read the attached article write at least
-at least 3 critique about this paper and
?	At least 3 questions that you have about the paper that is unclear 


This is an unformatted preview. Please download the attached document for the original format.
Ecological Applications, 18(3), 2008, pp. 789–804
Ó 2008 by the Ecological Society of America

EFFECTS OF STREAM RESTORATION ON DENITRIFICATION
IN AN URBANIZING WATERSHED
SUJAY S. KAUSHAL,1,5 PETER M. GROFFMAN,2 PAUL M. MAYER,3 ELISE STRIZ,3

AND

ARTHUR J. GOLD4

1

University of Maryland, Center for Environmental Science, Appalachian Laboratory,
301 Braddock Road, Frostburg, Maryland 21532 USA
Institute of Ecosystem Studies, Box AB, Route 44 A, Millbrook, New York 12545 USA
3
Of?ce of Research and Development, National Risk Research Management Laboratory,
U.S. Environmental Protection Agency, Ada, Oklahoma 74820 USA
4
University of Rhode Island, Department of Natural Resources, Kingston, Rhode Island 02881 USA
2

Abstract. Increased delivery of nitrogen due to urbanization and stream ecosystem
degradation is contributing to eutrophication in coastal regions of the eastern United States.
We tested whether geomorphic restoration involving hydrologic ‘‘reconnection’’ of a stream to
its ?oodplain could increase rates of denitri?cation at the riparian-zone–stream interface of an
urban stream in Baltimore, Maryland. Rates of denitri?cation measured using in situ 15N
tracer additions were spatially variable across sites and years and ranged from undetectable to
.200 lg NÁ(kg sediment)À1ÁdÀ1. Mean rates of denitri?cation were signi?cantly greater in the
restored reach of the stream at 77.4 6 12.6 lg NÁkgÀ1ÁdÀ1 (mean 6 SE) as compared to the
unrestored reach at 34.8 6 8.0 lg NÁkgÀ1ÁdÀ1. Concentrations of nitrate-N in groundwater and
stream water in the restored reach were also signi?cantly lower than in the unrestored reach,
but this may have also been associated with differences in sources and hydrologic ?ow paths.
Riparian areas with low, hydrologically ‘‘connected’’ streambanks designed to promote
?ooding and dissipation of erosive force for storm water management had substantially higher
rates of denitri?cation than restored high ‘‘nonconnected’’ banks and both unrestored low and
high banks. Coupled measurements of hyporheic groundwater ?ow and in situ denitri?cation
rates indicated that up to 1.16 mg NO3À-N could be removed per liter of groundwater ?ow
through one cubic meter of sediment at the riparian-zone–stream interface over a mean
residence time of 4.97 d in the unrestored reach, and estimates of mass removal of nitrate-N in
the restored reach were also considerable. Mass removal of nitrate-N appeared to be strongly
in?uenced by hydrologic residence time in unrestored and restored reaches. Our results suggest
that stream restoration designed to ‘‘reconnect’’ stream channels with ?oodplains can increase
denitri?cation rates, that there can be substantial variability in the ef?cacy of stream
restoration designs, and that more work is necessary to elucidate which designs can be effective
in conjunction with watershed strategies to reduce nitrate-N sources to streams.
Key words: Chesapeake Bay, USA; eutrophication; nitrogen; stream restoration; urbanization.

INTRODUCTION
Many coastal water bodies in the United States now
receive large loads of nitrogen as a result of land use
change (e.g., Howarth et al. 1996, Vitousek et al. 1997,
Paul and Meyer 2001). In the Chesapeake Bay
watershed, rapid expansion of urban, suburban, and
exurban land (Brown et al. 2005, Jantz et al. 2005) has
coincided with increases in eutrophication, hypoxia, and
harmful algal blooms in coastal waters (e.g., Boesch et
al. 2001, Howarth et al. 2002, Kemp et al. 2005). Land
use change has complicated efforts to identify sources
and ‘‘sinks’’ of nitrogen in this watershed (Boesch et al.
Manuscript received 13 July 2007; revised 5 September 2007;
accepted 11 September 2007. Corresponding Editor: J. S.
Baron.
5 Present address: University of Maryland, Center for
Environmental Science, Chesapeake Biological Laboratory, 1
Williams Street, P.O. Box 38, Solomons, Maryland 20688
USA. E-mail: kaushal@cbl.umccs.edu

2001), and many river miles of suburban and urban
streams are now being restored in the Chesapeake Bay
watershed and other areas of the United States with
ancillary objectives of improving water quality (Bernhardt et al. 2005, Hassett et al. 2005). Despite the
billions of dollars currently invested in the .37 000
stream restoration projects in the United States, there
are few actual measurements of the effects of stream
restoration on denitri?cation (Bernhardt et al. 2005,
Hassett et al. 2005). We quanti?ed rates of denitri?cation at the riparian-zone–stream interface of an urbanizing watershed using an in situ stable isotope approach
and investigated the potential for stream restoration
associated with storm water management to increase
rates of denitri?cation in riparian sediments.
Although ?uxes of total nitrogen from river basins
have doubled globally since pre-industrial times (Green
et al. 2004), it is estimated that only between 20% and
30% of the nitrogen that is added to large watersheds of
the eastern United States is delivered to coastal waters

789

790

Ecological Applications
Vol. 18, No. 3

SUJAY S. KAUSHAL ET AL.

(Howarth et al. 1996, Boyer et al. 2002). A considerable
amount of nitrogen appears to be removed via
denitri?cation, the microbial conversion of nitrate to
N2 and N2O gases (e.g., Alexander et al. 2000, Seitzinger
et al. 2002). From a landscape perspective, the interface
of riparian zones and streams may be a ‘‘hot spot’’ of
denitri?cation due to low oxygen and subsidies of
bioavailable organic carbon from streams and upland
sources in these zones (e.g., Peterjohn and Correll 1984,
Hedin et al. 1998, McClain et al. 2003). Small streams
can constitute up to 85% of total stream length within
drainage networks (Peterson et al. 2001), and rates of
denitri?cation have been shown to increase in riparian
groundwater with increasing proximity to streams
(Hedin et al. 1998, Kellogg et al. 2005). Recent in situ
measurements have shown that denitri?cation removes
;15% of nitrate in forest streams with very low
concentrations of nitrate (Mulholland et al. 2004) and
can remove 34–50% of nitrate in agricultural streams
with higher concentrations of nitrogen (Bohlke et al.
¨
2004, Pribyl et al. 2005).
In many urbanizing areas, the capacity of streams and
rivers to process and remove nitrogen has been impaired
due to the effects of surrounding land use change
(Groffman et al. 2002, Grimm et al. 2005, Kaushal et al.
2006). Increased inputs of nitrogen to headwater streams
from anthropogenic sources and domestic wastewater
can saturate biological demand, leading to increased
downstream transport of nitrogen (Bernot and Dodds
2005, Gucker and Pusch 2006). Channel incision as a
¨
result of erosive runoff from impervious surfaces (Wolman 1967, Henshaw and Booth 2000) can reduce
contact between water in the channel and the streambed
and subsurface zones of nitrogen uptake and removal in
sediments (Paul and Meyer 2001, Groffman et al. 2002,
Sweeney et al. 2004). Reduced in?ltration capacity in
watersheds due to impervious surfaces such as roadways
and parking lots, in combination with stream incision,
can also lead to marked reductions in riparian water
tables and soil moisture levels. Reductions in riparian
water tables decrease hydrologic ‘‘connectivity’’ between
riparian groundwater and the stream (Groffman et al.
2002, Walsh et al. 2004) and contribute to transmission
of nitrate to streams via deeper hydrologic ?ow paths in
riparian zones (Groffman et al. 2002, 2003, Bohlke et al.
¨
2007). These hydrologic changes can reduce denitri?cation rates in riparian zones and also cause incoming
nitrate from groundwater to bypass active sites of
riparian denitri?cation (Groffman et al. 2003, Bohlke
¨
et al. 2007).
In response to widespread urban stream degradation,
many suburban and urban streams and rivers are being
restored in the Chesapeake Bay region and the United
States (Bernhardt et al. 2005, Hassett et al. 2005, Wohl
et al. 2005). The primary goals of most of these
restorations are promoting geomorphic stability and/or
improved storm water management practices, but water
quality improvement is often listed as an ancillary goal

