influent opening, was used for this study.
Once at the beginning of the study, and once at the end, water was poured into a bucket
placed at a higher elevation than the cell outlet (see Figure 15). The bucket had a hose barb that
allowed for tubing to be connected to the bottom of the wetland cell via a quick disconnect. The
water level was kept at a constant head in the bucket above the column by manually pouring
water in at a steady flow. The effluent (from the hydraulic conductivity outlet) was measured in a
graduated cylinder for a given time. Using the flow rate, cross-sectional area of the cell, length
of flow path, and change in head between the bucket and the outlet, hydraulic conductivity was
found using Darcy’s Law. See Appendix E for more detail on calculation.

48

Figure 15: Hydraulic Conductivity Set-Up

Since the lab-scale wetland was based off of the field wetland in mid-Michigan, the
progression of clogging at the site was also examined. Because the media in the field wetland is
pea gravel, a traditional percolation test with a double-ring infiltrometer was impractical. Instead,
researchers dug into the media in order to look for visual evidence of clogging. No evidence was

49

found, but it should be remembered that the field site is very over-sized. See Appendix E for
more information regarding the examination of the field site for clogging.

Statistical Methods
Linear mixed effect modelling with the statistical software R developed by R Core Team
(2013) was used to determine statistical significance of cell treatment and phase change. The
specific package within R that was used for linear mixed effect modelling was developed by
Bates et al. (2014) and is called lme4. Data was not used in the statistical analyses while the
wetland was not determined to be in optimal health or in transition from one phase to another.
For example, some repairs required draining the saturated cells, and the wetland required time to
recover before testing began again. Typically, one to two weeks were needed for adjustment
from repairs or transitioning from phase to phase.
Linear mixed effect models, or mixed effect models, take into consideration both fixed
and random effects. Linear mixed effect modelling was chosen due to the nested data structure.
(Seltman 2013)
Although “random” and “fixed” effects are the common terminology for multi-level
modelling, there is not a common definition for these terms. Gelman (2007) described “random”
and “fixed” effects as “often misleading and confusing” terms, and outlined five distinct
definitions for those terms from different sources that were not only pointing to different things
as “random effects”, but even contradictory. Perhaps the mathematical definitions speak truest to
the model, which tend to be the standard definition in the literature. However, the actual or
practical definitions are less clear. One definition outlined by Gelman, also used by Seltman
(2013), is that fixed effects are constant across individuals, but random effects vary. Another

50

definition identified random effects as the realized effects of random variables, while yet another
described fixed effects as independently interesting but random effects as interesting in the
context of an underlying population (Gelman 2007).
The mathematical definition used by Gelman describes fixed effects as found through
least squares regression but random effects with shrinkage. This definition is “mathematically
precise”, but leaves some doubt which observed effects should be called fixed and which should
be called random (Gelman 2007). Depending on the definition, the same effect could qualify as
random or fixed, even though random and fixed effects are seen as contrasting (Gelman 2007).
Random effects can be described graphically in regards to characterizing models, and later in this
section, the various representations of random effects are shown in more detail. Finally, it is
important to keep in mind that random effects are not necessarily “random” in the traditional
statistical sense. Rather, they can be considered “random outcomes of a process identified with
the model that is predicting them” (Gelman 2007).
Figure 16 illustrates the “nested” data structure and why a linear mixed effect model was
the best choice for this study.

51

Figure 16: Nested Data Diagram

52

In Figure 16, the branches represent the data structure of the project. For example, a
single measurement point might be nested within the first phase of the second cell in the manurecontaminated wetland. A model with nested values gives numbers more context within the
model, so that the model takes into account that numbers within each grouping should be more
similar than numbers outside of the grouping. Values from one wash water will be most similar,
and of those values, the effluent of Cell 3 are more similar to each other. The linear mixed effect
model can take these multiple effects from groupings into account so the model is most efficient
and accurate.
Another important reason that the linear mixed effect model was chosen was because the
study is longitudinal. Again, the linear mixed effect model can put longitudinal values in context
and when making predictions, knows that the MC Wetland’s Cell 1 data points are related to
each other through time.
Since the study was longitudinal and had a nested data structure, a model that took the
numbers in context was necessary. For these reasons, the linear mixed effect model was selected.
The generalized formula to describe the statistical effects in this study is as follows:
𝑦 = 𝛽0 + 𝛽1 𝑆𝑡𝑎𝑔𝑒1 + 𝛽2 𝑆𝑡𝑎𝑔𝑒2 + 𝛽3 𝑆𝑡𝑎𝑔𝑒3 + 𝛽4 𝑃ℎ𝑎𝑠𝑒1 + 𝛽5 𝑃ℎ𝑎𝑠𝑒2

[1]

+ 𝛽6 𝑃ℎ𝑎𝑠𝑒3 + 𝛽7 𝑃ℎ𝑎𝑠𝑒4 + 𝛽8 𝑃ℎ𝑎𝑠𝑒5 + 𝛽9 𝑊𝑎𝑠𝑡𝑒
𝛽0 ~𝑁(𝑚𝑒𝑎𝑛0 , 𝜎02 )
𝛽1 ~𝑁(𝑚𝑒𝑎𝑛1 , 𝜎12 )
𝛽2 ~𝑁(𝑚𝑒𝑎𝑛2 , 𝜎22 )
… 𝑒𝑡𝑐, 𝑓𝑜𝑟 𝑎𝑙𝑙 𝛽
Where y refers to the value of either TP, COD, TN, nitrate, ammonia, or pH as dependent
variables. Independent variables are as follows: Stage refers to the stage of treatment (Cell 1, 2,

53

or 3); Phase refers to operational conditions; and Waste refers to the type of wash water. β0 is the
intercept, and the other β are the corresponding slopes of the parameters. As such, the mean is
assumed to be 0 for all β. The generalized formula is written in R as
y~factor(Stage)+factor(Phase)+Waste+(x|Date). In this case, (x|Date) refers to random effects by
testing events in a general sense. For actual statistical modelling, x was represented in 36
possible combinations involving the stage of testing. All 36 iterations of x were run, and the best
model was selected to continue analysis. Below are some simplified illustrations to show how x
affected the model.
For (1|Date) (Figure 17), the intercept, or influent value, is random, but treatment
efficiency (slope) remains constant (lines on graph represent testing events).

Figure 17: Random Effect (1|Date)

In this case, if the model was best fit for a random effect of (1|Date), it would mean that
the composition of the influent wastewater was random but the percent removal remained
constant.
For (Stage|Date) (Figure 18), the intercept is constant, but the slope, or efficacy of
treatment, is variable.
54

Figure 18: Random Effect (Stage|Date)

In this case, if the model was best fit for a random effect of (Stage|Date), it would mean
that the composition of the influent wastewater was statistically constant but the percent removal
varied randomly from week to week.
For (1+Stage|Date) (Figure 19), both the intercept and slope are variable.

Figure 19: Random Effect (1+Stage|Date)

In this case, if the model was best fit for a random effect of (1+Stage|Date), it would
mean that the composition of the influent wastewater and the percent removal varied randomly
55

from throughout the experiment. Random effects show variability as opposed to fixed effects,
which represent the effect of a variable on treatment.
The reason there were 36 models instead of three was because each stage of treatment
was considered for having an effect individually on the model, and all combinations of stage on
the model were tested.
As previously mentioned, the best model was selected for each variable. This was done
by comparing the AIC and BIC values for each model. AIC stands for Akaike Information
Criterion and BIC stands for Bayesian Information Criterion. AIC/BIC were developed to
determine the quality of a statistical model; the lower the values, the better the model. The model
with the lowest AIC and BIC values were selected: in the incidence that AIC and BIC values
were very similar for two models, the model was chosen theoretically (ex. graphical
representation of influent phosphorus indicated that intercept did not change between testing
events).
Statistical significance is usually determined with p-value, but R does not compute pvalue for linear mixed effect modelling (Bates 2006). SAS does compute a p-value for linear
mixed effect modelling, but there is debate in the academic community whether doing so is
valid. The creator of R’s linear mixed effect modelling program (lmer), Bates, does not believe
it is appropriate to compute p-value, because some assumptions on degrees of freedom would be
made in the modeling process, which he does not think are valid. The t-value given by R (in lieu
of a p-value) is found as a square root of an F ratio that has a single denominator value. SAS
assumes varying degrees of freedom for the singular value to compute p-values, which is not
considered correct by R. Academics are split on both sides as well. As a general rule of thumb,

56

t-values less than -2 or greater than 2 are considered to be roughly equivalent to a p-value of 0.05
and are seen as statistically significant.
R was chosen for linear mixed effect modelling instead of SAS or other programs for
three reasons: it was recommended by Gelman (2007), most familiar to the researcher, and easily
accessible due its free cost.

57

RESULTS AND DISCUSSION
Manure-contaminated (MC) and manure-free (MF) wash water streams were studied.
The MF wash water came from a milking parlor that is completely separated from holding or
standing areas that are prone to manure contamination. Therefore, the wash water was made up
of cleaning chemicals and milk. Any manure in the wash water is incidental from tracking if on
shoes from the milking parlor onto the floor of the milk house or using the sink. Conversely, the
MC wash water came from a milking parlor that did have holding areas that contributed to
manure contamination.
In order to progress to the results portion is it is necessary to explain the starting
conditions and phases that the system underwent in the course of the study. Afterwards,
analytical results of this study are categorized by nutrient: first examining individual wash water
streams, and next comparing them. Phosphorus is discussed first, followed by COD, and finally
nitrogen. Next, results of the hydraulic conductivity experiment are examined with respect to
clogging and the life of the system. Finally, the information on loading and clogging are utilized
to discuss sizing and design.

Starting Conditions
Each system was fed twice a day, from 9 am to 10 am and from 9 pm to 10 pm. The
starting flow rate was 250 ml/hour, and since the systems were fed two hours a day, the daily
flow rate was 500 ml/day, which translates to a hydraulic loading of 1.5 gal/day/ft2. Both
systems also had a recycle ratio of 3:1 for the first cell. In other words, the effluent of Cell 1
went back to be applied as influent for Cell 1 again or forward to be applied as influent of Cell 2
at a rate of 3:1. According to the literature, the 3:1 ratio is ideal for wetland microbial systems
58

(Ladu et al. 2013). The system was run under ambient temperatures at first to allow for the
development of microbial communities.

Phases
In order to compare different conditions with only two wetlands, both MC and MF
systems were tested under four different operating conditions, called phases. The goal of phase
changing was to alter one dependent variable at a time in each wetland in order to observe its
effects on independent variables. Measured nutrient values were the independent variables and
dependent variables and included time, phase, and influent values (loading). Time was a
dependent variable that could not be isolated, but according to the literature review, all nutrient
changes except phosphorus depend on biological equilibrium. Therefore, the time component
was not considered as a variable except for phosphorus. Influent values stayed constant
throughout some phases, but not all; these cases were considered with care.
The first two phases were the same for both lab-scale wetland, and the last two phases
were unique to each wetland. A summary table of the phases is at the end of the section (Table
6). Phase 1, ambient condition, operated at room temperatures to allow the microbial populations
to develop and establish a baseline for comparison. The temperature was s mostly in the 70s
(Fahrenheit).
Phase 2 for both wetlands initiated winter temperatures (cold phase). Each wetland was
cooled to approximately 40°F (See Appendix C for monitoring data) and all other dependent
variables were kept constant. On three separate occasions in the course of two winters,
temperature profiles were recorded at the field site, which confirmed an approximate consistent

59

40°F at depth. The cold phase allowed microbial communities to adjust to cold temperatures and
serve as a cold baseline for comparison to the operational adjustments.
Phase 3 was different for each wetland. Because the resultant data from previous phases
already indicated that separate designs were needed for MF and MC wetlands, phase 3 did not
compare results between the two. Again, each wetland was meant to hold all variables constant
except for the one controlled by phase. For the MF wetland, the loading was increased by a third
to a flow rate of 667 ml/day. This was done because the system was performing well and it was
time to start increasing stress to find its limits. The MC wetland was not adequately reducing the
level of nitrogen so the third cell was saturated. It was hoped that more anaerobic conditions
would lead to more treatment of the nutrient.
The fourth phase was unique to each wetland as well. Performance for the MF wetland
was still good, so the change was an increase in loading to twice its original value for a flow rate
of 1000 ml/day. The MC wetland still had poor performance, and developed leaking issues in the
third cell. As a result, the third cell was converted back to aerobic conditions. To increase
performance, the recycle ratio was decreased from 3:1 to 2:1, because it was believed the lack of
treatment in the second and third cells was due to a lack of carbon in those cells.

Phase

Dates

Days

1
5/1/13-8/1/13
0-93
2
8/1/13-10/2/13
93-155
3
10/2/13-1/10/14 155-255
4
1/10/14-2/24/14 255-300
Table 6: Operational Conditions

Manure-Contaminated
(MC)
Ambient Conditions
Cold Conditions
1.3x Loading
Cell 3 Saturation
2x Loading
2:1 Recycle Ratio

Manure-Free (MF)

60

To compare phases and cells, graphs were prepared and the statistical significance
(equivalent to p< 0.05, indicated by an asterisk in the tables) for the treatment efficiency of each
parameter was determined. In addition to showing the change for each cell, the changes in the
influent from phase to phase is listed and the “Overall” change (influent compared to Cell 3
effluent) is also provided. The numbers in the statistical tables indicate average percent change,
either from phase to phase or from the influent to effluent of a cell (also considered percent
treatment, or reduction). In the comparison tables, the average values for the appropriate
categories were computed for both wetlands and the numbers in the tables represent the percent
differences between those values.

Phosphorus
The analytical results of the MF wetland for phosphorus are shown in Figure 20 and
Tables 7 and 8 For this wetland, the three phase changes: ambient to cold, cold to 1.3 x loading,
and 1.3 x to 2 x loading.

70

P (mg/L)

60
50

MF Influent

40

MF Cell 1

30

MF Cell 2

20
MF Cell 3

10
0

Days since start of Experiment

Figure 20: Graph of Analytically Measured Phosphorus Levels for MF Wetland

61

Figure 20 shows that measured phosphorus levels were fairly consistent in the earlier
phases, only to increase dramatically in later phases. Table 7 shows that transitioning from
ambient to cold temperatures (Phase Change 1) resulted in a significant decrease in phosphorus
removal overall and in Cells 2 and 3. This is congruent with Stefanakis and Tsihrintzis’s 2012
study that found that phosphorus treatment is more effective in warm temperatures. It is not
clear why there was not significant change for Cell 1 as well.

Phase Change-MF
Condition
1
2
3
Overall
-29* 13
-119*
Influent
2.9
11
-66*
Cell 1
11
18
-129*
Cell 2
-81* 27
-125*
Cell 3
-48* -4.0* -155*
Table 7: Percent Difference between Phases for Phosphorus in MF Wetland

According to Table 7, influent remained constant throughout the first two phase changes,
but there was a significant change between the third and fourth phases. Influent increased 66%
percent during the last phase change. As seen in Figure 20, at first this resulted in a greater
reduction overall and for each cell compared to Phase Change 2 (1.3 x to 2 x loading). Since
phosphorus reduction is equilibrium driven, a higher level of influent does increase removal so
the results are predictable. However, as seen in Figure 20 and Table 7, the effluent mg P/L
leaving each cell increased dramatically overall as the removal capacity became depleted.
During the Phase 4, Cell 1 no longer provided significant treatment (Table 8).
Although the characteristics of the influent waste changed during the last phase, which
was not a planned change, the change was acceptable in terms of examining results. For the last
phase change, the plan was for loading to be increased and all other variables (besides time) kept
62

constant. Indeed, loading rate increased, but since higher influent phosphorus has the same
effect, the effect was cumulative. Phosphorus loading was increased more than planned, but the
results are still useful to look at in terms of the effects of increased loading.

Cell Treatment-MF
Condition
1
2
3
Overall
16* 25* 29*
Phase 1
15* 55* 34*
Phase 2
22* 8*
46*
Phase 3
28* 18
23*
Phase 4
0.02 20* 13*
Table 8: Percent Removal per Cell for Phosphorus in MF Wetland

From Table 8, phosphorus levels decreased with progression through each cell, which
remained true in all phases. The only exception was for Cell 2 for Phases 3 and Cell 1 for Phase
4. This is expected because the capacity of the media to physically sorb phosphorus was
continually being exhausted. The second cell discontinued removing phosphorus likely because
it has the longest retention time and its capacity was exhausted before the other cells. Cell 1
came into contact with the highest concentration of phosphorus, so the capacity was also stressed
with time.
While phosphorus treatment was adequate in the MF wetland, the capacity to treat
phosphorus is diminished with time. With the current design, there are no long term phosphorus
treatment mechanisms. For implementation, users may want to provide additional phosphorus
treatment if necessary.
The analytical results of the MC wetland for phosphorus are shown in Figure 21. In the
MC wetland, Phase Change 1 was ambient to cold, Phase Change 2 was cold to an anoxic Cell 3,

63

and Phase Change 3 was changing the anoxic Cell 3 back to aerobic and decreasing the

P (mg/L)

recirculation ratio from 3:1 to 2:1.

45
40
35
30
25
20
15
10
5
0

MC Influent
MC Cell 1
MC Cell 2
MC Cell 3

Days since start of Experiment

Figure 21: Graph of Analytically Measured Phosphorus Levels for MC Wetland

As seen in Figure 21 and Table 9, influent phosphorus remained constant throughout the
duration of the study, which implies changes in phosphorus treatment throughout the study were
based on changes within the system.
Phosphorus treatment removal was statistically significant only for the first phase change,
from ambient to cold conditions. During this phase change, all variables except time and the
internal temperature of the cell remained constant. Therefore, statistical changes in phosphorus
removal are likely due to temperature or time change. The statistical change from Phases 1 and 2
was evident in Cells 2 and 3. Like in the MF wetland, it is unclear why the change in Cell 1 was
not significant for the first phase change. However, a change due to temperature makes sense,
because it is what was found in the MF wetland as well as Stefanakis and Tsihrintzis’s 2012
study that phosphorus treatment is more effective in warm temperatures.
64

Since the other MC phase changes were not statistically significant, it is likely that the
other changes did not impact the mechanism for phosphorus removal.

