MARINE ECOLOGY PROGRESS SERIESMar Ecol Prog Ser

Vol. 491: 199–219, 2013doi: 10.3354/meps10466

Published October 2

INTRODUCTION

Estuaries provide nursery habitat for many econo -mically and ecologically important fishes (Weinstein1979, Able 1999, Minello et al. 2003). The nurseryfunction of estuaries derives from physical character-istics, such as proximity to freshwater runoff (Hollandet al. 2004) and semi-enclosed geomorphology(Kjerf ve & Magill 1989). These attributes foster anenvironment with higher temperatures (Day et al.

1989), higher productivity (Deegan 2002), and, often,lower predation risk (Paterson & Whitfield 2000,Manderson et al. 2004) than adjacent marine habi-tats. However, these same attributes create a verydynamic physicochemical environment, whereinabio tic factors such as temperature, salinity, and dis-solved oxygen (DO) vary considerably both tempo-rally and spatially (Tyler et al. 2009).

Anthropogenic impacts can disrupt natural pat-terns of estuarine variability to the detriment of resi-

© Inter-Research 2013 · www.int-res.com*Email: damian.brady@maine.edu

Movement of juvenile weakfish Cynoscion regalisand spot Leiostomus xanthurus in relation to

diel-cycling hypoxia in an estuarine tidal tributary

Damian C. Brady1,2,*, Timothy E. Targett1

1University of Delaware, School of Marine Science and Policy, Lewes, Delaware 19958, USA2Present address: University of Maine, School of Marine Sciences, Walpole, Maine 043573, USA

ABSTRACT: Fish movement and the spatial and temporal dynamics of hypoxia determine hypoxiaexposure and the effect of poor water quality on nursery habitat function. Although water qualitycriteria for dissolved oxygen (DO) are well defined, hypoxia exposure of juvenile estuary-depen-dent fishes in situ is largely unknown. Thirty-one juvenile weakfish Cynoscion regalis and spotLeiostomus xanthurus were implanted with acoustic tags. Fish were acclimated for 5 d to eitherDO saturation or diel-cycling hypoxia (cycling between 11.0 and 2.0 mg O2 l−1). Fish were releasedduring summer into Pepper Creek, Delaware, an estuarine tributary. A logistic generalized addi-tive model with generalized estimating equations was used to determine which environmentalcovariates significantly discriminated between movement types. Individual fish tracks were over-lain on spatiotemporal contour plots of DO that highlight behavioral avoidance thresholds deter-mined in the laboratory. Most models showed that DO, tide, and the spatial DO gradient were sig-nificant predictors of movement. Saturation-acclimated fish generally avoided DO

Mar Ecol Prog Ser 491: 199–219, 2013

dent populations (Breitburg 2002). For example,human-mediated nutrient enrichment (eutrophica-tion) can result in excessive primary production andcan lead to severe hypoxia (0.1 to 2.0 mg O2 l−1)or anoxia (Reckhow & Chapra 1999). Diel-cyclinghypoxia occurs within the photic zone (extending tothe bottom in shallow systems and above the pycno-cline in deeper systems). It is driven primarily by thedaytime-nighttime (diel) cycling of DO productionand consumption by living algae (Kemp & Boynton1980, D’Avanzo & Kremer 1994). DO fluctuates pre-dictably over the 24 h period, with lowest concentra-tions usually occurring around dawn following night-time re spiration and highest concentrations in thelate afternoon following maximum daytime photo-synthesis (D’Avanzo & Kremer 1994, Tyler et al.2009). Tyler et al. (2009) linked severe hypoxiaevents of greatest intensity, duration, and spatialextent with the following series of events: a pulse ofrain followed by a series of sunny days followed byone or more cloudy days.

Estuarine areas of low flushing, salinity, and DO aswell as high productivity are often simultaneouslyareas of high fish abundance and densities (Meng etal. 2001, Ross 2003, Yamashita et al. 2003, Meng et al.2004). When dispersal and settlement are based on asimple set of rules (e.g. traverse to productive head-waters), and environmental change (e.g. eutrophica-tion) uncouples the cues for settlement from habitatquality, then the potential for an ecological trap exists(Kokko & Lopez-Sepulcre 2006). An ecological trap isa low-quality habitat that is preferred over a higherquality habitat (Battin 2004). Any human activitysuch as eutrophication, and resulting diel-cyclinghypoxia, may cause headwater habitats to becomeecological traps.

Ultimately, if nursery areas are disproportionatelyused by juvenile fishes and impacted by diel-cyclinghypoxia, determining how fishes utilize these areasin both space and time and on ecologically relevantscales (i.e. measurable effects on growth and mortal-ity) is critical to assessing the impact of eutrophica-tion on essential fish habitat (Beck et al. 2001). Manystudies of fish–hypoxia interactions have utilizedtrawl surveys to determine presence/absence andrelative abundance of fishes based on proximate DOconditions (Howell & Simpson 1994, Eby & Crowder2002, Bell & Eggleston 2005, Tyler & Targett 2007).For example, Tyler & Targett (2007) demonstratedthat juvenile weakfish Cynoscion regalis were ab -sent from the upper portion of Pepper Creek, a tidaltributary of the Delaware Coastal Bays, when diel-cycling DO was below ~2.0 mg O2 l−1. However,

when DO was >2.0 mg O2 l−1, juvenile weakfish weredisproportionally more abundant in the upper tribu-tary than farther down the system.

Determining hypoxia avoidance mechanisms andthresholds will enhance predictive models. Basicquestions remain in relation to hypoxia avoidanceincluding: What DO levels induce the beginning ofavoidance behavior, how fast do fish move duringavoidance, and how far do fish move away fromhypoxic zones? Fish modeling studies consistentlyrecognize the importance of movement parameter-ization in calibration and validation (Tyler & Rose1997, Railsback & Harvey 2002, Wildhaber & Lam-berson 2004, Aumann et al. 2006). The scope ofmovement is particularly important in this regard.For example, how far a fish can move in onemodel time step will ultimately, in conjunctionwith the spatial extent of hypoxia, determine DOexposure.

Few studies have performed an explicit evaluationof individual fish movements in relation to hypoxiczones in estuaries. Emigration rates of 53 acousticallytagged summer flounder (235 to 535 mm total length[TL]) from the Mullica River-Great Bay Estuary, NJ,USA were negatively correlated with DO (Sackett etal. 2007). However, the range of DO values measured(i.e. 4.3 to 7.5 mg O2 l−1 at 18 to 26°C) were not lethalor even growth limiting (Stierhoff et al. 2006). More-over, all the more direct observations of individualsin hypoxic environments have been conducted inenvironments with water column stratification-induced hypoxia. No study has examined the move-ment of individual fish in relation to diel-cyclinghypoxia in shallow nursery environments.

The objectives of this study were to trackjuvenile weakfish and spot, 2 estuary-dependentsciaenid fishes, in an estuarine tributary nurseryarea in re lation to abiotic variables. Specifically,we examined the degree, duration, and spatialextent of diel-cycling hypoxia that elicited anavoidance response and characterized that re -sponse. We also characterized the role of diel-cycling hypoxia acclimation in habitat use. Thisstudy utilized a gated acoustic array along the axisof a tidal tributary that is consistently hypoxic dur-ing the summer in the early morning to early after-noon (Tyler et al. 2009). The high spatial and tem-poral resolution of DO monitoring and passivetracking allowed a comparison of the spatial scaleof fish movement with the spatial scale of hypoxicevents and a determination of what environmentalDO scenarios may be most problematic for estuary-dependent fishes.