of these efforts (Bernhardt et al. 2005, Hassett et al.
2005). Previous work has suggested that stream restoration has the potential to improve water quality (e.g.,
Stanley and Doyle 2002, Mayer et al. 2003, Groffman et
al. 2005), and recent work suggests that stream
restoration techniques may in?uence N retention via
alteration of hydrologic residence time (e.g., Kasahara
and Hill 2006, Boulton 2007, Bukaveckas 2007, Roberts
et al. 2007). Little is currently known, however, about
the speci?c types of restoration that are most effective
and the relative role of denitri?cation in N uptake and
transformation, and more empirical data is needed to
evaluate ef?cacy and variability across restoration sites,
given that the number of restoration projects is rapidly
growing (Wohl et al. 2005, Palmer and Bernhardt 2006).
In this study, we quanti?ed the effects of geomorphic
stream restoration on rates of in situ N removal via
denitri?cation using 15N-based ‘‘push-pull’’ methods
(e.g., Istok et al. 1997, Addy et al. 2002, Whitmire and
Hamilton 2005) along the riparian-zone–stream interface of a coastal stream in Baltimore, Maryland, USA.
This stream had been partially restored based on
geomorphic reconstruction of the stream channel
following design models of Rosgen (1996) using high
armored banks and also an experimental ?oodplain
subreach that did not use rigid structures to ?x the
stream channel in place. We hypothesized that stream
restoration has the potential to increase denitri?cation
rates at the riparian-zone–stream interface in an
urbanizing watershed, but that restoration designs
promoting low banks with increased hydrologic connectivity at the riparian-zone–stream interface would
show the highest rates of denitri?cation. A secondary
objective was to investigate the potential importance of
the riparian-zone–stream interface as a site for mass
removal of nitrate-N by coupling measured in situ
denitri?cation rates with estimates of groundwater ?ow.
Results of the present study provide estimates of in situ
denitri?cation along the riparian-zone–stream interface
of a restored stream and make a further contribution
toward the investigation of the importance of denitri?cation rates associated with degraded urban ecosystems
and certain forms of stream restoration.
METHODS
Site description
Minebank Run is a low-order stream with a
watershed area of ;8.47 km2 located in Baltimore
County, Maryland (latitude 39824 0 43 00 N, longitude
76833 0 12.5 00 W; Fig. 1). It lies in the Piedmont
physiographic province of the eastern United States
with its headwaters originating from a storm drain in a
densely urbanized section of Towson, Maryland, and
drains into the Gun Powder River, a tributary of the
Chesapeake Bay. The segment of the Minebank Run
watershed related to the present study was developed in
the 1950s and 1960s, prior to implementation of storm

April 2008

STREAM RESTORATION AND DENITRIFICATION

791

FIG. 1. Location of the Minebank Run study area, Baltimore County, Maryland, USA. Detailed locations are indicated in the
unrestored and restored reaches of Minebank Run for groundwater wells used in the in situ denitri?cation measurements, sampling
sites for stream water chemistry, and piezometer transects used for groundwater chemistry and hydrologic ?ow measurements. The
?gure is reprinted from Doheny et al. (2006).

water management regulations. Approximately 30–35%
of the drainage area is covered by impervious surfaces,
with 81% of the land area classi?ed as urban/suburban,
17% as forest/open space, and 2% as agriculture/farmland (Doheny et al. 2006). The watershed ranges in
elevation from ;122–152 m above sea level at the
drainage boundaries to ;46–122 m above sea level in the

stream valley. Stream slopes are ;1% in most locations,
but tend to be somewhat larger in the headwater areas.
The combination of fairly large stream slopes, signi?cant
relief, and impervious surfaces in the headwater areas
cause the stream stage to increase and decrease very
rapidly during storms (Doheny et al. 2006; Biohabitats,
Inc., et al., unpublished manuscript).

792

SUJAY S. KAUSHAL ET AL.

The long-term annual mean precipitation for Baltimore is 106.7 cm, and annual precipitation at Minebank
Run was 83.7 cm, 163.0 cm, and 131.3 cm, respectively,
during water years 2002, 2003, and 2004 (Doheny et al.
2006). During water years 1997 to 2004, annual runoff
averaged ;41.9 cm and annual mean discharge averaged
;0.099 m3/s (Doheny et al. 2006). Instantaneous ?ood
discharges exceeded 42.475 m3/s in response to storm
events occasionally during years from 1997 to 1998 with
substantial erosive force in the stream channel leading to
severe degradation of banks, exposed sewer lines, and
fractures in concrete structures (Doheny et al. 2006).
Stream restoration
The headwater reach of Minebank Run, draining 2.07
km2 of the watershed, was restored in 1998 and 1999,
and a lower reach of the stream draining 6.40 km2 of the
watershed was left unrestored and slated for restoration.
The Baltimore County Department of Environmental
Protection and Resource Management (DEPRM) used
geomorphic reconstruction techniques to remediate
severe stream incision from erosion and increase
geomorphic stability in the headwater reach (Fig. 2A).
Restoration efforts included ?lling the channel with
sediment, cobbles, and boulders and constructing point
bars, rif?es, and meander features along the reach and
creating step-pool sequences. In the riparian zone,
dominant planted trees included Acer saccarum, Fagus
grandifolia, Liriodendron tulipifera, Quercus alba, and Q.
rubra, and planted shrubs and herbs included dominant
varieties such as Kalmia latifolia, Andropogon gerardii,
and Panicum virgatum. Banks were stabilized in some
reaches by employing one or more of the following
techniques: reshaping slopes to reconnect the channel to
the ?oodplain, embedding root wads, planting cover
vegetation, and covering with erosion mats. In some
areas, incised, high banks prone to erosion were
armored with rocks and channelized to keep water in
the stream and rapidly transport water away from
commercial properties, with less potential for overbank
?ooding (Fig. 2B). A ‘‘nonconnected’’ armored and
channelized subreach was selected for the present study,
;50 m in length. In other areas away from commercial
properties, low banks were engineered to promote
?ooding over the banks and dissipation of erosive force,
creating low, hydrologically ‘‘connected’’ riparian areas
(Fig. 2C). This hydrologically connected subreach in the
present study was ;150 m in length. Mean bank height
in the entire restored reach was signi?cantly lower at
77.0 6 11.0 cm (mean 6 SE; N ¼ 12 replicates) than
mean bank height in the entire unrestored reach at 114.7
6 7.4 cm (N ¼ 21 replicates; two-sample t test, t ¼ 2.928,
P ¼ 0.006).
In the unrestored reach, steeply eroded banks and
general geomorphic instability were common, with
incision of up to 2–3 m revealing the bedrock in some
places. Riparian zones along the unrestored section of
Minebank Run consist of mixed hardwood, second

Ecological Applications
Vol. 18, No. 3

growth forest interspersed among mowed grass areas in
a county park.
Stream and groundwater sampling
Surface water samples from Minebank Run were
collected approximately every two weeks from April
2003 to December 2005 in both the unrestored and the
restored reaches of Minebank Run. Time series samples
for nitrate concentrations were collected at U.S.
Geological Survey (USGS) gauged station 0158397925,
Minebank Run at Intervale Court near Towson,
Maryland, since June of 2004 and USGS station
0158397967, Minebank Run near Glen Arm, Maryland,
since July of 2002 (information on site locations,
descriptions, and data from the USGS stations is
available online).6,7 Groundwater was sampled in the
unrestored and restored reaches during April, July, and
October 2003 and May 2004. Groundwater was
collected with a peristaltic pump through a ?ow cell
and Hydrolab (Hach, Loveland, Colorado, USA) from
piezometer nests installed in the stream channel and
stream banks. The network of piezometers was designed
to quantify spatial and temporal variability of hydrology
and biogeochemistry among stream features that were
altered signi?cantly by the restoration. Piezometers for
groundwater chemistry were installed along three
perpendicular transects in a severely incised and eroding
section of the unrestored reach and along two perpendicular transects in the restored reach where restored
channel features included low, hydrologically connected
banks for spreading of water over the ?oodplain and
dissipation of erosive force.
Piezometers consisted either of 2.5 cm diameter
stainless steel pipes or 0.95-cm polypropylene tubing
screened at the lower 15 cm with 0.25-mm stainless steel
mesh. Piezometers were arranged in nests of three wells
placed ;1 m apart with screens positioned 61, 122, and
183 cm below the surface of the streambed. Stream bank
piezometers were installed 7.7 6 1.9 m from the stream
channel thalweg at depths that matched the mean
elevation of the channel piezometers. One additional
piezometer nest was installed along the middle transect
of the unrestored reach to account for a large meander
in the stream at that position. Piezometers in the
restored reach were located between two automated
stream gauges operated by the U.S. Geological Survey
that recorded stream ?ow at ?ve-minute intervals.
Distances between transects were 38 m in the restored
reach and 72 m and 45 m in the unrestored reach. A
total of 33 piezometers and four surface water stations
(located between the piezometer transects) were sampled
at the unrestored reach. Eighteen piezometers and three
surface water stations (located between the piezometer
transects) were sampled at the restored reach. Samples
6 hh t t p : / / n w i s . w a t e r d a t a . u s g s . g o v / n w i s / n w i s m a n /
?site_no¼0158397925i
7 hhttp://waterdata.usgs.gov/nwis/uv?0158397967i

