Content uploaded by Alexander Joseph Reisinger
Author content
All content in this area was uploaded by Alexander Joseph Reisinger on Apr 06, 2017
Content may be subject to copyright.
Recovery and resilience of urban stream metabolism
following Superstorm Sandy and other floods
ALEXANDER J. REISINGER ,
1,
EMMA J. ROSI,
1
HEATHER A. BECHTOLD,
2
THOMAS R. DOODY,
3
SUJAY S. KAUSHAL,
3
AND PETER M. GROFFMAN
1,4
1
Cary Institute of Ecosystem Studies, Millbrook, New York 12545 USA
2
Department of Biological Sciences, Lock Haven University, Lock Haven, Pennsylvania 17745 USA
3
Department of Geology, Earth System Science Interdisciplinary Center, University of Maryland, College Park, Maryland 20742 USA
4
Department of Earth and Environmental Sciences, Brooklyn College, City University of New York Advanced Science Research Center,
New York, New York 10031 USA
Citation: Reisinger, A. J., E. J. Rosi, H. A. Bechtold, T. R. Doody, S. S. Kaushal, and P. M. Groffman. 2017. Recovery and
resilience of urban stream metabolism following Superstorm Sandy and other floods. Ecosphere 8(4):e01776. 10.1002/
ecs2.1776
Abstract. Urban streams are exposed to multiple different stressors on a regular basis, with increased
hydrological flashiness representing a common urban stream stressor. Stream metabolism, the coupled
ecosystem functions of gross primary production (GPP) and ecosystem respiration (ER), controls
numerous other ecosystem functions and integrates multiple processes occurring within streams. We
examined the effect of one large (Superstorm Sandy) and multiple small and moderately sized flood events
in Baltimore, Maryland, to quantify the response and recovery of urban stream GPP and ER before and
after floods of different magnitudes. We also compared GPP and ER before and after Superstorm Sandy to
literature values. We found that both GPP and ER decreased dramatically immediately following floods of
varying magnitudes, but on average GPP was more reduced than ER (80% and 66% average reduction in
GPP and ER, respectively). Both GPP and ER recovered rapidly following floods within 4–18 d, and recov-
ery intervals did not differ significantly between GPP and ER. During the two-week recovery following
Superstorm Sandy, two urban streams exhibited a range of metabolic activity equivalent to ~15% of the
entire range of GPP and ER reported in a recent meta-analysis of stream metabolism. Urban streams
exhibit a substantial proportion of the natural variation in metabolism found across stream ecosystems
over relatively short time scales. Not only does urbanization cause increased hydrological flashiness, it
appears that metabolic activity in urban streams may be less resistant, but also more resilient to floods than
in other streams draining undeveloped watersheds, which have been more studied. Our results show that
antecedent conditions must be accounted for when drawing conclusions about stream metabolism
measurements, and the rapid recovery and resilience of urban streams should be considered in watershed
management and stream restoration strategies targeting ecosystem functions and services.
Key words: disturbance; ecosystem respiration; flood; gross primary production; recovery; recurrence interval;
resilience; resistance; Superstorm Sandy.
Received 8 March 2017; accepted 9 March 2017. Corresponding Editor: Debra P. C. Peters.
Copyright: ©2017 Reisinger et al. This is an open access article under the terms of the Creative Commons Attribution
License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.
E-mail: reisingera@caryinstitute.org
INTRODUCTION
Urban stream ecosystems exhibit a multitude
of physicochemical and biological changes in
response to urban development (Walsh et al. 2005,
Wenger et al. 2009, Kaushal and Belt 2012).
These changes typically include geomorphic
alterations such as highly incised banks and
homogenous channel shape (Wolman 1967, Hen-
shaw and Booth 2000, Vietz et al. 2016), elevated
❖www.esajournals.org 1April 2017 ❖Volume 8(4) ❖Article e01776
nutrients and other chemical contaminants (Hatt
et al. 2004, Carle et al. 2005), and biotic commu-
nities with decreased intolerant and increased
tolerant species (Paul and Meyer 2001). These
physicochemical and biological changes to urban
streams are speculated to result in reductions in
stream ecosystem functions such as metabolism
and nutrient uptake, despite a limited amount of
empirical evidence (Walsh et al. 2005, Wenger
et al. 2009). In fact, a recent review found that
stream nitrogen uptake rates were similar
between reference and urban streams across a
range of urban settings (Reisinger et al. 2016),
and urban stream networks are dynamic trans-
formers of materials and energy, particularly
during baseflow (Kaushal et al. 2014a,b).
Stream metabolism is an integrative metric of
stream biological activity, and it represents the
fundamental ecosystem functions of gross pri-
mary production (GPP) and ecosystem respira-
tion (ER; Odum 1956, Hoellein et al. 2013). Both
GPP and ER can be simultaneously considered a
driver of other ecosystem processes (e.g., meta-
bolism controls nutrient uptake; Hall and Tank
2003) or as a dependent variable characterizing
the response of different streams to extrinsic dri-
vers (Mulholland et al. 2001, Bernot et al. 2010).
The development of new technologies and mod-
eling approaches (e.g., Grace et al. 2015, Hall
et al. 2016) has allowed for rapid expansion of
stream metabolism datasets and continuous
records, allowing us to test the response of this
fundamental ecosystem function to a range of
environmental drivers.
Long-term stream metabolism datasets now
show that metabolic activity within a stream var-
ies on multiple temporal scales, including sea-
sonal, daily, and episodic (e.g., storms, spates)
scales (Acu~
na et al. 2004, Roberts and Mulholland
2007, Beaulieu et al. 2013). Heavily impacted
stream ecosystems can vary over these different
time periods. For example, highly managed agri-
cultural ditches exhibit extreme day-to-day vari-
ability in metabolic activity over a given year, as a
single agricultural ditch had a wider annual range
of both GPP and ER than the entire range of all
literature data on stream metabolism from a
recent review (Hoellein et al. 2013, Roley et al.