200

Brady & Targett: Movement of fish around hypoxia

MATERIALS AND METHODS

Study area and field data collection

This study was conducted primarily in PepperCreek, a tributary of the Delaware Coastal Bays(Fig. 1a). Pepper Creek is a mesohaline-to-poly -haline tidal tributary with DO ranging from anoxiato supersaturation (often 300% saturation; Tyler etal. 2009). Depth at low tide ranges from 1.5 m in thechannel to

Mar Ecol Prog Ser 491: 199–219, 2013

sions (simultaneous tag transmissions result in onlyone tag detection). A 40 to 120 s nominal delay waschosen to balance these 2 sources of lost detections.Range testing was conducted during the summer of2007 at 3 sites (PC2, PC5, and PC6; Fig. 1c) over 3consecutive days. Receiver range was within 200 mduring over 90% of detections with a rare maximumof 600 m. When a fish was detected on a receiver, weassumed its location was identical to the receiver’slocation. Because the detection efficiency of the re -ceiver could be up to 600 m away, there is uncer-tainty in the match between fish location and abioticvariable. Generally, range was negatively correlatedwith boat traffic (i.e. range increased at night and onlow tides).

Tagging and release

Juvenile weakfish and spot (Table 1) were collec -ted from nursery areas in the Delaware Coastal Baysin 2005 and 2006 using an otter trawl. The mean(±SD) standard length (SL) of tagged fish was 137 ±16 mm (Table 1). Fish were maintained in recirculat-ing aquaria at constant temperature (25°C), salinity(20 ppt), and a 14 h light:10 h dark photoperiod for atleast 5 d before tagging. Fish were fed frozen mysidshrimp Mysis relicta. The protocol used to implantacoustic tags was developed by Harms (2005) andHarms & Lewbart (2000) and had been used success-fully on juvenile spot by K. Craig in the Nuese River,NC, USA (pers. comm.). Fish were anesthetized in aninduction bath of ~170 ppm MS-222 and were placedon a piece of seawater-saturated foam while a main-tenance dose of 100 ppm MS-222 was perfused overtheir gills. A tag was inserted through a small inci-sion on the ventral side midway between the analand pectoral fins.

One objective of this study was to understand therole of diel-cycling hypoxia acclimation in habitatuse. Six weakfish were acclimated to diel-cycling DOin a computer controlled DO system (Grecay & Stier-hoff 2002) for 5 d before release (see Table 1) to exa -mine the effect of diel-cycling hypoxia acclimationon the behavioral response of weakfish. Weakfishwere used for the acclimation experiment becauseBrady et al. (2009) demonstrated that weakfish altertheir behavior as a result of acclimation in the labora-tory, but this had not been demonstrated in situ. During acclimation, minimum (2.0 mg O2 l−1) andmaximum (11.0 mg O2 l−1) DO concentrations coin-cided with the beginning of the light (07:00 h) anddark (21:00 h) periods, respectively, to reflect DO

conditions observed in the field (Tyler et al. 2009). Allother fish were held in the laboratory for 5 d undersaturated DO conditions before being released in 6batches of 1 to 8 fish at site PC2 (Fig. 1b,c). No fishwith signs of infection near the incision were re -leased which accounts for uneven numbers of fish ineach batch release. Finally, there was no way to becertain that fish did not die during a track; however,Eguiluz & Wong (2005) reported maximum tidalflows of 0.1 m s−1 in the Delaware Coastal Bays andall individuals analyzed made at least one movementgreater than this velocity.

Data analysis

Temperature, salinity, and DO were linearly inter-polated, spatially and temporally, to create contourplots. The color map for DO was tuned to laboratorybehavioral thresholds (Brady et al. 2009). Brady et al.(2009) exposed juvenile weakfish to a step-wise de -crease in DO from 7.0 to 0.4 mg O2 l−1. Maximumswimming speed occurred when DO was 2.8 mg O2l−1. Angular correlation between successive movesdecreased precipitously as DO decreased from1.4 mg O2 l−1 to 0.8 mg O2 l−1. These 2 values werediscretized in the color map as blue (>1.4 mg O2 l−1

and ≤2.8 mg O2 l−1) and black (≤1.4 mg O2 l−1). Fishtracks were plotted on the DO contour plots to visual-ize if these laboratory-derived behavioral thresholdstranslated to important thresholds for activity in thefield.

To associate each estimated location with its atten-dant abiotic conditions, increased temporal and spa-tial resolution was needed. Therefore, a piecewisecubic hermite interpolating polynomial (PCHIP) wasused to interpolate DO, temperature, and salinity inboth time and space. This interpolation techniquewas preferable to spline interpolation in this contextbecause it offers smooth interpolation betweenpoints while not overshooting data to create artificialrelative maximums and minimums (Kahaner et al.1989). PCHIP was used to determine exposure if theestimated location occurred during the 15 min periodbetween sonde measurements. Additionally, detec-tions of the same individual by 2 acoustic receiverswithin less than 5 min of each other were assumed toresult from overlapping receiver ranges. In this case,the fish’s estimated location was assumed to bebetween the receivers and PCHIP interpolation wasused to assign abiotic conditions to this estimatedlocation. Overlapping receiver detections were rela-tively rare (on average,

Brady & Targett: Movement of fish around hypoxia 203

Bat

ch

Tag

Sp

ecie

sA

ccli

mat

ion

Sta

nd

ard

Wet

Det

ec-

Tim

e of

fir

stT

ime

of l

ast

Du

rati

onC

reek

% o

fre

leas

e ID

len

gth

wei

gh

tti

ons

det

ecti

ond

etec

tion

of t

rack

resi

den

cyti

me

inn

o.(m

m)

(g)

(d)

(d)

cree

k

113

3C

. reg

alis

Sat

ura

tion

172

64.2

314

08/2

6/05

18:

4008

/28/

05 0

0:34

1.

251.

1289

.61

137

L. x

anth

uru

sS

atu

rati

on13

253

.055

008

/26/

05 1

8:40

08/2

8/05

15:

591.

890.

5126

.81

144

L. x

anth

uru

sS

atu

rati

on12

544

.010

4408

/26/

05 1

8:47

08/3

0/05

07:

523.

540.

4913

.91

149

L. x

anth

uru

sS

atu

rati

on13

448

.216

108

/26/

05 1

8:49

08/2

8/05

11:

071.

681.

4888

.11

139

C. r

egal

isS

atu

rati

on18

064

.846

708

/26/

05 1

8:58

08/2

8/05

08:

271.

561.

4693

.41

142

L. x

anth

uru

sS

atu

rati

on12

037

.236

408

/26/

05 1

9:07

08/2

9/05

22:

273.

141.

4847

.11

148

L. x

anth

uru

sS

atu

rati

on12

139

.763

708

/26/

05 1

9:16

08/2

9/05

09:

142.

581.

4756

.92

140

C. r

egal

isS

atu

rati

on12

025

.825

08/3

0/05

16:

0308

/31/

05 1

4:43

0.94

0.33

35.2

314

6C

. reg

alis

Sat

ura

tion

161

54.9

1186

09/0

1/05

13:

0409

/06/

05 1

1:45

4.94

4.94

100.