Phase Change-MC
1
2
3
Overall -40*
-7.7 -5.8
Influent 3.1
-6.4 6.8
Cell 1
6.2
-8.1 -4.3
Cell 2
-67*
-7.6 -17
Cell 3
-102* -8.6 -8.3
Table 9: Percent Difference between Phases for Phosphorus in MC Wetland

The removal for each cell (Table 10) showed similar results to the MF wetland, as
expected. All cells except Cell 2 had significant removal throughout the study. Cell 2
maintained significant removal until Phase 4. The reason for the depletion of Cell 2 is likely
because sorption removal is a physical process that has a limited capacity. Since the MC
wetland had a lower loading than the MF wetland, it was able to maintain treatment in each cell
for longer.

Cell Treatment-MC
1
2
3
Overall 21* 31*
27*
Phase 1 22* 54*
36*
Phase 2 24* 18*
22*
Phase 3 23* 19*
21*
Phase 4 14* 8.5
27*
Table 10: Percent Removal per Cell for Phosphorus in MC Wetland

65

For the MC wetland, phosphorus was treated adequately in the study. In terms of design,
all three cells were utilized for phosphorus depletion, but as noted, phosphorus removal is limited
with time.
A comparison table of the two wetlands is presented in Table 11. The wetlands had
statistically identical treatment overall, except during Phase 1, which was ambient conditions, for
unclear reasons. The third and fourth phases were not compared because the systems were run
under different operating conditions. It was expected that the wetlands would have similar
removal since loading and influent concentrations were statistically undistinguishable. Perhaps
not all of the transient data for the first phase were removed properly, and skewed results for that
phase.
Statistical Differences Between MF and MC Wetlands
Overall Influent
Cell 1 Cell 2 Cell 3 Phase 1
Phase 2
Phosphorus -7.8
5.4
11
17
15
-3.2*
15
Table 11: Differences in Phosphorus Levels between MF and MC Wetlands

Both wetlands were able to effectively remove phosphorus, but both also showed a
decreasing capacity with time. This was expected because it is believed that the removal of
phosphorus is due to physical sorption. Also as expected, there is a negative relationship between
temperature and sorption (first phase change) for the reasons previously presented.
Although phosphorus is adequately removed in the short term, if this technology is
implemented on farms there may be some concern for phosphorus removal in the long term. An
additional treatment stage for phosphorus may be required for the long term. For wetlands, the
only sustainable phosphorus removal technique is through phytoremediation/harvesting, which,

66

while not very effective during warm periods, isn’t even applicable for this technology’s
wintering periods, since the wash water is applied below the active root zone.

COD
The analytical results of the MF wetland for COD are shown in Figure 22 and Tables 12
and 13. For this wetland, the three phase changes: ambient to cold, cold to 1.3 x loading, and

COD (mg/L)

1.3 x to 2 x loading.

2000
1800
1600
1400
1200
1000
800
600
400
200
0

MF Influent
MF Cell 1
MF Cell 2
MF Cell 3

Days since start of Experiment

Figure 22: Graph of Analytically Measured COD Levels for MF Wetland

Percent differences between phases is shown in Table 12 for the MF wetland. As seen in
Figure 22, influent COD concentrations were very high in the MF wastewater and also varied for
Phase Changes 2 and 3 (Table 12). Although all conditions (including influent concentrations)
were supposed to be kept constant, the factors of interest for the MF wetland during the last

67

phases were related to higher loadings. Since a higher concentration of COD contributes to a
higher loading, the change in influent concentration was acceptable for interpreting results.
Since the high influent loading occurred during the third and fourth phases, it
corresponded to cold atmospheric winter temperatures outside of the controlled experiment.
Before the influent is distributed to the wetland, it sits in holding tanks at both farms, and it is
believed that winter temperatures at these sites caused less biological activity and COD
degradation in the tanks. Therefore, there are higher influent COD levels in the collected
samples during the winter season. However, this did not have a significant effect on treatment
because Cell 1 was able to adequately remove most COD in all phases (see Table 13 for percent
change between cells).

Phase Change-MF
1
2
3
Overall 15
-17
-79
Influent 45
-102* -83*
Cell 1
19
15
-117
Cell 2
-8.3 -8.4
-58
Cell 3
5.5 29
-59
Table 12: Percent Difference between Phases for COD in MF Wetland

According to Table 13, the drop in COD level for all phases was primarily due to
treatment in the first cell, although Cell 2 also had an overall statistical impact. Despite
increasing loading to more than twice that of the field site, these results indicate the cells are
oversized for organic removal in the tested range. However, the results are expected, since
constructed wetlands have been shown to treat high COD loadings adequately (Stefanakis and

68

Tsihrintzis 2012). For COD, it is questionable whether Cell 2 or especially Cell 3 is necessary for
treatment.

Cell Treatment-MF
1
2
3
50*
-15
Overall 90*
62
-55
Phase 1 89*
49
-35
Phase 2 83*
36
12
Phase 3 93*
92*
53
11
Phase 4
Table 13: Percent Removal per Cell for COD in MF Wetland

The analytical data for the MC wetland is summarized in Figure 23. In the MC wetland,
Phase Change 1 was ambient to cold, Phase Change 2 was cold to an anoxic Cell 3, and Phase
Change 3 was changing the anoxic Cell 3 back to aerobic and decreasing the recirculation ratio

COD (mg/L)

from 3:1 to 2:1.

2000
1800
1600
1400
1200
1000
800
600
400
200
0

MC Influent
MC Cell 1
MC Cell 2
MC Cell 3

Days since start of Experiment

Figure 23: Graph of Analytically Measured COD Levels for MC Wetland

69

As shown in Figure 23, the MC wetland had shifting influent COD levels that were
believed to be affected by the atmospheric temperature at the collection sites. This was true for
the MF wetland as well, and since the only change that both influents underwent during the last
two phases was the chilling atmospheric conditions, the change is attributed to the outdoor
winter conditions.
Tables 14 shows the percent change between phases for each category. As expected from
Figure 23, the influent COD is significantly different for the second and third phase changes.
Unlike the MF wetland, which had no significant changes outside of the influent, Cells 2 and 3 in
the MC wetland were affected by a phase change (the second). In the second phase change, the
only planned change was to make the third cell anoxic. For this situation, all variables were kept
constant except time, influent COD, and Cell 3. Since biological processes in this instance aren’t
expected to change with time, and only the third cell was changed for the phase, the significant
change in Cell 2 is attributed to the higher influent COD.
For Cell 3, the significance of the second phase change could be anoxic conditions or a
higher influent loading. The reason for the significance is unclear looking at Table 14 alone, but
in Table 15 (which shows the percent change in treatment for each cell) it is observed that Cell 3
only becomes significant in treatment for the last phase. The last phase is defined by the
cumulatively higher organic loading in latter cells from increasing influent and decreasing
recycle ratio. Therefore, it is likely that high organic loading had the greater impact on treatment
instead of anoxic conditions.

70

Phase Change-MC
1
2
3
Overall
-26
-12*
-42
Influent -39
-108* -50*
Cell 1
6.5
-38
-29
Cell 2
-45
47*
-55
Cell 3
-27
51*
-34
Table 14: Percent Difference between Phases for COD in MC Wetland

According to Table 15, COD changed significantly for the first cell in every cell for the
MC wetland. Cell 2 significantly reduced COD overall and in Phases 3 and 4, but Cell 3 only
significantly impacted the reduction during Phase 4. It is believed that the higher COD loading
into the second cell during the last two phases caused the significant treatment of COD in Cell 2,
since all other factors besides time remained constant. During the last phase, COD loading was
increased from two sources: higher influent concentration and lower recycle ratio. The higher
loading caused more COD in Cells 2 and 3. The highest COD loading of all was during Phase 4,
when Cell 3 had an impact on treatment. It is likely enough COD leaked out of Cell 2 into Cell 3
during this time to allow Cell 3 to have an impact.
Since all three cells had an impact on COD treatment during the last phase, the capacity
of the cells for the MC wetland were reached. Unlike the MC wetland, in which Cells 2 and 3
were not impacting overall treatment, the MF wetland utilized all cells for treatment. Therefore,
for COD treatment, at least three cells should be designed.

71

Cell Treatment-MC
1
2
3
Overall 81* 45* 1.7
Phase 1 66* 42
-12
Phase 2 77* 11
1.5
Phase 3 85* 66* 7.8
Phase 4 87* 59* 20*
Table 15: Percent Removal per Cell for COD in MC Wetland

Table 16 shows the statistical differences between wetlands for COD. Influent COD was
not statistically different between wash water streams, but there was a difference in treatment for
all three cells. During the ambient phase, the COD removed from each wetland was not
statistically different, but a statistical difference developed in the second phase, before operating
conditions became divergent. According to tables 13 and 15, the difference between the systems
is seen clearly in Cell 2, where the MF wetland is still treating COD, but the MC wetland’s
treatment has stagnated. The reason for the difference in treatment is likely due to nitrogen, since
the wetlands were virtually identical otherwise. High nitrogen in the MC wetland depleted the
COD too quickly, but the MF wetland had an appropriate C:N ratio that was sustained into Cell
2. The relationship between COD and Nitrogen removal is discussed further in later sections.

Statistical Differences Between MF and MC Wetlands
Overall Influent
Cell 1 Cell 2 Cell 3 Phase 1
Phase 2
COD
4.9
-10
-108* -134* -100*
17
10*
Table 16: Differences in COD Levels between MF and MC Wetlands

Despite variable and high loadings, the wetlands were able to effectively treat COD. Due
to such variance in COD, percent removal for the first cell was considered in terms of influent

72

(Figure 24). COD removal (percent-wise) seems to peak at an influent of approximately 800
mg/l. The higher loadings are associated with a fairly level removal rate, meaning more grams of
COD left Cell 1. This follows a typical Monod kinetic relationship in that at a lower level of
COD, concentration is limited while at a higher level, metabolism is limiting (Doran 1995).

100
95

Percent Removal (%)

90
85

Phase 1-MF

80

Phase 2-MF

75

Phase 3-MF

70

Phase 4-MF

65

Phase 1-MC

60

Phase 2/3-MC

55

Phase 4-MC

50
0

200

400

600

800

1000

1200

1400

1600

1800

2000

Influent COD

Figure 24: Cell 1 COD Removal

According to the data, vertical treatment wetlands are capable of treating high and
variable loadings of COD, although the third cell might be superfluous. In later phases, the MC
wetland did start to see COD entering into Cells 2 and 3. More research is needed to find the
limits of COD treatment, since this study was effective at all operating conditions tested.

73

Nitrogen
The nitrogen section includes data from total nitrogen measurements as well as nitrate
and ammonia. Total nitrogen, nitrate, and ammonia will be discussed for the MF wetland first,
followed by the discussion of the MC wetland.
The graphical results of the MF wetland TN series is shown in Figure 25. For this
wetland, the three phase changes are ambient to cold, cold to 1.3 x loading, and 1.3 x to 2 x
loading. As seen in Figure 25, TN treatment remained fairly constant throughout each phase.

70
60

N (mg/L)

50
40

MF Influent

30

MF Cell 1

20

MF Cell 2
MF Cell 3

10
0

Days since start of Experiment

Figure 25: Graph of Analytically Measured TN Levels for MF Wetland

In Table 17, the lack of change between phases for TN removal is verified, as no
statistical significance was found for any phase change. This means that TN was not affected by
changing temperature, and the TN loading limits were not reached even in the final phase, where
the loading had doubled. Consequently, the MF wetland had adequate capacity even at the
highest loading tested.
74

Unlike COD, the influent TN did not vary between phases. This is expected as changes
are not likely under storage conditions, where high COD and low oxygen levels exist. Therefore,
the study was successful in isolating changes to the variable of interest and results may be
examined straight-forwardly.

Phase Change-MF
1
2
3
Overall -32
-14 11
Influent -13
-19 -2.7
Cell 1
-12
27
-5.5
Cell 2
-101 10
14
Cell 3
-1.4 -76 37
Table 17: Percent Difference between Phases for TN in MF Wetland

According to Table 18, only Cells 1 and 2 removed TN, with Cell 1 providing the
greatest contribution. During Phase 4, the TN in Cell 3 actually statistically decreased as well,
indicating a potential use for a third cell in design. Since the only change for nitrogen during this
time period was related to higher loading, it is likely the third cell only becomes useful when
adequate TN and COD leak into the cell from Cell 2. However, because no statistical
significance was found in Cell 3 for the last phase change, the treatment in Cell 3 during Phase 3
is similar to that in Phase 4. Therefore, while Cell 3 may play a role in some instances, it may not
be very important for treatment in general.

75

Cell Treatment-MF
1
2
3
12
Overall 64* 53*
-18
Phase 1 55* 73*
41
Phase 2 55* 51*
-15
Phase 3 72* 39
15*
Phase 4 72* 51*
Table 18: Percent Removal per Cell for TN in MF Wetland

The results indicate that TN treatment in a MF wetland is very consistent even at high
loadings. Most likely only two cells are needed for adequate treatment, but Cell 3 can play a role
under some circumstances and may be desired as part of a conservative design.
Figure 26 and Tables 19 and 20 are the MF data for ammonia.

30

NH3-N (mg/L)

25
20
MF Influent

15

MF Cell 1
MF Cell 2

10

MF Cell 3
5
0

Days since start of Experiment

Figure 26: Graph of Analytically Measured Ammonia Levels for MF Wetland

As seen in Figure 26, ammonia treatment remained fairly consistent throughout the
duration of the study. The first cell was responsible for most removal, and by the end of
76

treatment, ammonia levels were very low. As seen from Table 19, there were no significant
differences for any cells for any phase change. This is expected, since TN did not change
significantly with phase change either.

Phase Change-MF
1
2
3
Overall 33
-32 -43
Influent -6
-4.6 31
Cell 1
100
--- -299
Cell 2
-63 -60 -5.3
Cell 3
100
--- 100
Table 19: Percent Difference between Phases for Ammonia in MF Wetland

Like seen in Figure 26, Table 20 shows that Cell 1 was responsible for the great majority
of ammonia removal, sometimes depleting ammonia below measurable levels (100 percent
removal in Phase 2). For the most part, according to Table 20, Cells 2 and 3 are less helpful, but
again, as with TN, were utilized for treatment in the last phase. Therefore, while Cell 1 is the
only necessary treatment stage for the most part, under high loadings, Cells 2 and 3 may be
useful as well.

Cell Treatment-MF
1
2
3
Overall 97* -345* 65*
65
Phase 1 96* -56
100
Phase 2 100* --18
Phase 3 99* -925
Phase 4 91* -171* 100*
Table 20: Percent Removal per Cell for Ammonia in MF Wetland

Figure 27 and Tables 21 and 21 show the nitrate data for the MF wetland.
77

45
40

NO3-N (mg/L)

35
MF Influent

30
25

MF Cell 1

20

MF Cell 2

15

MF Cell 3

10
5
0
Days since start of Experiment

Figure 27: Graph of Analytically Measured Nitrate Levels for MF Wetland

As seen in Figure 27 and Table 21, nitrate was not significantly altered by phase change
in the MF wetland. This was as expected because TN and ammonia treatment was also
unaffected.

Phase Change-MF
1
2
3
Overall 18
41
-56
Influent 10
-8
-91
Cell 1
-34 53
28
Cell 2
30
79
-81
Cell 3
68
39
-81
Table 21: Percent Difference between Phases for Nitrate in MF Wetland

According to Table 22, in all phases, Cell 1 statistically increased nitrate levels, as
expected due to nitrification. Cell 2 decreased nitrate overall and in Phases 3 and 4. Decreases
in Phases 1 and 2 are evident in Figure 27, but were not statistically significant. This is also
expected as Cell 2 was anoxic and designed to denitrify. However, this function was not as
78

important until higher TN loadings were realized. It does appear, graphically and statistically,
that in the last phase, Cell 3 was nitrifying.

Cell Treatment-MF
1
2
3
Overall -1221* 65* -6
-33
Phase 1 -1501* 43
39
Phase 2 -2297* 70
87* -76
Phase 3 -945*
-296*
67* -76*
Phase 4
Table 22: Percent Removal per Cell for Nitrate in MF Wetland

Both the first and second cell were useful for treatment in the MF wetland for nitrate.
The third cell was significant during the last phase, but in this instance, nitrate was created,
which may contaminate ground water.
An important factor in nitrate removal capacity is an adequate C:N ratio. The C:N for the

C:N Ratio

MF wetland is represented graphically in Figure 28.

50
45
40
35
30
25
20
15
10
5
0

MF Influent
MF Cell 1
MF Cell 2
MF Cell 3

Days since start of experiment

Figure 28: C:N Ratio for MF Wetland

79

As the graph shows, the C:N ratio was fairly high throughout the duration of the study,
and increased with the increasing carbon levels. The effluent of Cell 1 had a C:N ratio of
approximately 5:1 or 10:1 in later phases. With these ratios, the MF wetland was able to perform
adequately in nitrogen and COD removal. During the last two phases, carbon was found to be
significantly higher, which led to higher C:N ratios in the wetland. For the MF wetland, the C:N
ratio and treatment was satisfactory throughout the experiment, but in the latter phases, the third
cell was able to contribute to removal as well, for unclear reasons, as nothing was significantly
different between the third and fourth phases.
The pH levels were recorded at each stage of treatment throughout the study, since pH
can indicate if nitrification/denitrification is occurring. Results for the pH study are shown in
Figure 29 and Tables 23 and 24.

