EPA/600/A-94/167
PERFORMANCE OF GRAVEL BED WETLANDS
IN THE UNITED STATES
by
Sherwood C. Reed* and Donald- S. Brown**
Environmental Engineering Consultants
^ RR I, Box 572, Norwich, VT 05055
Risk Reduction Engineering Laboratory
U.S. Environmental Protection Agency
Cincinnati, OH 45268
Contract No. 68-CO-0027
Work Assignment Manager
Donald S. Brown
Water and Hazardous Waste Research Division
Risk Reduction Engineering Laboratory
Cincinnati, OH 452S8
RISK REDUCTION ENGINEERING LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
CINCINNATI, OH 45268
To Be Presented at the IAWPRC Specialist Conference
"Wetland Systems in Water Pollution Control"
Sydney, Australia, Nov. 30 - Dec. 3, 1992
-------
-------
DISCLAIMER
The work described in this paper has been funded by the United
States Environmental Protection Agency.;through the Agency's Risk
Reduction Engineering Laboratory, Cincinnati, Ohio. - However, this
paper has not been subject to the Agency's review and therefore
does not necessarily reflect the views of the Agency, and no
official endorsement should be inferred. ;<- >- :
-------
-------
PERFORMANCE OF GRAVEL BED WETLANDS IN THE UNITED STATES
by: Sherwood C. Reed, P. E.. '
. Environmental Engineering Consultants
RR 1, Box 572
Norwich, VT 05055 U.S.A.
Donald Brown
US EPA RREL
Cincinnati, OH 45268 U.S.A.
INTRODUCTION
iim* Hec!r?] hundn:d 9rave1 bed. °r subsurface flow (SF) wetland systems exist in the
United States ranging in size from single family dwellings to municipal systems
tS npr? FIJI0!8 S0J° n'°°° m /CL Most of these systems have been constructed in
the period 1988 to 1992 without the benefit of a consensus on design, construction Sr
operational procedures. The U.S. Environmental Protection Agency (EPA) commence I a
continuing series of studies in 1989 to identify the critical issues and to dewlSp
S^r^iE0:?8 ,?rUeH1a fy th1S C0ncept' These efforts have incl"d<* a defied
S yi< tS and Performance evaluations, and special data collection and
evaluation programs at selected sites. Two of these special studies were ' coajlitod
mmer 9?1f and tw° are Presently underway. This paper 1s based n the
The focus ^of the paper is on the capability of these systems to remove biochemical
nnth6mand (B?S5)l t0tal susPended solids CTSS), and ammonia nitrogen ^(NH ,1s N^
since these are the major water quality parameters controlled by the regulatory
" A'0'31 °f U fUl1 scal« ^rational systems we re^IeS to develop
.data so"rces are not Identified in the graphical presentations but
* ™
Table 1. Data Sources for Performance Evaluation
Location
Green leaves, LA
Degussa Co., MS
Bear Creek, AL
Monterey, VA
Denham Springs, LA
Benton, LA
Haughton, LA
Carvllle, LA
Mandeville, LA
Benton, KY
Hardln, KY*
Hardln, KY*
Utica, MS8
Utica. MS"
Wastewater
Type
Municipal
Industrial
Domestic
Municipal
Municipal
Municipal
Municipal
Hospital
Municipal
Municipal
Municipal
Municipal
Municipal
Municipal
Design Flow
m3/d
. 564
6737
59
83
6548
378
380
465
4633
685
236
186
189
416
Treatment. Area
ha
0.44
OOQ
. oy
OOrt
• ell
0.02
615
0.61
0.61
0.26
1 AR
0.32-
0.32
Ofii
0.81
* Phragmites bed, * Scirpus bed, 8 North system, Q South system
-------
-------
PERFORMANCE EVALUATIOTT
BOD Removal ^
Input versus output BOD5 data for the 14 systems listed in Table 1 are shown 1n
Figure 1. All effluent values are well below the typical 20 mg/L effluent standard and
this has been achieved regardless of the Input concentration (within the range sihown)
The low values 1n the lower left corner of the graph Illustrate a minor limitation of'
these systems. These systems will typically export an effluent BOD in the range of 5
mg/L due to decomposition of natural organic materials in the system.
30
2O
+ +
+ +
1O 2O 30 40 60 60
»OO M>UT
2
g
i
i
§
100
90
ao
TO
60
eo
to
30
20
10
0
(
* •
• .