03

135

C. r

egal

isS

atu

rati

on13

027

.825

09/0

1/05

13:

0609

/02/

05 0

6:45

0.74

0.74

100.

04

131

C. r

egal

isH

ypox

ia13

636

.051

708

/28/

06 1

5:24

09/0

5/06

04:

357.

551.

9525

.94

147

C. r

egal

isH

ypox

ia12

636

.615

4508

/28/

06 1

5:28

08/3

1/06

10:

292.

792.

6092

.94

130

C. r

egal

isH

ypox

ia18

082

.878

908

/28/

06 1

5:35

10/0

3/06

03:

1635

.49

0.70

2.0

413

4C

. reg

alis

Hyp

oxia

148

52.9

372

08/2

8/06

15:

5008

/31/

06 0

1:03

2.38

1.71

71.7

414

5C

. reg

alis

Hyp

oxia

132

40.5

108

/29/

06 0

0:36

08/2

9/06

00:

360.

000.

00N

A5

316

C. r

egal

isS

atu

rati

on13

544

.418

908

/31/

06 1

1:21

09/2

3/06

19:

3223

.34

1.36

5.8

530

5C

. reg

alis

Sat

ura

tion

132

34.6

1014

08/3

1/06

11:

2109

/13/

06 0

1:46

12.6

02.

4519

.55

320

C. r

egal

isS

atu

rati

on13

341

.328

708

/31/

06 1

1:23

10/0

3/06

02:

3432

.63

0.86

2.6

513

6C

. reg

alis

Hyp

oxia

135

36.2

1218

08/3

1/06

11:

2410

/02/

06 1

8:48

32.3

10.

581.

85

143

C. r

egal

isS

atu

rati

on12

832

.917

08/3

1/06

11:

2408

/31/

06 2

0:58

0.40

0.35

88.2

532

1C

. reg

alis

Sat

ura

tion

145

51.7

3808

/31/

06 1

1:27

08/3

1/06

17:

370.

260.

2610

0.0

531

5C

. reg

alis

Sat

ura

tion

137

46.3

5908

/31/

06 1

2:55

08/3

1/06

21:

590.

380.

3078

.45

319

C. r

egal

isS

atu

rati

on13

035

.512

7508

/31/

06 1

3:19

09/0

8/06

01:

117.

492.

2630

.26

308

C. r

egal

isS

atu

rati

on13

033

.622

309

/04/

06 1

2:30

09/0

7/06

04:

592.

692.

6910

0.0

631

2C

. reg

alis

Sat

ura

tion

133

41.0

369

09/0

4/06

12:

3109

/06/

06 0

9:59

1.90

1.90

100.

06

309

C. r

egal

isS

atu

rati

on13

235

.57

09/0

4/06

12:

3109

/04/

06 1

8:39

0.26

0.26

100.

06

300

C. r

egal

isS

atu

rati

on12

429

.167

009

/04/

06 1

2:33

09/0

7/06

14:

353.

082.

4378

.76

307

C. r

egal

isS

atu

rati

on13

339

.165

309

/04/

06 1

2:38

09/1

0/06

01:

165.

532.

3141

.96

306

C. r

egal

isS

atu

rati

on13

236

.218

309

/04/

06 1

2:38

09/1

0/06

16:

316.

162.

8646

.56

313

C. r

egal

isS

atu

rati

on13

341

.213

709

/04/

06 1

2:41

09/0

7/06

19:

423.

292.

4474

.16

314

C. r

egal

isS

atu

rati

on13

135

.977

509

/04/

06 1

2:41

09/1

0/06

03:

365.

623.

5463

.0

Tab

le 1

. C

ynos

cion

reg

alis

and

Lei

osto

mu

s xa

nth

uru

s. C

har

acte

rist

ics

of t

agg

ed f

ish

in

clu

din

g s

pec

ies,

acc

lim

atio

n h

isto

ry (

hyp

oxia

in

dic

ates

5 d

of

exp

osu

re t

o d

iel-

cycl

ing

hyp

oxia

: cyc

lin

g b

etw

een

2.0

an

d 1

1 m

g O

2l−

1 ), l

eng

th a

nd

wei

gh

t an

d c

har

acte

rist

ics

of e

ach

in

div

idu

al’s

tra

ck i

ncl

ud

ing

tim

e of

fir

st a

nd

las

t d

etec

tion

, d

ura

tion

of

trac

k a

nd

cre

ek r

esid

ency

. Tim

es a

re m

m/d

d/y

y h

:min

Mar Ecol Prog Ser 491: 199–219, 2013

interpolation between sites located approximately500 m apart may also introduce an additional sourceof uncertainly when associating environmental co -variates with estimated locations.

To characterize the local abiotic gradients thatcould influence fish movement, the derivative of thePCHIP at every estimated location was solved bothtemporally and spatially (i.e. along the axis of thecreek). For example, a negative spatial gradient inDO means that DO was higher in the upper tributarythan the fish’s current location and a negative tempo-ral derivative indicates DO was decreasing during adetection. Insolation at the time of each estimatedlocation was used to detect diel-periodicity in move-ment patterns. The rate of tide height change wasdetermined for each estimated location by solving forthe derivative of PCHIP interpolated tidal amplitude.

Swimming speed was determined by dividing thedistance between estimated locations by the time be -tween estimated locations. There are several un -known factors which could affect estimation of swim-ming speed. The largest source of underestimation isthe assumption that fish move in a straight line fromstation to station. Any deviation from a straight linepath means the actual swimming speed was under-estimated. On the other hand, there are also 2 un -known sources of overestimation: receiver rangevariability and tidal flow. The present study assumedthat fish moved from the center of the range of onereceiver to the center of the range of another re -ceiver. If fish moved from edge to edge of receiverranges, then the actual distance the fish traversedwas less than estimated. Also, tidal flow can aidadvection of fish if movement is tidally synchronized.Tidal flow in this system is 0.1 m s−1 at maximum flow(Eguiluz & Wong 2005). However, a relative compar-ison between individuals and groups released at thesame time is instructive and whether tidally aided ornot, this estimation represents the speed fish need toattain to avoid hypoxia. Additionally, the more con-secutive estimated locations at different receivers ob -served, the more accurate the estimation of speed.Speed estimation using consecutive estimated loca-tions at 3 receivers were used to estimate swimmingspeed whenever possible.

Fish movement in relation to diel-cycling hypoxiawas initially characterized by Batch Release (BR; allfish released at the same time were pooled), due tothe potential influence of initial conditions on fishmovement. Each estimated location of a given fishwas compared to its previous estimated location todetermine if the fish had remained at that receiver(hereafter referred to as ‘stay-put’), or moved up -

stream or downstream. A binomial-based gene -ralized additive model (GAM) with a logit link func-tion was used to identify environmental variables—i.e. DO, temperature (T), salinity, the spatial andtemporal gradients in DO (d[DO] dx−1 and d[DO] dt−1,re spectively) and temperature (d[T] dx−1 and d[T]dt−1, respectively), insolation, and the rate of tideheight change (d[tide] dt−1)—that were predictors ofmovement (Diggle et al. 1995, Pirotta et al. 2011). Themethod of generalized estimating equations (GEEs)was used to account for correlations among estimatedlocations. Briefly, the GEE method uses the residualsbetween observations and a model that assumes in-dependence within each fish, and then uses the cor-relation estimates to obtain new coefficient estimates.The process is repeated until the change between 2successive estimates is very small. The wor king cor-relation matrix was assumed to be auto regressive;however, GEEs are relatively robust to misspecifica-tion of working correlation matrices (Pan & Connett2002). Models were run with standardized (by themean) and unstandardized covariates. Gradient vari-ables were log transformed after the absolute value ofthe most negative gradient variable was ad ded to allestimated locations because the derivatives of thetime series and spatial gradient were ex treme ly largewhen the decrease or increase was maximum. Logtransformation condensed the range of values foranalysis and decreased the leverage of extremelylarge values in the model. The distribution of re -siduals was checked to ensure normal distributions.