2014). Therefore, this agricultural ditch appears as
variable across days as across seasons, likely due
to an open canopy and high nutrient conditions
across seasons. Streams affected by urbanization
can also exhibit high variability in metabolic
parameters over a range of environmental condi-
tions such as streamflow, day length, and nutrient
status (Smith and Kaushal 2015). For example, a
two-year semi-continuous record of metabolism
in a suburban stream revealed that light availabil-
ity, temperature, and disturbance all controlled
variation in stream metabolism at different tem-
poral scales (Beaulieu et al. 2013).
Although often considered biologically impaired
due to low macroinvertebrate and vertebrate
biodiversity (Moore and Palmer 2005, Walsh et al.
2005, Walters et al. 2009), urban streams can
exhibit rates of ecosystem functions equivalent to,
or higher than, agricultural or reference streams
(Mulholland et al. 2008, Bernot et al. 2010,
Reisinger et al. 2016). For example, GPP and ER in
streams draining predominately urban watersheds
were significantly higher than in nearby streams
located within a forested state park throughout the
year (Kaushal et al. 2014a). Similarly, streams
draining a range of forested and urban water-
sheds, and under variable restoration strategies, all
exhibited similar GPP and ER in both summer and
winter (Sudduth et al. 2011). Despite the multiple
physicochemical stressors present in urban streams
that represent a constant “press”disturbance
(sensu Lake 2000), fundamental ecosystem func-
tions like stream metabolism and nutrient uptake
(Mulholland et al. 2008, Bernot et al. 2010,
Reisinger et al. 2016) may actually be more resis-
tant to urbanization than the biodiversity of stream
communities. From a practical perspective, this
suggests that ecosystem functions related to urban
water quality may be more resilient to disturbance
in urban streams than previously appreciated, and
thismaybedrivenbyhighfunctionalredundancy
within microbial communities driving these
ecosystem functions (Utz et al. 2016).
One of the most ubiquitous symptoms of urban-
ization is an increase in flow variability and magni-
tude, increasing the rate and magnitude of erosion,
scouring, and sediment transport (Hawley and
Vietz 2016). This increase in hydrologic “flashi-
ness”is caused by a combination of impervious
surfaces on the landscape and stormwater drains
increasing the efficiency of runoff from the water-
shed to the stream (Walsh et al. 2005, Kaushal and
Belt 2012, Vietz et al. 2016). High-flow events have
long been recognized as a controlling factor for
❖www.esajournals.org 2April 2017 ❖Volume 8(4) ❖Article e01776
REISINGER ET AL.
stream ecosystems (Poff et al. 1997). Indeed, recov-
ery and succession of streams following flooding
has been the focus of numerous classic studies
(Fisher et al. 1982, Grimm 1987). In naturally fla-
shy streams, algal biomass and primary produc-
tion recover to pre-floodlevelswithindaysto
weeks (Fisher et al. 1982, Grimm 1987), and ER
may be less affected by floods than GPP, leading to
an increase in heterotrophy following floods
(Uehlinger 2000, Uehlinger et al. 2003). Although
stream ecosystems are typically thought of as
highly resilient to floods (Lake 2000), increased
flashiness caused by urbanization (in terms of both
frequency and magnitude of flood events), coupled
with multiple other stressors that impair the
biological community within urban streams, may
combine the pulse and press disturbances of
flooding and urbanization to drastically limit the
ability of urban stream metabolism to recover from
flooding.
Here, we investigated the response of urban
stream metabolism to flood events of various
magnitudes. We quantified metabolism before
and after floods and the recovery rate following
the flood. We had three objectives: (1) quantify
recovery rates of stream metabolism across a
range of flood magnitudes, (2) establish whether
recovery is controlled by flood magnitude, and
(3) compare metabolic activity of urban streams
during a post-flood recovery to previously mea-
sured metabolic rates from the literature. These
three objectives aimed to expand our under-
standing of both metabolism, a key ecosystem
function driving multiple processes within
streams, and the response of urban streams to
disturbance. These objectives are consistent with
several key questions in urban stream ecology
(specifically questions 1, 4, 5, and 6) identified by
Wenger et al. (2009).
METHODS
Study sites
We initially selected two streams located within
Baltimore County and Baltimore City (Maryland,
USA) that are components of the U.S. National
Science Foundation-funded urban Long Term Eco-
logical Research network Baltimore Ecosystem
Study project. One stream was located near the
urban core of Baltimore (Gwynns Falls at Carroll
Park, hereafter the “urban site”), and the other
stream was located in a more suburban area
(Gwynns Falls at Gwynnbrook, hereafter the “sub-
urban site”; Table 1). At each of these sites, we
deployed dissolved oxygen (DO) and temperature
sensors (miniDOT; Precision Measurement Engi-
neering, Vista, California, USA) and photosynthet-
ically active radiation (PAR) sensors (Odyssey
PAR R e corder; D a taflow Systems Limited,
Christchurch, New Zealand) programmed to log
DO, temperature, and PAR every five minutes on
or before 22 October 2012. We collected the sensors
on or after 12 November 2012. This deployment
period included Superstorm Sandy (hereafter
“Sandy”), a large storm event (see Storm events)
that affected much of the northeastern United
States. Although Baltimore was less affected than
areas further north, large-scale flooding did occur,
allowing us to test the effect of a storm on stream
metabolism and its recovery in urban streams.
In addition to these two sites, we analyzed
metabolism data from four additional urban
streams located within Baltimore City or
County in 2015. We deployed miniDOTs which
Table 1. Characteristics of study stream watersheds (WS) and flood dates.
Site Lat. long.
Percent
developed
Percent
ISC
WS area
(km
2
) Flood dates
GFCP (referred to
as “urban”)
39°16′17.4″N, 76°38′54.8″W 79 28 165 29 October 2012
GFGB (referred to
as “suburban”)
39°26′34.6″N, 76°47′00.3″W 82 17 11 29 October 2012
SLB 39°22′25.7″N, 76°47′41.5″W 92 29 1 10 September
MBU 39°24′43.0″N, 76°33′12.5″W 82 23 1 18 May, 24 August, 10 September
MBD 39°24′34.6″N, 76°33′26.1″W 73 21 5 13 July, 24 August, 10 September
STN 39°21′22.2″N, 76°37′49.3″W 90 28 2 27 June, 24 August
Notes: ISC, impervious surface cover; WS area, watershed area; GFCP, Gwynns Falls at Carroll Park; GFGB, Gwynns Falls at
Gwynnbrook; SLB, Scotts Level Branch; MBU, Minebank Run—Upstream; MBD, Minebank Run—Downstream; STN, Stony
Run. Flood dates are from 2015 unless otherwise noted.
