TREATMENT WETLANDS - CHAPTER 15 pdf

The purpose of Part II of Treatment Wetlands, Second Edi-tion, is to provide information on how to design, construct, and operate a wetland for the purpose of water quality improvement

Part II

Implementation

The purpose of Part II of Treatment Wetlands, Second

Edi-tion, is to provide information on how to design, construct,

and operate a wetland for the purpose of water quality

improvement It is recognized that wetlands are almost never

stand-alone treatment devices but rather form part of a

treat-ment train Other components may be mechanical, such as

clarifiers or filters, or more natural, such as settling basins or

lagoons More than one type of wetland may be involved at

the same site—FWS, VF, HSSF, or biosolids systems In this

chapter, historical sizing methods are reviewed and placed in

perspective

Design of treatment wetlands may be roughly divided

into two categories: sizing calculations and physical

specifi-cations Sizing procedures may in turn be predicated on the

particular application, whether the purpose is for treatment of

flows intended for surface water discharge, or for conditioning

water for groundwater discharge, or for retaining and

treat-ing stormwater runoff Further, local regulations may require

compliance with specific numerical standards

(performance-based) or may simply require certain design specifications

according to prescriptive criteria (technology-based)

The physical aspects of treatment wetland design seem

deceptively simple Indeed, the basic elements have been on

record for over a hundred years

From Brian Mackney (Australia) via Hans Brix (1994d),

an excerpt from an essay written by Nemo to the head of the

Hornsby Literary Institute in 1904 reads:

Anyone who has a little ground about his house can dispose of

his dirty water as follows:

Dig up a plot of ground thoroughly to a depth of fifteen to

eighteen inches Cut a channel leading from the kitchen and

washhouse into the highest side of the plot and let all the dirty

water drain into it Plant the plot with plants that grow rapidly

and require a great deal of water such as Arum Lilies, for

instance The dirty water will be all absorbed by the roots

of the plants and a most luxuriant garden will be produced

which will defy the hottest weather and will always be green

and beautiful By this means a curse will be transformed

into a blessing Twenty or thirty feet square properly worked

would be enough for any ordinary family.

Of course, Nemo did not conceive of the difficulties of

scale-up from a single household to all the wastewater from a city

of 100,000 population For even modest-sized systems, there

is a need to consider the system layout, the number of

indi-vidual cells and their arrangement, and how to move water

through a large wetland complex Questions of hydraulic and

vegetation efficiency have important economic consequences

Therefore, the subsequent chapters consider the engineering

of the wetland

The first documented use of a wetland within a deliber-ately engineered treatment vessel appears to belong to Cleo-phas Monjeau (1901), as shown in Figure 15.1

Claims for this United States patent (issued in 1901) include distributed vertical flow, a fluctuating water level, and aeration of the wastewater This would be considered a cutting- edge wetland design even in the 21st century

Design of wetland systems may be either performance-based or technology-performance-based, depending upon whether the per-formance expectations are explicit in an operating permit or implicit in a set of statutory design specifications Large sys-tems are often in the former group, whereas small on-site and urban stormwater systems are in the latter The primary focus

of this chapter is performance-based design

15.1 HISTORICAL PERSPECTIVES

Much of treatment wetland sizing has evolved along very simplistic lines, mostly adapted from civil and sanitary engi-neering However, it is become increasingly apparent over the last 20 years that treatment wetlands are much more bio-logically complex than treatment processes used in the sani-tary engineering field, and simplifying assumptions applied

to activated sludge, trickling filter, and pond systems do not adequately describe the internal biogeochemical processes that govern treatment performance in wetlands

Historically, mechanistic wetland models have been of little utility because of calibration difficulties General rules

of thumb for area requirements are used to good avail in conceptual design Input–output regressions allow calcu-lations of outlet concentrations for specified systems, and scatter plots are available to examine intersystem variabil-ity Several variants of first-order models are now in use Virtually all of these existing techniques fail to address sig-nificant aspects of wetland performance, including effects of hydraulics, meteorology, interactions of biota and treatment, supply-side constraints, and seasonal and stochastic variabil-ity Soil accretion and the sustainable storage and dilution

of irreducible constituents are other aspects not covered by existing design methods There is also great danger of using empirical relations outside the envelope for which they were developed, and of unintentionally exceeding the capabili-ties of the fundamental wetland processes Proper use of this imperfect set of design tools is a current design challenge Three principal themes have been prevalent in the his-tory of constructed treatment wetland design: consideration

of the pollutant and hydraulic loadings, first-order removal models, and regression equations First, some general history and concepts of wetland design are discussed, and then some specifics for the principal variants of the technology