April 2008

STREAM RESTORATION AND DENITRIFICATION

793

FIG. 2. (A) Unrestored reach of Minebank Run showing extreme incision of banks and the stream channel. The cross section
shifted on 10 December 2002 (dark line) and became more incised on 8 July 2003 (lighter line). (B) Restored ‘‘nonconnected’’ reach
showing armoring of banks with rocks and channelized stream cross sections that promote rapid drainage of surface water away
from commercial property and decrease connections between the stream and the riparian zone. (C) Restored ‘‘connected’’ reach
showing step pool sequences, meanders, and engineered stream cross sections that promote overland ?ooding to dissipate erosive
force.

were stored on ice and acidi?ed to pH 2 and/or ?ltered
with 0.45-lm ?lters before subsequent chemical analyses
for water chemistry.
Measurement of in situ denitri?cation rates
We measured rates of in situ denitri?cation rates in
both low and high banks of the restored and unrestored
reaches of Minebank Run using a push-pull method
(similar to Trudell et al. 1986, Istok et al. 1997, Addy et

al. 2002) in June 2003, November 2003, and June 2004.
Measurements of denitri?cation were taken during these
months to quantify N removal during spring before
summer base-?ow conditions and during autumn before
snow cover. Detailed descriptions of the method can be
found in Addy et al. (2002) and Kellogg et al. (2005).
Brie?y, we ‘‘pushed’’ (i.e., injected) 10 L of previously
collected groundwater amended with 15N-enriched
NO3À and SF6 into a single mini-piezometer and then

794

Ecological Applications
Vol. 18, No. 3

SUJAY S. KAUSHAL ET AL.

‘‘pulled’’ (i.e., extracted) groundwater from the same
mini-piezometer after an incubation period of 4 h. Prior
to injection, the groundwater was adjusted to ambient
dissolved oxygen concentrations to mimic aquifer
conditions. Groundwater samples were analyzed for
15
N-enriched denitri?cation gases (N2O and N2) and a
conservative tracer (SF6) to provide information on the
recovery of the introduced plume. Denitri?cation rates
were estimated from only the ‘‘core’’ of the plume (the
?rst 2 L of the plume pulled from the mini-piezometer)
after the incubation period to minimize the confounding
effects of dispersion and advection.
Mini-piezometers (similar to those described by
Winter et al. [1988]) consisting of small steel well points
(1.8-cm outer diameter, 2-cm screen length; AMS,
American Falls, Idaho, USA) attached to gas-impermeable Te?on tubing extending into the sediment were
installed in four locations to a depth of 0.5 m below the
adjacent streambed. In the unrestored reach, two minipiezometers were installed on a high bank affected by
deep channel incision and three were installed on a low
opposite bank. In the restored reach, two minipiezomenters were established in an area where low
banks were engineered to promote ?ooding over the
banks (i.e., a low, hydrologically ‘‘connected’’ riparian
area), and two were established in a high bank that was
armored and stabilized by stones (i.e., a hydrologically
‘‘nonconnected’’ area). In both unrestored and restored
reaches, the two to three wells on a given bank were
placed .5 m apart. Locations of wells were assumed to
be independent because previous work has shown that
denitri?cation rates can be ‘‘patchy’’ over smaller spatial
scales such as these (Gold et al. 1998). In all four
locations, mini-piezometers were established ;0.5 m
from the stream bank.
To prepare for in situ NO3À removal push-pull tests,
we collected 10 L of groundwater from each minipiezometer ;24–48 h before performing isotopic pushpull tests on each mini-piezometer (10 L of water was
drawn from each well before being pushed back into the
same piezometer for denitri?cation measurements).
Groundwater from each piezometer was stored at 48C
(maximum of two-day storage in a carboy) until the
push-pull test. Each individual dosing solution speci?c
to a piezometer consisted of 10 L of ambient groundwater enriched with 32 mg BrÀ/L (as KBr) and 32 mg
NO3À-N/L (as KNO3; isotopically enriched [20 atom
percent 15N]). Because nitrate concentrations were
enriched relative to background, this may have contributed to elevated denitri?cation rates representing potential rates instead of actual rates. Previous work has
suggested, however, that denitri?cation may be less
limited by nitrate concentrations when background
nitrate concentrations are high and may be limited
instead by microbial diffusional constraints (Myrold and
Tiedje 1985). SF6 (100 lL/L) balanced in helium
(Matheson Trigas, Gloucester, Massachusetts, USA)
was bubbled into the dosing solution to saturate the

solution with SF6 and lower the dissolved oxygen (DO)
to ambient levels (;20 min per solution).
Samples of the dosing solution were taken for
dissolved solute and gas analysis (NO3À, N2, N2O,
15
N2, 15N2O) during each push phase. The 10-L dosing
solution was pushed into mini-piezometers with a
peristaltic pump at a slow rate (;10 L/h) to minimize
changes in the hydraulic potential surrounding the minipiezometer. After a 4-h incubation period, we obtained
samples (the pull phase) of groundwater from each minipiezometer. Groundwater from the mini-piezometers
was pumped slowly (9–13 L/h) to avoid generating air
bubbles within the tubing. Groundwater and gas
samples were collected at periodic intervals throughout
the pull phase. All groundwater samples for water
chemistry were stored at 48C until analysis.
Groundwater samples to be analyzed for dissolved
gases (N2, N2O, 15N2, 15N2O, SF6) were collected with a
syringe attached to an airtight sampling apparatus made
of stainless steel tubing connected to the peristaltic
pump. These groundwater samples were injected into an
evacuated serum bottle, and the headspace was ?lled
with high purity He gas. After incubating overnight at
48C and shaking, we sampled the bottle headspace to
extract SF6 and the gases produced by denitri?cation
(N2 and N2O) (Lemon 1981, Davidson and Firestone
1988).
Analytical methods
Dissolved oxygen and temperature of groundwater
were measured with a Model 55 DO/temperature meter
(YSI, Yellow Springs, Ohio, USA). Groundwater and
stream samples were analyzed for NO3À-N using an
automated cadmium reduction method on an Alpkem
Rapid Flow Autoanalyzer or Lachat Flow Injection
Analyzer (Hach, Loveland, Colorado, USA). Dissolved
organic carbon (DOC) was analyzed by high-temperature oxidation in the presence of a catalyst by a
Shimadzu TOC 5000 autoanalyzer (Shimadzu, Columbia, Maryland, USA). Concentrations and isotopic
composition of N2 and N2O gases were determined on
a PDZ Europa 20-20 continuous ?ow isotope ratio mass
spectrometer coupled to a PDZ Europa TGII trace gas
analyzer (Sercon, Cheshire, UK) at the Stable Isotope
Facility, University of California, Davis, California,
USA. Concentrations of N2O and SF6 gases were
analyzed by electron-capture gas chromatography on a
Tracor Model 540 (ThermoFinnigan, Austin, Texas) at
the Institute of Ecosystem Studies in Millbrook, New
York, USA.
Denitri?cation rate calculations
We calculated the generation rate of the denitri?cation gases (N2 and N2O) using the three ?eld replicate
gas samples with the highest tracer recovery (of six
within-sample replicates) similar to Addy et al. (2005)
and Kellogg et al. (2005), thus minimizing error from
dilution and dispersion. In order to calculate the masses