8.5
8

pH

7.5
MF Influent

7

MF Cell 1

6.5

MF Cell 2

6

MF Cell 3

Days since start of experiment

Figure 29: Graph of Analytically Measured pH for MF Wetland

Figure 29 shows that pH was fairly consistent throughout the study, though it was a little
significantly different from Phase 1 to Phase 2, as Table 23 shows. The only difference was that

80

Cell 1 showed a significant increase during Phase 1 (Table 23). It is unclear why this shift
occurred, though perhaps the microbial communities in Cell 1 were not fully adjusted during
Phase 1. In addition, a different pH probe was used in the beginning of the study, one that hadn’t
been calibrated.

Phase Change-MF
1
2
3
Overall -1.8* 1.8* 2.2
Influent -1.3
3.2 1.9
Cell 1
-3.9* 2.4 0.50
Cell 2
1.0
0.81 3.1
Cell 3
-2.9
0.78 3.3
Table 23: Percent Difference between Phases for pH in MF Wetland

According to Table 24, the pH increased overall in Phases 1 and 2. Cell 2 showed an
increase in all phases and overall. This is unexpected as the literature shows that nitrification
decreases pH (American Water Works Association 2002). However, denitrification increases pH
according to the literature (Ebeling et al. 2003), and an examination of TN shows that significant
denitrification must have been occurring in Cell 1, as well as the other two.

Cell Treatment-MF
1
2
3
Overall -4.6* -4.8* -1.0
-8.3* 1.8
Phase 1 -1.2
-3.2* -2.1*
Phase 2 -4.8
-4.8* -2.1*
Phase 3 5.6*
Phase 4 -7.2* -2.1* -1.9*
Table 24: Percent Removal per Cell for pH in MF Wetland

81

Even though the pH performance in the MF wetland was unexpected in some ways, the
performance of the wetland in terms of nitrogen treatment was excellent, so the pH levels in each
cell were acceptable.
Overall, the MF wetland treated nitrogen species effectively in two cells, showing
capacity for nitrification in the first cell, and denitrification in the first and second cells. There
was no statistical significance of phase change except for pH, though there was evidence that
under higher loadings, the system began to rely more heavily on latter cells, even utilizing the
third cell when necessary. The study concludes that nitrogen removal is effective under the tested
loadings, but more research is needed to find the system’s loading limit. It was found that the
field site’s footprint was at least twice as large as it needed to be. For now, including a third cell
for nitrogen treatment is recommended for design as a conservative judgment.
The results of the TN study for the MC wetland are shown in Figure 30 and Tables 25
and 26. Again, the first phase change for this wetland was from ambient to cold temperatures, the
second change converted Cell 3 to anoxic conditions, and the final phase change returned Cell 3
to aerobic conditions and decreased the recycle ratio from 3:1 to 2:1.

82

180
160

N (mg/L)

140
120
100

MC Influent

80

MC Cell 1

60

MC Cell 2

40

MC Cell 3

20
0
Days since start of Experiment

Figure 30: Graph of Analytically Measured TN Levels for MC Wetland

As seen in Figure 30, nitrogen treatment was not as consistent in the MC wetland
compared to the MF wetland. For the first two phases, it is clear that nitrogen removal was very
inconsistent and poor. Therefore, for the latter two phases, the variable changes were based on
trying to improve nitrogen treatment. In Figure 30, the third and fourth phases show good
nitrogen removal.
Table 25 shows the significant percent changes between phases for each cell. In the
phase change from cold conditions to a saturated third cell, there was a significant decrease in
nitrogen leaving each cell, even though influent nitrogen remained statistically constant. For the
third phase change, the researchers kept all factors constant except the saturation of the third cell.
Yet, the first and second cells had significantly better treatment. Even though influent nitrogen
remained the same, the influent COD was significantly higher for this phase. Since high COD
was the only changing factor going into Cells 1 and 2, with all other factors constant, it was
determined that the higher COD was the cause of the better treatment, by creating a more
favorable C:N ratio.
83

For the third cell, which also experienced improved treatment, it is unclear how treatment
was affected by saturation and changing influent into Cell 3. However, both variables were
compounding instead of crossing, and according to Table 26, Cell 3 does not have a significant
effect on treatment, therefore neither variable was significant enough to be concerned with.
As seen in Table 25, influent nitrogen increased significantly from Phase 3 to Phase 4.
Although influent nitrogen increased, treatment was statistically constant, and, moreover, it was
still adequate in terms of treatment. For the last phase, the recycle ratio was decreased from 3:1
to 2:1, which increased loading. Influent increased in nitrogen concentrations had a
compounding effect with a decreased recycle ratio: both increased loading. Therefore, even
though there were two variables, both have the same effect, and the effect was isolated.

Phase Change-MC
1
2
3
Overall -5.2 46* 5.9
Influent -2.8 8.6 -23*
Cell 1
0.42 43* -2.25
Cell 2
-5.6 72* 19
Cell 3
-13 60* 29
Table 25: Percent Difference between Phases for TN in MC Wetland

In Table 26, the percent removal in each cell and the corresponding significance is
shown. Throughout the study, only Cell 1 and 2 had an effect on treatment. Even though there
was no chance in significance with time, by looking at percent removals, it is clear that Phase 3
and Phase 4 had superior treatment. Cell 1 went from removing approximately 30% of the
nitrogen to 50-60%. Cell 2 went from removing about 25% to 60-70%. Treatment was

84

significantly better, even though treatment was still statistically visible in the earlier phases. Cell
3 had no significant effect throughout the duration of the study.

Cell Treatment-MC
1
2
3
Overall 38* 36* 26
Phase 1 24* 27* 31
Phase 2 27* 22* 27
Phase 3 54* 62* -3.3
Phase 4 62* 70* 9.6
Table 26: Percent Removal per Cell for TN in MC Wetland

In conclusion, TN was not affected by the first phase change, but the second was
significant for all cells, even though influent nitrogen remained the same. The last phase was not
statistically significant compared to the third phase (Table 25), indicating similar treatment in the
two phases, even though the influent was statistically different. For all phases, the first two cells
were effective in treatment (Table 26). Therefore, treatment was best with a higher C:N ratio,
and the recommended design may not need a third cell for TN removal.
Ammonia data is shown in Figure 31 and Tables 27 and 28.

85

250

NH3-N (mg/L)

200
150

MC Influent
MC Cell 1

100

MC Cell 2
MC Cell 3

50
0

Days since start of Experiment

Figure 31: Graph of Analytically Measured Ammonia Levels for MC Wetland

According to Figure 31, ammonia treatment seems fairly consistent throughout the study,
but according to Table 27, there were some significant differences. The general trend observed
throughout the study had high influent ammonia that was virtually eliminated in the first cell.
Next, ammonia was created in Cell 2, only to be eliminated in Cell 3. In fact, for most of the
experiment, Cells 2 and 3 were statistically indistinguishable.
Even though results in Table 27 show an overall significant difference between the first
and second phase changes, it is not clear where the difference is coming from, since no
individual cells show a significant difference. Overall, there is an increase in ammonia from the
first phase to the second, a decrease from the second to the third. The second phase, therefore,
appears to be an anomaly in treatment, and the phenomena can be observed in Figure 31. Cold
temperatures decrease nitrification kinetics (Stefanakis 2012) explaining the reduction in
ammonia removal for Phase Change 1, however it is unknown why the difference was temporary
and recovered for the third phase.

86

During the last phase change, Cells 2 and 3 underwent a significant change. For this
change, the level of ammonia was higher in Cell 2. Probably, the altered recycle ratio caused
more nitrogen and carbon to leak into latter cells, which then made the latter cells become more
significant. As for Cell 3, no percent change could be found because the average ammonia
leaving Cell 3 was 0 mg/L during Phase 3. The difference between the average of 0 and 1.2 mg/L
was significant statistically, but not practically.

Phase Change-MC
1
2
3
Overall -360* 63* -287
Influent -21
-4.2 7.1
Cell 1
-224
87
-458
Cell 2
-432
67
-412*
Cell 3
-765
100 ---*
Table 27: Percent Difference between Phases for Ammonia in MC Wetland

Table 28 shows the efficacy of each cell in ammonia treatment. As seen in the table, Cell
1 was very capable of removing the majority of ammonia. Cell 1 was significant in all phases. In
Phase 3, Cells 2 and 3 showed some significance. However, since Cell 1 removed 99% of
ammonia, the slight changes in ammonia in latter cells may be statistically significant, but not
practically significant. The average level of ammonia leaving Cell 1 during Phase 3 was 1.8
mg/L. For Cell 2, it was 6.0 mg/L, and for Cell 3, it was 0 mg/L. While the differences in percent
removals look large, in terms of the actual numbers, the difference is quite small.

87

Cell Treatment-MC
1
2
3
Overall 94* -82
87*
Phase 1 94* 18
81
Phase 2 89* -35
70
Phase 3 99* -250* 100*
Phase 4 92* -213
96
Table 28: Percent Removal per Cell for Ammonia in MC Wetland

Overall, ammonia changed significantly in the first and third cells but not the second.
Mostly, ammonia was treated exclusively with the first cell (Table 28). The most significant
reduction was expected in Cell 1 and the other significant changes were very low in magnitude.
Consequently, for ammonia removal, only Cell 1 is needed.
Nitrate data for the MC wetland is shown in Figure 32 and Tables 29 and 30.

120

NO3-N (mg/L)

100
80
MC Influent
60

MC Cell 1

40

MC Cell 2
MC Cell 3

20
0

Days since start of Experiment

Figure 32: Graph of Analytically Measured Nitrate Levels for MC Wetland

The difference between the latter two phases compared to the first two phases is starkly
observed in Figure 32 for nitrate removal. Although influent nitrate was consistently low, the
88

effluent nitrate for each cell was obviously different. The level of nitrate is lower coming out of
Cell 1, despite steady levels of influent TN, which indicates some denitrification is occurring in
the cell alongside nitrification. This is supported by the literature, which found that oxygen was
used up within inches of the top of media surface (Stefanakis 2012).
According to the statistical model, nitrate treatment was significantly changed with the
second phase for all cells (Table 29). The second phase changed the third cell to saturated
conditions. Only the third cell was altered, yet there was a significant change in Cells 1 and 2.
All variables were held constant for those cells except for influent COD. Therefore, the
improvement in treatment was due a higher C:N ratio and not the alteration of the third cell.
Improved treatment in the last cell might have been due to saturation or lower influent nitrogen,
however, treatment was not significant in the third cell at this time.

Phase Change-MC
1
2
3
Overall -2.4 47
7.2
Influent -47 7.3 -31
Cell 1
22
44* -4.8
Cell 2
18
67* 61
Cell 3
2.0 72* 3.2
Table 29: Percent Difference between Phases for Nitrate in MC Wetland

According to Table 30, Cell 1 significantly increased nitrate in every phase. This is
expected due to nitrification. Cell 2, which was responsible for denitrification, did not
significantly denitrify in Phase 1 or 2, presumably because the carbon supply was too low, since
denitrification requires carbon. Cell 2 began to contribute to treatment in the third phase, when
the carbon source was more adequate. The third cell was only significantly in the fourth phase,

89

when the loading increased. Again, this was likely caused by the availability of COD resulting
from the changed operational conditions, because denitrification requires an external carbon
source (Vymazal 1998). Higher nitrogen and COD as well as a decreased recycle ratio led to
higher COD loading which allowed the third cell to participate in treatment.

Cell Treatment-MC
1
2
3
Overall -6610*
37* 2.5
Phase 1 -12118* 28 14
Phase 2 -6364*
24 -6.5
Phase 3 -3789*
55* 9.0
Phase 4 -3023*
83* -125*
Table 30: Percent Removal per Cell for Nitrate in MC Wetland

As a result of this study, it is clear that an adequate C:N ratio is required for
denitrification. In addition, it was found that Cell 1 was able to denitrify as well. Cells 1 and 2
were vital for dentrification, but Cell 3 would nitrify at higher loadings, creating more nitrate. In
that light, Cell 3 could be detrimental to the nitrate treatment. However, Cell 3 might have
capacity to denitrify like Cell 1 if there is enough carbon left over.
Influent TN remained consistent throughout the duration of the experiment except for the
last phase change. However, despite this, there was a significant change in treatment for the
second phase change. The phase change should have only affected the third cell, which became
saturated, but the first and second cells were affected as well. Influent COD did change during
this time, and the resultant change in C:N ratio (Figure 33) might have contributed to the sudden
improvement in treatment in all cells.

90

16
14

C:N Ratio

12
10
MC Influent

8

MC Cell 1

6

MC Cell 2

4

MC Cell 3

2
0

Days since start of Experiment

Figure 33: C:N Ratio for MC Wetland

As this graph shows, the C:N ratio was much higher in the last two phases, which most
likely caused the improved denitrification, and as a result, there was greater overall TN removal.
The higher carbon levels in the first cell meant that more carbon was getting into the latter cells,
creating better nitrogen removal in the system at all stages of treatment. Consequently, with
higher carbon, as was seen in the last two phases of operation, nitrogen treatment is viable in a
MC wetland. Otherwise, treatment is lacking, especially for nitrates.
The pH levels were recorded at each stage of treatment are shown in Figure 34 and
Tables 31 and 32.

91

9
8.5

pH

8
MC Influent
7.5

MC Cell 1

7

MC Cell 2

6.5

MC Cell 3

6

Days since start of Experiment

Figure 34: Graph of Analytically Measured pH for MC Wetland

Although the pH measurements look fairly consistent according to Figure 34 for the MC
wetland, the first cell was affected by the first two phase changes (Table 31). It is unclear why
this shift occurred, though perhaps the microbial communities in Cell 1 were not fully acclimated
during Phase 1. In addition, a different pH probe was used in the beginning of the study, one that
hadn’t been calibrated. The MF wetland also experienced change in the beginning of the study,
which supports the idea that the changing pH probe affected the measurements. However,
temperature affects nitrification kinetics, and the system was cooled down in the second phase,
which is also applicable to the MF wetland.

Phase Change-MC
1
2
3
Overall -1.3
0.31 0.59
Influent -3.2
2.5 0.68
Cell 1
-4.0* 2.8* 1.2
Cell 2
2.1
-1.7 1.5
Cell 3
-0.26 -2.3 -1.0
Table 31: Percent Difference between Phases for pH in MC Wetland

92

As expected from literature, pH decreased in the first cell due to nitrification and
increased in the second cell due to denitrification (see Table 32). The third cell also experienced
an increase in pH. In the third phase, Cell 1 was significant, but not in other phases. However,
Phase 3 and 4 were the best phases for the MC wetland and represent the ideal operating
conditions, which corresponds to the pH during that time. Overall, pH made sense in the MC
wetland and reveal that the processes of nitrification and denitrification were occurring. A
change of pH in the third cell may be indicative of treatment in that cell, supporting the presence
of the third cell in design.

Cell Treatment-MC
1
2
3
Overall 3.0* -8.1* -2.6*
Phase 1 3.3
-11* -0.66
Phase 2 2.5
-4.0* -3.1*
Phase 3 2.8* -8.9* -3.7*
Phase 4 3.3
-8.6* -6.3*
Table 32: Percent Removal per Cell for pH in MC Wetland

To summarize nitrogen treatment in the MC wetland, a carbon source is required for
adequate treatment. Cell 1 is useful for nitrification and some denitrification. Cell 2 is required
for adequate for denitrification. It is unclear whether Cell 3 is truly useful, but can decrease TN
in some situations. Therefore, it might be useful to include Cell 3 as a conservative design. In
some situations, the third cell can increase nitrate, which might be a concern for groundwater
discharges.
A comparison table of nitrogen species in both wetlands is shown in Table 33. In
addition, a graphical comparison (Figure 35) is also provided.

93

As observed in Table 33, nitrogen in the two systems was very different, as compared to
COD. The MC wastewater had an influent TN approximately three times higher than the MF
wash water. TN was statistically different in every stage of treatment and in every comparable
phase. Nitrate and ammonia were also different with respect to the wastewater streams. Influent
nitrate is low in both wash water, but after it is created in the first cell, nitrate levels diverge for
every stage and phase of treatment. Ammonia was mostly relevant in the first and third cells,
where it is considered different between wetlands. Although treatment in the MF wetland was
seen as more effective for nitrogen, pH levels in the MC wetland were consistent with a
successful multi-cell treatment process.

Statistical Differences Between MF and MC Wetlands
Overall Influent Cell 1 Cell 2 Cell 3 Phase 1
Phase 2
Nitrogen 17*
-257*
-521* -744* -613* 28*
33*
Nitrate
-9.3*
-12
-469* -926* -841* -5.7*
-14*
Ammonia -3.1*
-533*
-1380* -500
-117
-1.7*
4.5*
pH
29
-3.2
4.3*
1.3
-0.28
12
56*
Table 33: Differences in Nitrogen Speciation between MF and MC Wetlands

Interestingly, it also appears that in MF wash water, there is proportionally more organic
nitrogen (as seen by the stratification of total nitrogen and ammonia), whereas the nitrogen in the
MC wash water is almost entirely made up of ammonia (Figure 35). The representation of data in
the side-by-side graphs also emphasizes the efficacy of nitrogen removal in the MF wetland
compared to the MC wetland.

94

Figure 35: Nitrogen Speciation Comparison across Cells

95

The representation of data in the side-by-side graphs also emphasizes the efficacy of
nitrogen removal in the MF wetland compared to the MC wetland. It appears that with
increasing COD, there was decreasing total nitrogen leaving Cell 1, leading to lower influent
nitrogen going into Cell 2, that, when combined with higher carbon in the cell, is removed
effectively. Figure 36 shows the importance of the C:N ratio on nitrogen removal in the first
cell.

90
80

% Removal

70
Phase 1-MF

60

Phase 2-MF

50

Phase 3-MF

40

Phase 4-MF

30

Phase 1-MC

20

Phase 2/3-MC

10

Phase 4-MC

0
0

5

10

15

20

25

30

35

40

45

50

C:N Ratio

Figure 36: Percent TN Removal in Cell 1 with Increasing C:N Ratio

According to Figure 36, an increasing C:N ratio increases nitrogen removal. This occurs
until a C:N of approximately 10:1, after which improvement slows with increasing C:N. At a
C:N ratio of 10:1, percent removal is approximately 60%.