F IRST -O RDER M ODELING

In the mid-1980s, simple first-order models of pollutant

removal in wetlands made their appearance in the treatment

wetland literature (U.S EPA, 1983b; Boon, 1985; Kadlec,

1985; Reed et al., 1988) This was, in part, a natural

exten-sion of models that had earlier been developed for waste

stabilization ponds (Marais and Shaw, 1961; Marais, 1974;

Thirumurthi, 1974) Some of these early models were founded

on shaky ground because they borrowed extensively from

nonwetland results and were not well-calibrated to actual

treatment wetland data Over the past two decades,

consid-erable progress in understanding the behavior of treatment

wetlands has allowed major improvements in the way that

first-order models are constructed and implemented It is the

premise of this book that first-order modeling remains one of

the most effective wetland design tools but that other

infor-mation can and should also be brought to bear on design

The information in Part I sets forth the many processes

affecting pollutant removal in wetlands Nearly all of the

compo-nent processes are first-order within some range of constraints

that are often met within wetland ecosystems Therefore, it is not

surprising that longitudinal profiles are virtually all describ-able by such a model, provided proper account is taken of background concentrations and the actual flow patterns observed in treatment wetlands Further, calibrated first-order models are capable of adequately representing the trends in time series of pollutant concentrations exiting wetland systems However, it is also the case that the many varieties of treatment wetlands, together with the variations of a given type, should not be expected to produce identical results Plant varieties and their fractional coverage, media depth, aspect ratio, com-partmentalization, and environmental factors vary from

wet-land to wetwet-land A distribution of k-values is to be expected,

and has been observed for all the common pollutants (see Part I) Intersystem variability is thus a fact of life for wet-land systems that are intimately connected to their climate and surrounding environment

It is also true that, without exception, effluent con-centrations exhibit random variation Environmental and biological factors drive behavior on hour-to-hour and day-to-day time scales Any model capable of describ-ing such rapid phenomena would be quite complex, and

in fact no such models currently exist Therefore, effects such as those of wind, rain, and animal invasions must be dealt with through an understanding of the intrasystem variability present in the performance of a treatment wet-land Because such random effects are a large proportion of the wetland response, that variability must be understood, quantified, and included in design Deterministic relations, either a rate-constant model or a loading relationship, will produce only the general trend of outlet concentrations and not the random variability

Previous Criticism of First-Order Modeling

Reliance on first-order modeling, although nearly universally accepted, has had one severe critic Because that criticism appears in a U.S government publication advocating a spe-cific treatment wetland design procedure, it is necessary to examine this lone voice of opposition in more detail

U.S EPA (2000a) contended that “ … the development

of a rigorous model of the process, or parts of it, has not been achieved as is true with many of the wastewater treatment processes designed today.”

This statement is no longer correct Credible treatment wetland models exist at all the levels that they do for activated sludge processes, which are the most advanced in terms of modeling The work of McBride and Tanner (2000),

Langer-graber (2001), Wynn and Liehr (2001), Howell et al (2005)

and others has carried wetland modeling to an advanced level of detail for municipal wastewater treatment

Stormwa-ter wetland models abound (Fitz et al., 1996; Moustafa and Hamrick, 2002; Munson et al., 2002; Walker and Kadlec,

2005) However, to our knowledge, these complex models are not often in use for purposes of design, the exception being the dynamic model for stormwater treatment areas (DMSTA) (Walker and Kadlec, 2005)

FIGURE 15.1 1901 U.S patent for a treatment wetland system

(From U.S Patent 681,884.)

Fig 1.

Vegetation

Fig 2.

Fig 3.

B C

d1

d

w D

x

x

C C

C

P P

P

F A

Indeed, U.S EPA (2000a) went on to conclude that the

design should be based on loading, an approach that:

… has been used for many decades by environmental

engi-neers in the design of highly complex unit processes,

includ-ing the activated sludge process and waste stabilization ponds

Only when a carefully designed series of iterative studies

have been conducted, and data based on quality-controlled

specifications have been analyzed, can rigorous models be

provided for use in wetland system design.

It is almost unbelievable that this statement should have been

made on the heels of a carefully designed series of iterative

studies sponsored by U.S EPA for the purpose of

develop-ment of design equations (George et al., 1998) for treatdevelop-ment

wetland systems

A compromise position is the use of highly aggregated

models that contain major features of the observed

behav-ior of treatment wetlands That compromise is a first-order

model of some sort, possibly with temperature- or

season-dependent parameters A primary message of the U.S EPA

(2000a) is that such first-order removal models should not be

used, because these are prominently absent from that manual

That is paradoxical, because the wetland publications of all

of the major contributors/authors advocate the use of

first-order removal models (Reed et al., 1988; Crites, 1994; Reed

et al., 1995; Kadlec and Knight, 1996; Kemp and George,

1997; Crites and Tchobanoglous, 1998; Campbell and Ogden,

1999)

In recent years, first-order models have been reaffirmed as a

useful tool in wetland design Rousseau et al (2004) performed

a review of model-based design of horizontal subsurface-

flow treatment wetlands, and concluded:

At present, the state-of-the-art k-C* model seems to be the

best available design tool if the designer makes sure that all

the assumptions are fulfilled and if he is aware of the pitfalls

in the model.