• •
1 * '
•
•
•
•
1 1 1 I 1 | t
> 1 2 3 4 5 6 7 t>
WT M
Fig. 1 BOD5 input vs output
Fig. 2 BOD removal vsi HRT
Figure 2 presents the removal of BOD versus the actual hydraulic residence time
(HRT) in these systems. The removal of BOD 1s strongly dependent on HRT up to about 1 d
but improves only slightly thereafter, up to an HRT of 7.5 d. The 60 to 65 percent
removals at about 1 d HRT are not due to ineffective removal capability but rather to
relatively low input levels.
It has been suggested that these SF wetland systems should be constructed with a
high aspect ratio (L:W) to insure plug flow conditions and high levels of performance
Figure 3 tests that hypothesis with L:W from less than 2:1 to over 17:1. As
demonstrated by the figure there does not appear to be any relationship between aspect
ratio and BOD removal. The very low removal associated with the highest L:W in the plot
is due to the very low BOD input-at this-system (< 5 mg/L) and not due to any
relationship with aspect ratio.
Figure 4 illustrates the relationship between mass organic loading rates and mass
removal rates in SF constructed wetlands. A linear relationship is confirmed by the
relatively high r£ value.
90
SO
TO
60
SO
4O
30
20
10
A V
e a 10 12 14
ASPECT RATIO LlW
16 18 20
O 10 20 30 40 SO 00 70 80 90 100110 (20130140 ISO
MO UU* LCMDMft
Fig 3. Bod Removal vs Aspect Ratio
Fig 4. BOD removal vs BOD loading
-------
-------
A first order plug flow model for BOD removal has Seen suggested for design of
these systems in Australia, Europe and the U.S. (Bavor, H.J. et al, 1988, Reed, 1933
Boon, A.G. 1985, US EPA, 1988, WPCF, 1990, Conley, L.M., et al.,^1991).
A plug flow rate constant was calculated for each of the systems listed in Table
1. These results are plotted versus organic loading on Figure 5. The relationship
between the rate constant and the organic loading has an r * of 0.95 indicating an
excellent correlation.
For comparative purposes, the same relationship for facultative lagoons, as
derived by Neel, et al (1961) is also shown on Figure 5. Preliminary work with data
from free water surface wetlands (FWS) indicates a similar relationship with the curve
fall ng about midway between facultative lagoons and SF wetlands. An extension of this
Sit !Tfy analys1^ locates that an FWS wetland might be up to 75% larger than an SF
wetland for comparable flow and BOD removal goals. The choice between the two concepts
may then depend on the availability and cost of land and the cost for the SF media in
tne local area.
wPti«iL1S btl1e^d. ^at the rate constant for SF wetlands is higher than that for FWS
wetlands or facultative lagoons because the media 1n the SF wetland provides more
specific surface area and opportunity for retention of the organisms which contribute
the biological treatment responses. In open water, continuous flow reactors 1t 1s
usually necessary _to provide for sludge recycle and a high "sludge age" to obtain
1
TJ
Z i
CONST;
w
^
tr
1.C
1.4-
1.2-
1.O-
O.6-
0.6-
O.4-
0.2-
0.0
FACULTATIVE LAGOON
O 20 40 60 80 100 12O 14O 16O
ORGANIC LOADING (Kg/ha/d)
Fig. 5 Plug Flow Rate Constant vs Organic Loading
comparable reaction rates. The relationship for SF wetlands shown on Figure 5 suggests
that these systems can be reliably designed for organic loadings up to at least 1IDO
H( u u Tt!ls prov1des support for use of a rate constant in the neighborhood of l.o d~1
which has been used in the plug flow models mentioned above.
The relationship shown on Figure 5 also suggests that many of the existing-systems
are larger than they need to be for effective BOD removal. This has probably occurred
because the designers made conservative estimates for input BOD and the actual input
BOD has been significantly less. '
-------
-------
TrF
OT.
results or a tracer study, using lithium chloride, ^oriducted
m -vcrrm i c , ;ur> . '
>•»
caci i CTn
.
the four such studies conducted under the EP& program, Issentially 100 percent of the
tracer was accounted for 1n the effluent so it can be considered a valid study. The
centrold of the curve at 48 hr 1s Identical to the theoretical HRT for the system. It
clearly does not exhibit ideal plug flow responses but is similar to curves for "lagoons
and other reactors designed with plug flow kinetics. The plug flow model seems to give
reasonably accurate estimates of system performance. As additional data is obtained the
use of alternative models such as a series of continuously stirred reactors may be
adopted, but the plug flow model is likely to continue 1n use for the near term.