❖www.esajournals.org 3April 2017 ❖Volume 8(4) ❖Article e01776
REISINGER ET AL.
continuously logged DO and temperature at these
four sites (beginning in April 2015), and PAR sen-
sors at three of four sites. The site without a PAR
sensor was located near a site with a PAR sensor,
and we therefore used PAR from this nearby site
for modeling. We quantified watershed character-
istics (area, percent developed, percent impervious
surface cover [ISC]) by first delineating watersheds
using 30-m resolution digital elevation models
from the United States Geological Survey (USGS)
and then extracting land-use categories using stan-
dard methods in ArcGIS (Version 10.3.1, Esri Cor-
poration, Redlands, California, USA). Both percent
developed and percent ISC were extracted from
the 2011 National Land Cover Database (NLCD;
Homer et al. 2015) to provide a general representa-
tion of the land use in the watersheds of these sites
(Table 1). We note that percent developed repre-
sents the sum of the four different developed land-
use classifications from NLCD. These streams are
representative of urban streams throughout the
Baltimore area, with closed canopies, substrate
ranging from sand to cobble, stormwater drain
inlets, and elevated background nutrient concen-
trations (Kaushal et al. 2014a).
Storm events
On 29 October 2012, Hurricane Sandy made
landfall in the northeastern United States. Altho-
ugh the majority of the storm surge, property loss,
and economic harm occurred further north, rain-
fall was heaviest in eastern Maryland. For exam-
ple, 16.8 cm of precipitation fell at the Baltimore-
Washington Airport during the course of the storm
(Blake et al. 2013), and this caused major flooding
in streams throughout the Baltimore area. In less
than 24 h, discharge increased from baseflow val-
ues of 0.07 and 0.83 m
3
/s to 18 and 291 m
3
/s for
the suburban and urban streams, respectively.
For the storm events using the four additional
sites, we sorted through a daily discharge and
metabolism record spanning April–November
2015 to identify high-flow events with enough
time at baseflow between events to allow for
metabolic recovery. We only included events
with at least four days of baseflow prior to the
flood with relatively stable GPP and ER, coupled
with enough time following the flood for GPP
and ER to recover to pre-flood rates. We selected
nine stream–storm events to include in addition
to the Sandy data (Table 1).
We calculated the flood recurrence intervals
(RIs, Eq. 1) for each of these stream–flood events
(n=11; two from Sandy and nine from 2015)
using two approaches. For sites with USGS
gages, we used the maximum annual flow from
USGS records for every year available and pro-
duced flood RIs using the approach outlined
below. For sites that did not have USGS gaging
stations, we identified nearby gaging stations
located on the same stream (but further down-
stream in the network). We used USGS data from
these downstream gaging stations to estimate
flood RIs for our ungaged sites, which provides a
conservative estimate for site-specificflood RIs.
Flood RIs were calculated by first compiling
the maximum flow for each year on record from
the USGS gages. We then ranked each year by
the magnitude of maximum annual flow, with
the highest flow year being ranked first. We cal-
culated RIs as
Flood RI ¼Nþ1
Ranki
(1)
where N is the total number of years in the USGS
record, and Rank
i
is the rank for year i.Wethen
regressed log
10
-transformed flood RI vs. peak flow
for each year (L/s) to provide an equation for cal-
culating RIs for each flood event in our analysis.
Metabolism estimation
We used the single-station open-channel O
2
exchange approach to estimate stream metabo-
lism. Our modeling approach was based upon a
modification of the daytime regression approach
(Atkinson et al. 2008, Grace et al. 2015) in which
we modeled GPP and ER as
½DOtþ1¼½DOtþAIp
tRhðTtTmeanÞ
þKDO 1:0241ðTtTmeanÞ
½DOsat;t½DOmodeled;t
(2)
where tis the timestep; AI
p
is the primary pro-
duction term (mg O
2
L
1
d
1
), where Ais a con-
stant, Iis surface irradiance, and pis an exponent
accounting for photo-saturation; Ris respiration
(mg O
2
L
1
d
1
); his the temperature depen-
dence of respiration; T
t
and T
mean
are water tem-
perature at time tand average daily water
temperature; KO2is the aeration coefficient (d
1
);
and sat and modeled refer to [DO] at saturation
❖www.esajournals.org 4April 2017 ❖Volume 8(4) ❖Article e01776
REISINGER ET AL.
and modeled concentrations, respectively. To
carry out this metabolism estimation, we used an
updated version of the Bayesian single-station
estimation (BASE) modeling approach (Grace
et al. 2015), which has been modified based on
recommendations of Song et al. (2016) to esti-
mate daily GPP and ER. The updated BASE
approach employs Eq. 2 to use direct concentra-
tion of DO rather than a stepwise approach, and
uses modeled DO concentration rather than
measured concentration to estimate oxygen defi-
ciency for aeration rates. The updated BASE
model (BASE v2.0) can be accessed online
(https://github.com/dgiling/BASE).
Due to high levels of diel temperature fluctua-
tion, we used BASE to simultaneously model
GPP, ER, K,p, and h. Output from BASE provides
GPP and ER in volumetric units. In order to com-
pare metabolic rates across sites and with previ-
ous literature values, we multiplied volumetric
rates by mean daily stream depth to convert
from volumetric to areal rates (g O
2
m
2
d
1
).
We also calculated net ecosystem productivity
(NEP; g O
2
m
2
d
1
)as
NEP ¼GPP þjERj(3)
where |ER|is the absolute value of ER, which is
traditionally expressed as a negative value. For
each site, we used daily discharge coupled with
an empirically derived discharge–depth relation-
ship (unique for each site; data not shown) to
estimate mean daily stream depth.
Analysis
After estimating GPP and ER for each stream
before and after flood events, we calculated the
percentage of reduction due to flooding as
Reduction ð%Þ¼1Ratepost
Ratepre
100%(4)
where Rate
post
is GPP or ER on the first modelable
day after the flood event, and Rate
pre
is the aver-
age GPP or ER at baseflow conditions prior to the
flood event. We were unable to model metabolism
on the day of peak flow or often for at least one
day following a flood due to the reliance of the
BASE model on an assumption of constant dis-
charge (Grace et al. 2015), which is a common
assumption for whole-stream metabolism models.