Crites et al (2006) reiterate the basic features of both areal

and volumetric first-order models for free water surface

(FWS) wetlands, and acknowledge that “… they do not

directly account for the complex reactions and interactions

that occur in wetlands.” Crites et al (2006) go on to state:

Such an approach is the best that can be done with the

cur-rently available database and understanding of wetland

processes.

In this book, the first-order model is retained as the primary

tool in the design process However, the loading approach

is also retained, as a means of accounting for intersystem

variability

L OADING S PECIFICATIONS

An early source of guidance utilized by wetland analysts was

the methods used by wastewater stabilization pond

design-ers A common method was the prescription of a specified areal-loading rate for biochemical oxygen demand (BOD),

on the order of 40 kg/ha·d to achieve an outflow

concentra-tion of 30 mg/L (Reed et al., 1988) BOD loading is still

the recommended design method for ponds, although with modifications, that acknowledge temperature effects (Shilton and Mara, 2005; Shilton, 2005) At 3°C, the recommended loading rate is 40 kg/ha·d, which is the winter condition for cold climate systems As the minimum winter temperature goes up, say, to 20°C for a tropical climate, the allowable

loading goes up as well to about 250 kg/ha·d Reed et al.

(1988) suggested using the BOD-loading rate to the wetland

as a constraint, not for design sizing, with an upper limit of

100 kg/ha·d It is perhaps coincidental that U.S EPA (2000a) concluded that the allowable design loading to a fully veg-etated FWS wetland should be 40 kg/ha·d, identical to that for temperate pond design

The areal-loading procedure may easily be converted to a population equivalence (PE), if the person equivalence of the particular wastewater is known For instance, the per-person BOD generation rate in the United States is approximately

63 g/PE·d (U.S EPA, 2002c) and in Austria is specified as

60 g/PE·d (Haberl et al., 1998) In the United States, BOD

generation is also taken to be 60 g/PE·d, but a septic tank (or other primary treatment device) is virtually always used as pretreatment for domestic wastes, and therefore 40 g/PE·d is used as the influent to a wetland (Wallace and Knight, 2006)

So, for example, if a design specification is 5 m2/PE for a HSSF wetland treating primary effluent, then the equivalent loading specification is 40/5  8 g/m2·d 80 kg/ha·d

To compare first-order calculations to loading criteria, the hydraulic loading to the wetland must be known The rate-constant approach separates the effect of inlet concentration from that of flow rate The water flow per person equivalent is

to some extent independent of the amounts of pollutants gen-erated Water usage in a typical household varies from 50 to

200 L/PE·d, depending on whether it is situated in an arid or wet location, or in a developed or developing country Small communities have slightly greater water usage per person The relationship between rate-constant specification and areal-loading specification is straightforward, but not unique: the rate approach is strength (concentration) dependent and the loading method is not The connection for a pollutant with zero background is

kC qC

i

¦¥

³ µ´

¤

¦¥

³ µ´

§

©

¨

¨

¶

¸

·

··

(15.1)

where

C

Coi

inlet concentration, mg/L outlet conce

rate constant, m/d hydrau

k q

 llic loading, m/d and

qCipollutant loading, g/m d2•

It is clear from Equation 15.1 that the outlet concentration

depends not only on inlet pollutant loading but also on inlet

concentration

Suppose it is required that Co 25 mg/L, and that k 

0.1 m/d For Ci 100 mg/L, the allowable hydraulic loading

is q 0.072 m/d The corresponding pollutant loading is 7.2

g/m2·d However, if the inlet concentration is Ci 200 mg/L,

the rate model suggests that the allowable hydraulic

load-ing is q 0.048 m/d and the corresponding pollutant

load-ing is 9.6 g/m2·d Thus, the rate approach gives different

area requirements, depending on waste strength To further

complicate the matter, we might consider that the two cases

differed only in water use, and that the same number of

pop-ulation equivalents (same mass load) was involved Suppose

PE  100 and the per capita generation was 50 g/PE·d, for a

total of 5,000 g/d For the weak strength of 100 mg/L (more

water usage), the area required would be 5,000/7.2  694 m2

or 6.9 m2/PE For the stronger strength of 200 mg/L (less

water usage), the area required would be 5,000/9.6  520 m2

or 5.2 m2/PE Thus, the loading method and the rate method

are not equivalent

R EGRESSION E QUATIONS

The principal variables that determine the outlet concentration

(Co) are the hydraulic loading (q), or the equivalent detention

time (h/q), and the inlet concentration (Ci) Other influences

include temperature, solar radiation, and pH There is,

there-fore, the possibility of simply postulating a functional

rela-tion, such as a product-power law, and regressing intersystem

wetland input–output data to fit that equation In so doing,

some of the power of intrawetland data is lost, because such

regression formulae may not represent the observed

longitu-dinal profiles for a specific wetland However, there can be no

doubt that an individual data set can be fit with a reasonable

guess at a fitting formula, because there are typically three

or more fitting parameters For example, Kadlec and Knight (1996) set forth linear regressions and power law regressions for different common pollutants A few examples are: Gumbricht (1993a), total nitrogen leaving a single FWS SAV system (R2 0.92):