Clogging
A hydraulic conductivity study was done once at the beginning of the study and once at
the end to determine the potential for clogging (see Appendix E for hydraulic conductivity
96

calculations). For the MC wetland, the hydraulic conductivity was found to be 3.1x10-3 m/s at
the beginning of the study, and 2.7 x10-3 m/s at the end of the study for a calculated decrease of
4 x10-4 m/s. Both measured hydraulic conductivities fall within the normal range for gravel
(Swiss Standard SN 670 010b), however it is unclear whether the decrease in hydraulic
conductivity is meaningful. The MF wetland found an initial hydraulic conductivity of 2.8 x10-3
m/s even though it should have been identical to the MC wetland. Therefore, the margin of error
is too large to allow for the small change in hydraulic conductivity to be statistically significant.
With this data it is impossible to reliably calculate the lifetime of the system, but since there is no
evidence a substantial change occured under higher loadings, the lifetime of the system will
likely be adequately long.

Sizing and Design
Since the system was tested at appropriate depth, only the surface area is sized. Sizing for
the system may rely on hydraulic loading rate, or the influent concentration of carbon or
nutrients. Potentially, the lifetime of the system due to clogging is also a factor. Sizing a system
based on these criteria is called performance-based design. In order to determine applicability of
the wetland design, effluent values were compared to typical effluent values of accepted
technology. Table 34 shows the expected effluent concentrations of both wetlands and other
technology.

97

Parameter

MF
Wetland

MC
Wetland

Septic
Tank

Aerobic
Treatment
Unit

Nitrogen (mg/L
4.5
17
30-50
5
N)
COD (mg/L)
44
58
100-250
650-700
Table 34: Effluent Concentration Comparison across Different Technologies

MI Filter
Mound
--205-815

From the table, it is clear the wetland is a competitive technology compared to septic tank
(University of Minnesota 2010), Aerobic Treatment Unit (Larson and Safferman 2009), and the
Michigan Filer Mound (Rathbun et al. 2012). To size a wetland system, loading values were
calculated for both MF and MC wetlands (Table 35). The reason a traditional drainfield cannot
be used is because the organic material and milk fat in the waste will cause clogging and prevent
infiltration through the media (Christopherson et al. 2007).

Loading
Hydraulic Loading (gal/day/ft2)

MF Wetland
0.504

MC Wetland
0.251

Nitrogen Loading (lb-N/day/ft2)

2.87x10-4

1.03x10-3

COD Loading (lb-COD/day/ft2)
Table 35: Loading Values

1.06x10-2

5.4310-3

By knowing gal/day or lb-N/day, calculating the required foot print for the wetland is
simple: divide by the loadings in the table. This value represents the total surface area, and
should be divided by three for the total surface area of each wetland cell. The largest calculated
footprint should be used because it will ensure adequate treatment of each parameter. It is
important to note that a C:N ratio of 10:1 or higher is recommended.
There is flexibility in the design in that the recirculation ratio can be adjusted onsite. With
proper monitoring, if it is determined that a 3:1 ratio is unneeded to achieve adequate treatment,
98

the ratio can be adjusted to minimize energy costs. In addition, if there is a high concern for
clogging, the size can be increased.

99

CONCLUSIONS
Overall, the wetland used for treating milking facility wash water without manure
contamination performed very well throughout the duration of the experiment. Stress was placed
on the system, but a loading that caused failure was not realized. The manure-contaminated
wastewater was harder to treat, and adequate treatment did not occur until a more favorable C:N
ratio was present. Consequently, the flexibility of the tested wetland design demonstrated
treatment can be customized to meet site-specific wash water characteristics. Further, greater
efficiencies can be realized as compared to the field system by reducing the size and/or
recirculation ratio.
Specific, significant findings are summarized below.


Nitrogen and COD are treatable with VFCW (96% and 88% respective removals for MF,
and 96% and 90% for MC during the last phase of treatment), but phosphorus removal is
limited in the long term (45% and 34% difference in removal rates from the first phase to
the last phase for MF and MC, respectively).



Influent phosphorus and nitrogen remained statistically constant during the duration of
the study, but a statistical significance difference in influent COD from summer to
winter.



Statistical analysis showed that influent COD and phosphorus were not significantly
different between MF and MC waste types, but influent nitrogen was approximately three
times more in the MC wash water, which was statistically significant (influent nitrogen
was 34 mg-N/L for MF versus 122 mg-N/L for MC).



Nitrogen removal worked best with a C:N ratio of 10:1.

100



Cell 1 was responsible for nitrification (1220% increase in nitrate for MF and 6610% for
MC) and removed the majority of COD (90% decrease in COD for MF, and 81% for
MC).



Cell 2 was responsible for denitrification (65% decrease in nitrate for MF, and 37% for
MC) and removed the majority of phosphorus (55% decrease in phosphorus for MF, and
54% for MC, during the first phase).



Cell 3 was found to be ineffective in most scenarios except for phosphorus removal (29%
decrease in phosphorus for MF, and 27% for MC).



Wetland performance between the two systems was divergent for nitrogen and COD
(93% and 86% respective removals for MF in first phase, versus 78% and 62% for MC),
but similar for phosphorus (74% removal for MF in first phase, versus 77% for MC).



There is evidence of a change in hydraulic conductivity with time, but the margin of error
for the test was too large to be sure. Regardless, this change was minimum (-4.0x104
m/s), indicating clogging was not significantly occurring.



Sizing a system for use will be significantly different depending on the presence of
manure contamination, due to the significant difference in influent nitrogen. If not for
nitrogen differences, the waste would be virtually identical, and, therefore, treatment
design would have been the same.



Sizing a MF wetland is controlled by hydraulic or COD loading, depending on site
conditions. The design criteria is 0.504 gal/day/ft2 and 1.06x10-2 lb-COD/day/ft2 for
hydraulic and COD loading, respectively.

101



Sizing a MC wetland is dependent on nitrogen loading. The design criteria is 1.03x10-3
lb-N/day/ft2 for nitrogen loading.

In order to determine if the objectives of the study were met, the objectives are re-printed
below for reference.
1. Collect the necessary data to develop an interim design standard, which is a
mechanism that allows the NRCS to field test new technology (NRCS 2009).
2. Create a treatment design that is more cost efficient than the demonstration field site.
3. Design for operational control flexibility for the technology to adapt for field specific
conditions or further optimization.
4. Determine if the characteristics of wash water (manure-free/manure-contaminated)
are significant enough to be considered separately for design purposes.
The first objective was met; the necessary data requested by the NRCS were influent and
effluent values for each nutrient, and sizing criteria, which were all found in the course of the
study. In terms of the second objective, it was easily achieved for the MF wash water, which
would require a wetland less than half the size of the field site. However, the MC wash water
was found to be effective at a hydraulic loading that was only slightly less than the current
loading at the field site. The third objective was successful as well; the design allows for
adjustment of the recycle ratio on-site to meet site-specific field conditions. Finally, for the last
objective it is necessary to design separately for MC and MF treatment wetlands.

While the system has proven to be effective, the design is still not optimized. Suggestions
for future study are below.
102



Testing the MF wetland under higher loadings until failure.



Experimenting with recycling directly back into the septic tank in order to denitrify
before the wash water enters the wetland.



Utilizing wood chips or another source of carbon in the second cell to test its effect on
improving denitrification.



Testing the system for its capacity to handle shock loading.



Optimizing the recycle ratio in the first cell.



Monitoring the development of clogging in the wetland cells by running more replicates.



Studying total solids at each stage of treatment in order to better document and
understand the progression of clogging in a multi-celled wetland. Measuring volatile
solids of the gravel would also help determine clogging potential.



Testing efficacy of system in different applications.



Experimenting with different technologies/substrate for more capacity for phosphorus
removal such as reactive media.



Further testing to determine the necessity of the third cell or if it can be used as an
alternating first cell to extend system life (by limiting clogging) effectively.

103

APPENDICES

104

APPENDIX A:

DETAILED RESULTS OF MICHIGAN FIELD SITE

105

Alkalinity (mg/L as CaCO3)
Cell 1
Cell 2
Cell 3
Removal
Date
Septic Alkalinity
%
Alkalinity
%
Alkalinity
%
for
Sampled
Tank
(mg/L) Removal (mg/L) Removal (mg/L) Removal System
6/16/2009
370
310
16%
508
-64%
337
34%
9%
7/14/2009
303
390
-29%
265
32%
267
-1%
12%
9/25/2009
238
285
-20%
293
-3%
278
5%
-17%
10/20/2009
335
395
-18%
370
6%
365
1%
-9%
12/3/2009
400
415
-4%
400
4%
400
0%
0%
12/28/2009
260
360
-38%
340
6%
335
1%
-29%
1/15/2010
280
320
-14%
350
-9%
375
-7%
-34%
1/21/2010
275
350
-27%
355
-1%
360
-1%
-31%
2/20/2010
335
390
-16%
355
9%
340
4%
-1%
3/6/2010
265
330
-25%
360
-9%
390
-8%
-47%
3/11/2010
275
325
-18%
375
-15%
380
-1%
-38%
3/27/2010
1045
940
10%
870
7%
790
9%
24%
4/10/2010
535
500
7%
485
3%
460
5%
14%
4/24/2010
420
430
-2%
390
9%
375
4%
11%
5/21/2010
560
500
11%
475
5%
440
7%
21%
6/4/2010
640
600
6%
470
22%
465
1%
27%
Std. Dev
207
160
17%
139
21%
118
9%
24%
Table 36: Field Site Alkalinity Removal

106

1200

1000

800

600

400

200

0

Septic Tank

Cell 1

Cell 2

Cell 3

Figure 37: Field Site Alkalinity Graph
Note: a load of milk was released into the system creating a visible peak in all of the graphs.
While milk is not recommended to put in the system, the recovery was remarkable.

107

Ammonia
Cell 1
Cell 2
Cell 3
Removal
Date
Septic Ammonia
%
Ammonia
%
Ammonia
%
for
Sampled Tank
(mg/L) Removal (mg/L) Removal (mg/L) Removal System
6/16/2009
23
100%
11
2
82%
7/14/2009
42
6
86%
2.5
58%
0.5
80%
99%
8/7/2009
14
3
79%
1.5
50%
0.5
67%
96%
9/25/2009
9.5
1.5
84%
1
33%
0
100%
100%
10/20/2009
13
6
54%
3.5
42%
3.5
0%
73%
11/12/2009
10
4
60%
1.5
63%
0
100%
100%
11/19/2009
16
13.5
16%
2.5
81%
2
20%
88%
12/18/2009
2.5
7.5
-200%
5
33%
0.5
90%
80%
12/28/2009
13
10.5
19%
6.5
38%
2
69%
85%
1/21/2010
7
4.5
36%
4.5
0%
2
56%
71%
2/20/2010
3.5
6
-71%
4
33%
1
75%
71%
3/6/2010
7
11
-57%
3.5
68%
0.5
86%
93%
3/11/2010
5
9.5
-90%
3
68%
0.5
83%
90%
3/27/2010
24
15
38%
7.5
50%
2
73%
92%
4/10/2010
18
11
39%
5
55%
0.5
90%
97%
4/24/2010
15
8
47%
4.5
44%
0.5
89%
97%
5/21/2010
19.7
6.1
69%
3.1
49%
0.8
75%
96%
6/4/2010
22.4
6.7
70%
2.3
66%
1.0
55%
95%
Std. Dev.
9.5
3.7
78%
2.4
19%
0.9
26%
10%
Table 37: Field Site Ammonia Removal

108

45
40
35

mg/L

30
25
20
15
10
5
0

Sample Date
Septic Tank

Cell 1

Cell 2

Cell 3

Figure 38: Field Site Ammonia Graph
Note: a load of milk was released into the system creating a visible peak in all of the graphs.
While milk is not recommended to put in the system, the recovery was remarkable.

109

BOD (mg/L)
Date Sampled Septic Tank Cell 1
Cell 2
Cell 3
2/20/2010
450
60
20 out of range
3/11/2010
570 out of range out of range
25
6/4/2010
726
25 under limit under limit
Table 38: Field Site BOD Measurements

Complications with BOD testing have (until recently) led to inaccurate readings.
Values from the last two reliable tests are shown above.

110

Chemical Oxygen Demand
Cell 1

Cell 2

Cell 3

Date
Septic COD
%
COD
%
COD
%
Sampled
Tank (mg/L) Removal (mg/L) Removal (mg/L) Removal
6/16/2009
1396
72
95%
220 -206%
54
75%
7/14/2009
1791
129
93%
36
72%
36
0%
8/7/2009
1219
67
95%
31
54%
27
13%
9/25/2009
1224
56
95%
44
21%
60
-36%
11/12/2009
1352
86
94%
62
28%
20
68%
11/19/2009
4052
431
89%
34
92%
19
44%
12/28/2009
2932
200
93%
69
66%
24
65%
1/15/2010
1072
35
8
77%
2/20/2010
1148
152
87%
72
53%
36
50%
3/6/2010
2032
436
79%
428
2%
68
84%
3/11/2010
1992
300
85%
78
74%
36
54%
3/27/2010 66450
7150
89%
3167
56%
375
88%
4/10/2010 30400
3724
88%
1524
59%
74
95%
4/24/2010 14700
1495
90%
423
72%
69
84%
5/21/2010
1527
147
90%
42
71%
19
55%
6/4/2010
1338
119
91%
35
71%
17
51%
Std. Dev.
17268
1958
4%
829
72%
87
35%
Table 39: Field Site COD Removal

111

Removal
for
System
96%
98%
98%
95%
99%
100%
99%
99%
97%
97%
98%
99%
100%
100%
99%
99%
1%

70000
60000

COD mg/L

50000
40000
30000
20000
10000
0

Septic Tank

Date
COD (mg/L)

Cell 2

Cell 3

Figure 39: Field Site COD Graph

Note: a load of milk was released into the system creating a visible peak in all of the graphs.
While milk is not recommended to put in the system, the recovery was remarkable.

112

Nitrate
Cell 1
Cell 2
Date
Septic Nitrate
%
Nitrate
%
Sampled
Tank
(mg/L) Removal (mg/L) Removal
6/16/2009
12.5
4.7
62%
1.8
62%
7/14/2009
20.5
0.27
99%
1.2
-344%
8/7/2009
30
7.5
75%
20
-167%
9/25/2009
19.2
3.5
82%
7.2
-106%
11/12/2009
15.2
3.6
76%
2.6
28%
11/19/2009
12
4
67%
2.2
45%
1/15/2010
67.8
2.6
96%
2.2
15%
3/27/2010
31.2
7.6
76%
3.2
58%
4/10/2010
24.3
4.2
83%
3.3
21%
4/24/2010
19
3.9
79%
2.5
36%
5/21/2010
25
5.2
79%
2.8
46%
6/4/2010
24
5
79%
12
-140%
Std. Dev.
14.8
2.0
10%
5.6
126%
Table 40: Field Site Nitrate Removal

113

Cell 3
Removal
Nitrate
%
for
(mg/L) Removal System
6.6
-267%
47%
2.8
-133%
86%
13
35%
57%
0.6
92%
97%
2.6
0%
83%
2.8
-27%
77%
1.6
27%
98%
2.9
9%
91%
1.8
45%
93%
1.7
32%
91%
2.6
7%
90%
16
-33%
33%
4.9
96%
21%

80
70

mg/L NO3

60
50
40
30
20
10
0

Date of Sample
Septic Tank

Cell 1

Cell 2

Cell 3

Figure 40: Field Site Nitrate Graph
Note: a load of milk was released into the system creating a visible peak in all of the graphs.
While milk is not recommended to put in the system, the recovery was remarkable.

114

pH
Date
Septic
Sampled
Tank
Cell 1
Cell 2
6/16/2009
6.1
7.09
6.86
7/14/2009
5.94
7.1
6.85
8/7/2009
6.23
6.76
7.42
9/25/2009
6.36
7.37
7.4
10/20/2009
6.62
6.91
7.24
11/12/2009
6.77
7.03
7.37
11/19/2009
6.01
6.84
7.33
12/3/2009
6.49
6.83
7.37
12/28/2009
6.29
6.99
7.4
1/15/2010
6.66
6.86
7.09
1/21/2010
6.66
7.11
6.94
2/20/2010
7.64
7.61
7.71
3/6/2010
6.4
6.6
7.13
3/11/2010
6.5
6.52
7.04
3/27/2010
6.7
6.72
6.84
4/10/2010
6.7
6.83
7.02
4/24/2010
6.54
6.78
7.12
5/21/2010
6.35
6.94
7.3
6/4/2010
6.29
6.79
7.29
Std. Dev
0.37
0.26
0.24
Table 41: Field Site pH Measurements

Cell 3
7.17
7.17
7.21
7.61
7.4
7.43
7.57
7.48
7.42
7.2
7.15
7.74
7.3
7.2
6.92
7.31
7.4
7.34
7.2
0.19

115

8
7.5

pH

7
6.5
6
5.5
5

Sample Date
Septic Tank

Cell 1

Cell 2

Cell 3

Figure 41: Field Site pH Graph
Note: a load of milk was released into the system creating a visible peak in all of the graphs.
While milk is not recommended to put in the system, the recovery was remarkable.