April 2008

STREAM RESTORATION AND DENITRIFICATION

of N2O-N and N2 gases (in micrograms) in our
headspace extraction samples, we used equations and
constants provided by Tiedje (1982) and Mosier and
Klemedtsson (1994). The masses of N2O-N and N2 were
transformed to the mass of 15N2O-N and 15N2 by
multiplying by their respective 15N sample enrichment
proportion (the ratio of pulled atom percent to pushed
atom percent, both corrected for ambient atom percent).
Sample 15N2O-N and 15N2 gas production rates were
expressed as micrograms of N per kilogram per day
(total mass of 15N2O-N or 15N2 per volume of water
pulled/[dry mass of sediment per volume of water pulled
3 incubation period]). Each pulled sample represented 1
L of groundwater that occupied 4.37 kg of sediment
(bulk density, ;1.65 g/cm3 from ?eld measurements;
porosity, 0.38). Denitri?cation rates were calculated as
the sum of 15N2O-N and 15N2 generation rates.
Groundwater ?ow through the
riparian-zone–stream interface
Temporal and spatial variability in hydrologic ?ow
and residence time of groundwater at the riparian-zone–
stream interface were characterized across three piezometer transects of the unrestored stream reach (transects
1–3), and they were characterized across one transect
closest to denitri?cation measurements in the restored
reach nearest the restored, low, connected bank where
groundwater ?ow data was available (transect 4) (Fig.
3). In both unrestored and restored reaches, piezometer
networks used for characterization of hydrologic ?ow
and residence time were located near push-pull measurements. Groundwater elevation values in piezometers
were measured on dates coinciding with push-pull
denitri?cation measurements or following the push-pull
denitri?cation measurements, and simple assumptions
using continuous water level data were made to provide
an estimate of range in ?ow and hydrologic residence
time along the reach. Groundwater ?ow was split into
left and right bank compartments of dimensions 1.5 3
1.5 3 1.5 m on both sides of the stream thalweg. Flow in
each compartment was estimated by calculating vertical
and horizontal gradients based on measurements of
water levels in piezometers and applying Darcy’s law:


]h~ ]h ~
]h
q ¼ Àk
iþ k ¼k
]x
]z
]s
where k was the hydraulic conductivity measured by slug
tests at each site. The horizontal gradient, ]h/]x, was
calculated as the difference between the piezometer
water levels in the stream bank and streambeds.
Concurrently, the vertical gradient, ]h/]z, was determined using the water levels in streambed piezometers.
The ?ow vector on either side of the center line of the
stream thalweg was then de?ned by determining its
gradient, ]h/]s, and direction counterclockwise from
horizontal, h, using the following two equations (Freeze
and Cherry 1979):

795

v?????????????????????????????????????
  u" 2  2 #
]h u ]h
]h
 ¼t
þ
 ]s 
]x
]z

tan h ¼

!
]h ]h
=
:
]z ]x

Once the magnitude of the flow vector and its direction
were known, it was possible to calculate the length of the
flow path line, s, through the vertical depth of the
compartment and the associated travel time, t, along
that path line using s ¼ depth/cos(90 À h) and t ¼ s/(q/n),
where n was a field-measured value of porosity. Finally,
a volumetric flow for the compartment was also
determined:
Q ¼ qA
where q is the velocity of flow through the compartment
expressed as distance per time and A is the cross
sectional area of the compartment perpendicular to the
flow vector.
Nitrate removal at the riparian-zone–stream interface
Estimates of groundwater ?ow on 10 June 2003, 21
November 2003, and 28 May 2004 (as close to dates for
push-pull tracer tests as possible) were coupled with
coinciding measurements of in situ denitri?cation rates
to investigate the importance of mass nitrate removal at
the riparian-zone–stream interface in the unrestored
reach. We could only estimate groundwater ?ow on
dates directly coinciding with the push-pull measurements in the unrestored reach. Instead, groundwater
?ow was only measured in the restored reach following
push-pull measurements, and mean denitri?cation rates
were coupled with the range of observed ?ow conditions
from August to November 2004 to estimate a potential
range of N removal. We avoided direct comparisons
between mass removal rates between reaches (unlike
comparisons of in situ dentri?cation rates) because
measurements of groundwater ?ow were not directly
concurrent in the unrestored and restored reaches,
although hydrologic ?ow in the restored reach has
showed very little variation across multiple years of
monitoring (E. Striz, unpublished data). Therefore,
differences in denitri?cation rates across sites were
compared by actual measurements using the isotopic
15
N push-pull methods, and estimates of mass removal
were only used in elucidating the potential role of the
riparian-zone–stream interface in mass nitrate-N removal in both unrestored and restored reaches.
Estimates of mass removal were calculated using the
following assumptions: (1) the relatively small 1.5 3 1.5
3 1.5 m streambed compartment acted as a steady state
complete mix reactor with constant groundwater ?ow,
(2) the denitri?cation process followed zero order
kinetics such that the rates measured from push-pull
tests were constant in space and time for the compartment, (3) the denitri?cation rate was constant with

796

Ecological Applications
Vol. 18, No. 3

SUJAY S. KAUSHAL ET AL.

assumptions were supported given the short distances
(meters) and time frames (hours) considered in this
study.
In order to calculate the nitrate mass removal in the
compartment, the measured denitri?cation rate from the
push-pull samples was ?rst converted to a compartment
?uid volume basis using ?eld-measured values of bulk
density, qb, and porosity, n:
nitrate mass removal ¼ rateðqb =nÞ
where the nitrate mass removal is measured in micrograms of NO3À per cubic meter of ?uid per day, the rate
is measured in micrograms of NO3À per kilogram of soil
per day, and porosity is measured in kilograms per cubic
meter. The total mass of nitrate removed per cubic meter
of groundwater was determined by multiplying the
converted denitri?cation rate by the residence time, t,
and total volume of the stream compartment, V, where t
¼ nV/Q and V ¼ wlh. Using the groundwater ?ow rates,
Q, from the ?eld measured conductivity and hydraulic
gradients and the seasonal measurements of in situ
denitri?cation rates, the mass removal of nitrate for each
of the transect compartments was calculated on dates
with coinciding push-pull denitri?cation measurements
and information on groundwater ?ow.
Statistical analysis

FIG. 3. (A) Concentrations of NO3À in stream water in the
restored reach and unrestored reach sampled monthly from
May 2004 to October 2005. The restored reach had signi?cantly
lower concentrations of NO3À in stream water than the
unrestored reach (paired t test, P , 0.05, t ¼ 5.06). (B) Mean
NO3À concentrations in groundwater in the restored reach
(closest to the hydrologically connected subreach) and unrestored reach of Minebank Run across four dates between April
2003 and May 2004. The center vertical line of the box-andwhisker plot marks the median of the sample. The length of
each box shows the range within which the central 50% of the
values fall. Box edges indicate the ?rst and third quartiles. Stars
($) and circles (*) represent outside values: stars denote values
1.53 the interquartile range, and the circles denote values that
are 33 the interquartile range. The restored reach had
signi?cantly lower concentrations of NO3À in groundwater
than the unrestored reach averaged over all sampling dates (t
test, P , 0.001).

concentration and temperature, and (4) there was
suf?cient input of nitrate into the compartment to
support the mass removal rates measured with the pushpull method. Although it is dif?cult to determine a rate
constant from single well measurements (Schroth and
Istok 2006), zero order kinetics and a mixed reactor
model were chosen in this case, similar to Haggerty et al.
(1998) and Whitmire and Hamilton (2005). These