HAK AT »« HR
cannon. HRT » 41 m
100* u Btcovm
20
40 60
TKE fexn)
—I • 1—
80 100
3O
20
10
2O ing/L
A
-J 1 1_
120
Fig 6. Lithium Tracer Study
10 20 30 40 50 60 70 60 90 .00 11O 120
«u»w»»eo MUM HHIT 50 mg/L) have facultative lagoons for preliminary
treatment and the high solids are due to algal carry-over from the lagoon. The;'removal
and decomposition of these solids in the SF bed has an impact on the NH, status 1n the
system as discussed in the next section of this paper.
The relationship between TSS removal and HRT in the system is similar to the BOD ''
results shown on Figure 2. After about 1 day HRT there is little significant *
Improvement. The relationship between TSS removal and aspect ratio is also similar to
the BOD results shown on Figure 3, indicating that there is no correlation between
system aspect ratio and TSS removal.
Concerns have been expressed over the potential for gradual clogging of these beds
with TSS or dead root/rhizome material. The special EPA studies have excavated pits in
four systems in Louisiana ranging in age from 2 to 5 years and minimal clogging has
been observed in all cases. In three cases, the solids represent less than two percent
of the available void spaces; 1n the worst case the solids approached six percent of
the available void spaces. In all cases, these trapped solids were composed of at least
80 percent Inorganic matter. .Discussions with the operators and review of the
construction histories Indicate that these materials may have been delivered to the
site during construction either as fines with the media or as soil on the tires of the
trucks and other equipment, therefore,- further accumulation may be minimal
The surface flow which can be observed on many of these systems has been
attributed to clogging of the media. A more likely explanation is provision of an
Inadequate hydraulic gradient in the hydraulic design of the system. Many of these
early systems had L:W ratios approaching 10:1, had a flat bottom on the bed, and had
the outlet ports in the effluent manifold near the top of the bed.
-------
-------
The ma^or concern is the removal of unoxidized ammonia (NH3) to _
I
!-
I
100
60
60
40
BO
0
-so
-40
HSO
-W
_aiV\
D
a
»
,4 *
«
i (I
4
A
' » *
a POJ^TS eeLow -100 «)
Fig 8. NH3 Input vs Output
MVOAAUUO HtMMMQK TM |O9
Fig 9. NH3 Removal vs HRT
oxygen to oxidize this ammonia to nitrate. Support for this hypothesis is shown on
Figure 10 which presents ammonia removal versus HRT,
Most c the systems shown on Figure 9 display a marginal or a negative ammonia
removal rr regardless of detention time. Two of the systems, shown as open squares in
thlfigure .splay veVy high removal rates at comparable HRT levels. The difference for
these ?wo - ata sets is that the root zone was fully developed in the bed so thai plant
available '.ygen could support nitrification. In one case (Bear Creek, AL) the bed is
only 03m .eep and supports a stand of Typha, with the roots penetrating to th.» bottom
f the'fine gravel bed (Watson, 1990). Ammonia removal at this system reaches 80
percent with an HRT of 3.9 d. The second case is the Sdrpus bed at the Santee, CA
pilot system, where the roots also penetrated to the bottom of the 0.76 m bed and
ammonia removals of 94 percent were achieved with an HRT of 7d with primary treated
wastewater as Input (Qersberg, et al, 1985).