116

Phosphorus
Cell 1
Cell 2
Cell 3
Removal
Date
Septic Phos.
%
Phos.
%
Phos.
%
for
Sampled Tank (mg/L) Removal (mg/L) Removal (mg/L) Removal System
6/16/2009 27.17
4.3
84%
7.36
-71%
1.5
80%
94%
7/14/2009
33.5
14.5
57%
1.26
91%
0
100%
100%
9/25/2009
33
11.5
65%
7
39%
3
57%
91%
10/20/2009
23
0.5
98%
6.5 -1200%
2.5
62%
89%
11/12/2009
33.5
16.5
51%
7
58%
4.5
36%
87%
11/19/2009
34
15.5
54%
5.5
65%
4
27%
88%
12/3/2009
26
35.5
-37%
3
92%
3
0%
88%
12/18/2009
29.5
19
36%
18
5%
6
67%
80%
12/28/2009
31.5
15.5
51%
36.5
-135%
15
59%
52%
1/21/2010
27
9
67%
12
-33%
8.5
29%
69%
2/20/2010
35
15
57%
11.5
23%
9
22%
74%
3/6/2010
45.5
16.5
64%
15
9%
7.5
50%
84%
3/11/2010
36
15.5
57%
12.5
19%
7
44%
81%
3/27/2010
52
19.5
63%
14.5
26%
9.5
34%
82%
4/10/2010
43.5
17
61%
13
24%
8.5
35%
80%
4/24/2010
37
15.5
58%
11.5
26%
8
30%
78%
5/21/2010
42
15.8
62%
6.2
61%
1.5
76%
96%
6/4/2010
116
20.6
82%
4.1
80%
0
100%
100%
Std. Dev.
20.5
7.1
27%
7.9
294%
4.0
27%
12%
Table 42: Field Site Phosphorus Removal

117

140
120

mg/L

100
80
60
40
20
0

Septic Tank

Date
Cell 1

Cell 2

Cell 3

Figure 42: Field Site Phosphorus Graph
Note: a load of milk was released into the system creating a visible peak in all of the graphs.
While milk is not recommended to put in the system, the recovery was remarkable.

118

Total Solids (TS) and Volatile Solids (VS)
Date

6/16/2009
9/25/2009
11/19/2009
12/18/2009
12/28/2009
2/20/2010
3/6/2010
3/11/2010
3/27/2010
4/10/2010
4/24/2010
5/21/2010
Std. Dev

TS
ST
Cell 1
Cell 2
Cell 3
(mg/L) (mg/L) %Reduction (mg/L) %Reduction (mg/L) %Reduction
1443
560
61.2%
820
-46.4%
585
29%
1237
736
40.5%
688
6.5%
700
-2%
2924
972
66.8%
732
24.7%
728
1%
1196
668
44.1%
662
0.9%
644
3%
2383
860
63.9%
634
26.3%
604
5%
1150
1184
-3.0%
632
46.6%
562
11%
5500
1616
70.6%
1996
-23.5%
1644
18%
3188
1672
47.6%
1878
-12.3%
1584
16%
30468
10466
65.6%
2956
71.8%
1675
43%
16425
4025
75.5%
1994
50.5%
1064
47%
5984
1246
79.2%
624
49.9%
710
-14%
1560
640
59.0%
354
44.7%
240
32%
8778
2810
22%
820
35%
482
19%

Removal
for
System
59%
43%
75%
46%
75%
51%
70%
50%
95%
94%
88%
85%
19%

Table 43: Field Site TS Removal
Date

VS
Removal
ST
Cell 1
Cell 2
Cell 3
for
(mg/L) (mg/L) %Reduction (mg/L) %Reduction (mg/L) %Reduction System
6/16/2009
704
97
86%
180
-86%
77
57%
89%
9/25/2009
600
118
122
-3%
11/19/2009
1268
324
74%
212
35%
180
15%
86%
12/18/2009
684
252
63%
164
35%
218
-33%
68%
12/28/2009
1723
236
86%
170
28%
154
9%
91%
2/20/2010
696
658
5%
268
59%
294
-10%
58%
3/6/2010
4474
536
88%
1016
-90%
634
38%
86%
3/11/2010
1936
448
77%
208
54%
568
-173%
71%
3/27/2010
24506
8046
67%
2266
72%
784
65%
97%
4/10/2010
11640
2402
79%
1009
58%
626
38%
95%
4/24/2010
5046
1142
77%
324
72%
420
-30%
92%
5/21/2010
768
325
58%
180
45%
188
-4%
76%
Std. Dev
7295
2226
23%
637
58%
241
62%
13%

Table 44: Field Site VS Removal

119

35000
30000

TS (mg/L)

25000
20000
15000
10000
5000
0

Date
ST

Cell 1

Figure 43: Field Site TS Graph

120

Cell 2

Cell 3

30000
25000

VS (mg/L)

20000
15000
10000
5000
0

Date
ST

Cell 1

Cell 2

Figure 44: Field Site VS Graph

121

Cell 3

APPENDIX B:

MONITORING SHEET

122

Figure 45: Monitoring Sheet (Front)

123

Figure 46: Monitoring Sheet (Back)

124

APPENDIX C:

TEMPERATURE MONITORING DATA

125

Date
Influent
7/31/2013
--8/1/2013
--8/2/2013
39
8/4/2013
39
8/5/2013
38
8/6/2013
38
8/7/2013
38
8/9/2013
38
8/10/2013
40
8/11/2013
36
8/12/2013
38
8/13/2013
36
8/14/2013
36
8/15/2013
35
8/16/2013
35
8/18/2013
36
8/19/2013
36
8/21/2013
36
8/22/2013
36
8/23/2013
38
8/24/2013
39
8/25/2013
36
8/26/2013
38
8/27/2013
36
8/28/2013
36
8/29/2013
37
8/30/2013
37
8/31/2013
38
9/1/2013
38
9/4/2013
37
9/5/2013
36
9/6/2013
35
9/7/2013
33
9/8/2013
35
9/9/2013
36
9/10/2013
33
9/13/2013
33
9/14/2013
38

Manure Free
Cell 1
Cell 2
Cell 3
73
73
72
73
72
73
43.5
43
43
43
43
43
43
43
42
43
42
43
43
42
42
42
41
41
43
43
43
42
41
42
42
41
42
41
40
40
42
41
41
41
40
41
43
42
42
43
41
42
41
40
40
39
38
40
39
38
39
42
41
44
43
42
42
42
40
41
42
41
42
40
40
40
40
41
41
41
41
42
42
40
42
41
41
42
41
41
41
41
40
41
40
40
41
39
39
39
37
36
36
39
39
39
42
40
41
40
37
38
36
33
34
42
42
42

Table 45: Temperature Monitoring Data

126

Manure Contaminated
Cell 1
Cell 2
Cell 3
73
72
73
73
72
73
44
43
43
43
42
43
43
43
42
43
43
42
43
43
42
42
42
41
43
44
43
42
41
40
42
41
40
40
40
39
41
41
40
42
41
41
42
42
40
42
40
41
40
40
39
39
39
39
38
38
38
41
41
42
42
42
42
42
42
42
42
41
41
41
40
40
42
41
40
42
41
41
42
40
40
42
42
41
42
41
41
41
40
40
41
41
41
40
39
39
37
36
36
40
39
39
42
41
41
39
37
37
35
33
34
43
42
42

Table 45 (cont’d)
Date
Influent
9/15/2013
40
9/16/2013
33
9/18/2013
37
9/19/2013
39
9/20/2013
38
9/21/2013
38
9/23/2013
39
9/24/2013
40
9/25/2013
39
9/26/2013
40
9/27/2013
36
9/28/2013
39
9/29/2013
38
9/30/2013
37
10/1/2013
38
10/3/2013
36
10/4/2013
36
10/6/2013
35
10/7/2013
36
10/8/2013
35
10/9/2013
35
10/10/2013
33
10/11/2013
37
10/22/2013
39
10/23/2013
36
10/24/2013
39
10/25/2013
35
10/27/2013
35
10/28/2013
33
10/29/2013
36
10/30/2013
33
10/31/2013
38
11/1/2013
42
11/3/2013
37
11/4/2013
35
11/5/2013
38
11/7/2013
38
11/8/2013
46

Manure Free
Cell 1
Cell 2
Cell 3
48
43
44
37
35
36
40
40
40
45
44
45
45
43
45
43
43
45
44
43
44
45
44
44
44
43
43
43
43
41
41
39
39
43
43
44
42
42
43
41
40
41
42
41
41
38
37
37
40
40
41
36
36
36
39
39
39
37
37
37
37
37
37
37
36
37
40
41
41
43
41
42
38
37
38
43
41
42
45
37
38
38
37
38
36
35
35
41
40
40
37
35
36
44
41
41
45
44
45
40
38
39
39
38
39
39
39
40
41
41
40
49
46
48

127

Manure Contaminated
Cell 1
Cell 2
Cell 3
45
44
44
37
37
36
40
40
40
45
45
45
45
45
45
46
46
44
44
43
44
44
44
44
44
43
43
43
43
43
40
39
39
44
43
44
43
43
43
41
40
40
42
41
40
39
37
37
41
41
44
37
36
36
40
40
40
38
38
38
38
37
37
37
36
36
41
41
41
43
41
43
39
37
38
43
42
44
39
37
38
38
37
38
36
34
36
41
41
40
37
36
45
42
41
42
45
45
45
40
39
39
40
39
39
40
39
39
42
42
42
48
46
50

Table 45 (cont’d)

Date
Influent
11/11/2013
41
11/12/2013
41
11/13/2013
37
11/14/2013
37
11/15/2013
36
11/17/2013
40
11/18/2013
39
11/19/2013
39
11/20/2013
36
11/21/2013
34
11/22/2013
36
11/26/2013
44
11/27/2013
44
11/28/2013
44
11/29/2013
45
11/30/2013
46
12/1/2013
46
12/2/2013
46
12/3/2013
43
12/6/2013
44

Manure Free
Cell 1
Cell 2
Cell 3
45
44
45
48
47
50
41
40
40
49
4
41
40
39
40
44
45
46
46
43
45
41
41
40
40
38
40
37
38
37
40
39
39
41
38
39
42
39
41
40
39
40
40
39
40
40
38
39
42
38
40
41
39
41
41
39
40
42
38
42

128

Manure Contaminated
Cell 1
Cell 2
Cell 3
46
45
46
48
47
49
41
41
41
41
40
42
42
40
43
46
45
45
44
42
45
39
40
41
39
38
39
38
38
38
40
39
40
40
38
39
43
41
43
43
41
39
43
39
40
39
39
40
40
38
40
41
39
41
40
39
41
42
39
41

APPENDIX D:

ANALYTICAL RECORDING SHEET

129

Figure 47: Analytical Recording Sheet

130

APPENDIX E:

HYDRAULIC CONDUCTIVITY CALCULATION

131

Darcy’s law:
𝑄 = −𝐾 ∗ 𝑖 ∗ 𝐴
Where:
𝑄 = 𝑑𝑖𝑠𝑐ℎ𝑎𝑟𝑔𝑒
𝐾 = ℎ𝑦𝑑𝑟𝑎𝑢𝑙𝑖𝑐 𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑖𝑣𝑖𝑡𝑦
𝑖 = ℎ𝑦𝑑𝑟𝑎𝑢𝑙𝑖𝑐 𝑔𝑟𝑎𝑑𝑖𝑒𝑛𝑡 =

∆𝐻
∆𝑙

∆𝐻 = 𝑐ℎ𝑎𝑛𝑔𝑒 𝑖𝑛 ℎ𝑒𝑎𝑑
∆𝑙 = 𝑙𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑓𝑙𝑜𝑤 𝑝𝑎𝑡ℎ
𝐴 = 𝑐𝑟𝑜𝑠𝑠 𝑠𝑒𝑐𝑡𝑖𝑜𝑛𝑎𝑙 𝑎𝑟𝑒𝑎

Figure 48: Location of Darcy’s Law Measurements
Testing event 1:
Trial

Volume
Time (s) Q (ml/s)
(ml)
1
225
10
22.5
2
210
10
21
3
205
10
20.5
4
200
10
20
5
210
10
21
AVG 210
10
21
Table 46: Trial 1 for Hydraulic Conductivity Measurement
𝐴 = 𝜋𝑟 2 = 𝜋22 = 12.57 𝑖𝑛2 = 81.07 𝑐𝑚2
𝑖=

∆𝐻 26
=
= 0.84
∆𝑙
31
132

−𝑄

3
−21𝑐𝑚 ⁄

𝑠
𝑐𝑚⁄ = - 3.1E-03 m/s
𝐾 = 𝑖∗𝐴 = 0.84∗81.07𝑐𝑚
𝑠
2 = −0.31

Testing event 2:
Trial

Volume
Time (s) Q (ml/s)
(ml)
1
190
10
19
2
180
10
18
3
185
10
18.5
4
170
10
17
5
180
10
18
6
175
10
17.5
7
185
10
18.5
AVG 181
10
18.1
Table 47: Trial 2 for Hydraulic Conductivity Measurement
−𝑄

3
−18.1𝑐𝑚 ⁄

𝐾 = 𝑖∗𝐴 = 0.84∗81.07𝑐𝑚𝑠 2 = −0.27 𝑐𝑚⁄𝑠 = -2.7E-03 m/s

∆𝐾 = −3.1𝑥10−3 − (−2.7𝑥10−3 ) = −4.0𝑥10−4

𝑚
𝑠

133

APPENDIX F:

WETLAND FIELD SITE CLOGGING EXAMINATION

134

At the field wetland site, the gravel bed was examined for clogging. Results are shown below.
From surface to 2 ft depth:
Gravel was extremely loose and could not hold shape. No buildup of material, organic or
otherwise, could be seen or felt.

Figure 49: Picture of Field Site Media

135

Greater than 2 ft (0.6 m) below surface:
Note: at this depth, gravel comes into contact with both summer and winter wash water
applications. As a result, the media is often wet and sometimes saturated. Gravel was black in
color due to microbial activity. The gravel was not clumpy, indicating the wetland is still
operating without any signs of clogging.

Figure 50: Picture of Sampling Hole

136

APPENDIX G:

ANALYTICAL RESULTS

137

Table 48: MF Wetland Analytical Data

138

Table 48 (cont’d)

139

Table 49: MC Wetland Analytical Data

140

Table 49 (cont’d)

141

APPENDIX H:

QAQC

142

The relative percent difference between a random set of replicates was calculated for
each testing event. Results of this practice are summarized in the table below, separated by phase
and nutrient. The COD test had the most variability, something the researchers noticed and tried
repeatedly to remediate, to little avail. Total nitrogen also had some variability, probably due to
the difficult and complicated testing procedure associated with the nutrient.

Phase 1 Phase 2 Phase 3 Phase 4
Phosphorus
6.7
4.5
9.4
11.6
COD
29.9
8.6
36.9
42.4
Total Nitrogen
8.9
8.1
21.0
13.3
Nitrate
10.2
3.7
4.4
4.6
Ammonia
3.6
4.3
4.9
0.8
pH
2.3
2.7
3.2
4.7
AVE
10.3
5.3
13.3
12.9
Table 50: Relative Percent Difference between Replicates

AVE
8.1
29.4
12.9
5.7
3.4
3.2
10.4

The relative percent difference between a tested standard and its supposed value was
calculated for each testing event. Results of this practice are summarized in the table below,
separated by phase and nutrient.

Phase 1 Phase 2 Phase 3 Phase 4
8.6
6.0
10.3
8.9
23.7
12.5
25.2
34.8

Phosphorus
COD
Total
7.6
5.3
2.4
2.8
Nitrogen
Nitrate
5.5
3.9
7.7
2.6
Ammonia
4.6
8.2
9.5
10.8
10.0
7.2
11.0
12.0
AVE
Table 51: Relative Percent Difference for Standards

143

AVE
8.4
24.0
4.5
4.9
8.3
10.0

APPENDIX I:

STATISTICS RESULTS

144

Phosphorus: Overall
Phosphorus~factor(StageC)+factor(Phase)+WD+(S0+S1|
Equation
Date)
AIC
BIC
Model Fit
1213.691 1269.505
Estimate
Std. Error t-value
(Intercept)
25.7865
1.6982 15.185
factor(StageC)1
-5.9354
0.9757 -6.083
factor(StageC)2
-12.865
1.3913 -9.247
factor(StageC)3
-17.8519
1.3887 -12.855
Fixed Effects factor(Phase)1
6.5609
1.3612
4.82
factor(Phase)2
3.7249
1.6546
2.251
factor(Phase)3
7.381
1.7916
4.12
factor(Phase)4
25.5832
1.8291 13.987
factor(Phase)5
10.0824
1.7233
5.851
WD
-0.8342
0.8658 -0.963
Variance Std.Dev.
Random
(Intercept)
2.83
1.682
Effects
S0
32.05
5.661
S1
15.78
3.973
Table 52: Statistical Data for Overall Models

145

COD: Overall
COD~factor(StageC)+factor(Phase)+WD+(S0|Date)
AIC

BIC
2374.012 2419.977
Estimate
Std. Error t value
(Intercept)
666.859
77.124
8.647
factor(StageC)1
-572.206
76.22 -7.507
factor(StageC)2
-618.248
76.196 -8.114
factor(StageC)3
-616.503
76.196 -8.091
factor(Phase)1
-1.737
19.227
-0.09
factor(Phase)2
-47.919
25.477 -1.881
factor(Phase)3
30.644
28.298
1.083
factor(Phase)4
-20.094
28.72
-0.7
factor(Phase)5
47.26
26.453
1.787
WD
17.344
16.672
1.04
Variance Std.Dev.
(Intercept)
25.04
5.004
S0
164154
405.159

Table 52 (cont’d)

Equation

Total Nitrogen: Overall
Nitrogen~factor(StageC)+factor(Phase)+WD+(1|Date)

BIC
AIC
BIC
1683.536 1722.689
1678.613 1718.252
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
48.9773
4.1435
11.82 (Intercept)
-10.338
3.219 -3.211
factor(StageC)1
-33.0079
4.1781
-7.9 factor(StageC)1
30.983
3.319
9.334
factor(StageC)2
-49.8637
4.1781 -11.934 factor(StageC)2
17.907
3.336
5.367
factor(StageC)3
-56.7243
4.1781 -13.577 factor(StageC)3
17.797
3.355
5.305
Fixed Effects factor(Phase)1
1.1778
3.8439
0.306 factor(Phase)1
-2.736
3.17 -0.863
factor(Phase)2
1.3154
5.3504
0.246 factor(Phase)2
-3.796
4.318 -0.879
factor(Phase)3
-30.132
5.9889 -5.031 factor(Phase)3
-27.938
4.99 -5.599
factor(Phase)4
0.8139
6.1191
0.133 factor(Phase)4
-3.676
4.977 -0.739
factor(Phase)5
-25.9764
5.5296 -4.698 factor(Phase)5
-29.902
4.608
-6.49
WD
66.6631
3.8214 17.445 WD
37.502
3.012 12.449
Variance Std.Dev.
Variance Std.Dev.
Random
(Intercept)
0
0
(Intercept)
3.287
1.813
Effects
Model Fit