Because there were 3 categories of movement (i.e.stay-put, upstream, and downstream), 2 logistic mod-els (i.e. upstream versus stay-put and downstreamversus stay-put) were compared using stay-put as thereference outcome. Quasi-likelihood information cri-teria were used to determine the significance of eachcovariate and whether each covariate significantlyimproved model prediction (Pan 2001b,a, Aarts et al.2008). The odds ratio (OR) was also used to comparethe influence of multiple covariates on movement.The OR for each outcome/covariate combination(e.g. ORupstream/DO) represents the change in the prob-ability (P) of that outcome given a one unit increase inthe covariate:

(1)

where j represents the outcome (i.e. upstream ordownstream), p represents the covariate, and a and brepresent 2 values of the covariate (e.g. 0 and 1 if theodds ratio is describing the change in probability fora one unit increase in the covariate). The units of gra-

ORstay-put

j p a bP Y j x a P Y x aP Y j x/

( , )( | ) / ( | )( |

= = = = == =bb P Y x b) / ( | )= =stay-put

204

Brady & Targett: Movement of fish around hypoxia

dient covariates (e.g. d[DO] dx−1 and d[DO] dt−1)resulted in relatively small values; therefore, stan-dardized ORs were computed for these covariates forsimpler interpretation. For example, the ORs ford(DO)/dx represent the increased probability of amovement for a one standard deviation (SD) changein d(DO) dx−1. The 95% confidence intervals (CI) foreach OR were calculated. If the CI included 1, thenthat covariate was assumed to have little impact oninducing movement. When OR = 1, a change in 1 unitor SD in the covariate has no influence on the out-come. Conversely, if OR = 0.5 or 2, a 1 unit or SDchange in the covariate results in halving or doublingof the probability of that movement type, respec-tively. Finally, the Wald χ2 test statistic was obtainedby comparing the quasi-likelihood estimate of thecoefficients (βs) to an estimate of the standard error.Relatively high Wald χ2 values indicate those covari-ates are significant contributors to the outcome (i.e.upstream or downstream movement). All statisticalanalyses were performed using SAS.

RESULTS

Attributing causation to a relationship between anyindividual environmental covariate and movement iscomplicated by unmeasured biotic factors such aspredation and the prey field. However, an assump-tion supported by many laboratory studies of weak-fish and spot (e.g. McNatt & Rice 2004, Brady etal. 2009, Stierhoff et al. 2009b) is that when DOapproaches lethal levels (below 2 mg O2 l−1), fish pri-oritize hypo xia avoidance above other biotic and abi-otic stimuli. Therefore, the analysis breaks downeach batch release according to statistically relevantcovariates that were correlated with movement types(i.e. up stream and downstream) and a focus on themovement of individual fish during hypoxic periodswhen appropriate. The average duration of a track in2005 (BR 1−3) and 2006 (BR 3−6) was 2.23 and 8.86 d,respectively (Table 1). The track duration differenceis partially attributable to increased spatial coveragein 2006 (Fig. 1).

A majority of the GAM logits included the rate oftide height change, DO, and the spatial gradient inDO (Table 2). These statistical results were corrobo-rated by graphical analyses of tracks. Upstream anddownstream movements were clearly linked to floodand ebb tides, respectively. The rate of tide heightchange was 4.23 × 10−5 and −8.47 × 10−5 m s−1 for allupstream and downstream movements, respectively.Additionally, fish tended to avoid DO levels below

2.8 mg O2 l−1. Significant correlations between thespatial DO gradient and movement in most GAMlogits suggest that fish may utilize this gradient aspart of their avoidance mechanism. Finally, diel-cycling hypoxia acclimated fish avoided DO levelslower than saturation-acclimated fish, based onlonger exposure times to DO levels in between 2.8and 1.4 mg O2 l−1.

Batch releases 1 and 2

BR 1 and BR 2 occurred on August 26 and 30, 2005,respectively (Table 1). Meteorological conditionsduring the week following the BR 1 and 3 d of BR 2(N = 5 saturation-acclimated spot and N = 2 satu -ration-acclimated weakfish) were hot (i.e. 27.3°C)and cloudy (Fig. 2b) with calm winds (averaging only1.9 m s−1). High temperature, low insolation, and lowwind speed during BR 1 resulted in the longest last-ing and most spatially extensive hypoxic conditions(Fig. 2a) of any batch release period.

Of the 3072 estimated locations during BR 1, 65were upstream movements and 92 were down-stream. The GAM logit describing upstream move-ments indicated that particularly important variables(in order of Wald χ2 magnitude in Table 2) in predict-ing upstream movements were temperature, the spa-tial gradient in temperature, salinity, the spatial gra-dient in DO, DO, rate of tide height change, andinsolation. Model results were heavily influenced byenvironmental conditions during a 12 h period be -tween midnight on August 27 and noon on August28. Almost half of upstream movements occurredduring these 12 h. Hypoxia occurred throughout thistime period and extended from the head of tide tonear the mouth of the tributary at PC5 and PC6(Fig. 2a). Upstream movements during this time weremade by smaller spot that had recently moved 5 kmdownstream to avoid hypoxia (

Mar Ecol Prog Ser 491: 199–219, 2013

a negative DO gradient (Fig. 4a,e). Ultimately, thestron gest predictors of upstream movements weretemperature and the spatial gradient in temperature(Table 2). The upstream GAM logit indicates thatupstream movements were ~3 times (i.e. the inverseof the OR) more likely given a 1 SD decrease in thetemperature gradient (i.e. the upstream location wascharacterized by higher temperatures; Table 2).

Overall, temperature related covariates were help-ful predictors of upstream and downstream move-ments in BR 1. Temperature, the rate of temperaturechange, and insolation (in order of Wald χ2 magni-tude in Table 2) were significant predictors of down-

stream movements. Downstream movements werecharacterized by a 0.44°C s−1 decrease in tempera-ture, compared to a 0.0062°C s−1 decrease observedduring stay-put estimated locations, consistent withthe finding that most downstream movements weremade at night. A decrease in 1 SD in insolationresulted in a doubling of the probability of a down-stream movement (OR = 0.51, Table 2). Another sim-ilarity between up and downstream movements wasthat the sign of the coefficient for d(DO) dx−1 was positive for all movements, suggesting that weakfishand spot use the local spatial gradient in DO to moveto areas of higher DO.