We also calculated recovery intervals for GPP and
ER, which allowed us to compare the time it took
for a site to recover to pre-flood rates while
accounting for differences in base metabolic activ-
ity due to non-flood-related factors (e.g., variable
canopy, water chemistry, stream size).
Metabolic recovery intervals were calculated
in a multi-step approach. First, we quantified
mean GPP and ER at baseflow before the flood.
Next, we calculated response ratios for post-
flood GPP and ER as
RateRR ¼Ratei
RatePre
(5)
where Rate
RR
is the response ratio for either GPP
or ER and Rate
i
is GPP or ER on day ipost-flood.
We applied Eq. 5 to each day post-flood until
response ratios reached a plateau, suggesting a
new equilibrium had been reached following the
flood. After calculating response ratios for both
GPP and ER on each day post-flood, we
regressed response ratios vs. time since the flood
event (d), providing us with a linear regression
equation. The slope of the regression line repre-
sents the recovery rate. After modeling recovery
trajectories from each site, we set y=1 for each
unique regression equation and solved for x,
which equates to how much time was required
following the flood for Rate
post
to equal Rate
pre
(alternatively, when did Rate
RR
=1 based on the
regression). We will refer to this length of time as
the recovery interval (d). We compared recovery
intervals for GPP and ER using a paired ttest,
and we compared seasonal differences in recov-
ery intervals between summer (1 June–31
August; n=5) and autumn (1 September–30
November; n=5) flood events using indepen-
dent ttests. One flood event was excluded from
the seasonal analysis as it was a spring flood
occurring on 18 May (Table 1).
We quantified the effect of flood magnitude on
metabolic recovery using simple linear regres-
sion with recovery intervals for either GPP or ER
as the response variable and flood RI as the pre-
dictor variable. This approach, coupled with the
fact that recovery times are based upon response
ratios and therefore account for differences in
baseflow metabolic activity across space and
time, allows us to isolate the effect of flood mag-
nitude on metabolic recovery across the range of
streams. Finally, to compare urban stream meta-
bolic activity to other stream ecosystems, we
compared GPP vs. ER for the suburban and
❖www.esajournals.org 5April 2017 ❖Volume 8(4) ❖Article e01776
REISINGER ET AL.
urban streams in response to Sandy to literature
values of stream metabolism taken from a recent
review (Hoellein et al. 2013).
RESULTS
Response to Superstorm Sandy
Prior to Superstorm Sandy, both the suburban
and urban streams were heterotrophic, with GPP
averaging 0.70 and 1.80 g O
2
m
2
d
1
and ER
averaging 4.12 and 2.62 g O
2
m
2
d
1
for the sub-
urban and urban streams, respectively (Fig. 1).
The second day following peak flow from Sandy
was the first day which provided an acceptable
model fit. Both the suburban and urban streams
exhibited a larger reduction in GPP (84% and 92%
reductions, respectively) than in ER (72% and
86% reductions, respectively), although the differ-
ences in GPP and ER reduction were modest for
both sites (Fig. 1). Following Sandy, both sites
exhibited a reduction in metabolic activity, sug-
gesting a disturbance, but as time since the flood
progressed, metabolic activity exhibited a post-
flood recovery (Fig. 2). Both the urban and subur-
ban streams exhibited a large amount of variation
in GPP (coefficient of variation [CV] =59.8% and
101.8%, respectively) and ER (CV =74.8% and
40.3%, respectively) following Sandy (Figs. 2, 3).
Indeed, over the two-week recovery trajectory, the
Fig. 1. Gross primary production (GPP; dark gray;
C, D) and ecosystem respiration (ER; light gray; C, D)
recover following Superstorm Sandy (A, B) at both the
urban (A, C) and suburban (B, D) sites. Vertical lines
represent the day with peak flow for Superstorm
Sandy.
Fig. 2. Gross primary production (GPP) and ecosys-
tem respiration (ER) decreased immediately following
Superstorm Sandy, but increased to equivalent or
higher rates over time. Black triangles denote pre-flood
rates, whereas circles are post-flood rates, with the
lightest shading immediately following Sandy and
shading growing darker as metabolic rates recover.
❖www.esajournals.org 6April 2017 ❖Volume 8(4) ❖Article e01776
REISINGER ET AL.
range of metabolic activity in the suburban and
urban streams was equal to 12% and 17% of the
literature range for GPP, and 15% and 18% of the
literature range for ER, respectively (Fig. 3; Hoel-
lein et al. 2013).
Metabolic recovery across multiple flood events
Across the 11 stream–flood events, GPP showed
a greater percentage of reduction on average than
ER, with GPP being reduced by 79% (range
17–99%) and ER reduced by 72% (range 11–89%),
but this was not statistically significant (paired t
test; t=1.46, df =10, P=0.17). Our response
ratio-based regression approach for calculating
recovery intervals yielded significant regressions
for GPP response ratios at each stream–flood com-
bination (P<0.05, average r
2
=0.78, n=11;
Table 2) and all but three stream–flood combina-
tions had a significant regression for ER response
ratios (P<0.05, average r
2
=0.64, n=8; Table 2),
suggesting minimal resistance (but potentially
high resilience) to floods. Two stream–flood
events exhibited a marginal recovery trend for ER
response ratios (GFCP-10/29/2012 and MBD-7/13/
2015; P=0.078 and 0.070, r
2
=0.34 and 0.45,
respectively), and one stream–floodeventdidnot
show any relationship between ER response ratios
and time since peak flow (MBU-08/24/2015;
P>0.1). For subsequent analyses, we used GPP
(n=11) and ER (n=10) recovery intervals based
upon significant or marginally significant (P<0.1)
regressions.