Co1 63 0 78 Ci 1 7 ln(Tnom) (15.2)

where

C

Coi

inlet concentration, mg/L outlet conce

nominal detention time, nom

Kadlec and Knight (1996), nitrate leaving several FWS sys-tems of different types (R2 0.35):

Co 0 093Ci0 474q0 745

where

q hydraulic loading rate, cm/d

Tunçsiper et al (2006), ammonia leaving one square meter

HSSF mesocosms with various plants (R2 0.67):

i

i pH

¤

¦

³ µ

´ 0 371 0 0048 0 080 2 2886

(15.4) The operational data and multiple linear regression models are shown in Figure 15.2, along with the rate-constant model

Tunçsiper et al (2006) concluded that the multiple linear model

was better, but that is not supported by the sum of the squared errors derived from data and models (SSQE  0.30 for rate-constant model, and SSQE  1.98 for multiple linear model) Great care must be exercised in defining the variable ranges of such regression models The potential extremes of

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

0 20 40 60 80 100 120 140 160 180 200 220 240 260

Time (days)

Data Rate Constant Model Multiple Linear Model

FIGURE 15.2 Time series of ammonia concentrations exiting HSSF mesocosms, together with rate constant and multiple linear regression

models (From Tunçsiper et al (2006) Water Science and Technology 53(12): 111–120 Reprinted with permission.)

variables can produce unintended anomalies that are a

poten-tial trap for designers For instance, the Gumbricht (1993a)

regression (Equation 15.2) produces negative concentrations

for long-detention times, a fact that might go unnoticed by

a potential user of this equation set The Kadlec and Knight

(1996) regression (Equation 15.3) suggests that Co increases

with decreasing Ci for Ci  1.0 mg/L, which is a highly

improbable circumstance The Tunçsiper et al (2006)

regres-sion (Equation 15.4) shows that removal goes up with inlet

concentration and, in fact, for Ci 10 mg/L, more than 100%

removal is forecast That Ci is higher than the calibration

range, but an unsuspecting designer could easily misapply

Equation 15.4 to that condition

These example regressions contain 3, 3, and 4

param-eters, respectively Any possibility of generalization, and use

in design, is hindered by the need to predict all of those

coef-ficients in a new situation Consequently, regression

equa-tions have not found use in treatment wetland design

15.2 FREE WATER SURFACE WETLANDS

Early FWS treatment wetlands were often natural wetlands

employed to improve water quality Although those

ecosys-tems were preexisting, there was no issue of sizing, but rather

an issue of how much water could be put into the wetland

Recipes existed, and continue to exist, covering the

appropri-ate hydraulic loading to natural systems (e.g., for Florida) In

contrast, FWS constructed wetland design sizing for

continu-ous flows has historically been virtually entirely performance

based

Unfortunately, there has been considerable

misinter-pretation and misuse of these simplistic models One of the

problems has been unsupported pronouncements of model

parameters When plug flow and C* 0 are presumed, there

is only one parameter remaining, which is here designated

as k1 (or kV1) For instance, there is an oft-repeated value of

kV1 0.678 d−1 in U.S literature for BOD reduction in FWS

wetlands at 20°C (Reed et al., 1995; Water Environment Federation, 2001; Crites et al., 2006) This value is intended

for use in an exponential (plug flow) model with a zero back-ground concentration Although no basis for this number has ever been presented, it is now possible to place it in the per-spective of calibrations for large numbers of FWS systems The suggested value, for a system 40 cm deep, corresponds to

k  100 m/year This ranks at the 98th percentile of k1 -values for 386 wetland-years of data for FWS wetlands (Figure 15.3) From Figure 15.3, it is clear that the

expo-nential plug-flow model with C*  0 has the wrong shape compared to intersystem data, and therefore a different

cali-bration kV cannot fix it Most of the time this particular single-parameter model will overestimate removals, as well as the model providing incorrect trends

15.3 STORMWATER WETLANDS

Stormwater wetlands are also FWS systems, but these differ markedly in design philosophy Early in the brief history of urban stormwater wetlands, data was collected from a variety

of constructed and natural systems, and the central perfor-mance tendency of these treatment wetlands was determined