206

Batch Direction Insolation DO d(DO) dx−1

release no. Estimate OR Wald χ2 Estimate OR Wald χ2 Estimate OR Wald χ2

1 Upstream −0.79* 0.45 6.02 −0.59** 0.55 8.84 1.13** 3.10 8.89Downstream −0.67** 0.51 6.85 0.10 1.10 0.57 0.21 0.21 0.89

3 Downstream 0.79** 2.19 435 0.96** 2.60 312 −1.02** 0.36 473

4 Upstream 0.082 1.08 0.23 −0.066 0.94 0.02 0.84** 2.32 22.3Downstream 0.17 1.19 1.42 −0.49* 0.61 5.29 – – –

5 Upstream – – – 1.16** 3.20 45.0 −0.83** 0.43 27.1Downstream −0.23 0.80 1.21 0.56** 1.74 34.3 −0.54** 0.58 47.8

6 Downstream 0.39* 1.47 3.72 0.099 1.10 0.86 −0.095 0.91 0.43

Batch Direction d(DO) dt−1 Tide Temperaturerelease no. Estimate OR Wald χ2 Estimate OR Wald χ2 Estimate OR Wald χ2

1 Upstream – – – 0.69* 2.00 6.61 −1.13** 0.27 19.5Downstream 0.012 1.01 1.31 0.16 1.18 0.66 −0.94** 0.39 26.4

3 Downstream – – – 0.65** 1.93 3568 – – –

4 Upstream −0.29* 0.74 5.14 0.030 1.03 0.05 −0.54* 0.58 0.90Downstream −0.13 0.87 0.74 −0.33* 0.72 4.44 – – –

5 Upstream – – – 0.40** 1.49 83.3 −0.57** 0.57 11.65Downstream −0.039 0.96 0.42 −0.24** 0.78 8.54 – – –

6 Downstream −2.25** 0.10 18.4 −0.49** 0.61 9.87 – – –

Batch Direction d(Temp.) dx−1 d(Temp.) dt−1 Salinityrelease no. Estimate OR Wald χ2 Estimate OR Wald χ2 Estimate OR Wald χ2

1 Upstream −1.29** 0.27 15.92 – – – −0.84** 0.43 10.91Downstream – – – −0.25** 0.78 13.00 – – –

3 Downstream 2.94** 18.90 37827.00 – – – 0.91** 2.48 3116.00

4 Upstream 0.11 1.12 0.78 −0.11* 0.89 3.95 −0.20 0.81 2.25Downstream −0.28 0.75 2.65 – – – −0.18 0.83 0.80

5 Upstream – – – – – – −0.62* 0.54 5.10Downstream – – – −0.12** 0.89 130.71 – – –

6 Downstream −0.33 0.72 3.05 1.57** 4.79 19.35 −0.12 0.89 0.55

Table 2. Results of the generalized estimating equations for each batch release of fish. Each model compares upstream and downstreammovements to stay-puts. Estimates are the coefficients of the model, odds ratios (OR) for each outcome/covariate combination (e.g. ORupstream/DO) represent the change in the probability of that outcome given a 1 SD increase in the covariate, and the Wald χ2 test statistic isthe squared ratio of the estimate to the SE of the respective predictor. –: covariate was not included in the model because it did not improve

model fit as calculated with the quasi-likelihood information criterion. Significance: *p < 0.05, **p < 0.01

Brady & Targett: Movement of fish around hypoxia

Batch releases 1 and 2: hypoxia avoidance

Of the 7 fish in BR 1, 2 spot (spot nos. 137 and 144,red and green squares, respectively, in Fig. 2a) vacatedthe creek during the first night when DO was below2.8 mg O2 l−1 throughout the creek except at site PC3(~2 km from the head of tide, Figs. 1 & 2a). Site PC3 islocated at a shallower portion of the creek where meanlow water is ~1 m. The other 5 fish (2 weakfish and 3spot) vacated the creek on the second night when DO

was ≤1.4 mg O2 l−1 along the axis of the creek and theupper portion of the tributary was severely hypoxic forover 12 h. Fish that stayed in the creek during the firstnight moved downstream to PC3, which was the onlyarea in the creek with DO conditions >2.8 mg O2 l−1

(see Fig. 4). Both fish that left the upper tributaryduring the first night after release remained at themouth of the tributary until the next night, at whichpoint they attempted to re-enter the tributary (Figs. 2a& 3e). Only spot nos. 144 and 142 (green and gray

207

Fig. 2. Cynoscion regalis and Leiostomus xanthurus. Batch Release 1 and 2. (a) Dissolved oxygen (DO) plot and tracks of allfish and (b) insolation for 5 d following release on 08/26/2005 (dates are given as mm/dd/yy). Markers represent the hourlylocations along the creek axis for weakfish nos. 133, 139, and 140 (red, green, and white circles, respectively) and spot nos.137, 144, 149, 142, and 148 (red, green, white, gray, and cyan squares, respectively). The estimated locations have been randomly jiggered ±200 m (approximate receiver range). The color scale for DO reflects laboratory-derived behavioral thresh-olds of saturation-acclimated weakfish (see ‘Materials and methods’): swimming speed peaked at 2.8 mg O2 l−1 and angular

correlation between successive movements decreased markedly at

Mar Ecol Prog Ser 491: 199–219, 2013208

Fig

. 3.

Cyn

osci

on r

egal

isan

d L

eios

tom

us

xan

thu

rus.

Dis

solv

ed o

xyg

en (

DO

) p

lot,

tra

ck,

and

mov

emen

t ve

loci

ty f

or f

ish

ID

no.

[b

atch

rel

ease

no.

]: (

a) s

pot

no.

148

[1]

,(b

−d

) w

eak

fish

nos

. 147

[4]

, 130

[4]

, an

d 1

31 [

4], (

e) s

pot

no.

144

[1]

, (f−

g)

wea

kfi

sh n

os. 1

39 [

1] a

nd

133

[1]

, an

d (

h)

spot

no.

149

[1]

. Tra

cks

rep

rese

nt

hyp

oxia

avo

idan

cere

spon

ses

at 3

con

secu

tive

rec

eive

rs. T

he

esti

mat

ed l

ocat

ion

s h

ave

bee

n r

and

omly

jig

ger

ed ±

200

m (

app

roxi

mat

e re

ceiv

er r

ang

e). T

he

colo

r sc

ale

for

DO

ref

lect

s la

b-

orat

ory-

der

ived

beh

avio

ral

thre

shol

ds

of s

atu

rati

on-a

ccli

mat

ed w

eak

fish

(se

e ‘M

ater

ials

an

d m

eth

ods’

): s

wim

min

g s

pee

d p

eak

ed a

t 2.

8 m

g O

2l−

1an

d a

ng

ula

r co

rrel

atio

n b

etw

een

su

cces

sive

mov

emen

ts d

ecre

ased

mar

ked

ly a

t <

1.4

mg

O2

l−1

Brady & Targett: Movement of fish around hypoxia 209

squares, respectively, in Fig. 2a) remained in theacoustic array until the third night, which was alsocharacterized by spatially ex tensive hypoxia.

Only 1 fish (weakfish no. 140, Figs. 2a [white circles]& 5b) was released in BR 2 and that fish was detected25 times over the course of ~1 d. The behavior ofweakfish no. 140 was similar to the fish re leased dur-ing BR 1. Downstream movements were made duringthe first night when hypoxia (≤2.8 mg O2 l−1) extendedfrom the headwaters to PC3 and PC4 (2 to 3 km fromthe head of tide; Fig. 5b). The next night was charac-terized by an even more extensive hypo xic zone, andweakfish no. 140 vacated the acoustic array. As inBR 1, movements were characterized by lower DO

(4.41 mg O2 l−1) than stay-put detections (6.88 mg O2l−1). Additionally, downstream movements were madeafter the spatial gradient in DO transitioned from ex-tremely negative (higher in the upper tributary) to aneutral level and then to a positive gradient on anebbing tide.