Recovery intervals for GPP (mean =9.2,
range =4.3–18.2 d; Fig. 4) and ER (11.3, 6.9–15.7 d;
Fig. 4) varied across the 11 stream–flood periods
(Table 2). Recovery intervals for GPP and ER did
not differ (paired ttest, t=1.62, df =9, P=0.14;
Fig. 3. Gross primary production (GPP) and ecosys-
tem respiration (ER) from the suburban (dark gray tri-
angles) and urban (gray squares) streams spanned a
wide range of the variation present in the literature
(light gray circles; from Hoellein et al. 2013). Dashed
line represents GPP =ER.
Table 2. Magnitude of flood events and metabolic recovery from floods in urban and suburban streams across
Baltimore City and County.
Site Flood date
Flood
RI (yr)
GPP ER
Percentage of
reduction Pr
2
Recovery
interval (d)
Percentage of
reduction Pr
2
Recovery
interval (d)
GFCP 29 October 2012 2.75 92 <0.0001 0.71 18.2 86 0.0775 0.34 15.7
GFGB 29 October 2012 1.82 84 0.0004 0.73 7.2 72 0.0005 0.79 10.3
MBU 18 May 2015 1.36 99 <0.0001 0.9 5.4 88 0.0045 0.66 6.9
STN 27 June 2015 1.81 99 <0.0001 0.93 10.1 81 0.0442 0.38 14.1
MBD 13 July 2015 3.57 53 0.0257 0.48 7.1 89 0.0702 0.45 11.2
MBD 24 August 2015 1.42 94 <0.0001 0.87 7.6 79 0.0001 0.78 13.1
MBU 24 August 2015 1.42 71 <0.0001 0.87 4.3 11 0.1324 ns ns
STN 24 August 2015 0.97 88 <0.0001 0.85 6.9 70 <0.0001 0.83 11.2
MBD 10 September 2015 0.77 97 <0.0001 0.74 9.0 84 0.0036 0.52 8.8
MBU 10 September 2015 0.77 83 <0.0001 0.84 13.8 20 0.0002 0.73 9.9
SLB 10 September 2015 0.82 17 0.0003 0.63 11.3 50 0.0094 0.39 11.7
Notes: GPP, gross primary production; ER, ecosystem respiration; RI, recurrence interval. Site abbreviations are provided in
Table 1. ns indicates no significant or marginal recovery regression (P>0.1).
❖www.esajournals.org 7April 2017 ❖Volume 8(4) ❖Article e01776
REISINGER ET AL.
Fig. 4), and there was no difference in the percent
reduction in GPP or ER between summer and
autumn (independent ttest; t=0.59 and 0.40,
df =6.10 and 8.80, P=0.58 and 0.70, respec-
tively). However, there was a marginal difference
in recovery intervals between summer and
autumn for GPP (independent ttest; t=2.19,
df =5.74, P=0.07), but not ER (independent t
test; t=0.81, df =6.27, P=0.45), with a trend
for shorter recovery intervals (faster recovery) in
summer. Additionally, recovery intervals of GPP
and ER were not related to flood magnitude
(Fig. 4).
DISCUSSION
Extreme flow events have long been recog-
nized as drivers of ecosystem structure and func-
tion (Poff et al. 1997), and stream ecosystems
may be adapted to flashy hydrographs in areas
where extreme flows are a regular occurrence
(Fisher et al. 1982, Grimm 1987, Mart
ıet al.
1997). However, urban streams are stressed by a
myriad of physicochemical factors (Walsh et al.
2005, Vietz et al. 2016) in addition to the flashy
hydrographs caused by urbanization. In this
study, we found that both GPP and ER were
diminished immediately following high flows,
but both recovered in 4–18 d in these urban
streams. A recovery period of approximately two
weeks is similar to recovery times found for algal
biomass and stream metabolism in naturally
flashy desert streams (Fisher et al. 1982, Grimm
1987). The similarities between urban and desert
streams in response to a major disturbance, in
spite of drastically different abiotic conditions,
confirm the idea that hydrological disturbances
are a fundamental driver of ecosystem function-
ing across a range of stream types (Poff et al.
1997) and point to the need for a better under-
standing of disturbance ecology within urban
ecosystems (Grimm et al. 2017).
The degree of reduction in GPP (17–99%) and
ER (11–89%) from the 11 stream–flood events
exceeded values measured in an alpine river
(Uehlinger 2000), experimental alpine floods
(Uehlinger et al. 2003), and a forested headwater
stream (Roberts et al. 2007). In the current study,
each urban stream eventually recovered to pre-
flood rates despite greater reductions in meta-
bolic activity than has been found previously
(Uehlinger 2000, Roberts et al. 2007, Beaulieu
et al. 2013). Although there was a slight trend for
increasing metabolic recovery intervals with
increased flood RIs, these were not significant.
Therefore, based upon the current data, it does
not appear that flood magnitude is a primary
driver of either the degree of reduction or the
metabolic recovery interval. The lack of a rela-
tionship, coupled with similar metabolic recov-
ery intervals compared to desert streams (Fisher
et al. 1982, Grimm 1987), suggests that stream
metabolism in flashy environments may be
adapted to, and therefore highly resilient to,
high-flow events. The resilience that we observed
is similar to previous work in a suburban stream
that found that GPP was resilient to desiccation,
with metabolic activity returning within three
days of rewetting (Beaulieu et al. 2013).
Although resilience of stream metabolism may
be a characteristic of flashy streams, the role of
urban physicochemical stressors in driving both
baseline metabolic activity and the ability of the
stream to recover must also be considered. The
multiple physicochemical stressors occurring in
urban streams, collectively referred to as “the
Fig. 4. Metabolic recovery intervals for gross pri-
mary production (GPP; dark gray circles) and ecosys-
tem respiration (ER; light gray triangles) did not differ
from each other and were similar across a range of
flood recurrence intervals.
❖www.esajournals.org 8April 2017 ❖Volume 8(4) ❖Article e01776
REISINGER ET AL.
urban stream syndrome”(Walsh et al. 2005,
Wenger et al. 2009), have well-established effects
on stream biodiversity (Paul and Meyer 2001,
Walsh and Webb 2016), but a broader under-
standing of urbanization effects on stream bio-
geochemistry (including primary production)
has been identified as a priority for research
(Wenger et al. 2009). In this study, we found that
metabolic activity of a set of urban streams was
indistinguishable from a suite of reference, agri-
cultural, and other urban streams (Fig. 3; Hoel-
lein et al. 2013). This result is consistent with a
recent review of nitrogen transformations in
urban environments, which found no difference
in process rates between urban and reference
environments (Reisinger et al. 2016). Taken
together, these results suggest that biogeochemi-
cal process rates may be more resistant to urban-
ization effects than biodiversity (Lake 2000, Utz
et al. 2016).