AIC

Nitrate: Overall
Nitrate~factor(StageC)+factor(Phase)+WD+(1|Date)

146

Table 52 (cont’d)

Equation

Ammonia: Overall
Ammonia~factor(StageC)+factor(Phase)+WD+(1|Date)

BIC
AIC
BIC
1744.307 1783.459
52.52219 97.83238
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
48.5236
4.7271 10.265 (Intercept)
7.33624
0.06138 119.52
factor(StageC)1
-60.0883
4.9591 -12.117 factor(StageC)1
0.0196
0.06051
0.32
factor(StageC)2
-56.8789
4.9582 -11.472 factor(StageC)2
0.49802
0.06051
8.23
factor(StageC)3
-63.3806
4.9333 -12.848 factor(StageC)3
0.631
0.06051
10.43
Fixed Effects factor(Phase)1
5.5216
4.6393
1.19 factor(Phase)1
0.14564
0.05928
2.46
factor(Phase)2
2.86
6.1311
0.466 factor(Phase)2
0.0473
0.07215
0.66
factor(Phase)3
3.9065
7.0829
0.552 factor(Phase)3
0.10077
0.08084
1.25
factor(Phase)4
0.9849
7.0627
0.139 factor(Phase)4
-0.12484
0.09225
-1.35
factor(Phase)5
10.1829
6.5415
1.557 factor(Phase)5
0.05854
0.09225
0.63
WD
26.7622
4.5004
5.947 WD
-0.04796
0.03927
-1.22
Variance Std.Dev.
Variance Std.Dev.
Random
(Intercept)
0
0
(Intercept)
0.01037
0.1018
Effects
S0
0.04641
0.2154
Model Fit

AIC

pH: Overall
pH~factor(StageC)+factor(Phase)+WD+(S0|Date)

147

Phosphorus: Manure Free

COD: Manure Free
COD~factor(StageC)+factor(Phase)+(S0|Date)

Equation

Phosphorus~factor(StageC)+factor(Phase)+(S0+S1|Date)
AIC
BIC
Model Fit
632.7169 669.3286
Estimate
Std. Error t value
(Intercept)
27.5631
2.4893 11.072
factor(StageC)2
-5.506
1.2879 -4.275
factor(StageC)3
-12.4027
2.1304 -5.822
Fixed Effects
factor(StageC)4
-18.0877
2.1242 -8.515
factor(Phase)2
3.5286
1.9136
1.844
factor(Phase)3
0.2272
1.999
0.114
factor(Phase)4
23.7845
2.2725 10.466
Variance Std.Dev.
Random
(Intercept)
16.76
4.094
Effects
S0
91.78
9.58
S1
46.76
6.838
Table 53: Statistical Data for Manure Free Models

148

AIC

BIC
1065.767 1094.534
Estimate
Std. Error t value
(Intercept)
652.85
69.29
9.421
factor(StageC)2
-575.99
68.91 -8.359
factor(StageC)3
-608.39
68.9
-8.83
factor(StageC)4
-603.95
68.9 -8.766
factor(Phase)2
-15.27
14.14 -1.081
factor(Phase)3
-20.57
14.77 -1.393
factor(Phase)4
10.87
16.8
0.647
Variance Std.Dev.
(Intercept)
615.4
24.81
S0
127179.3
356.62

Table 53 (cont’d)
Total Nitrogen: Manure Free
Nitrogen~factor(StageC)+factor(Phase)+(S0|Date)
Equation

BIC
AIC
BIC
617.2597 645.5815
565.083 602.2385
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
35.77
2.329 15.361 (Intercept)
0.52543
0.48261
1.089
factor(StageC)2
-20.245
2.125 -9.528 factor(StageC)2
8.58046
1.57933
5.433
factor(StageC)3
-26.879
2.125 -12.648 factor(StageC)3
2.54952
1.06283
2.399
Fixed Effects
factor(StageC)4
-28.047
2.125 -13.197 factor(StageC)4
2.93415
1.06603
2.752
factor(Phase)2
-2.35
1.929 -1.218 factor(Phase)2
0.21093
0.74695
0.282
factor(Phase)3
-2.602
2.01 -1.295 factor(Phase)3
0.02354
0.78269
0.03
factor(Phase)4
-3.536
2.27 -1.558 factor(Phase)4
0.58699
0.89665
0.655
Variance Std.Dev.
Variance Std.Dev.
Random
(Intercept)
7.155
2.675
(Intercept)
27.003
5.196
Effects
S0
73.828
8.592
S0
24.98
4.998
S1
47.467
6.89

Model Fit

AIC

Nitrate: Manure Free
Nitrate~factor(StageC)+factor(Phase)+(S0+S1|Date)

149

Table 53 (cont’d)
Ammonia: Manure Free
Ammonia~factor(StageC)+factor(Phase)+(S0+S1|Date)
Equation

BIC
AIC
BIC
473.0134
509.625
21.51817 57.99056
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
18.3659
1.0888 16.868 (Intercept)
7.19009
0.07986
90.04
factor(StageC)2
-16.8573
1.0603 -15.898 factor(StageC)2
0.2236
0.09209
2.43
factor(StageC)3
-15.2911
1.0448 -14.636 factor(StageC)3
0.6042
0.08654
6.98
Fixed Effects
factor(StageC)4
-16.7799
1.0448 -16.061 factor(StageC)4
0.6736
0.09209
7.31
factor(Phase)2
-1.3529
0.8123 -1.665 factor(Phase)2
0.25722
0.06531
3.94
factor(Phase)3
-0.8406
0.8123 -1.035 factor(Phase)3
0.1105
0.0683
1.62
factor(Phase)4
-0.7723
0.9303
-0.83 factor(Phase)4
-0.04358
0.08639
-0.5
Variance Std.Dev.
Variance Std.Dev.
Random
(Intercept)
5.908
2.431
(Intercept)
0.02397
0.1548
Effects
S0
24.715
4.971
S0
0.18474
0.4298
S1
3.819
1.954
S2
0.02519
0.1587

Model Fit

AIC

pH: Manure Free
pH~factor(StageC)+factor(Phase)+(S0+S2|Date

150

Phosphorus: Manure Contaminated

COD: Manure Contaminated
COD~factor(StageC)+factor(Phase)+(S0+S3|Date)

Equation

Phosphorus~factor(StageC)+factor(Phase)+(S2+S3|Date)
AIC
BIC
Model Fit
493.6355 529.5364
Estimate
Std. Error t value
(Intercept)
25.0811
1.3432 18.673
factor(StageC)2
-6.5107
0.5273 -12.348
factor(StageC)3
-13.4173
1.2736 -10.535
Fixed Effects
factor(StageC)4
-17.9348
1.1388 -15.749
factor(Phase)2
7.167
1.171
6.12
factor(Phase)4
8.4994
1.3415
6.336
factor(Phase)6
10.5295
1.242
8.478
Variance Std.Dev.
Random
(Intercept)
31.401
5.604
Effects
S2
31.575
5.619
S3
23.442
4.842
Table 54: Statistical Data for Manure Contaminated Models

151

AIC

BIC
1098.29
1134.19
Estimate
Std. Error t value
(Intercept)
758.96
107.9
7.034
factor(StageC)2
-604.44
98.88 -6.113
factor(StageC)3
-671.11
99.08 -6.773
factor(StageC)4
-669.85
106.13 -6.311
factor(Phase)2
10.81
15.5
0.697
factor(Phase)4
-43.6
17.66 -2.469
factor(Phase)6
-44.44
16.4
-2.71
Variance Std.Dev.
(Intercept)
2259
47.53
S0
234949
484.72
S3
1447
38.04

Table 54 (cont’d)
Total Nitrogen: Manure Contaminated
Nitrogen~factor(StageC)+factor(Phase)+(S0+S3|Date)
Equation

BIC
AIC
BIC
846.5426 882.4435
787.7519 823.6527
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
129.059
6.027 21.414 (Intercept)
4.381
2.536
1.728
factor(StageC)2
-46.429
7.485 -6.203 factor(StageC)2
56.185
7.027
7.995
factor(StageC)3
-73.288
7.485 -9.791 factor(StageC)3
35.127
5.118
6.863
Fixed Effects
factor(StageC)4
-85.842
7.783 -11.029 factor(StageC)4
34.23
5.118
6.688
factor(Phase)2
3.497
6.442
0.543 factor(Phase)2
-1.193
3.621 -0.329
factor(Phase)4
-24.839
7.345 -3.382 factor(Phase)4
-8.263
4.129 -2.001
factor(Phase)6
-18.251
6.817 -2.677 factor(Phase)6
-8.909
3.832 -2.325
Name
Variance Std.Dev.
Variance Std.Dev.
Random
(Intercept)
399.4
19.98
(Intercept)
380.14
19.497
Effects
S0
1052.7
32.45
S0
492.26
22.187
S3
713.7
26.72
S1
713.46
26.711

Model Fit

AIC

Nitrate: Manure Contaminated
Nitrate~factor(StageC)+factor(Phase)+(S0+S1|Date)

152

Table 54 (cont’d)
Ammonia: Manure Contaminated
Ammonia~factor(StageC)+factor(Phase)+(S0+S2|Date)
Equation

BIC
727.2287 762.5338
Estimate
Std. Error
(Intercept)
113.248
6.042
factor(StageC)2
-108.47
5.829
factor(StageC)3
-102.31
8.445
Fixed Effects
factor(StageC)4
-113.844
5.812
factor(Phase)2
7.562
2.645
factor(Phase)4
-1.525
2.85
factor(Phase)6
4.579
2.645
Variance Std.Dev.
Random
(Intercept)
3.611
1.9
Effects
S0
692.791
26.321
S2
509.172
22.565

Model Fit

AIC

pH: Manure Contaminated
pH~factor(StageC)+factor(Phase)+(1|Date)
AIC

BIC
10.3383 32.63433
t value
Estimate
Std. Error t value
18.743 (Intercept)
7.41925
0.05748 129.06
-18.609 factor(StageC)2
-0.22232
0.05649
-3.94
-12.115 factor(StageC)3
0.36727
0.05649
6.5
-19.587 factor(StageC)4
0.5725
0.05649
10.13
2.858 factor(Phase)2
0.09406
0.07259
1.3
-0.535 factor(Phase)4
0.07451
0.08277
0.9
1.731 factor(Phase)6
0.03139
0.09182
0.34
Variance Std.Dev.
(Intercept)
0.0102
0.101
0.0351
0.1873

153

Phosphorus: Influent
COD: Influent
Equation
Phosphorus~factor(Phase)+WD+(1|Date)
COD~factor(Phase)+WD+(1|Date)
AIC
BIC
AIC
BIC
Model Fit
325.25715 342.09796
637.80529 655.19172
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
31.4079
2.7559 11.397 (Intercept)
511.11
70.62
7.238
factor(Phase)1
-0.7818
3.4166 -0.229 factor(Phase)1
-70.83
95.84 -0.739
factor(Phase)2
-4.1877
4.5085 -0.929 factor(Phase)2
179.14
121.31
1.477
Fixed Effects
factor(Phase)3
0.449
5.1564
0.087 factor(Phase)3
564.33
139.82
4.036
factor(Phase)4
13.9567
5.1604
2.705 factor(Phase)4
766.61
139.09
5.512
factor(Phase)5
-1.1817
4.7825 -0.247 factor(Phase)5
1085.96
130.46
8.324
WD
-0.3357
3.0832 -0.109 WD
-81.86
74.44
-1.1
Random
Variance Std.Dev.
Variance Std.Dev.
Effects
(Intercept)
9.774
3.126
(Intercept)
16555
128.7
Table 55: Statistical Data for Influent Models

154

Table 55 (cont’d)
Total Nitrogen: Influent
Nitrate: Influent
Equation
Nitrogen~factor(Phase)+WD+(1|Date)
Nitrate~factor(Phase)+WD+(1|Date)
AIC
BIC
AIC
BIC
Model Fit
382.96135 399.98773
100.40916 117.79559
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
30.387
5.043
6.026 (Intercept)
0.4563
0.1649
2.767
factor(Phase)1
2.41
6.224
0.387 factor(Phase)1
0.1969
0.2286
0.862
factor(Phase)2
7.372
8.237
0.895 factor(Phase)2
0.2068
0.279
0.741
Fixed Effects
factor(Phase)3
-6.857
9.309 -0.737 factor(Phase)3
0.18
0.3172
0.568
factor(Phase)4
9.106
9.424
0.966 factor(Phase)4
0.8214
0.3172
2.59
factor(Phase)5
17.596
8.623
2.041 factor(Phase)5
0.4221
0.2995
1.41
WD
88.618
5.464 16.218 WD
0.2506
0.1537
1.63
Random
Variance Std.Dev.
Variance Std.Dev.
Effects
(Intercept)
35.8
5.983
(Intercept)
0.1252
0.3538

155

Table 55 (cont’d)
Ammonia: Influent
pH: Influent
Equation
Ammonia~factor(Phase)+WD+(1|Date)
pH~factor(Phase)+WD+(1|Date)
AIC
BIC
AIC
BIC
Model Fit
394.13405 410.97486
39.48399 56.13532
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
14.3034
6.1151
2.339 (Intercept)
7.35015
0.08807
83.45
factor(Phase)1
10.3058
7.9555
1.295 factor(Phase)1
0.09269
0.12015
0.77
factor(Phase)2
5.1883
10.2681
0.505 factor(Phase)2
-0.21099
0.14569
-1.45
Fixed Effects
factor(Phase)3
22.5861
11.8015
1.914 factor(Phase)3
0.01431
0.16296
0.09
factor(Phase)4
-0.5338
11.8045 -0.045 factor(Phase)4
-0.35015
0.18627
-1.88
factor(Phase)5
13.4286
10.9107
1.231 factor(Phase)5
-0.05557
0.18627
-0.3
WD
91.7881
7.2831 12.603 WD
0.09208
0.07826
1.18
Random
Variance Std.Dev.
Variance Std.Dev.
Effects
(Intercept)
29.67
5.447
(Intercept)
0.03693
0.1922

156

Phosphorus: Cell 1
COD: Cell 1
Equation
Phosphorus~factor(Phase)+WD+(1|Date)
COD~factor(Phase)+WD+(1|Date)
AIC
BIC
AIC
BIC
Model Fit
288.44411 305.28492
484.09289 500.9337
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
25.4019
1.8272 13.902 (Intercept)
60.75
20.55
2.956
factor(Phase)2
-1.5448
2.386 -0.647 factor(Phase)2
-6.3
25.67 -0.245
factor(Phase)3
-5.6978
2.9641 -1.922 factor(Phase)3
-11.83
32.55 -0.364
Fixed Effects
factor(Phase)4
-0.2986
3.3421 -0.089 factor(Phase)4
36.81
35.85
1.027
factor(Phase)5
20.1516
3.3624
5.993 factor(Phase)5
54.32
36.85
1.474
factor(Phase)6
1.4562
3.1425
0.463 factor(Phase)6
78.87
33.57
2.35
WD
-1.1705
1.7874 -0.655 WD
58.98
19.82
2.976
Random
Variance Std.Dev.
Variance Std.Dev.
Effects
(Intercept)
11.55
3.398
(Intercept)
1220
34.94
Table 56: Statistical Data for Cell 1 Models

157

Table 56 (cont’d)
Total Nitrogen: Cell 1
Nitrate: Cell 1
Equation
Nitrogen~factor(Phase)+WD+(1|Date)
Nitrate~factor(Phase)+WD+(1|Date)
AIC
BIC
AIC
BIC
Model Fit
387.73798 404.57879
425.80563 443.19206
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
16.903
6.089
2.776 (Intercept)
12.983
6.055
2.144
factor(Phase)2
-1.885
6.998 -0.269 factor(Phase)2
-4.642
7.668 -0.605
factor(Phase)3
-6.575
9.748 -0.674 factor(Phase)3
-6.056
10.597 -0.571
Fixed Effects
factor(Phase)4
-39.431
10.86 -3.631 factor(Phase)4
-38.746
12.374 -3.131
factor(Phase)5
-6.01
11.136
-0.54 factor(Phase)5
-7.97
12.253
-0.65
factor(Phase)6
-38.281
10.028 -3.818 factor(Phase)6
-37.129
11.421 -3.251
WD
73.677
6.965 10.579 WD
59.406
7.547
7.871
Random
Variance Std.Dev.
Variance Std.Dev.
Effects
(Intercept)
0
0
(Intercept)
0
0

158

Table 56 (cont’d)
Ammonia: Cell 1
pH: Cell 1
Equation
Ammonia~factor(Phase)+WD+(1|Date)
pH~factor(Phase)+WD+(1|Date)
AIC
BIC
AIC
BIC
Model Fit
306.73087 323.57168
13.45293 30.10426
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
-0.2223
2.0027 -0.111 (Intercept)
7.34583
0.05869 125.17
factor(Phase)2
3.3602
2.6589
1.264 factor(Phase)2
0.34785
0.07428
4.68
factor(Phase)3
0.5123
3.4721
0.148 factor(Phase)3
0.19667
0.09902
1.99
Fixed Effects
factor(Phase)4
-4.8185
4.1018 -1.175 factor(Phase)4
0.10854
0.11377
0.95
factor(Phase)5
1.3798
4.0097
0.344 factor(Phase)5
0.15917
0.12714
1.25
factor(Phase)6
3.3018
3.7963
0.87 factor(Phase)6
0.01713
0.12714
0.13
WD
6.8145
2.5942
2.627 WD
-0.22296
0.06682
-3.34
Random
Variance Std.Dev.
Variance Std.Dev.
Effects
(Intercept)
0
0
(Intercept)
0.004265
0.06531