Swimming speed was particularly high during theavoidance of conditions that developed on the nightof August 27 and early morning of August 28, wherespeed increased to 1−2 m s−1 (Fig. 3a,e−h). The mostaccurate determination of speed in this array is whena fish passes more than 2 receivers in one avoidanceresponse. Both weakfish and spot consistently movebe tween 1 and 2 m s−1 when exposed to oxygen levels

Fig. 4. Cynoscion regalis and Leiostomus xanthurus. (a) Dissolved oxygen (DO) plot and fish track for saturation-acclimatedweakfish no. 139 in batch release 1, (b) movement velocity (positive for downstream speed and negative for upstream speed)(c) rate of tide height change (positive for flooding tides and negative for ebbing tides), (d) DO at the fish’s location, and (e)spatial DO gradient associated with each detection (positive when DO is greater downstream and negative when DO isgreater upstream). The estimated locations have been randomly jiggered ±200 m (approximate receiver range). The colorscale for DO reflects laboratory-derived behavioral thresholds of saturation-acclimated weakfish (see ‘Materials and meth-ods’): swimming speed peaked at 2.8 mg O2 l−1 and angular correlation between successive movements decreased markedly

at

Mar Ecol Prog Ser 491: 199–219, 2013210

Fig. 5. Cynoscion regalis and Leiostomus xanthurus. (a–c) Dissolved oxygen (DO) plot and tracks for (fish ID no. and [batchrelease no.]) saturation-acclimated weakfish nos. (a) 319 [5], (b) 140 [2], and (c) 139 [1]. (d–f) DO plot and tracks for diel-cyclinghypoxia acclimated (d) weakfish nos. 147 [4], (e) 130 [4], and (f) 131 [4]. The estimated locations have been randomly jiggered±200 m (approximate receiver range). The color scale for DO reflects laboratory-derived behavioral thresholds of saturation-acclimated weakfish (see ‘Materials and methods’): swimming speed peaked at 2.8 mg O2 l−1 and angular correlation between

successive movements decreased markedly at

Brady & Targett: Movement of fish around hypoxia 211

below 2.8 mg O2 l−1 (see Fig. 3 for 8 instances of hy-poxia avoidance at 3 receivers). For a 137 mm fish(mean size of all released fish), 1 m s−1 converts to 7.3BL (body length) s−1.

Batch release 3

BR 3 occurred on September 1, 2005 and consistedof 2 weakfish (Table 1). Meteorological conditionsduring the following week were generally unfavor-able for development of spatially extensive hypoxia(i.e. high insolation). Daytime temperatures were stillhigh, >28°C. There were 1208 estimated locations,1186 of which were from weakfish no. 146 (Table 1).Neither individual made any upstream movementsand only 4 downstream movements were detected,thus only the downstream GAM logit was calculated(Table 2). Significant covariates in the downstreamGAM logit were the spatial temperature gradient,rate of tide height change, salinity, the spatial DOgradient, insolation, and DO (in order of Wald χ2

magnitude in Table 2). Downstream movementswere almost 19 times more likely for a 1 SD increasein the temperature gradient (i.e. indicates movementto higher temperatures downstream).

Batch release 3: hypoxia avoidance

Weakfish no. 146 generally stayed at the leadingedge of the hypoxic zone (Fig. 6a); only making 3downstream movements as the hypoxic zone(

Mar Ecol Prog Ser 491: 199–219, 2013

vided by Fig. 5 (panels a−c [saturation-acclimated]versus d−f [diel-cycling hypoxia acclimated]). Whereassaturation-acclimated and diel-cycling hy poxia accli-mated weakfish were not released concurrently,there is no definitive evidence of differential be -havioral tolerance of hypoxia. However, diel-cyclinghypoxia acclimated fish were consistently in DO conditions (

Brady & Targett: Movement of fish around hypoxia

other weakfish left Pepper Creek on the first ebb tidebut were subsequently detected in the open bay untilearly October. Only 2 weakfish of the original 8 werein Pepper Creek during the only hypoxic period ofthis release, and both individuals were detectedbefore and after this event but never within thehypoxic zone (e.g. Fig. 5a).

Batch release 6

The last batch release, BR 6, occurred on September4, 2006 and consisted of 8 saturation-acclimated weak -fish (Table 1) detected 2996 times (38 downstream andno upstream movements). Rain and cloudiness char-acterized September 5 and consequently DO at PC1

213

Fig. 7. Cynoscion regalis and Leiostomus xanthurus. Batch releases 4 and 5. (a) Dissolved oxygen (DO) plot and tracks of all fish,(b) insolation, (c) precipitation, and (d) wind speed for a 5 d following release on 08/28/2006. Markers represent the hourly loca-tions along the creek axis for diel-cycling hypoxia acclimated (circles) weakfish nos. 131 (red), 147 (green), 130 (white), 134(gray), and 136 (cyan) and saturation-acclimated (squares) weakfish nos. 316 (red), 305 (green), 320 (white), 143 (gray), 321(cyan), 315 (magenta), and 319 (blue). The color scale for DO reflects laboratory-derived behavioral thresholds of saturation-acclimated weakfish (see ‘Materials and methods’): swimming speed peaked at 2.8 mg O2 l−1 and angular correlation between

successive movements decreased markedly at

Mar Ecol Prog Ser 491: 199–219, 2013

dropped below 2.8 mg O2 l−1 for ~3 h. In general, hy-poxia was limited throughout the week, likely due tolow temperatures, regardless of low insolation.

Particularly important covariates discriminatingdown stream movements from stay-put estimated lo-cations were (in order of Wald χ2 magnitude in Table2): the rate of temperature change, DO, tide heightchange, and insolation. Downstream movements co-incided with large temperature changes, either posi-tive or negative. The mean (±SD) rate of temperaturechange was −0.17 ± 5.47°C s−1 during downstreammovements, suggesting that extreme temperaturechanges elicited movement to more stable tempera-ture regimes in the open bay (~5 km from the head oftide). In fact, a downstream movement was almost 5times more likely given a 1 SD increase in the rate oftemperature change. Downstream movements alsousually followed extreme decreases in DO as evi-denced by the fact that a 1 SD decrease in the rate ofDO change resulted in a 9.5-fold increase in the prob-ability of a downstream movement (Table 2). Finally,downstream movements were more likely to occur onebbing tides, and during the day. In contrast to BR 1,downstream movements were 8.84-fold more likely tooccur for a 1 kW m−2 increase in insolation. Thus, inthis batch release during limited hypoxia, weakfishwere more active during the day, indicating that hy-poxia may induce more nocturnal movement thanwould be expected during normoxic conditions.

Batch release 6: hypoxia avoidance

The only moderately hypoxic event (

Brady & Targett: Movement of fish around hypoxia 215

the open bay on the subsequent flood tide. It is alsopossible that this was a species-specific difference;however, there is no evidence in the literature forspecies-specific differences in hypoxia tolerance(Wannamaker & Rice 2000, McNatt & Rice 2004,Shimps et al. 2005, Stierhoff et al. 2009b) and therewere not enough coincident releases of spot andweakfish to definitively make this comparison.