Our results may be confounded by our com-
parison of floods of different magnitudes from
different urban streams. We attempted to mini-
mize these confounding effects by calculating
metabolic reductions and recovery intervals by
comparing post-flood to pre-flood activity from
the same stream. Despite our analytical
approach, physicochemical factors likely play a
large role in constraining not only absolute meta-
bolic rates, but also recovery trajectories. A larger
number of floods of various sizes from the same
stream are needed to clarify the effect of flood
magnitude on the reduction and recovery of
urban stream metabolism. As the technology for
collecting and analyzing data for whole-stream
metabolism continues to improve, it is likely that
these data will be available soon. Nonetheless,
we observed clear patterns in metabolic reduc-
tions and resiliency following floods in nearly all
stream–flood events.
Reduction and recovery of stream metabolism
has previously been shown to differ among sea-
sons. For example, in a forested headwater
stream, a spring storm caused a decrease in GPP
and ER, whereas an autumn storm caused an
increase in GPP but a decrease in ER. This sea-
sonal difference was attributed to the storm
washing leaves out of the stream in autumn and
allowing light to reach benthic autotrophs
(Roberts et al. 2007). A separate study found that
GPP generally recovered more quickly than ER
in two rivers across seasons, but that recovery
rates were faster in summer than in winter for
both GPP and ER (Uehlinger 2000). In contrast to
these previous studies, we found no seasonal dif-
ferences in recovery intervals between summer
and autumn. This lack of a difference may be
due to the increased frequency of storm flows in
urban streams (Walsh et al. 2005, Beaulieu et al.
2013, Smith and Kaushal 2015), which could
overwhelm other important drivers of metabo-
lism which vary seasonally such as canopy cover
(light), nutrient availability, and temperature.
Drastic reductions in stream metabolism fol-
lowing extreme flow events are not surprising.
However, many of the storms we analyzed were
not very extreme. In fact, of the 11 stream–flood
events we analyzed, four had a RI <1yr. We
acknowledge that the discharge record used to
develop our rating curves is relatively modest,
but extending the discharge record for a longer
period would have a minimal effect on non-
extreme RIs. Urbanization increases hydrological
flashiness (Walsh et al. 2005, Roy et al. 2009), and
scouring events can occur in response to modest
(<1.5 cm) rainfall events in urban streams (Mur-
dock et al. 2004, Hawley and Vietz 2016, Vietz
et al. 2016). It is possible that non-scouring events
still disturb stream communities and reduce
metabolism, suggesting that even smaller events
may impact stream ecosystem function. Overall,
the frequency of precipitation events coupled with
the metabolic response to even small floods
(Beaulieu et al. 2013, Table 2) suggests that urban
stream metabolism may be in a frequent if not
near-constant state of recovery.
Although metabolic activity was drastically
reduced by floods, the two streams studied during
Sandy were typically heterotrophic (GPP <ER)
before and after Sandy. In fact, we found no signif-
icant differences in reduction in GPP and ER in
response to floods, and similar recovery intervals
for both metabolic rates. The similar reductions
and recovery intervals between GPP and ER sug-
gest that most of the ER recovery is driven by
autotrophic respiration in urban streams (Beaulieu
et al. 2013), whereas heterotrophic respiration
may be less impacted by disturbance. Further
study of the differential effects of storm events
on autotrophic and heterotrophic processes is
warranted. Additionally, these differential effects
may alter other biogeochemical processes (e.g.,
❖www.esajournals.org 9April 2017 ❖Volume 8(4) ❖Article e01776
REISINGER ET AL.
denitrification) and change food web dynamics.
These differential effects, as well as general meta-
bolic responses of urban streams to flooding,
merit attention by stormwater managers, as
changes to the flow regime due to stormwater
management will likely have different outcomes
for biogeochemical and biodiversity responses
within urban streams.
The balance between GPP and ER has implica-
tions for the urban carbon cycle, which is known
to be altered by changes in anthropogenic inputs
(e.g., wastewater treatment plant effluent, sewage
leaks, organic compounds) and hydrologic con-
nectivity (Kaushal et al. 2014b). Urban stream res-
piration may rely on higher-quality carbon during
baseflow conditions than reference forested
streams (Kaushal et al. 2014a). This higher-quality
carbon also likely enters urban streams during
high-flow events (either from upstream or from
terrestrial sources), and may explain the appar-
ently smaller response of ER to storms compared
to GPP. Ultimately, differences in autotrophic and
heterotrophic responses to storm events may be
important, as headwater streams account for a
disproportionately high proportion of global CO
2
emissions (Raymond et al. 2013). Given this
importance, the responsiveness of GPP and ER in
urban streams to even minor storm events, and
the current and projected future level of urbaniza-
tion occurring globally (Grimm et al. 2008, United
Nations 2010, Pickett et al. 2011), there is a clear
need for more research on metabolic activity in
urban streams at multiple spatial and temporal
scales.
CONCLUSIONS
Urban stream ecosystems are subject to hydro-
logically extreme events in addition to a multitude
of other physicochemical stressors. Based upon
our results, it appears that hydrological extremes
may override other environmental stressors in
controlling stream metabolic activity, but the mag-
nitude of flooding does not affect metabolic recov-
ery. Reduction and subsequent recovery of GPP
and ER was common across a range of urban
streams and flood events, suggesting both high
sensitivity—higher than non-flashy stream ecosys-
tems such as an alpine river (Uehlinger 2000, Ueh-
linger et al. 2003) or a forested headwater stream
(Roberts et al. 2007)—and high resilience to
flooding disturbance. However, the recovery inter-
vals exhibited by these urban streams were similar
to those found in naturally flashy desert streams
(Fisher et al. 1982, Grimm 1987), suggesting that
urban streams can adapt to floods and other dis-
turbances in spite of myriad environmental stres-
sors (Grimm et al. 2017). Overall, we found that
urban streams can exhibit high metabolic activity
despite a multitude of physicochemical stressors
and that this metabolism rapidly recovers from
floods, regardless of magnitude.