(Strecker et al., 1992; Schueler, 1992) The time-variable

nature of stormwater wetlands means that the information was and is quite sparse Source characterization is difficult and often the result of semiquantitative rules for the amounts

of water and pollutants that are washed off of various types

of watersheds

Given this meager performance base, it is not surpris-ing that the earliest sizsurpris-ing rules were empirical One method involves the specification of the wetland area compared to the contributing watershed area, the wetland-to-watershed area ratio (WWAR) This is a scaling rule, comparable to those

in use for continuous-flow FWS wetlands The implied per-formance of a particular WWAR is inferred from the perfor-mance data from operating systems similar to WWAR

FIGURE 15.3 A presumptive model line on the BOD-loading graph for an inlet concentration of 50 mg/L The model is the first-order

volumetric with kV1 0.678 d −1 and with C* 0, but truncated at BOD  5 mg/L per literature recommendation A free water depth of

45 cm is presumed for the model.


















A second approach relies on the capture and detention of

a specified amount of runoff, which is then subject to

treat-ment during the interevent period For instance, the

antici-pated runoff from a rain event of specified return frequency

may be used to set the wetland water volume Runoff from

larger events is partially bypassed or runs quickly through the

wetland Again, the implied degree of treatment is that

asso-ciated with holding the water during the interim periods

A third approach involves adapting the continuous flow

rate-constant method to event-driven systems The use of the

k-C* model, with average flow rates, was suggested by Kadlec

and Knight (1996) and by Wong and Geiger (1997) A number

of calibrations for stormwater systems were provided in the

summary of Carleton et al (2001) More recently, the concept

has been extended to the P-k-C* model by Wong et al (2006).

A drawback of the use of P-k-C* calculations for

event-driven wetlands is that rate constants for continuous-flow

situations do not necessarily carry over to the averaged

transient-flow situation (Kadlec, 2001a) As a result, dynamic

modeling is to be preferred However, the complexity of

dynamic modeling is daunting, partly because of the need

to construct input time sequences for a significant period of

operation Nonetheless, for large and costly projects, dynamic

simulations may be warranted Currently, these are in use for

the design of the phosphorus removal wetlands in South

Florida (Walker and Kadlec, 2005)

15.4 HORIZONTAL SUBSURFACE FLOW

WETLANDS

Over the last 50 years, multiple variations of horizontal

sub-surface flow (HSSF) wetlands have been developed, and in

more recent years, HSSF wetland technology has made

sig-nificant advances As a result, there are a large number of

ref-erence documents on HSSF wetlands that contain outmoded

concepts and should no longer be used for design purposes

This section provides an overview of HSSF wetland design

and how the technology has evolved

T HE R OOT -Z ONE M ETHOD

In 1952, a German scientist, Kathe Seidel, began

investigat-ing the water purification capabilities of bulrush

(Schoeno-plectus lacustris) grown in artificial-rooting environments

In the early 1960s, in collaboration with Seidel, Reinhold Kickuth at the University of Göttingen, Germany, developed

a wetland treatment process known as the root-zone method

(Wurzelraumentsorgung) (Brix, 1994d) Root-zone method wetlands are comprised of a soil media (typically clay loam

to sandy clay), and calcium, iron, or aluminum additives were sometimes proposed as additives to the media to increase phosphorus adsorption (Kickuth, 1977) The soil was seen

as a reactive media, a concept that has resurfaced with HSSF wetlands using expanded shale and clay aggregates (see Chapter 10) Although no longer considered a modern design concept, many wetlands designed on the principles of the root-zone method exist in Europe

In most of the root-zone method wetlands, the exclusive

use of the common reed (Phragmites australis) was based on

the theory that root formation would increase the hydraulic conductivity of the soil matrix from approximately 10−5 m/s

to 10−3 m/s within two to four years (Cooper and Boon, 1987) The intended hydraulic conductivity of the soil (10−3m/s) required many of these wetlands to be considerably wider than their length, and the beds were often tipped at a slope of 1% or greater to increase the gradient in the (mistaken) belief that the tipped beds would force subsurface flow through the beds (Figure 15.4)

After a few years, the collaboration between Seidel and Kickuth ended for personal reasons Both scientists and their respective schools became rivals The information produced

by the two groups during this period was often conflict-ing, and resulted in confusion among practicing wastewater

engineers and the governing regulatory authorities (Börner

et al., 1998) By the early 1980s, most wetlands in Germany

were being constructed on the root-zone method, although examples of the MPIP (Seidel) system were constructed in

St Bohaire, France (Liénard et al., 1990), and Oaklands

Park, United Kingdom (Burka and Lawrence, 1990)

E VOLUTION OF HSSF W ETLAND D ESIGN IN E UROPE

In 1984, water utilities in the United Kingdom began to investigate the root-zone method as an alternative for small treatment works (Cooper and Boon, 1987) An effort was made to coordinate research efforts with Denmark, Germany,

FIGURE 15.4 Example of an early root-zone method HSSF wetland (From Cooper and Boon (1987) In Aquatic Plants for Water

Treat-ment and Resource Recovery Reddy and Smith (Eds.), Magnolia Publishing, Orlando, Florida, pp 153–174 Reprinted with permission.)