Another important observation was that fish did notmake upstream and downstream movements duringevery tidal cycle and would remain at one site overseveral tidal cycles when DO was >2.8 mg O2 l−1 (seeFigs. 6 & 7). Ultimately, the smaller spot in BR 1 thatbecame retained in the hypoxic creek during theflood tide may have been exhausted after moving atleast 5 km (assuming a completely straight swimmingpath) in 8 h as DO in the water column de crea sed. Acombination of swimming at speeds above ebbingtidal velocity during the initial avoidance responseand swimming against the subsequent flooding tidemay have induced exhaustion. Exhaustion was alsoevident in similar sized weakfish in a mesocosm whenthe angular correlation between movements declinedprecipitously below 1.4 mg O2 l−1 (Brady et al. 2009).

Exhaustion following avoidance of hypoxia hasbeen observed in acoustically tagged spot in theNeuse River, NC, USA (K. Craig pers. comm.). Hyp -oxia in the Neuse River occurs during periods whenwind induced mixing ceases and the bottom of thestratified water column becomes hypoxic. Duringperiods of low wind and spatially extensive hypoxia,spot increase swimming speed (on average 5-fold)and decrease path sinuosity until they reach refugesalong shore. Following long ‘searches’ for normoxicrefugia (7 to 24 h), spot have been observed to re -main quiescent for 8 to 24 h, presumably to recover.Smaller fish in the current study may have been inthe same situation. After moving at speeds of up to1−2 m s−1 (8 to 16 BL s−1 for a 120 mm fish) duringavoidance, the smaller fish remained at the mouth ofPepper Creek and during this recovery period wereforced to remain in hypoxia.

Another possibility is that smaller fish stay at themouth of Pepper Creek because of potential preda-tion in the open bay. Many studies have demon-strated that predators forage in shoal habitats adja-cent to subtidal creeks (Szedlmayer & Able 1993,Rountree & Able 1997, Rypel et al. 2007), and Clark(2001) generally captured adult estuary-dependentfish (e.g. summer flounder, bluefish, and stripedbass) in greater abundance at the mouth of the Dela -ware Coastal Bay tributaries than farther up the trib-utaries. Szedlmayer & Able (1993) found that tagged

summer flounder (210 to 254 mm SL) released in sub-tidal creeks in New Jersey, USA preferred to remainat the mouth of those creeks. In the present study,predator avoidance may have had a higher prioritythan hypoxia avoidance for smaller fish as theyapproached the open bay. Whether hypoxia expo-sure was controlled by exhaustion or predation risk,the identification of scenarios wherein fish can beexposed to growth-limiting and even lethally low DOconcentrations is important, especially consideringthat early juvenile spot and weakfish with lowerswimming speeds than the late juvenile fish taggedin this study could be caught in spatially extensivehypoxic zones during flood tides.

Support for the exhaustion-induced hypoxic expo-sure can be found in the Brady et al. (2009) examina-tion of the behavioral responses of weakfish to diel-cycling hypoxia in the laboratory. In that work,saturation-acclimated weakfish were exposed to de -creasing DO from 7.0 to 0.4 mg O2 l−1 (at ~2.5 mg O2l−1 h−1) and a subsequent increase from 0.4 to 7.0 mgO2 l−1 at the same rate. Even at 7.0 mg O2 l−1 after theDO recovery phase, saturation-acclimated weakfishonly recovered 60% of their original swimmingspeed (Brady et al. 2009). Several other aspects of thepresent study correspond with the results of the lab-oratory investigation. Saturation-acclimated weak-fish generally avoided 2.8 mg O2 l−1 as evidenced byvery few estimated locations under these conditions,and increased swimming speed as DO decreased. Inthe laboratory, saturation-acclimated weakfish in -creased swimming speed by 46% as DO decreasedfrom 7.0 to 2.8 mg O2 l−1. The present study showsthat this active response occurs in the field as welland may function to allow the best use of hypoxichabitats.

Additional similarity between the laboratory andfield studies on juvenile weakfish can be seen inthe response of diel-cycling hypoxia acclimated fishto hypoxia. Unlike saturation-acclimated fish, diel- cycling hypoxia acclimated weakfish in the presentstudy did not avoid DO levels from 1.4 to 2.8 mg O2 l−1,and only increased swimming speed and vacated ar-eas when DO decreased below 1.4 mg O2 l−1. Like-wise, diel-cycling hypoxia acclimated weakfish in thelaboratory were also less responsive to decreasing DOthan were saturation-acclimated fish (Brady et al.2009). The only evidence of an active response in diel-cycling hypoxia acclimated weakfish in the laboratorywas an increase in swimming path straightness, a rel-atively non-strenuous response (Brady et al. 2009). In-terestingly, both the laboratory- and field-observedbehavioral responses of diel-cycling hypoxia accli-

Mar Ecol Prog Ser 491: 199–219, 2013

mated weakfish occurred only when DO was belowthe lowest acclimation DO level (2.0 mg O2 l−1). Thatis, while acclimation tended to mute the escape re-sponse, avoidance behavior was still enacted duringexposure to DO levels below the lowest levels experi-enced during the acclimation phase.

Miller (2010) tagged and released summer flounderusing the same equipment in the same system (PepperCreek) and a direct comparison of more active species(i.e. weakfish and spot) with summer flounder (N = 17;197 to 301 mm SL) illustrates some interesting spe-cies-specific behavioral dynamics. For instance, in-stead of generally avoiding DO

Brady & Targett: Movement of fish around hypoxia

be noted that the weakfish and summer floundercaptured in Stierhoff et al. (2009a) were significantlysmaller (54 to 85 mm SL for weakfish and 43 to104 mm SL for summer flounder) than the weak -fish tagged in the present study and the summerflounder tagged by Miller (2010). Smaller individuals(2.8 mg O2 l−1) as it developed through thenight as long as hypoxia was not spatially extensive,suggesting that secondary/primary productivity ben-efits outweighed risks of hypoxia exposure.

In conclusion, juvenile weakfish and spot avoidedsevere hypoxia (

Mar Ecol Prog Ser 491: 199–219, 2013

Acknowledgements. The authors thank S. Baker, D. Tuzzo -lino, K. Stierhoff, D. Thompson, and R. Tyler for their help inthe field and insights during data analysis. The authors alsothank 4 anonymous reviewers for their thoughtful insight.This research was supported by funding from NOAA,National Centers for Coastal Ocean Science, Center forSponsored Coastal Ocean Research, through the CoastalHypoxia Research Program: funding provided to the Univer-sity of Delaware, (CHRP Grant # NA05NOS4781199 toT.E.T. and D. M. Di Toro) and Virginia Institute of MarineScience (CHRP Grant No. NA05NOS4781202 to R. J. Diaz).This is CHRP Contribution No. 167.