Many of the early investigators and designers assumed that the plant roots would
always penetrate to their full potential depth so the entire depth of the bed would be
an active root zone with at least some oxygen available. The recent EPA studies
indicate that root/rhizome penetration is limited to about 0.3 m regardless of the
plant species used. Deeper roots can be found ne.ir the edges of the bed and other dead
soots" for flow but in general the active root zone is 0.3 m or less in most systems in
the u S As a result, a significant portion of the flow at these systems never comes in
contact with the root zone and has no opportunity to utilize the available oxygen for
nitrification. At the Bear Creek and Santee systems shown on Figure 9 all of the flow
comes in contact with the root zone and there is sufficient residence time to complete
the nitrification reactions. There is no consensus on how much oxygen might be
available at the plant roots to support nitrification. Estimates range from zero to
about 45 gm 09/m2/d (Cooper, et a! 1990). Based on the experience at Santee and at the
Bear Creek site it seems likely that some plant produced oxygen is available in the
root zone, on the surfaces of the root hairs. It can be calculated from the ammonia
removal data at these sites that the available oxygen is about 7.5 gm/m /d in the
active root zone for Typha, Scirpus, and Phragmites. Assuming a 0.6 m deep active root
zone this would translate to 4.5 gm 02/m2/d of wetland surflcial area, which 1s; near
-------
-------
the low end of available oxygen values rin t&e ,literatuirea^u> "" '. -
Based on the work at Santee and in Europe the maximum potential root zone depths
are: Typha 0.3 m, Scirpus 0.8 m, Phragmites 0.6 m. Using these values and the 7.5
g/m-Vd oxygen availability, it is possible to show that nitrification to low levels of
ammonia ($ 2 mg/L) will require at least 5 to 6 days residence time in the system
during the warm weather growing seasoii.
Other models are available for predicting ammonia removal (Bavor, 1988, WPCF, 1990)
in SF constructed wetlands, but these predict even larger land areas than the previous
case. The reason is believed to be that these models were derived from systems which
were generally deficient in oxygen and do not reflect the potential nitrification if the
root zone is fully developed. These models may adequately describe the performance of
the present generation of systems in the U.S. which tend to have short detention times
and inadequate root zone development.
However, the costs of the land and the media for the SF bed in a system with six
basis te
i?t unsa*urated vertical flow reactors for nitrification such as
o + I a"d/eci Bating sand filters has been common practice in
wastewater treatment for a number of years. The use of vertical flow wetland eel 'Is
been demonstrated and discussed in the literature (Cooper, it al 1990^
A vertical flow recirculating filter composed of fine gravel has been desianed for
" IC§ntUCky WMch W3S haVln* Prob^eL Beting the 9'
lS-S^" ^ IC§ntUCky WMch W3S haVln* Probe eting the
fS^1^iSr4SJ^F^^rS^ 2£ 5^5 $3 twos70seKt? S
high hydraulic conductivity. Based on nitrification experience with other attach S
growth processes it appears that about 1230 m* of this secific surfa
surace area
required to oxidize 1 kg of ammonia nitrogen per day with a possible recycle ratio of
up to 3:1. The depth of the bed to be used is 0.6 m! The ni?r1? ca?ion bed Jm L
superimposed on top of the existing SF. constructed, wet! and? at the head of thllll
. , o
ltr rWi }l ^Z the T irculated ^fluent from the SF cell to the ?op of S
nnt^C; I nitrified percolate is expected to drain through the media and mix with the
untreated wastewater in the bed. Denitrification of the nitrate 1s then expected £
occur. This system was under construction 1n April 1992 and long term data
'
and
ar0a This.reci rc"!atl°n exponent combined with a normal SF bed should need less land
for nJ&JEJtS T^3" a SySt!m dSSl9ned for total reliance °» Plant amiable oJygen
h?L ow t I ": 1S approach may be tne ««»t cost effective method for ach eving
high levels of nitrogen removal 1n these constructed wetland systems. acrnevin9
Phosphorus Removal
Phosphorus removal was not consistent in the systems Included in this studv Tn
associated with the gravel may have provided the adsorption sites needed for phosphorus
-------
-------
removal.
Role of Vegetation
°" ***** Jjvest1gat1ons described in this paper it appears the major
r?h ?6 r9f ati?" 1n these SF systems 1s Serv1ce of the root/rhizome
?^strate for microbial activity and as a limited oxygen source for
I I suggests that if the plant is expected to play a major role the
it 5? ld ?0t 6XCeed the P°tential ro°t development for the plant species
ma ntn'rnot ™ f9?ests £hat a management plan will be necessary to induce and
maintain root penetrations below the 0.3 m depth commonly found in most systems in the
in thIhtonth«rn°M I?"*1™ harvesting is somewhat controversial in the U.S. Some systems
if £ ?h2 U'f; °?ndUCt a rout1ne annua1 harvest regardless of the plant species
ial r^l ± L°Plni°n that SUch a program is not necessary when plants such
Ai2S ? Phragmites are used, but may be necessary when soft tissue
plants are the dominant species.