ACKNOWLEDGMENTS
We thank H. Wellard-Kelly for help collecting and
organizing Superstorm Sandy data; D. Locke,
J. Lagrosa, and C. Welty for help with GIS analyses;
and participants in the Coastal SEES “sweet spots”pro-
ject for help identifying study sites used in this study.
A. Reisinger conceived of study, performed field rese-
arch, analyzed data, and wrote first draft of paper;
E. Rosi and H. Bechtold conceived of study, performed
field research, and wrote the paper; S. Kaushal and
P. Groffman conceived of study and wrote the paper;
and T. Doody performed field research and wrote the
paper. The authors declare no conflicts of interest. This
project was supported by the National Science Founda-
tion Coastal SEES Grant EAR-1426819 Award to
P. Groffman, E. Rosi, and S. Kaushal, as well as support
from the Baltimore Ecosystem Study (http://www.be
slter.org), a Long Term Ecological Research Station in
Baltimore, Maryland (NSF Grant DEB-1027188).
LITERATURE CITED
Acu~
na, V., A. Giorgi, I. Munoz, U. Uehlinger, and
S. Sabater. 2004. Flow extremes and benthic organic
matter shape the metabolism of a headwater Mediter-
ranean stream. Freshwater Biology 49:960–971.
Atkinson, B. L., M. R. Grace, B. T. Hart, and K. E. N.
Vanderkruk. 2008. Sediment instability affects the
rate and location of primary production and respi-
ration in a sand-bed stream. Journal of the North
American Benthological Society 27:581–592.
Beaulieu, J. J., C. P. Arango, D. A. Balz, and W. D.
Shuster. 2013. Continuous monitoring reveals
multiple controls on ecosystem metabolism in a
suburban stream. Freshwater Biology 58:918–937.
Bernot, M. J., et al. 2010. Inter-regional comparison of
land-use effects on stream metabolism. Freshwater
Biology 55:1874–1890.
Blake, E. S., T. B. Kimberlain, R. J. Berg, J. P. Cangialosi,
and J. L. Beven II. 2013. Tropical Cyclone Report
Hurricane Sandy (AL182012) 22–29 October 2012.
❖www.esajournals.org 10 April 2017 ❖Volume 8(4) ❖Article e01776
REISINGER ET AL.
National Oceanic and Atmospheric Administration,
National Hurricane Center, Miami, Florida, USA.
Carle, M. V., P. N. Halpin, and C. A. Stow. 2005.
Patterns of watershed urbanization and impacts on
water quality. Journal of the American Water
Resources Association 41:693–708.
Fisher, S. G., L. J. Gray, N. B. Grimm, and D. E. Busch.
1982. Temporal succession in a desert stream
ecosystem following flash flooding. Ecological
Monographs 52:93–110.
Grace, M. R., D. P. Giling, S. Hladyz, V. Caron, R. M.
Thompson, and R. Mac Nally. 2015. Fast processing
of diel oxygen curves: estimating stream metabolism
with BASE (BAyesian Single-station Estimation).
Limnology and Oceanography: Methods 13:103–114.
Grimm, N. B. 1987. Nitrogen dynamics during succes-
sion in a desert stream. Ecology 68:1157–1170.
Grimm, N. B., S. H. Faeth, N. E. Golubiewski, C. L.
Redman, J. Wu, X. Bai, and J. M. Briggs. 2008.
Global change and the ecology of cities. Science
319:756–760.
Grimm, N. B., S. T. A. Pickett, R. L. Hale, and M. L.
Cadenasso. 2017. Does the ecological concept of
disturbance have utility in urban social-ecological-
technological systems? Ecosystem Health and Sus-
tainability 3:e01255.
Hall, R. O., and J. L. Tank. 2003. Ecosystem metabo-
lism controls nitrogen uptake in streams in Grand
Teton National Park, Wyoming. Limnology and
Oceanography 48:1120–1128.
Hall, R. O., J. L. Tank, M. A. Baker, E. J. Rosi-Marshall,
and E. R. Hotchkiss. 2016. Metabolism, gas
exchange, and carbon spiraling in rivers. Ecosys-
tems 19:73–86.
Hatt, B. E., T. D. Fletcher, C. J. Walsh, and S. L. Taylor.
2004. The influence of urban density and drainage
infrastructure on the concentrations and loads of
pollutants in small streams. Environmental Man-
agement 34:112–124.
Hawley, R. J., and G. J. Vietz. 2016. Addressing the
urban stream disturbance regime. Freshwater
Science 35:278–292.
Henshaw, P. C., and D. B. Booth. 2000. Natural restabi-
lization of stream channels in urban watersheds.
Journal of the American Water Resources Associa-
tion 36:1219–1236.
Hoellein, T. J., D. A. Bruesewitz, and D. C. Richardson.
2013. Revisiting Odum (1956): a synthesis of aquatic
ecosystem metabolism. Limnology and Oceanogra-
phy 58:2089–2100.
Homer, C. G., J. A. Dewitz, L. Yang, S. Jin, P. Danielso,
G. Xian, J. Coulston, N. D. Herold, J. D. Wickham,
and K. Megown. 2015. Completion of the 2011
National Land Cover Database for the contermi-
nous United States-Representing a decade of land
cover change information. Photogrammetric Engi-
neering and Remote Sensing 81:345–354.
Kaushal, S. S., and K. T. Belt. 2012. The Urban Water-
shed Continuum: evolving spatial and temporal
dimensions. Urban Ecosystems 15:409–435.
Kaushal, S. S., K. Delaney-Newcomb, S. E. G. Findlay,
T. A. Newcomer, S. Duan, M. J. Pennino, G. M.
Sivirichi, A. M. Sides-Raley, M. R. Walbridge, and
K. T. Belt. 2014a. Longitudinal patterns in carbon
and nitrogen fluxes and stream metabolism along
an urban watershed continuum. Biogeochemistry
121:23–44.
Kaushal, S. S., W. H. McDowell, and W. M. Wollheim.
2014b. Tracking evolution of urban biogeochemical
cycles: past, present, and future. Biogeochemistry
121:1–21.