Phragmites



France, Spain, and the Netherlands Early treatment wetlands

in Denmark and the United Kingdom had designs based on

the root-zone method principles The claim of hydraulic-

conductivity improvement from 10−5 m/s to 10−3 m/s did not

hold true Instead, the hydraulic conductivity either remained

stable or decreased with time, remaining on the order of 10−5 m/s

(Brix and Schierup, 1989a; Netter and Bischofsberger, 1990;

Findlater et al., 1990; Coombes, 1990; Bucksteeg, 1990).

Because the anticipated increase in hydraulic

conductiv-ity did not occur, the HSSF beds did not have enough cross-

sectional area to allow for subsurface flow As a result, water

would pond at the inlet and flow across the surface of the bed

Thus, the hydraulically failed mode of operation for these

wet-lands was overland flow—essentially a FWS wetland with a

very shallow water depth As it has turned out, there is not a large

difference in treatment performance between FWS and HSSF

wetlands, so the overland flow mode of operation has often

been tolerated because it provides an acceptable level of

treat-ment, even though the systems do not function as intended

However, some of these systems were constructed with

sloping beds (top and bottom surfaces both sloped) When

overland flow occurred, the water would channelize, further

reducing treatment performance These experiences led to

the creation of design standards for subsurface-flow wetlands

based on coarser bed materials and beds with flat upper

sur-faces (only bottom surface sloped) (WRc, 1989)

Other root-zone method predictions also turned out to

be inaccurate The actual oxygen transfer from the plant root

systems was much lower than anticipated (Brix and Schierup,

1990) In addition, actual BOD removal was slower (observed

rate removal coefficient of 0.10 m/d versus the predicted 0.19

m/d), resulting in a trend towards larger bed areas (5 m2/PE

versus 2.2 m2/PE) (EC/EWPCA Emergent Hydrophyte

Treat-ment Systems Expert Contact Group and Water Research

Centre, 1990)

Regional differences in HSSF wetland designs still exist

The practice in Germany is to use relatively fine sand (0.2 mm

to 1.0 mm) (Gesellschaft zur Förderung der Abwassertechnik d.V [GFA], 1998) because it is thought to be the best com-promise between available surface area for biofilm growth, suitability as a rooting media, and hydraulic conductivity (Geller, 1996) Because they are designed in accordance with Darcy’s law, these beds are much wider than they are long This design tactic minimizes the organic loading across the inlet cross-sectional area German authorities report that the decomposition of most organic substances is essentially com-plete within 1 to 2 m of the inlet zone (Geller, 1997) Figure 15.5

is a schematic representation of a HSSF wetland in Germany

(Geller et al., 1990).

The HSSF wetland systems in Austria, the Czech Republic, and the United Kingdom are usually designed with the media

of similar diameter Austrian standards recommend the use of a gravel media between 4 and 8 mm (ÖNORM B 2505, 1997) HSSF wetlands in the Czech Republic tend to use sand and gravel less than 20 mm (Vymazal, 1996) In the United King-dom, bed materials of 3 to 6 mm and 5 to 10 mm have been recommended (EC/EWPCA Emergent Hydrophyte Treat-ment Systems Expert Contact Group and Water Research Centre, 1990) HSSF wetlands using coarser materials are often constructed with a length-to-width (L:W) ratio

of greater than 1.0 For example, the average L:W ratio was 1.76:1 for 28 HSSF wetlands in the Czech Republic (Vymazal, 1996)

Wetland systems throughout Scandinavia are typically dimensioned using EC/EWPCA criteria In Norway, the use

of light-expanded clay aggregate (LECA) has been used as

an alternative to gravel (Jenssen et al., 1994b) Assessment

of the performance of these wetlands indicates that LECA

has a high-phosphorus sorption capacity (Jenssen et al.,

1996) compared to natural aggregates As the technology has gained acceptance in Scandinavia, LECA remains one of the preferred media materials because of its availability and

its ability to bind phosphorus (Jenssen et al., 2002)

Phos-phorus removal in LECA wetlands occurs through chemical

Shaft

Regulation of water-level

in the beds Inlet drainage pipe (bottom)

Inlet drainage pipe (bottom) Aboveground inlet

50 m

to IDM

Maisach-river

Aboveground inlet Main

distribution

From settling-basin

Outlet drainage pipe (bottom)