Statistical analyses were performed using Systat,
Version 11 (Systat, Richmond, California, USA).
Differences in nitrate concentrations in stream water
across the study reaches were evaluated using a paired t
test. Differences in nitrate concentrations in groundwater were evaluated using a two-sample t test with pooled
variance. Differences in DOC concentrations in groundwater were evaluated using a two-sample t test with
pooled variance. We evaluated differences in denitri?cation rates across restoration status, bank type, and
seasonality at the riparian-zone–stream interface using
repeated-measures ANOVA. A fully factorial model (all
interactions included) was used with the following three
factors: (1) restoration status (restored vs. unrestored),
(2) bank type (low bank vs. high bank), and (3) season.
In order to accurately represent variability in the pushpull method, ?eld replicates (i.e., repeated measures)
were taken that consisted of three individual measurements of denitri?cation at each well during a sampling
date. To compare the effect of restoration, ?ve push-pull
wells were sampled in the unrestored reach and four
push-pull wells were sampled in the restored reach. The
effects of bank type (low vs. high) were compared by
sampling the two to three wells on each of two to three
high and low banks in restored and unrestored reaches.
The effect of seasonality was compared among three
dates of push-pull measurements (May 2003, November
2003, and June 2004). Given the labor intensiveness of
the 15N push-pull method, sample sizes were limited to a
total of 26 15N tracer additions and push-pull measurements conducted during our study. The relationship

April 2008

STREAM RESTORATION AND DENITRIFICATION

797

TABLE 1. In situ groundwater denitri?cation rates (means 6 SE, N ¼ 3 replicates per well).
Groundwater denitrification rate (lg NÁkg soilÀ1ÁdÀ1)

Well
number

Site description

1
2
3
4
5
6
7
8
9

unrestored, low bank
unrestored, low bank
unrestored, low bank
unrestored, high bank
unrestored, high bank
restored, high, nonconnected bank
restored, high, nonconnected bank
restored, low, connected bank
restored, low, connected bank

June 2003
0.2
13.8
8.2
58.2
10.5
4.3
7.8
32.0
112.6

6
6
6
6
6
6
6
6
6

November 2003

0.1
5.2
7.9
47.8
8.3
4.2
3.5
2.9
21.1

0.1
5.2
0.1
0.2
7.8
20.0
22.9
31.4
121.2

6
6
6
6
6
6
6
6
6

0.0
2.9
0
0
4.9
5.3
3.4
7.0
18.5

June 2004
181.9 6 30.3
130.9 6 31.1
no datum
0.1 6 0.0
102.4 6 4.8
99.0 6 37.9
47.7 6 4.4
262.3 6 36.5
234.8 6 1.7

Pooled seasons
60.7
48.8
4.2
19.5
40.2
41.1
26.1
108.6
156.2

6
6
6
6
6
6
6
6
6

31.5
22.7
0.2
16.9
15.9
18.4
6.1
40.0
21.3

Note: The study was conducted in Minebank Run, a low-order stream with a watershed area of ;8.47 km2, located in Baltimore
County, Maryland, USA.

between mass removal of nitrate-N in the unrestored
reach and the scenario in the restored reach and
hydrologic residence time was evaluated by linear
regression analysis.
RESULTS
The nitrate-N concentration of stream water in the
unrestored reach, 1.47 6 0.05 mg/L, was signi?cantly
higher than the nitrate-N concentration of stream water
of the restored reach, 1.15 6 0.04 mg/L (t ¼ 5.06, N ¼ 62,
P , 0.01; Fig. 3A). The nitrate-N concentration in
groundwater in the unrestored reach, 1.55 6 0.08 mg/L,
also was signi?cantly higher than the nitrate-N concentration in the restored reach, 0.84 6 0.06 mg/L averaged
over all sampling dates (t ¼ 3.98, N ¼ 147, P , 0.001;
Fig. 3B). Mean concentrations of DOC across all dates
were also signi?cantly lower in the restored reach than
the unrestored reach (t ¼ 2.63, N ¼ 147, P , 0.01), and
DOC concentrations were typically low and ,1 mg/L in
both reaches.
During the sampling periods, riparian groundwater
denitri?cation rates ranged from ,1 to 262 lg
NÁkgÀ1ÁdÀ1 for all unrestored and restored sites (Table
1). The mean values for groundwater dissolved oxygen
(DO) and temperature for all push-pull wells ranged
from 3.1 mg/L to 5.3 mg/L and 14.58C to 17.68C,
respectively (Table 2), and showed no pattern with
denitri?cation rates. Denitri?cation rates varied widely
and were highest during spring 2004. Results from the

repeated-measures ANOVA using three factors (restoration, bank height, and season) showed signi?cant
differences in mean denitri?cation rates between restored and unrestored sites (F1,14 ¼ 8.8, P ¼ 0.01; Fig.
4A), denitri?cation rates varied across seasons (F2,14 ¼
25.1, P , 0.001; Fig. 4B), and denitri?cation rates were
higher at the low-bank, hydrologically connected sites
than at the high-bank sites (F1,14 ¼ 20.3, P , 0.001; Fig.
4B). Over all seasons, denitri?cation rates were 77.4 6
12.6 lg NÁkgÀ1ÁdÀ1 at restored sites and 34.8 6 8.0 lg
NÁkgÀ1ÁdÀ1 at unrestored sites (Fig. 4A). The hydrologically connected, low-bank, restored site consistently
had signi?cantly higher mean in situ rates of denitri?cation than the other sites (Fig. 4B), a result that also
produced a signi?cant interaction between restoration
and bank height (F1,14 ¼ 7.2, P ¼ 0.02). A signi?cant
interaction between bank height and season was a
function of very high denitri?cation rates in low banks
during the June 2004 sampling date (F2,14 ¼ 7.9, P ¼
0.005). There were no signi?cant differences in replicates
within subjects, indicating that the replicate measures in
wells did not trend upward or downward. Likewise, no
replicate by factor interactions were signi?cant, and no
trends occurred among the replicates within the factors
(all P values . 0.14).
Results from piezometer studies indicated that there
was active hydrologic exchange at the riparian-zone–
stream interface in the unrestored reach. Regional
shallow groundwater showed a general ?ow toward

TABLE 2. Dissolved oxygen and temperature of ambient groundwater in push-pull wells on June 2003, November 2004, and June
2004 (means 6 SE, N ¼ 3 sampling dates).
Well
number

Site description

1
2
3
4
5
6
7
8
9

unrestored, low bank
unrestored, low bank
unrestored, low bank
unrestored, high bank
unrestored, high bank
restored, high, nonconnected bank
restored, high, nonconnected bank
restored, low, connected bank
restored, low, connected bank

Dissolved oxygen (mg/L)
5.3
4.5
4.9
4.1
3.7
4.5
3.6
3.6
3.1

6
6
6
6
6
6
6
6
6

1.2
0.5
0.8
0.9
0.7
1.3
1.3
0.8
1.7

Temperature (8C)
15.6
15.9
16.2
15.2
15.5
17.6
15.9
14.5
15.8

6
6
6
6
6
6
6
6
6

1.7
1.1
1.5
0.6
0.5
3.4
1.9
2.1
1.7

798

Ecological Applications
Vol. 18, No. 3

SUJAY S. KAUSHAL ET AL.

FIG. 4. (A) Mean in situ denitri?cation rates over all sites in
both the restored and unrestored reaches of Minebank Run.
Values are means 6 SE of four or ?ve wells in each reach,
sampled three times between May 2003 and June 2004.
Restored sites showed a higher overall mean denitri?cation
rate than unrestored sites (ANOVA, P ¼ 0.01). (B) In situ
denitri?cation rates across four different riparian-zone–stream
interfaces in the unrestored and restored reaches. Values are
means 6 SE of two or three wells in each site, sampled three
times between May 2003 and June 2004. Denitri?cation rates
varied widely and were highest during spring 2004 with
hydrologically connected sites showing the highest denitri?cation rates across all seasons. Results from the repeatedmeasures ANOVA using three factors (restoration, bank
height, and season) showed signi?cant differences in mean
denitri?cation rates between restored and unrestored sites (F1,14
¼ 8.8, P ¼ 0.01), across seasons (F2,14 ¼ 25.1, P , 0.001), and
between high and low banks (F1,14 ¼ 20.3, P , 0.001).