Lake, P. S. 2000. Disturbance, patchiness, and diversity
in streams. Journal of the North American Bentho-
logical Society 19:573–592.
Mart
ı, E., N. B. Grimm, and S. G. Fisher. 1997. Pre- and
post-flood retention efficiency of nitrogen in a
Sonoran Desert stream. Journal of the North
American Benthological Society 16:805–819.
Moore, A. A., and M. A. Palmer. 2005. Invertebrate
biodiversity in agricultural and urban headwater
streams: implications for conservation and man-
agement. Ecological Applications 15:1169–1177.
Mulholland, P. J., et al. 2001. Inter-biome comparison
of factors controlling stream metabolism. Freshwa-
ter Biology 46:1503–1517.
Mulholland, P. J., et al. 2008. Stream denitrification
across biomes and its response to anthropogenic
nitrate loading. Nature 452:202–205.
Murdock, J., D. Roelke, and F. Gelwick. 2004. Interac-
tions between flow, periphyton, and nutrients in a
heavily impacted urban stream: implications for
stream restoration effectiveness. Ecological Engi-
neering 22:197–207.
Odum, H. T. 1956. Primary production in flowing
waters. Limnology and Oceanography 1:102–117.
Paul, M. J., and J. L. Meyer. 2001. Streams in the urban
landscape. Annual review of Ecology and System-
atics 32:333–365.
Pickett, S. T. A., et al. 2011. Urban ecological systems:
scientific foundations and a decade of progress.
Journal of Environmental Management 92:331–362.
Poff, N. L., J. D. Allan, M. B. Bain, J. R. Karr, K. L.
Prestegaard, B. D. Richter, R. E. Sparks, and J. C.
Stromberg. 1997. The natural flow regime. Bio-
Science 47:769–784.
Raymond, P. A., et al. 2013. Global carbon dioxide
emissions from inland waters. Nature 503:355–359.
Reisinger, A. J., P. M. Groffman, and E. J. Rosi-Marshall.
2016. Nitrogen cycling process rates across urban
ecosystems. FEMS Microbiology Ecology 92:fiw198.
❖www.esajournals.org 11 April 2017 ❖Volume 8(4) ❖Article e01776
REISINGER ET AL.
Roberts, B. J., and P. J. Mulholland. 2007. In-stream
biotic control on nutrient biogeochemistry in a
forested stream, West Fork of Walker Branch. Jour-
nal of Geophysical Research 112:G04002.
Roberts, B. J., P. J. Mulholland, and W. R. Hill. 2007.
Multiple scales of temporal variability in ecosystem
metabolism rates: results from 2 years of continu-
ous monitoring in a forested headwater stream.
Ecosystems 10:588–606.
Roley, S. S., J. L. Tank, N. A. Griffiths, R. O. Hall Jr.,
and R. T. Davis. 2014. The influence of floodplain
restoration on whole-stream metabolism in an agri-
cultural stream: insights from a 5-year continuous
data set. Freshwater Science 33:1043–1059.
Roy, A. H., A. L. Dybas, K. M. Fritz, and H. R. Lubbers.
2009. Urbanization affects the extent and hydrologic
permanence of headwater streams in a midwestern
US metropolitan area. Journal of the North Ameri-
can Benthological Society 28:911–928.
Smith, R. M., and S. S. Kaushal. 2015. Carbon cycle of
an urban watershed: exports, sources, and metabo-
lism. Biogeochemistry 126:173–195.
Song, C., W. K. Dodds, M. T. Trentman, J. R€
uegg, and
F. Ballantyne. 2016. Methods of approximation
influence aquatic ecosystem metabolism estimates.
Limnology and Oceanography: Methods 14:557–569.
Sudduth, E. B., B. A. Hassett, P. Cada, and E. S. Bern-
hardt. 2011. Testing the field of dreams hypothesis:
functional responses to urbanization and restora-
tion in stream ecosystems. Ecological Applications
21:1972–1988.
Uehlinger, U. 2000. Resistance and resilience of ecosys-
tem metabolism in a flood-prone river system.
Freshwater Biology 45:319–332.
Uehlinger, U., B. Kawecka, and C. T. Robinson. 2003.
Effects of experimental floods on periphyton and
stream metabolism below a high dam in the Swiss
Alps (River Spol). Aquatic Sciences 65:199–209.
United Nations. 2010. Water for sustainable urban
human settlements. Briefing Note. United Nations,
New York, New York, USA.
Utz, R. M., K. G. Hopkins, L. Beesley, D. B. Booth, R. J.
Hawley, M. E. Baker, M. C. Freeman, and K. L.
Jones. 2016. Ecological resistance in urban streams:
the role of natural and legacy attributes. Fresh-
water Science 35:380–397.
Vietz, G. J., C. J. Walsh, and T. D. Fletcher. 2016. Urban
hydrogeomorphology and the urban stream syn-
drome: treating the symptoms and causes of geo-
morphic change. Progress in Physical Geography
40:480–492.
Walsh, C. J., A. H. Roy, J. W. Feminella, P. D. Cotting-
ham, P. M. Groffman, and R. P. Morgan II. 2005.
The urban stream syndrome: current knowledge
and the search for a cure. Journal of the North
American Benthological Society 24:706–723.
Walsh, C. J., and J. A. Webb. 2016. Interactive effects of
stormwater drainage, land clearance, and flow
regime on stream macroinvertebrate assemblages
across a large metropolitan region. Freshwater
Science 35:324–339.
Walters, D. M., A. H. Roy, and D. S. Leigh. 2009. Envi-
ronmental indicators of macroinvertebrate and fish
assemblage integrity in urbanizing watersheds.
Ecological Indicators 9:1222–1233.
Wenger, S. J., et al. 2009. Twenty-six key research ques-
tions in urban stream ecology: an assessment of the
state of the science. Journal of the North American
Benthological Society 28:1080–1098.
Wolman, M. G. 1967. A cycle of sedimentation and
erosion in urban river channels. Geografiska
Annaler Series A: Physical Geography 49:255–265.
❖www.esajournals.org 12 April 2017 ❖Volume 8(4) ❖Article e01776
REISINGER ET AL.