Outlet drainage pipe (bottom) Dyke

Bed with alkali-fine-grained soil

Bed with alkali-coarse-grained soil

FIGURE 15.5 HSSF wetland in Germerswang, Germany (From Cooper and Findlater (1990) Constructed Wetlands in Water Pollution

Control Pergamon Press, New York Reprinted with permission.)

adsorption to the surface of the media (Zhu et al., 1997)

When the sorption sites are exhausted, phosphorus removal

ceases Studies in Sweden indicate that these adsorption

sites are primarily associated with iron and aluminum oxides

present on the surface of the particle, and are a function of

the parent material and manufacturing process (Johansson,

1997) Materials with similar surface characteristics also

dis-play comparable phosphorus removal capabilities (Brooks

et al., 2000) Phosphorus sorption capacities of various SSF

aggregates are discussed in detail in Chapter 10

The HSSF wetlands are recognized as being effective

for BOD removal However, due to the low-oxygen

trans-fer rates in HSSF wetlands, nitrification is limited Vertical

flow wetlands that are capable of nitrification are being used

increasingly in Europe (Cooper et al., 1997) Combinations

of vertical flow and horizontal flow wetlands are being

devel-oped (Platzer, 1996; Cooper, 1999) These hybrid systems are

designed with the goal of nitrification (in the vertical flow

wetland) and denitrification (in the horizontal flow wetland)

E VOLUTION OF HSSF W ETLAND

D ESIGN IN N ORTH A MERICA

In the United States between 1972 and 1976, Frederic Spangler

and colleagues studied artificial wetlands (Spangler et al., 1976b)

in conjunction with a study of the assimilation capabilities of

natural wetlands The earliest artificial marsh cells contained

wire frames to support the emergent wetland plants, but this

was quickly abandoned in favor of gravel media (19 to 25 mm)

covered with smaller pea gravel These early gravel-filled marsh

cells could be operated with the water above or below the gravel

surface, although subsurface flow was identified as the preferred

mode of operation

In 1973, Seidel obtained a U.S patent (Seidel, 1973),

and her concepts were more widely introduced in the United

States (Wolverton, 1987a) By 1981, B.C (Billy) Wolverton

and his colleagues at the National Aeronautics and Space Administration (NASA), who had been researching float-ing-plant treatment systems, began to focus on subsurface-flow wetlands (Wolverton, 1983) These rock–reed filters were very long and narrow and used a coarse gravel media

of crushed railroad ballast (Wolverton, 1987b) Although these HSSF wetlands were similar to the horizontal stages

in the Max Planck Institute Process (MPIP) systems, they operated without the preceding vertical flow wetland cells Figure 15.6 is a schematic representation of an early rock–reed filter for a single-family home (Wolverton and Wolverton, 2001)

Richard Gersberg and colleagues began to study nitro-gen transformations in pilot-scale subsurface-flow wetlands

(Gersberg et al., 1983) and larger demonstration-scale wet-lands (Gersberg et al., 1984) in Santee, California Early

studies treated nitrified secondary effluent, although primary effluents were used in later studies Subsequent work focused

on pathogen reduction in subsurface-flow wetlands (Gersberg

et al., 1987).

In 1988 the U.S Environmental Protection Agency pub-lished a design manual on constructed wetlands that outlined design procedures for subsurface-flow wetlands (U.S EPA, 1988b) This manual included a first-order plug-flow equa-tion for BOD reducequa-tion with a temperature- and

surface-area-adjustable removal rate Media sizes ranged from a d10 of

1 to 8 mm, with associated hydraulic conductivities of 4.9 r

10−3 to 5.8 r−3 m/s The 1988 manual was replaced with updated manuals in 1993 (U.S EPA, 1993c) and 2000 (U.S EPA, 2000a)

The Tennessee Valley Authority (TVA) began demonstra-tion of subsurface-flow constructed wetlands in 1986, based

on design information obtained from the work of Kickuth,

Wolverton, and Gersberg (Steiner et al., 1987) This led to

the development of a guide for small system and single- family home wetland system design in 1991, which was

Elephant ears Roses

Cattails

Septic tank

Plastic liner with rock filter Reeds

Purified wastewater enters lake

FIGURE 15.6 Early rock–reed filter concept for treatment of wastewater from a single-family home (From Wolverton and Wolverton

(2001) Growing Clean Water Wolverton Environmental Services, Picayune, Mississippi Reprinted with permission.)