the west (Fig. 5A), but vertical cross-sections of
groundwater equipotentials near the stream indicated
that ?ow could vary in direction and amount over the
scale of a few meters. Groundwater ?ow moved from the
bank into the stream or parallel to the stream in
piezometer transect 1 (Fig. 5B); ?ow moved from both
banks and strongly upward into the stream in piezometer transect 2 (Fig. 5C); and stream ?ow moved down
from the center of the bed and outward roughly equally
into both banks in piezometer transect 3 (Fig. 5D).
Across sampling periods and transects, it was estimated
that hydrologic ?ow through a 1.5 3 1.5 3 1.5 m
compartment at the riparian-zone–stream interface
ranged from 0.14 m3/d to 2.64 m3/d and hydrologic
residence time of this groundwater ranged from 0.5 to
9.31 d. Piezometer transect 3, where water ?owed from
the stream into the banks, showed the longest hydrologic residence time at the riparian-zone–stream inter-

face of 2.4–9.31 d, and rates of ?ow were very slow at
only 0.14–0.43 m3/d. Other transects (where water
?owed from the banks into the stream) had shorter
residence times of ,5 d and faster ?ow rates of 0.22–
2.64 m3/d. In contrast, hydrologic ?ows in the restored
reach in piezometer transect 4 showed a relatively
uniform discharge across seasons in the same direction
from the bank to the stream.
Mass removal of nitrate-N in groundwater at the
riparian-zone–stream interface of the unrestored reach
was considerable, but it varied based on hydrologic ?ow
paths and residence time (Table 3). For all sampling
periods and piezometer transects, estimated mass
removal in ?ow through the 1.5 3 1.5 3 1.5 m
compartment of sediment in the hyporheic–stream
interface ranged from 0.007 mg NO3À-N/L over 0.95 d
to 7 mg NO3À-N/L over 9.31 d. Piezometer transect 3,
which had the slowest hydrologic ?ow rates, showed the
highest rates of nitrate-N removal in the compartment
with a mean of 1.74 mg NO3À-N removed per liter of
groundwater ?ow over a mean residence time of 4.97 d.
Piezometer transects 1 and 2, which had higher ?ow
rates, had lower mean rates of nitrate-N removal of
0.20–0.57 mg NO3À-N/L over mean residence periods of
1.19–2.54 d, respectively.
In the restored reach, we estimated removal of nitrateN in a scenario by applying the mean in situ
denitri?cation rate measured in the restored, low,
connected bank during 2003–2004 to a range in
hydrologic conditions measured in the same subreach
shortly following the period of study (August 2004 to
November 2004). Under this scenario, the potential for
mass removal of NO3À would range from 1.46 mg
nitrate-N/L over a span of 1.91 d to 2.81 mg nitrate-N/L
over a span of 3.67 d (Table 4). Mass removal of nitrateN at the riparian-zone–stream interface in both the
unrestored reach and the scenario in the restored reach
suggested a linear relationship with hydrologic residence
time (unrestored reach, R2 ¼ 0.87, N ¼ 18, P , 0.05;
restored scenario, R2 ¼ 0.70, N ¼ 8, P , 0.05; Fig. 6).
DISCUSSION
The in situ groundwater denitri?cation rates measured
in the present study, from 4.1 to 156.2 lg NÁkgÀ1ÁdÀ1,
are similar to those reported for other riparian wetlands
in the northeastern United States using push-pull
methodology. Most comparably, Addy et al. (2002)
and Kellogg et al. (2005) observed denitri?cation rates
ranging from ,1 to 330 lg NÁkgÀ1ÁdÀ1 in riparian sites
in Rhode Island, USA. Push-pull measurements of
denitri?cation are increasing in the literature (e.g., Istok
et al. 1997, Addy et al. 2002, 2005, Kellogg et al. 2005,
Whitmire and Hamilton 2005) because this technique
allows determination of in situ rates of nitrogen removal
and denitri?cation. In sites with low NO3À concentrations, the method may arti?cially increase denitri?cation
rates by exposing microbial communities to increased
levels of nitrate (Kellogg et al. 2005) and by promoting

April 2008

STREAM RESTORATION AND DENITRIFICATION

799

FIG. 5. (A) Horizontal ?ow ?eld of groundwater in the unrestored reach of Minebank Run at the scale of hundreds of feet
(1 foot ¼ 0.304 m) on 10 June 2003. (B) Vertical ?ow ?eld of groundwater at piezometer transect 1 on 10 June 2003. (C) Vertical
?ow ?eld of groundwater at piezometer transect 2 on 10 June 2003. (D) Vertical ?ow ?eld of groundwater at piezometer transect 3
on 10 June 2003.

small-scale mixing within transport-limited reaction
zones (Smith et al. 2005). An additional problem with
the push-pull method is that it can be dif?cult to identify
a rate constant from single-well measurements, which
complicates calculations of mass removal rates (Whitmire and Hamilton 2005, Schroth and Istok 2006).
Furthermore, the labor and expense associated with
isotopic tracer additions such as the push-pull method
allow relatively fewer measurements of denitri?cation
than with other laboratory-based techniques (Groffman

et al. 2006). Despite these challenges, the relatively large
volume of groundwater and soil encompassed by the
method in the ?eld allows scaling up of in situ ?eld
observations to a level that may be potentially useful for
managers and helpful in identifying practices that
improve water quality.
Our results suggest that stream restoration associated
with storm water management that increases hydrologic
connectivity can increase denitri?cation rates, and it
supports the idea that the riparian-zone–stream interface

800

Ecological Applications
Vol. 18, No. 3

SUJAY S. KAUSHAL ET AL.

TABLE 3. Groundwater ?ow through a 1.5 3 1.5 3 1.5 m box adjacent to the unrestored reach of Minebank Run representing the
riparian-zone–stream interface.

Site and date

Q (m3/d)

Direction

Denitrification rate
(lg NÁkgÀ1ÁdÀ1)

Residence
time (d)

Nitrate removal
(lg N/m3)

Transect 1, south bank
10 June 2003
21 November 2003
28 May 2004

1.80
2.50
2.64

stream to bank
stream to bank
stream to bank

34.4
4.0
51.3

0.79
0.57
0.54

107.4
9.0
109.5

Transect 1, north bank
10 June 2003
21 November 2003
28 May 2004

1.33
1.82
1.93

bank to stream
bank to stream
stream to bank

7.4
1.8
156.0

2.17
1.59
1.50

61.9
11.0
901.0

Transect 2, south bank
10 June 2003
21 November 2003
28 May 2004

0.31
0.36
0.22

bank to stream
bank to stream
bank to stream

34.4
4.0
51.3

3.58
3.01
4.97

728.8
71.3
1510.0

Transect 2, north bank
10 June 2003
21 November 2003
28 May 2004

1.24
1.45
0.86

bank to stream
bank to stream
bank to stream

7.4
1.8
156.0

1.11
0.95
1.60

34.3
7.1
1044.1

Transect 3, south bank
10 June 2003
28 May 2004

0.30
0.14

stream to bank
stream to bank

34.4
51.3

3.50
7.41

764.9
2417.1

Transect 3, north bank
10 June 2003
21 November 2003
28 May 2004

0.30
0.42
0.14

stream to bank
stream to bank
stream to bank

7.4
1.8
156.0

4.20
2.97
9.31

149.8
25.8
7004.7

Note: Estimates of mass removal of nitrate (micrograms of N removed per cubic meter of groundwater ?ow) were obtained by
coupling measurements of in situ denitri?cation rates (micrograms of N removed per kilogram of soil per day) on different sides of
the banks coinciding with measurements of groundwater ?ow.

can be an active site for denitri?cation rates (Groffman
et al. 1996, Hill 1996, Hedin et al. 1998, Kellogg et al.
2005). We observed substantial denitri?cation rates in
both the unrestored and restored reaches of Minebank
Run. In particular, the restored, low, connected bank
had consistently and signi?cantly higher in situ denitri?cation rates measured by the push-pull tracer technique
across seasons. Wider channel width and decreased
channel incision with well-developed riparian vegetation
may have increased hydrologic connectivity between
groundwater and upper soil horizons and, thereby,
affected denitri?cation rates (Groffman et al. 2002,

2003). Increased hydrologic interaction with organicrich soils underlying well-developed vegetation in the
riparian zone can stimulate higher denitri?cation rates at
the riparian-zone–stream interface in the restored, low,
connected bank of Minebank Run by providing organic
carbon as a substrate for increased denitri?cation
activity (e.g., Groffman and Crawford 2003). Step-pool
sequences and meanders similar to those in the low,
connected subreach have also been shown to increase
hydrologic residence times and nitrogen retention in
transient storage zones at the riparian-zone–stream
interface (Malard et al. 2002, Kasahara and Hill 2006)

TABLE 4. Groundwater ?ow through a 1.5 3 1.5 3 1.5 m box adjacent to the restored reach of Minebank Run representing the
riparian-zone–stream interface.