Preliminary Treatment Requirement
«„«+ S0m6 f°*m °f .Preljminary treatment is necessary for these SF wetland systems The
most common form in the U.S. is facultative lagoons since in many cases the wetland
XsTre sunablf for* IHK^^99^ ** * P°11Shin3 ^ep Se^HanK S'Soff
J™1 * suitable for small to moderately sized systems. Larger size systems may need
SfTuInt? m6ChaniCal Pre11min^y treatment but only to the equivalent of pMmary
COSTS
Ihe C^ °J these systems are Presented in other papers at this conference and
h ah!er? iR6ed> S>C" S- ?r°Wn' 1992)' The^°'- i^ue of- concern Is^SirolatlveW
high cost to procure and place the media in the SF wetland bed. This one factor III
s^TSc50 I? 62 PefCent °f the t0tal Construction costs. A cost "caparison between
mod3^ ,Hetcand SYStemS ^"^ Depend- on -the- cost -of the land-and on the cost of ?he
media for the SF concept. Even though the FWS system is likely to require a larger land
cos? wh1^ ^cept will be the more cost effective became of SIse two
The SF concept offers other advantages over the FWS alternative in that the
subsurface flow provides positive control over odors and insec? vectors and leslens
public access concerns. These factors are particularly important when systems are to be
located adjacent to habitations or at public facilities. systems ars to be
CONCLUSIONS
5
suited for smal1 to moderate
U S lTrioia^TalJn m°St °^ th6 present seneration of operating SF systems in the
U.S. is deficient. The reason is believed to be the short detention time and the lack
of oxygen in the bed profile to support nitrification.
-------
-------
_.H^!ggi^sggm to be avaiiabletojn^ucgjiand maintain root zone development and the
Xy9e source:~"™rs WTTT require •wa^erneWrmsflaggBiarc'^ the bed utffff^
n , e e u
adjustable outlet mechanism, A system with "a fully developed root zone might stni
require up to six days to produce low levels of ammonia.
The surface flow observed on many systems in the U.S. is believed due to
inadequate hydraulic design and not to clogging of the bed. The use of an adjustable
outlet mechanism should correct these problems.
A recirculating nitrification filter in combination with an SF wetland seems to
offer promise as a cost effective method for achieving effective nitrogen removal in
these systems.
up JllvelTrfaf lEt'U'TSlSE'" d1SPUyS a I1near "'""-IP to mass ,oad.ng
A first order plug flow kinetic model seems to provide a reasonably accurate
estimate of BOD removal capability in SF constructed wetlands.
REFERENCES
Bavor, H.J. et al. (1988). Joint Study on Sewage Treatment Using Shallow
Lagoon - Aquatic Plant Systems, Water Research Laboratory, Hawkesbury
Agricultural College, Richmond, NSW, Australia.
for yaste Ha"S9eKnt ' Treatment
Boon, A.6. (1985) Report of a Visit by Members and Staff of WRC to Germany
to Investigate the Root Zone Method for Treatment of Wastewaters, Water
research Center, Stevenage, England.
US EPA, (1988) Design Manual - Constructed Wetlands and Aquatic Plant
Systems for Municipal Wastewater Treatment,, EPA 625/11-88/022, US EPA CERI
Cincinnati , OH. . '
WPCF, (1990) Natural Systems for Wastewater Treatment, ! Manual of Practice
FD-16, Water Pollution Control Federation, Alexandria, VA.
Conley, L.M., et al. (1991) An Assessment of the Root Zone Method of
Wastewater Treatment, Jour. WPCF (63)3,239-247.
Neel J.K., et al (1961). Experimental Lagooning of Raw Sewage, Jour. Water
Pollution Control Fed., 33(6)603-641.
•«-» i'l' i1"°l %LS.1£" and Performance of the Constructed Wetland Wastewater
Treatment System at Phillips High School, Bear Creek, AL, TVA/WR/WQ-90/5 , TVA
Chattanooga, TN. • *
Gersberg, R.M. , et al, (1985). Role of Aquatic Plants in Wastewater
Treatment by Artificial Wetlands, Water Research, 20:363-367.
Cooper, P.P.,. Find later, B.C., Editors, (1990) Constructed Wetlands in •
Water Pollution Control, Pergamon Press, New York, NY.
i!
Reed, S.C., D. Brown (1992). Constructed Wetland Design - The First
Generation, Jour. WEF, (in press).
-------
------- |