revised and updated in 1993 (Steiner and Watson, 1993) A

schematic representation from this early guide is shown in

Figure 15.7

The sizing criterion was effectively 11 m2/PE The TVA

guidelines recommended a gravel media of 3 to 6 mm,

pre-sumed to have hydraulic conductivity of 3 r 10−3 m/s, be used

to account for clogging in the gravel bed This guideline also

recommended that the cross-sectional organic loading be

lim-ited to less than 244 g BOD/m2·d to minimize the potential

for bed clogging and associated overland flow

In 1993, the U.S Environmental Protection Agency

pub-lished a technology assessment of subsurface-flow wetlands

(U.S EPA, 1993f) that compared the design methods of

Wolverton and TVA This report recommended a first-order

plug-flow equation for BOD reduction, corrected for

temper-ature Media sizes of 12 to 25 mm with a presumed hydraulic

conductivity of 1.2 r 10−2 m/d and an L:W ratio of 2:1 were

recommended Also in 1993, Region Six of the U.S

Environ-mental Protection Agency published design guidelines that

originally recommended 50- to 250-mm rock media (U.S

EPA, 1993c); this recommendation was later revised to be

in conformance with the 1993 Technology Assessment (U.S

EPA, 1993f)

In 1995, Sherwood Reed and his colleagues published

design recommendations for subsurface-flow wetlands (Reed

et al., 1995), which essentially expanded on an earlier work

(Reed et al., 1988) Reed’s work included thermal design,

a first-order plug-flow equation for BOD and ammonia removal, and an empirical relationship for total suspended solids (TSS) removal Recommended media sizes ranged

from a d10 of 2 to 128 mm, with presumed hydraulic con-ductivities between 1.2 r 10−3 m/s and 2.8 m/s Reed’s work also recommended that no more than 20% of the clean-bed hydraulic conductivity be used for design purposes

In 1996, Robert Knight and Robert Kadlec published design recommendations based on an areal-based, first-order equation set with nonzero background concentrations for BOD, total suspended solids (TSS), ammonia, nitrate, total nitrogen, phosphorus, and fecal coliform (Kadlec and Knight, 1996) This approach was later reported in an international guideline (IWA Specialist Group on Use of Macrophytes in Water Pollution Control, 2000)

Kadlec and Knight provided methods for determining hydraulic conductivity based on the grain size of the media, with the recommendation that only 10% of the clean-bed hydraulic conductivity be used for design purposes Methods

to calculate the water surface profile within the wetland were also provided These methods were applied in a forensic anal-ysis of a HSSF wetland with inlet ponding problems (Kadlec and Watson, 1993)

In 2000, the U.S Environmental Protection Agency published a new design manual on constructed wetlands





















&'*'%
"'%&
'
#"&'%('
*' "


.%&'

+

&%&
#
$ &'
) )&
#%

'&

,

&&
"
'
#''#!
#
'
.%&'

'#
#"&%)
*'%
"
$%#)
!#%



,
%##'&
#
'&
!%&
$ "'&
#%!

"&
!'
!#"
'
%)


*'%
'%
%#!
'
.%&'

$&&&
"'#
'
&#"

'%#(

$%#%'

&*)
&'"$$&
#'
"
#"%'
'"&
'
'
"
#



&'*'%

$$



&

+%
#
%)

#)%
*'
'#$&#

"
'"
!( 

 $"'
%
"
%%#*
 
,
*'%
"


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

%". 
&! %
'#
'#&
(&
*'
#")"'#"
&$'
'"&


#"&'%('
*' "&

'&
#"
%
"
( '

'%#(#('
'
"'#"
'#
" 
*&'*'%
%#!
!#&' +
&!
%(%

#!!("'&
"
#!&
*%
'%'#"
'%'!"'
&+&'!&
%


$%# !

&'*'%
/#*&
"'#
'
#"&'%('
*' "
%#!

&$'
'"
#%

#'%
'+$
#
$%!%+
'%'!"'
&+&'!
%
'
*&'*'%
&
)" +

&'%('
!#"
'
$ "'&
*%
!%##%"&!&
"
!

%'#"&
%
#*"
#%"
!'% &
"
$# ('"'&

#"&'%('
*' "&
$%#)
&!$ 
-')
"
#*
#&'

*&'*'%
'%'!"'
*"
#!$%
*'
#")"'#"
&+&'!&










 

FIGURE 15.7 HSSF wetland schematic from TVA wetland design manual (From Steiner and Watson (1993) General Design, Construction,

and Operation Guidelines: Constructed Wetlands Wastewater Treatment Systems for Small Users Including Individual Residences Second

Edition, TVA/WM-93/10, Tennessee Valley Authority Resource Group Water Management: Chattanooga, Tennessee.)

surface-area-adjustable removal... extended to the P-k-C* model by Wong et al (2006).

A drawback of the use of P-k-C* calculations for

event-driven wetlands is that rate constants for continuous-flow

situations... HSSF wetlands are recognized as being effective

for BOD removal However, due to the low-oxygen

trans-fer rates in HSSF wetlands, nitrification is limited Vertical

flow wetlands