Date

Q (m3/d)

Denitrification rate
(lg NÁkgÀ1ÁdÀ1)

Residence time (d)

Nitrate removal
(lg N/m3)

6 August
2 September
14 September
21 September
29 September
5 October
20 October
17 November

0.29
0.55
0.42
0.42
0.45
0.41
0.39
0.37

132.4
132.4
132.4
132.4
132.4
132.4
132.4
132.4

3.67
1.91
2.49
2.49
2.39
2.60
2.81
2.98

2806.5
1460.6
1904.1
1580.0
1516.6
1649.8
2202.7
2329.8

Note: The potential importance of estimates of mass removal of nitrate (in micrograms of N removed per cubic meter of
groundwater ?ow) was investigated by coupling an average measurement of in situ denitri?cation rate during the study (in
micrograms of N removed per kilogram of soil per day) on the south bank of transect 4 with a range of measurements of bank-tostream groundwater ?ow during a three-month period in 2004 following denitri?cation measurements.

April 2008

STREAM RESTORATION AND DENITRIFICATION

801

FIG. 6. Relationship between hydrologic residence time and mass removal of nitrate-N in the unrestored reach (obtained using
coinciding measurements of denitri?cation rates and groundwater ?ow) and the potential relationship between hydrologic residence
time and mass removal of nitrate-N in the restored reach (obtained from a scenario using mean denitri?cation rates in a restored,
low, connected bank during 2003–2004 and groundwater ?ow rates following denitri?cation measurements in 2004).

relative to straighter runs (e.g., Hill et al. 1998, Gucker
¨
and Boechat 2004). Therefore, hydrologic ?ow paths
also may have been important in fostering higher
denitri?cation rates in the low, connected bank, and
recent work at the same site with conservative tracer
injections has shown that lateral groundwater inputs
along the riparian-zone–stream interface can be substantial (C. Klocker, unpublished data). Restoration
activities focused on increasing hydrologic ‘‘connectivity’’ in riparian zones may be important in enhancing
denitri?cation rates via multiple mechanisms such as soil
organic carbon availability and hydrologic ?ow paths,
which deserve further research attention (e.g., Fennessy
and Cronk 1997, Groffman et al. 2003, Mayer et al.
2005, Boulton 2007).
The higher in situ denitri?cation rates in the restored
reach appeared to coincide with the lower nitrate-N
concentrations in groundwater wells in this reach. It is
possible, however, that the lower concentrations of
nitrate in groundwater were not entirely attributable to
the signi?cantly higher dentri?cation rates, but also to
differences in sources of nitrate delivery in this reach.
For example, there may have been downstream sources
of pollution that elevated the concentration of nitrate-N
in the groundwater of the unrestored reach relative to
the restored reach. The very high rates of in situ
denitri?cation and potential for mass removal at the
riparian interface, however, were suggestive that N
removal by microbes may have been important at
Minebank Run.

Coupling of the high denitri?cation rates with
groundwater ?ow rates suggested that mass removal of
nitrate-N at the riparian-zone–stream interface could be
substantial in both unrestored and restored reaches. In
the unrestored reach, the groundwater ?ow and the
hydrologic residence times that we measured may have
been potentially high enough to reduce NO3À concentrations by up to 0.20–1.74 mg NO3À-N/L, depending on
groundwater ?ow rates/residence time. These estimates
suggested that even in the unrestored reach, denitri?cation rates may have the potential to substantially
in?uence groundwater NO3À concentrations, which
typically range from 1 to 2 mg nitrate-N/L. Our
estimates of nitrate-N removal for the hydrologically
connected, low-bank, restored site based on a scenario
using in situ denitri?cation measurements and groundwater residence data following denitri?cation measurements also showed a high potential for mass removal of
nitrate-N (ranging from 1.46 mg nitrate-N/L over 1.91 d
to 2.81 mg nitrate-N/L over 3.67 d). Although the
estimates of mass removal between unrestored and
restored reaches were not directly comparable due to
differences in time periods of hydrologic measurements
(unlike measurements of in situ denitri?cation), they
both suggest that considerable amounts of nitrate-N
could be removed at the riparian-zone–stream interface
due to high denitri?cation rates coupled with hydrologic
?ow and that hydrologic residence time in the riparianzone–stream interface may be an important factor.
Nitrate-N removal appeared to be higher across a
similar range of hydrologic residence times than the

802

Ecological Applications
Vol. 18, No. 3

SUJAY S. KAUSHAL ET AL.

unrestored reach, suggesting that the effects of restoration on factors such as soil organic carbon and
hydrologic paths in riparian zones in Minebank Run
may have also been important (Groffman and Crawford
2003, Mayer et al. 2003). The estimates of mass removal
are consistent with other studies in riparian zones
showing that some forms of riparian buffer restoration
and management can potentially contribute to reducing
nitrate-N concentrations in riparian groundwater (e.g.,
Pinay et al. 1993, Fennessy and Cronk 1997, Tockner et
al. 1999).
A growing body of work now suggests that stream
restoration may have the potential to in?uence stream
hydrology and hydrologic residence time (e.g., Kasahara
and Hill 2006, Bukaveckas 2007, Loheide and Gorelick
2007, Roberts et al. 2007). In particular, hydrologic
exchange and increased residence time may foster
environmental conditions (i.e., reduced redox conditions
and metabolic activity) that promote higher in situ
denitri?cation rates at the riparian-zone–stream interface of channels and riparian zones (e.g., Hedin et al.
1998, Sobczak et al. 2003, Boulton 2007). Because
riparian organic matter, geomorphology, hydrologic
?ow paths, and underlying geology may all play roles
in explaining variations in denitri?cation rates (e.g.,
Alexander et al. 2000, Stanley and Doyle 2002, Groffman and Crawford 2003, Gucker and Boechat 2004,
¨
Wollheim et al. 2006), all of these factors should be
considered and further evaluated in the ef?cacy of
restoration designs aimed at increasing both denitri?cation rates and mass removal of nitrate-N in riparian
zones. In particular, further research on coupled
restoration practices with storm water management
may be useful because it may be desirable to create
conditions with high denitri?cation rates in urban areas
where water from the landscape is concentrated (Pouyat
et al. 2007).
Our results suggest that restoration practices for
storm water management that foster ‘‘connectivity’’
between the stream and the riparian zone can increase
rates of in situ denitri?cation in stream banks and that
mass nitrate-N removal may be substantial at the
riparian-zone–stream interface. Our results also suggest
that there can be substantial variability in denitri?cation
rates among restoration designs based on hydrological
connectivity and bank height and that continuing work
is necessary to identify which types of stream restoration
practices will be most effective at removing nitrogen
(e.g., Stanley and Doyle 2002, Kasahara and Hill 2006,
Palmer and Bernhardt 2006, Bukaveckas 2007, Roberts
et al. 2007). Because of uncertainties concerning the
magnitude and range of nitrate-N removal possible,
until more is known, stream restoration by itself is not
appropriate for compensatory mitigation but may
complement watershed-based management strategies
for reducing nitrate-N sources to streams. Key questions
relate to how the potential for denitri?cation and mass
nitrogen removal at the riparian-zone–stream interface