159

Phosphorus: Cell 2
COD: Cell 2
Equation
Phosphorus~factor(Phase)+WD+(1|Date)
COD~factor(Phase)+WD+(1|Date)
AIC
BIC
AIC
BIC
Model Fit
282.6361 299.84434
429.3745 446.40084
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
12.94
1.388
9.326 (Intercept)
31.57372
9.43684
3.346
factor(Phase)2
8.467
1.644
5.15 factor(Phase)2
10.87377
12.69711
0.856
factor(Phase)3
3.156
2.295
1.375 factor(Phase)3
-0.07372
15.27671 -0.005
Fixed Effects
factor(Phase)4
9.24
2.579
3.583 factor(Phase)4
-25.69534
16.77845 -1.531
factor(Phase)5
23.265
2.635
8.831 factor(Phase)5
14.64667
17.15256
0.854
factor(Phase)6
12.719
2.376
5.352 factor(Phase)6
4.68639
16.0344
0.292
WD
-2.145
1.61 -1.332 WD
45.33989
7.93958
5.711
Random
Variance Std.Dev.
Variance Std.Dev.
Effects
(Intercept)
0.2171
0.466
(Intercept)
443
21.05
Table 57: Statistical Data for Cell 2 Models

160

Table 57 (cont’d)
Total Nitrogen: Cell 2
Nitrate: Cell 2
Equation
Nitrogen~factor(Phase)+WD+(1|Date)
Nitrate~factor(Phase)+WD+(1|Date)
AIC
BIC
AIC
BIC
Model Fit
346.9026 363.7434
365.0839 382.29213
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
6.8952
3.7007
1.863 (Intercept)
6.347
3.463
1.833
factor(Phase)2
1.2824
4.253
0.302 factor(Phase)2
-4.663
4.254 -1.096
factor(Phase)3
-0.6352
5.9244 -0.107 factor(Phase)3
-5.419
5.893
-0.92
Fixed Effects
factor(Phase)4
-47.2912
6.6001 -7.165 factor(Phase)4
-38.202
6.79 -5.626
factor(Phase)5
-1.534
6.7676 -0.227 factor(Phase)5
-4.67
6.79 -0.688
factor(Phase)6
-51.0637
6.0943 -8.379 factor(Phase)6
-47.432
6.267 -7.568
WD
60.0085
4.2327 14.177 WD
47.045
4.201 11.199
Random
Variance Std.Dev.
Variance Std.Dev.
Effects
(Intercept)
0
0
(Intercept)
0
0

161

Table 57 (cont’d)
Ammonia: Cell 2
pH: Cell 2
Equation
Ammonia~factor(Phase)+WD+(1|Date)
pH~factor(Phase)+WD+(1|Date)
AIC
BIC
AIC
BIC
Model Fit
374.0468 390.88761
20.3956 37.04692
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
-0.8638
4.74 -0.182 (Intercept)
7.94494
0.06974 113.93
factor(Phase)2
7.0336
6.0477
1.163 factor(Phase)2
-0.04375
0.09542
-0.46
factor(Phase)3
3.8371
8.0004
0.48 factor(Phase)3
-0.0391
0.11528
-0.34
Fixed Effects
factor(Phase)4
-0.3202
9.2074 -0.035 factor(Phase)4
-0.0436
0.12869
-0.34
factor(Phase)5
3.995
9.2074
0.434 factor(Phase)5
-0.2816
0.14737
-1.91
factor(Phase)6
24.5803
8.5038
2.89 factor(Phase)6
-0.10806
0.14737
-0.73
WD
7.2314
5.8724
1.231 WD
-0.08354
0.06108
-1.37
Random
Variance Std.Dev.
Variance Std.Dev.
Effects
(Intercept)
0
0
(Intercept)
0.02391
0.1546

162

Phosphorus: Cell 3
COD: Cell 3
Equation
Phosphorus~factor(Phase)+WD+(1|Date)
COD~factor(Phase)+WD+(1|Date)
AIC
BIC
AIC
BIC
Model Fit
282.6361 299.84434
429.3745 446.40084
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
12.94
1.388
9.326 (Intercept)
31.57372
9.43684
3.346
factor(Phase)2
8.467
1.644
5.15 factor(Phase)2
10.87377
12.69711
0.856
factor(Phase)3
3.156
2.295
1.375 factor(Phase)3
-0.07372
15.27671 -0.005
Fixed Effects
factor(Phase)4
9.24
2.579
3.583 factor(Phase)4
-25.69534
16.77845 -1.531
factor(Phase)5
23.265
2.635
8.831 factor(Phase)5
14.64667
17.15256
0.854
factor(Phase)6
12.719
2.376
5.352 factor(Phase)6
4.68639
16.0344
0.292
WD
-2.145
1.61 -1.332 WD
45.33989
7.93958
5.711
Random
Variance Std.Dev.
Variance Std.Dev.
Effects
(Intercept)
0.2171
0.466
(Intercept)
443
21.05
Table 58: Statistical Data for Cell 3 Models

163

Table 58 (cont’d)
Total Nitrogen: Cell 3
Nitrate: Cell 3
Equation
Nitrogen~factor(Phase)+WD+(1|Date)
Nitrate~factor(Phase)+WD+(1|Date)
AIC
BIC
AIC
BIC
Model Fit
346.9026 363.7434
365.0839 382.29213
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
6.8952
3.7007
1.863 (Intercept)
6.347
3.463
1.833
factor(Phase)2
1.2824
4.253
0.302 factor(Phase)2
-4.663
4.254 -1.096
factor(Phase)3
-0.6352
5.9244 -0.107 factor(Phase)3
-5.419
5.893
-0.92
Fixed Effects
factor(Phase)4
-47.2912
6.6001 -7.165 factor(Phase)4
-38.202
6.79 -5.626
factor(Phase)5
-1.534
6.7676 -0.227 factor(Phase)5
-4.67
6.79 -0.688
factor(Phase)6
-51.0637
6.0943 -8.379 factor(Phase)6
-47.432
6.267 -7.568
WD
60.0085
4.2327 14.177 WD
47.045
4.201 11.199
Random
Variance Std.Dev.
Variance Std.Dev.
Effects
(Intercept)
0
0
(Intercept)
0
0

164

Table 58 (cont’d)
Ammonia: Cell 3
pH: Cell 3
Equation
Ammonia~factor(Phase)+WD+(1|Date)
pH~factor(Phase)+WD+(1|Date)
AIC
BIC
AIC
BIC
Model Fit
374.0468 390.88761
20.3956 37.04692
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
-0.8638
4.74 -0.182 (Intercept)
7.94494
0.06974 113.93
factor(Phase)2
7.0336
6.0477
1.163 factor(Phase)2
-0.04375
0.09542
-0.46
factor(Phase)3
3.8371
8.0004
0.48 factor(Phase)3
-0.0391
0.11528
-0.34
Fixed Effects
factor(Phase)4
-0.3202
9.2074 -0.035 factor(Phase)4
-0.0436
0.12869
-0.34
factor(Phase)5
3.995
9.2074
0.434 factor(Phase)5
-0.2816
0.14737
-1.91
factor(Phase)6
24.5803
8.5038
2.89 factor(Phase)6
-0.10806
0.14737
-0.73
WD
7.2314
5.8724
1.231 WD
-0.08354
0.06108
-1.37
Random
Variance Std.Dev.
Variance Std.Dev.
Effects
(Intercept)
0
0
(Intercept)
0.02391
0.1546

165

Phosphorus: Phase 1
COD: Phase 1
Equation
Phosphorus~factor(StageC)+WD+(S1|Date)
COD~factor(StageC)+WD+(S0|Date)
AIC
BIC
AIC
BIC
Model Fit
343.6852 363.7921
812.8625 832.9695
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
31.8979
0.8595
37.11 (Intercept)
68.54
25.97
2.64
factor(StageC)2
-6.3574
1.4774
-4.3 factor(StageC)2
43.03
32.07
1.342
Fixed Effects
factor(StageC)3
-19.2978
0.8064 -23.93 factor(StageC)1
416.31
58.34
7.135
factor(StageC)4
-23.0371
0.7973
-28.9 factor(StageC)4
12.28
31.55
0.389
WD
-1.3637
0.5874
-2.32 WD
-27.4
22.79 -1.202
Variance Std.Dev.
Variance Std.Dev.
Random
(Intercept)
3.106
1.762
(Intercept)
324.1
18
Effects
S1
15.15
3.892
S0
26670.4
163.31
Table 59: Statistical Data for Phase 1 Models

166

Table 59 (cont’d)
Total Nitrogen: Phase 1
Nitrate: Phase 1
Equation
Nitrogen~factor(StageC)+WD+(1|Date)
Nitrate~factor(StageC)+WD+(1|Date)
AIC
BIC
AIC
BIC
Model Fit
578.235 593.4557
637.7796 653.8128
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
42.719
6.606
6.467 (Intercept)
-18.584
5.576 -3.333
factor(StageC)1
-20.926
8.196 -2.553 factor(StageC)2
28.151
6.84
4.116
Fixed Effects
factor(StageC)2
-38.721
8.196 -4.724 factor(StageC)1
40.569
6.741
6.018
factor(StageC)3
-50.977
8.196
-6.22 factor(StageC)3
25.107
6.951
3.612
WD
64.905
5.887 11.025 WD
40.776
4.908
8.308
Variance Std.Dev.
Variance Std.Dev.
Random
(Intercept)
11.54
3.397
(Intercept)
30.4
5.514
Effects

167

Table 59 (cont’d)
Ammonia: Phase 1
pH: Phase 1
Equation
Ammonia~factor(StageC)+WD+(1|Date)
pH~factor(StageC)+WD+(S0|Date)
AIC
BIC
AIC
BIC
Model Fit
630.5639 646.5971
33.4513 53.94129
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
49.198
5.33
9.231 (Intercept)
7.397341
0.095932
77.11
factor(StageC)2
-57.343
6.748 -8.497 factor(StageC)2
0.505764
0.111204
4.55
Fixed Effects
factor(StageC)1
-56.077
6.753 -8.304 factor(StageC)1
-0.156736
0.111204
-1.41
factor(StageC)3
-59.917
6.66 -8.997 factor(StageC)3
0.483264
0.111204
4.35
WD
21.794
4.748
4.59 WD
-0.002783
0.050999
-0.05
Variance Std.Dev.
Variance Std.Dev.
Random
(Intercept)
0
0
(Intercept)
0.01459
0.1208
Effects
S0
0.07315
0.2705

168

Phosphorus: Phase 2
COD: Phase 2
Equation
Phosphorus~factor(StageC)+WD+(1|Date)
COD~factor(StageC)+WD+(1|Date)
AIC
BIC
AIC
BIC
Model Fit
286.00824 299.6669
573.59116 587.2499
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
30.6856
1.1653 26.332 (Intercept)
370.27
24.84
14.91
factor(StageC)3
-9.9991
1.4956 -6.686 factor(StageC)2
-325.54
31.88 -10.21
Fixed Effects
factor(StageC)2
-7.0472
1.4956 -4.712 factor(StageC)4
-341.42
31.88 -10.71
factor(StageC)4
-17.2947
1.4956 -11.564 factor(StageC)3
-346.19
31.88 -10.86
WD
-0.5805
1.0607 -0.547 WD
81.83
22.61
3.62
Variance Std.Dev.
Variance Std.Dev.
Random
(Intercept)
0
0
(Intercept)
0
0
Effects

Table 60: Statistical Data for Phase 2 Models

169

Table 60 (cont’d)
Total Nitrogen: Phase 2
Nitrate: Phase 2
Equation
Nitrogen~factor(StageC)+WD+(1|Date)
Nitrate~factor(StageC)+WD+(S1|Date)
AIC
BIC
AIC
BIC
Model Fit
416.87312 430.5318
427.816 445.3772
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
41.615
4.689
8.875 (Intercept)
-14.443
4.869 -2.966
factor(StageC)2
-24.354
6.018 -4.047 factor(StageC)2
34.782
7.696
4.52
Fixed Effects
factor(StageC)3
-37.503
6.018 -6.232 factor(StageC)3
22.635
6.2
3.651
factor(StageC)4
-47.556
6.018 -7.903 factor(StageC)4
23.094
6.2
3.725
WD
68.817
4.268 16.125 WD
33.002
4.41
7.483
Variance Std.Dev.
Variance Std.Dev.
Random
(Intercept)
0
0
(Intercept)
2.509
1.584
Effects
S1
157.694
12.558

170

Table 60 (cont’d)
Ammonia: Phase 2
pH: Phase 2
Equation
Ammonia~factor(StageC)+WD+(1|Date)
pH~factor(StageC)+WD+(1|Date)
AIC
BIC
AIC
BIC
Model Fit
378.8152 391.3045
3.3971 17.05581
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
50.254
7.391
6.8 (Intercept)
8.10935
0.06524
124.3
factor(StageC)2
-60.019
9.509 -6.312 factor(StageC)3
-0.2
0.06778
-2.95
Fixed Effects
factor(StageC)3
-56.798
9.509 -5.973 factor(StageC)2
-0.46892
0.06778
-6.92
factor(StageC)4
-63.71
9.509
-6.7 factor(StageC)1
-0.57231
0.06778
-8.44
WD
35.2
6.752
5.214 WD
-0.11501
0.05172
-2.22
Variance Std.Dev.
Variance Std.Dev.
Random
(Intercept)
0
0
(Intercept)
0.01107
0.1052
Effects

171

Phosphorus: Phase 3 (Manure Free)
COD: Phase 3 (Manure Free)
Equation
Phosphorus~factor(StageC)+(S0+S1|Date)
COD~factor(StageC)+(S0+S1|Date
AIC
BIC
AIC
BIC
Model Fit
129.3093 142.2678
238.8125 251.7711
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
27.22
7.392
3.682 (Intercept)
690.3
104
6.636
Fixed Effects factor(StageC)2
-7.516
3.342 -2.249 factor(StageC)4
-662.4
104.5 -6.337
factor(StageC)3
-11.125
7.021 -1.584 factor(StageC)3
-658.8
104.5 -6.302
factor(StageC)4
-14.853
7.021 -2.115 factor(StageC)2
-641.3
112.2 -5.716
Variance Std.Dev.
Variance Std.Dev.
Random
(Intercept)
0.9608
0.9802
(Intercept)
0
0
Effects
S0
291.5957
17.0762
S0
64286.8
253.55
S1
86.2132
9.2851
S1
440.2
20.98
Table 61: Statistical Data for MF Phase 3 Models

172

Table 61 (cont’d)
Total Nitrogen: Phase 3 (Manure Free)
Nitrate: Phase 3 (Manure Free)
Equation
Nitrogen~factor(StageC)+(S0+S2|Date)
Nitrate~factor(StageC)+(S1+S3|Date)
AIC
BIC
AIC
BIC
Model Fit
144.997 157.9556
77.94504 90.90364
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
37.758
5.949
6.347 (Intercept)
0.6631
0.2051
3.232
Fixed Effects factor(StageC)4
-30.53
5.3
-5.76 factor(StageC)4
0.9696
0.5439
1.783
factor(StageC)3
-31.498
6.684 -4.712 factor(StageC)3
0.2643
0.1525
1.733
factor(StageC)2
-27.43
5.3 -5.175 factor(StageC)2
6.2636
1.4279
4.386
Variance Std.Dev.
Variance Std.Dev.
Random
(Intercept)
16.8682
4.107
(Intercept)
0.1827
0.4274
Effects
S0
166.8448
12.917
S1
12.0945
3.4777
S2
34.0869
5.838
S3
1.6354
1.2788

173

Table 61 (cont’d)
Ammonia: Phase 3 (Manure Free)
pH: Phase 3 (Manure Free)
Equation
Ammonia~factor(StageC)+(S0+S1|Date)
pH~factor(StageC)+(S0|Date)
AIC
BIC
AIC
BIC
Model Fit
106.756 119.7146
11.79431 21.21874
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
19.492
1.094 17.825 (Intercept)
7.1392
0.1595
44.75
Fixed Effects factor(StageC)4
-17.06
2.063
-8.27 factor(StageC)3
0.7667
0.1696
4.52
factor(StageC)3
-16.518
2.063 -8.008 factor(StageC)2
0.4033
0.1696
2.38
factor(StageC)2
-19.202
1.162 -16.53 factor(StageC)4
0.9342
0.1696
5.51
Variance Std.Dev.
Variance Std.Dev.
Random
(Intercept)
17.9767
4.2399
(Intercept)
0
0
Effects
S0
24.5422
4.954
S0
0.13275
0.3644
S1
13.6356
3.6926

174

Phosphorus: Phase 3 (Manure Contaminated)
Equation
Phosphorus~factor(StageC)+(S0|Date)
AIC
BIC
Model Fit
71.03866 77.21936
Estimate
Std. Error t value
(Intercept)
32.077
1.998 16.057
factor(StageC)4
-16.267
2.545 -6.392
Fixed Effects
factor(StageC)3
-12.041
2.545 -4.732
factor(StageC)2
Random
Effects

(Intercept)
S0

-7.456
2.545
Variance Std.Dev.
1.178
1.085
22.201
4.712

COD: Phase 3 (Manure Contaminated)
COD~factor(StageC)+(S0+S1|Date)
AIC
BIC
153.751 162.2495
Estimate
Std. Error t value
(Intercept)
1008.8
152.9
6.595
factor(StageC)3
-956.1
153.2 -6.243
factor(StageC)2
-854.5
132.8 -6.435