changes with growth of new riparian vegetation and
changes in soil organic matter, responses to increased or
decreased storm water ?ows, and the relative contribution of denitri?cation at the riparian-zone–stream
interface to whole-stream denitri?cation. It is important
to recognize that restored streams in urban watersheds
may have a different capability of transforming nitrogen
at the riparian-zone–stream interface than streams in
undeveloped watersheds that have been much more
thoroughly studied (e.g., Peterson et al. 2001, Hall and
Tank 2003, Mulholland et al. 2004, Kaushal and Lewis
2005). Expansion of in situ measurements of denitri?cation and hydrologic ?uxes to other restoration sites
and comparisons with forest reference and suburban
ecosystems will be critical in determining and/or
establishing any effective standards for potential water
quality improvements associated with riparian and
stream restoration in urban areas.
ACKNOWLEDGMENTS
We gratefully thank Edward J. Doheny of the U.S.
Geological Survey for providing water quality data and channel
measurements. We thank Dan Dillon, Ken Jewell, Tara Krebs,
Sabrina LaFave, Dave Lewis, Russell Neill, and Roger Starsoneck for assistance with ?eld and laboratory work; Kelly Addy
and Dorothy Kellogg for guidance on push-pull experiments;
Karen Ogle and Steve Stewart of the Baltimore County
Department of Environmental Protection, Regulation, and
Management, and Bob Shedlock of the U.S. Geological Survey
for helpful discussions regarding characteristics and monitoring
of the restoration site; and Bill Stack of the Baltimore City
Department of Public Works and Gary Fisher of the U.S.
Geological Survey for comments on a previous version of the
manuscript. This research was supported by the U.S. Environmental Protection Agency through its Of?ce of Research and
Development under cooperative agreement CR829676 with the
Institute of Ecosystem Studies and by the U.S. National Science
Foundation Long-Term Ecological Research (LTER) program
(DEB-9714835). The research has not been subjected to Agency
review and therefore does not necessarily re?ect the views of
any of the funding agencies, and no of?cial endorsement should
be inferred.
LITERATURE CITED
Addy, K., A. Gold, B. Nowicki, J. McKenna, M. Stolt, and P.
Groffman. 2005. Denitri?cation capacity in a subterranean
estuary below a Rhode Island fringing salt marsh. Estuaries
28:896–908.
Addy, K. L., D. Q. Kellogg, A. J. Gold, P. M. Groffman, G.
Ferendo, and C. Sawyer. 2002. In situ push-pull method to
determine ground water denitri?cation in riparian zones.
Journal of Environmental Quality 31:1017–1024.
Alexander, R. B., R. A. Smith, and G. E. Schwartz. 2000. Effect
of stream channel size on the delivery of nitrogen to the Gulf
of Mexico. Nature 403:758–761.
Bernhardt, E. S., et al. 2005. Ecology—synthesizing U.S. river
restoration efforts. Science 308:636–637.
Bernot, M. J., and W. K. Dodds. 2005. Nitrogen retention,
removal, and saturation in lotic ecosystems. Ecosystems 8:
442–453.
Boesch, D. F., R. B. Brins?eld, and R. E. Magnien. 2001.
Chesapeake Bay eutrophication: scienti?c understanding,
ecosystem restoration, and challenges for agriculture. Journal
of Environmental Quality 30:303–320.
Bohlke, J. K., J. W. Harvey, and M. A. Voytek. 2004. Reach¨
scale isotope tracer experiment to quantify denitri?cation and

April 2008

STREAM RESTORATION AND DENITRIFICATION

related processes in a nitrate-rich stream, mid-continent
United States. Limnology and Oceanography 49:821–838.
Bohlke, J. K., M. E. O’Connell, and K. L. Prestegaard. 2007.
¨
Ground water strati?cation and delivery of nitrate to an
incised stream under varying ?ow conditions. Journal of
Environmental Quality 36:664–680.
Boulton, A. J. 2007. Hyporheic rehabilitation in rivers:
restoring vertical connectivity. Freshwater Biology 52:632–
650.
Boyer, E. W., C. L. Goodale, N. A. Jaworski, and R. W.
Howarth. 2002. Anthropogenic nitrogen sources and relationships to riverine nitrogen export in the northeastern
U.S.A. Biogeochemistry 57:137–169.
Brown, D. G., K. M. Johnson, T. R. Lovelan, and D. M.
Theobald. 2005. Rural land-use trends in the conterminous
United States, 1950–2000. Ecological Applications 15:1851–
1863.
Bukaveckas, P. A. 2007. Effects of channel restoration on water
velocity, transient storage, and nutrient uptake in a
channelized stream. Environmental Science and Technology
41:1570–1576.
Davidson, E. A., and M. K. Firestone. 1988. Measurement of
nitrous oxide dissolved in soil solution. Soil Science Society
of America Journal 52:1201–1203.
Doheny, E. J., R. J. Starsoneck, E. A. Striz, and P. M. Mayer.
2006. Watershed characteristics and pre-restoration surfacewater hydrology of Mine Ban Run, Baltimore County,
Maryland, water years 2002–2004. Scienti?c Investigations
Report 2006-5179. U.S. Geological Survey, Baltimore,
Maryland, USA.
Fennessy, M. S., and J. K. Cronk. 1997. The effectiveness and
restoration potential of riparian ecotones for the management of nonpoint source pollution, particularly nitrate.
Critical Reviews in Environmental Science and Technology
27:285–317.
Freeze, R. A., and J. A. Cherry. 1979. Groundwater. Prentice
Hall, Englewood Cliffs, New Jersey, USA.
Gold, A. J., P. A. Jacinthe, P. M. Groffman, W. R. Wright, and
R. H. Puffer. 1998. Patchiness in groundwater nitrate
removal in a riparian forest. Journal of Environmental
Quality 27:146–155.
Green, P. A., C. J. Vorosmarty, M. Meybeck, J. N. Galloway,
B. J. Peterson, and E. W. Boyer. 2004. Pre-industrial and
contemporary ?uxes of nitrogen through rivers: a global
assessment based on typology. Biogeochemistry 68:71–105.
Grimm, N. B., R. W. Sheibley, C. L. Crenshaw, C. N. Dahm,
and W. J. Roach. 2005. N retention and transformation in
urban streams. Journal of the North American Benthological
Society 24:626–642.
Groffman, P. M., M. A. Altabet, J. K. Bohlke, K. Butterbach¨
Bahl, M. B. David, M. K. Firestone, A. E. Giblin, T. M.
Kana, L. P. Nielsen, and M. A. Voytek. 2006. Methods for
measuring denitri?cation: diverse approaches to a dif?cult
problem. Ecological Applications 16:1091–2122.
Groffman, P. M., D. J. Bain, L. E. Band, K. T. Belt, G. S.
Brush, J. M. Grove, R. V. Pouyat, I. C. Yesilonis, and W. C.
Zipperer. 2003. Down by the riverside: urban riparian
ecology. Frontiers in Ecology and the Environment 6:315–
321.
Groffman, P. M., N. J. Boulware, W. C. Zipperer, R. V.
Pouyat, L. E. Band, and M. F. Colosimo. 2002. Soil nitrogen
cycling processes in urban riparian zones. Environmental
Science and Technology 36:4547–4552.
Groffman, P. M., and M. K. Crawford. 2003. Denitri?cation
potential in urban riparian zones. Journal of Environmental
Quality 32:1144–1149.
Groffman, P. M., A. M. Dorsey, and P. M. Mayer. 2005.
Nitrogen processing within geomorphic features in urban
streams. Journal of the North American Benthological
Society 24:613–625.

803

Groffman, P. M., G. Howard, A. J. Gold, and W. M. Nelson.
1996. Microbial nitrate processing in shallow groundwater in
a riparian forest. Journal of Environmental Quality 25:1309–
1316.
Gucker, B., and I. G. Boechat. 2004. Stream morphology
¨
controls ammonium retention in tropical headwaters. Ecology 85:2818–2827.
Gucker, B., and M. T. Pusch. 2006. Regulation of nutrient
¨
uptake in eutrophic lowland streams. Limnology and
Oceanography 51:1443–1453.
Haggerty, R., M. H. Schroth, and J. D. Istok. 1998. Simpli?ed
method of ‘‘push-pull’’ test data analysis for determining in
situ reaction rate coef?cients. Ground Water 36:314–324.
Hall, R. O., and J. L. Tank. 2003. Ecosystem metabolism
controls nitrogen uptake in streams in Grand Teton National
Park, Wyoming. Limnology and Oceanography 48:1120–
1128.
Hassett, B., M. Palmer, E. Bernhardt, S. Smith, J. Carr, and D.
Hart. 2005. Restoring watersheds project by pro