-2.93 factor(StageC)4
(Intercept)
S0
S1

Table 62: Statistical Data for MC Phase 3 Models

175

-960.3
153.2
Variance Std.Dev.
0
0
93314.8
305.47
5484.8
74.06

-6.27

Table 62 (cont’d)
Total Nitrogen: Phase 3 (Manure Contaminated)
Equation
Nitrogen~factor(StageC)+(S0+S1|Date)
AIC
BIC
Model Fit
101.0361 109.5346
Estimate
Std. Error t value
(Intercept)
111.425
7.117 15.657
-60.275
12.014 -5.017
Fixed Effects factor(StageC)2
factor(StageC)4
-91.175
5.743 -15.876
factor(StageC)3
Random
Effects

(Intercept)
S0
S1

Nitrate: Phase 3 (Manure Contaminated)
Nitrate~factor(StageC)+(S1+S2|Date)
AIC
BIC
89.704 98.20248
Estimate
Std. Error t value
(Intercept)
0.865
1.263
0.685
factor(StageC)3
14.325
4.52
3.169
factor(StageC)4
12.957
1.166 11.109

-91.812
5.743 -15.987 factor(StageC)2
Variance Std.Dev.
32.254
5.679
(Intercept)
120.867
10.994
S1
270.439
16.445
S2

176

32.778
10.562
Variance Std.Dev.
3.662
1.914
440.739
20.994
76.278
8.734

3.103

Table 62 (cont’d)
Ammonia: Phase 3 (Manure Contaminated)
pH: Phase 3 (Manure Contaminated)
Equation
Ammonia~factor(StageC)+(S0|Date)
pH~factor(StageC)+(1|Date)
AIC
BIC
AIC
BIC
Model Fit
91.18757 97.36828
-4.3013702 0.3341621
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
128.6
26.52
4.848 (Intercept)
7.4375
0.04887
152.2
-128.6
26.54 -4.845 factor(StageC)2
-0.21
0.06652
-3.16
Fixed Effects factor(StageC)4
factor(StageC)3
-122.55
26.54 -4.617 factor(StageC)3
0.4325
0.06652
6.5
factor(StageC)2
Random
Effects

(Intercept)
S0

-126.83
26.54
Variance Std.Dev.
0
0
2810.413
53.013

-4.778 factor(StageC)4
(Intercept)

177

0.72
0.06652
Variance Std.Dev.
0.0007014
0.02648

10.82

Phosphorus: Phase 4 (Manure Free)
COD: Phase 4 (Manure Free)
Equation
Phosphorus~factor(StageC)+(S0+S1|Date)
COD~factor(StageC)+(S0|Date)
AIC
BIC
AIC
BIC
Model Fit
91.86893 100.3674
158.5611 164.7418
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
45.072368
9.233936
4.881 (Intercept)
1263.6
198.6
6.363
Fixed Effects factor(StageC)2
-0.008158
3.411145 -0.002 factor(StageC)4
-1219.4
199.7 -6.107
factor(StageC)3
-8.875789
6.387424
-1.39 factor(StageC)3
-1213.9
199.7
-6.08
factor(StageC)4
-13.485
6.387424 -2.111 factor(StageC)2
-1157.4
199.7 -5.797
Variance Std.Dev.
Variance Std.Dev.
Random
(Intercept)
77.5967
8.8089
(Intercept)
0
0
Effects
S0
161.3393
12.7019
S0
156008
394.98
S1
48.0126
6.9291
Table 63: Statistical Data for MF Phase 4 Models

178

Table 63 (cont’d)
Total Nitrogen: Phase 4 (Manure Free)
Nitrate: Phase 4 (Manure Free)
Equation
Nitrogen~factor(StageC)+(S0+S1|Date)
Nitrate~factor(StageC)+(1|Date)
AIC
BIC
AIC
BIC
Model Fit
61.22545 69.72393
50.20839 54.84392
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
38.775
3.61 10.741 (Intercept)
1.2662
0.5426
2.334
Fixed Effects factor(StageC)4
-34.236
3.512 -9.749 factor(StageC)4
1.685
0.5478
3.076
factor(StageC)3
-33.414
3.512 -9.515 factor(StageC)3
0.41
0.5478
0.748
factor(StageC)2
-27.881
3.609 -7.726 factor(StageC)2
3.7463
0.5478
6.839
Variance Std.Dev.
Variance Std.Dev.
Random
(Intercept)
0.08148
0.2854
(Intercept)
0.5773
0.7598
Effects
S0
49.2069
7.0148
S1
5.37196
2.3177

179

Table 63 (cont’d)
Ammonia: Phase 4 (Manure Free)
pH: Phase 4 (Manure Free)
Equation
Ammonia~factor(StageC)+(S0+S3|Date)
pH~factor(StageC)+(1|Date)
AIC
BIC
AIC
BIC
Model Fit
45.94224 54.44072
6.361341 9.270781
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
13.398
3.468
3.863 (Intercept)
7
0.1126
62.18
Fixed Effects factor(StageC)4
-13.398
3.47 -3.861 factor(StageC)4
0.81
0.0753
10.76
factor(StageC)3
-10.266
3.105 -3.306 factor(StageC)2
0.505
0.0753
6.71
factor(StageC)2
-12.24
3.105 -3.942 factor(StageC)3
0.6633
0.0753
8.81
Variance Std.Dev.
Variance Std.Dev.
Random
(Intercept)
0.5345
0.7311
(Intercept)
0.029508
0.17178
Effects
S0
38.4779
6.2031
S3
0.5345
0.7311

180

Phosphorus: Phase 4 (Manure Contaminated)
Equation
Phosphorus~factor(StageC)+(S0+S1|Date)
AIC
BIC
Model Fit
106.9236 117.8767
Estimate
Std. Error t value
(Intercept)
29.891
3.765
7.94
Fixed Effects factor(StageC)2
-4.203
1.639 -2.564
factor(StageC)3
-6.376
2.949 -2.162
factor(StageC)4
-12.765
2.949 -4.328
Variance Std.Dev.
Random
(Intercept)
4.593
2.143
Effects
S0
36.948
6.078
S1
11.926
3.453
Table 64: Statistical Data for MC Phase 4 Models

181

COD: Phase 4 (Manure Contaminated)
COD~factor(StageC)+(S0+S1|Date)
AIC
BIC
190.85 201.8031
Estimate
Std. Error t value
(Intercept)
1515.2
95
15.95
factor(StageC)2
-1316.6
75.95 -17.33
factor(StageC)3
-1433.6
91.47 -15.67
factor(StageC)4
-1450.2
91.47 -15.85
Variance Std.Dev.
(Intercept)
699.5
26.448
S0
41656.21
204.099
S1
5420.6
73.625

Table 64 (cont’d)
Total Nitrogen: Phase 4 (Manure Contaminated)
Equation
Nitrogen~factor(StageC)+(S0+S1|Date)
AIC
BIC
Model Fit
130.8387 141.7918
Estimate
Std. Error t value
(Intercept)
136.6
12.217 11.181
Fixed Effects factor(StageC)2
-84.3
9.308 -9.057
factor(StageC)3
-120.76
12.287 -9.829
factor(StageC)4
-122.28
12.287 -9.952
Variance Std.Dev.
Random
(Intercept)
0
0
Effects
S0
737.807
27.163
S1
72.082
8.49

182

Nitrate: Phase 4 (Manure Contaminated)
Nitrate~factor(StageC)+(S0+S1|Date)
AIC
BIC
107.3724 118.3254
Estimate
Std. Error t value
(Intercept)
1.129
0.4742
2.381
factor(StageC)2
34.131
8.5736
3.981
factor(StageC)3
4.831
1.2313
3.923
factor(StageC)4
12.254
1.2313
9.952
Variance Std.Dev.
(Intercept)
5.146
2.268
S0
5.336
2.31
S1
332.767
18.242

Table 64 (cont’d)
Ammonia: Phase 4 (Manure Contaminated)
pH: Phase 4 (Manure Contaminated)
Equation
Ammonia~factor(StageC)+(S0+S2|Date)
pH~factor(StageC)+(S0+S2|Date)
AIC
BIC
AIC
BIC
Model Fit
161.093 172.0461
13.88637 19.22034
Estimate
Std. Error t value
Estimate
Std. Error t value
(Intercept)
119.52
11.44 10.447 (Intercept)
7.3867
0.1587
46.53
Fixed Effects factor(StageC)2
-109.63
10.94 -10.02 factor(StageC)2
-0.2467
0.1703
-1.45
factor(StageC)3
-88.57
32.18 -2.752 factor(StageC)3
0.3667
0.2677
1.37
factor(StageC)4
-118.36
10.94 -10.818 factor(StageC)4
0.8533
0.1703
5.01
Variance Std.Dev.
Variance Std.Dev.
Random
(Intercept)
42.09
6.488
(Intercept)
0.032471
0.1802
Effects
S0
446.21
21.124
S0
0.083955
0.28975
S2
2489.02
49.89
S2
0.036351
0.19066

183

REFERENCES

184

REFERENCES

American Water Works Association, and Economic and Engineering Services Inc.,
“Nitrification,” Prepared for USEPA Office of Water, Distribution System Issue Paper, Aug. 15,
2002.
D. Bates, “[R] lmer, p-values and all that,” https://stat.ethz.ch/pipermail/r-help/2006May/094765.html, May, 2006.
Douglas Bates, Martin Maechler, Ben Bolker and Steven Walker (2014). lme4: Linear mixedeffects models using Eigen and S4. R package version 1.0-6. http://CRAN.Rproject.org/package=lme4
Beach, Deborah N.H., John E. McCray, Kathryn S. Lowe, and Robert S. Siegrist. “Temporal
changes in hydraulic conductivity of sand porous media biofilters during wastewater infiltration
due to biomat formation.” Journal of Hydrology 311 (2005): 230-243.
C.D. Beal, E.A. Gardner, and N.W. Menzies. “Process, performance, and pollution potential: A
review of septic tank-soil absorption systems.” Australian Journal of Soil Research 43 (2005):
781-802.
C.D. Beal, E.A. Gardner, G. Kirchhoff, and N.W. Menzies. “Long-term flow rates and biomat
zone hydrology in soil columns receiving septic tank influent.” Water Research 40 (2006): 23272338.
J. Brisson, and F. Chazarenc, “Maximizing pollutant removal in constructed wetlands: Should
we pay more attention to macrophyte species selection?,” Science of the Total Environment, vol.
407, no. 13, Jun 15, 2009.
J. J. Chang, S. Q. Wu, Y. R. Dai, W. Liang, and Z. B. Wu, “Treatment performance of
integrated vertical-flow constructed wetland plots for domestic wastewater,” Ecological
Engineering, vol. 44, pp. 152-159, Jul, 2012.
J. H. Choi, H. Kang, and S. S. Park, “Comparison of Enzyme Activities in Vegetated and
Nonvegetated Sediments,” Journal of Environmental Engineering-Asce, vol. 135, no. 5, pp. 299305, May, 2009.
S.H. Christopherson, D.R. Schmidt, and K.A. Janni, “Recirculating Media Filter— A Milk
House Wastewater Treatment Option,” University of Minnesota Extension Fact Sheet,
Publication 1204, Sep, 2007.
L. H. Cui, J. K. Feng, Y. Ouyang, and P. W. Deng, “Removal of nutrients from septic effluent
with re-circulated hybrid tidal flow constructed wetland,” Ecological Engineering, vol. 46, pp.
112-115, Sep, 2012.
185

Pauline M. Doran, “Bioproces Engineering Principles,” Academic Press, Elsevier Ltd., London,
1995. ISBN: 978-0-12-22086030222
J. Ebeling, S. Tsukuda, J. Hankins, and C. Solomon, “Performance Evaluation of a Recirculating
Sand Filter and Peat Filter in West Virginia,” Small Flows Quarterly, vol. 4, no. 1, pp. 27-37,
Winter 2003.
M. S. Fountoulakis, S. Terzakis, A. Chatzinotas, H. Brix, N. Kalogerakis, and T. Manios, “Pilotscale comparison of constructed wetlands operated under high hydraulic loading rates and
attached biofilm reactors for domestic wastewater treatment,” Science of the Total Environment,
vol. 407, no. 8, pp. 2996-3003, Apr, 2009.
A. Gelman and J. Hill, “Data Analysis Using Regression and Multilevel/Hierarchical Models,”
Cambridge University Press, New York, 2007.
B. Hawkins, “Handling Milking Facility Washwater”, Factsheet: order no. 11-039W, AGDEX
720/410, 2011.
B. Holmes and S. Struss, “Milking Center Wastewater Guidelines,” A Companion Document to
Wisconsin NRCS Standard 629, http://socwisconsin.org/ accessed Jan, 2014.
M. S. M. Jetten, L. van Niftrik, M. Strous, B. Kartal, J. T. Keltjens, H. J. M Op den Camp,
“Biochemistry and molecular biology of anammox bacteria,” Critical Reviews in Biochemistry
and Molecular Biology, vol. 44, no. 2–3, pp. 65–84, Jun, 2009.
R.K. Kadlec, S.D. Wallace, “Treatment Wetlands”, Second edition, CRC Press, Taylor &
Francis Group, New York, 2009.
S. Kouki, F. M'Hiri, N. Saidi, S. Belaid, and A. Hassen, “Performances of a constructed wetland
treating domestic wastewaters during a macrophytes life cycle,” Desalination, vol. 246, no. 1-3,
pp. 452-467, Sep, 2009.
J. L. C. Ladu, M. Zheng, P. L. Demetry, X. Lu, “Effect of Hydraulic Retention Time and
Recycle Ratio on Anoxic/Oxic Bioreactor and Artificial Wetland Performance for Domestic
Wastewater Treatment,” Curr World Environ 2013;8(2). doi:
http://dx.doi.org/10.12944/CWE.8.2.01
G. Langergraber, “Modeling of processes in subsurface flow constructed wetlands: A review,”
Vadose Zone Journal, vol. 7, no. 2, pp. 830-842, May, 2008.
R. A. Larson and S. I. Safferman, “Aerobic Treatment Unit Performance on Milking Parlor
Wash Water,” Transactions of the ASABE, vol. 52, no. 3, pp. 825-834, May, 2009.

186

P. Mantovi, M. Marmiroli, E. Maestri, S. Tagliavini, S. Piccinini, N. Marmiroli, “Application of
a horizontal subsurface flow constructed wetland on treatment of dairy parlor wastewater”,
Bioresource Technology, vol. 88, pp. 85-94, 2003.
K. McKague, “Evaluation of a Subsurface Constructed Wetland as a Management Option for
Handling Milking Centre Washwater and Solid Manure Storage Runoff,” Environmental
Sustainability Research Program, Project 09A4, Final Report, Jun, 2012.
D. Payer and B. Holmes, “Controlling milking center wastewater: An overview,” Factsheet:
A3608, Milking Center Wastewater Management, University of Wisconsin-Extension.
R Core Team (2013). R: A language and environment for statistical computing. R Foundation for
Statistical Computing, Vienna, Austria. URL http://www.R-project.org/.
J. Rathbun, S. I. Safferman, S. S. Davis, T. Cleary, K. Foight, “Performance Evaluation of the
Michigan Filter Mound for Treating Milking Center Wastewater from a Small Dairy,”
Transactions of the ASABE, vol. 55, no. 3, pp. 995-1002, May, 2012.
S. I. Safferman, “Milking Facility Wash Water: Fact and Figures.” Michigan Dairy Review, vol.
13, no. 1, pp. 1-2, 2008.
H. J. Seltman, “Experimental Design and Analysis,” Department of Statistics, Carnegie Mellon
University, Dec, 2013
A.I. Stefanakis, C.S. Akratos, G.D. Gikas, V.A. Tsihrintzis, “Effluent quality improvement of
two pilot-scale, horizontal subsurface flow constructed wetlands using natural zeolite
(clinoptololite)”, Micropor. Mesopor. Mater. 124 (1–3) (2009) 131–143.
A.I. Stefanakis, V.A. Tsihrintzis, “Effects of loading, resting period, temperature, porous media,
and aeration on performance of pilot-scale vertical flow constructed wetlands”, Chemical
Engineering Journal 181– 182 (2012) 416– 43
U. Stottmeister, A. Weissner, P. Kuschk, U. Kappelmeyer, M. Kaestner, O. Bederski, R. A.
Mueller, M. Moormann, “Effects of plants and microorganisms in constructed wetlands for
wastewater treatment,” Biotechnology Advances, vol. 22, pp. 93-117, 2003.
Swiss Standard SN 670 010b, Characteristic Coefficients of Soils, Association of Swiss Road
and Traffic Engineers
University of Minnesota, “Parameters to Monitor Septic System Performance,” Onsite Sewage
Treatment Program Factsheets, Sep, 2010.

187

USDA-Natural Resources Conservation Service, GM_450_401_B - Subpart B - Conservation
Practice Standards, Oct, 2009.
USDA-Soil Conservation Service, 1992. National Engineering Handbook. Agricultural Waste
Management Field Handbook. Washington, DC (Chapter 4).
Jan Vymazal, Hans Brix, Paul F. Cooper, Raimund Haberl, Reinhard Perfler, Johannes Laber ,
“Removal mechanisms and types of constructed wetlands,” Constructed Wetlands for
Wastewater Treatment in Europe, pp. 17-66, 1998.
J. Vymazal, “Removal of nutrients in various types of constructed wetlands”, Sci. Total Environ.
380 (2007) 48–65.
J. Vymazal, “Constructed Wetlands for Wastewater Treatment: Five Decades of Experience,”
Environmental Science & Technology, vol. 45, no. 1, pp. 61-69, Jan, 2011.
G. Ying, Y. Xing, Z. Li, J. Pan, and X. Kuang, “Advantages of Psychrophiles in Improving BioTreatment Efficiency of Small Size Constructed Wetlands During Cold Weather,”
Environmental Progress & Sustainable Energy, vol. 29, no. 1, pp. 25-33, Apr, 2010.
P. Zhang, R. T. Hai, D. K. Zhou, Y. Q. He, and Z. Y. Bai, “Synergism of Novel Sequence Bioecological Process and Biological Aerated Filter for Sewage Treatment in Cold Climate,”
Chinese Journal of Chemical Engineering, vol. 19, no. 5, pp. 881-890, Oct, 2011.

188