LITERATURE CITED

Aarts G, MacKenzie M, McConnell B, Fedak M, Mat -thiopoulos J (2008) Estimating space-use and habitatpreference from wildlife telemetry data. Ecography 31: 140−160

Able KW (1999) Measures of juvenile fish habitat quality: examples from a National Estuarine Research Reserve.In: Benaka LR (ed) Fish habitat: essential fish habitat andrehabilitation, Book 22. American Fisheries Society,Bethesda, MD

Aumann CA, Eby LA, Fagan WF (2006) How transientpatches affect population dynamics: the case of hypoxiaand blue crabs. Ecol Monogr 76: 415−438

Battin J (2004) When good animals love bad habitats: eco-logical traps and the conservation of animal populations.Conserv Biol 18: 1482−1491

Beck MW, Heck KL, Able KW, Childers DL and others(2001) The identification, conservation, and manage-ment of estuarine and marine nurseries for fish andinvertebrates. Bioscience 51: 633−641

Bell GW, Eggleston DB (2005) Species-specific avoidanceresponses by blue crabs and fish to chronic and episodichypoxia. Mar Biol 146: 761−770

Brady DC, Tuzzolino DM, Targett TE (2009) Behavioralresponses of juvenile weakfish (Cynoscion regalis) todiel-cycling hypoxia: swimming speed, angular correla-tion, expected displacement and effects of hypoxia accli-mation. Can J Fish Aquat Sci 66: 415−424

Breitburg D (2002) Effects of hypoxia, and the balancebetween hypoxia and enrichment, on coastal fishes andfisheries. Estuaries 25: 767−781

Burton DT, Richardson LB, Moore CJ (1980) Effect of oxygenreduction rate and constant low dissolved oxygen con-centrations on two estuarine fish. Trans Am Fish Soc 109: 552−557

Cartamil DP, Vaudo JJ, Lowe CG, Wetherbee BM, HollandKN (2003) Diel movement patterns of the Hawaiianstingray, Dasyatis lata: implications for ecological inter-actions between sympatric elasmobranch species. MarBiol 142: 841−847

Casey JF, Doctor SB (2003) Investigation of Maryland’sAtlantic Ocean and coastal bay finfish stocks. MarylandDepartment of Natural Resources Fisheries Service,Annapolis, MD

Chao LN, Musick JA (1977) Life history, feeding habits, andfunctional morphology of juvenile sciaenid fishes in theYork River estuary, Virginia. Fish Bull 75: 657−702

Clark JH (2001) Fish habitat considerations for Delaware’stidal tributaries. Federal Aid in Fisheries Restoration Project F-56-R-5. Annual report. Delaware Division ofFish and Wildlife, Dover, DE

Craig JK, Crowder LB (2005) Hypoxia-induced habitat shiftsand energetic consequences in Atlantic croaker and

brown shrimp on the Gulf of Mexico shelf. Mar Ecol ProgSer 294: 79−94

Craig JK, Gillikin PC, Magelnicki MA, May LN (2010) Habi-tat use of cownose rays (Rhinoptera bonasus) in a highlyproductive, hypoxic continental shelf ecosystem. FishOceanogr 19: 301−317

D’Avanzo C, Kremer JN (1994) Diel oxygen dynamics andanoxic events in an eutrophic estuary of Waquoit Bay,Massachusetts. Estuaries 17: 131−139

Day JW, Hall CSS, Kemp WM, Yanez-Arancibia A (1989)Estuarine ecology. John Wiley & Sons, New York, NY

Deegan LA (2002) Lessons learned: the effects of nutrientenrichment on the support of nekton by seagrass and saltmarsh ecosystems. Estuaries 25: 727−742

Derickson WK, Price KS (1973) The fishes of the shore zoneof Rehoboth and Indian River Bays, Delaware. Trans AmFish Soc 3: 522−562

Diggle PJ, Liang KY, Seger SL (1995) Analysis of longitudi-nal data. Oxford University Press, New York, NY

Eby LA, Crowder LB (2002) Hypoxia-based habitat com-pression in the Neuse River Estuary: context-dependentshifts in behavioral avoidance thresholds. Can J FishAquat Sci 59: 952−965

Eby LA, Crowder LB, McClellan CM, Peterson CH, PowersMJ (2005) Habitat degradation from intermittent hypo -xia: impacts on demersal fishes. Mar Ecol Prog Ser 291: 249−261

Eguiluz A, Wong K (2005) Second order tidally induced flowin the inlet of a coastal lagoon. Estuar Coast Shelf Sci 64: 509−518

Fonds M, Cronie R, Vethaak AD, van der Puyl P (1992)Metabolism, food consumption and growth of plaice(Pleuronectes platessa) and flounder (Platichthys flesus)in relation to fish size and temperature. Neth J Sea Res29: 127−143

Gibson RN (2003) Go with the flow: tidal migration inmarine animals. Hydrobiologia 503: 153−161

Grecay PA, Stierhoff KL (2002) A device for simultaneouslycontrolling multiple treatment levels of dissolved oxygenin laboratory experiments. J Exp Mar Biol Ecol 280: 53−62

Grecay PA, Targett TE (1996) Effects of turbidity, light leveland prey concentration on feeding of juvenile weakfishCynoscion regalis. Mar Ecol Prog Ser 131: 11−16

Harms CA (2005) Surgery in fish research: common proce-dures and postoperative care. Lab Anim 34: 28−34

Harms CA, Lewbart GA (2000) Surgery in fish. Vet ClinNorth Am Exot Anim Pract 3: 759−774

Holland KN, Lowe CG, Wetherbee BM (1996) Movementsand dispersal patterns of blue trevally (Caranx melam -pygus) in a fisheries conservation zone. Fish Res 25: 279−292

Holland AF, Sanger DM, Gawle CP, Lerberg SB and others(2004) Linkages between tidal creek ecosystems and thelandscape and demographic attributes of their water-sheds. J Exp Mar Biol Ecol 298: 151−178

Howell P, Simpson D (1994) Abundance of marine resourcesin relation to dissolved oxygen in Long Island Sound.Estuaries 17: 394−402

Kahaner D, Moler C, Nash S (1989) Numerical methods andsoftware. Prentice Hall, Englewood Cliffs, NJ

Kemp WM, Boynton WR (1980) Influence of biological andphysical processes on dissolved oxygen dynamics in anestuarine system: implications for measurement of com-munity metabolism. Estuar Coast Mar Sci 11: 407−431

Kjerfve B, Magill KE (1989) Geographic and hydrodynamiccharacteristics of shallow coastal lagoons. Mar Geol 88: 187−199

218

http://dx.doi.org/10.1016/0025-3227(89)90097-2http://dx.doi.org/10.1016/S0302-3524(80)80065-Xhttp://dx.doi.org/10.2307/1352672http://dx.doi.org/10.1016/S0022-0981(03)00357-5http://dx.doi.org/10.1016/0165-7836(95)00442-4http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=11228930&dopt=Abstracthttp://dx.doi.org/10.1038/laban0105-28http://dx.doi.org/10.1016/j.jembe.2005.09.011http://dx.doi.org/10.3354/meps131011http://dx.doi.org/10.1016/S0022-0981(02)00379-9http://dx.doi.org/10.1023/B%3AHYDR.0000008488.33614.62http://dx.doi.org/10.1016/0077-7579(92)90014-6http://dx.doi.org/10.1016/j.ecss.2005.03.014http://dx.doi.org/10.3354/meps291249http://dx.doi.org/10.1139/f02-067http://dx.doi.org/10.1007/BF02804902http://dx.doi.org/10.2307/1352562http://dx.doi.org/10.1111/j.1365-2419.2010.00545.xhttp://dx.doi.org/10.3354/meps294079http://dx.doi.org/10.1577/1548-8659(1980)109%3C552%3AEOORRA%3E2.0.CO%3B2http://dx.doi.org/10.1007/BF02804904http://dx.doi.org/10.1139/F09-007http://dx.doi.org/10.1007/s00227-004-1483-7http://dx.doi.org/10.1641/0006-3568(2001)051[0633%3ATICAMO]2.0.CO%3B2http://dx.doi.org/10.1111/j.1523-1739.2004.00417.xhttp://dx.doi.org/10.1890/0012-9615(2006)076[0415%3AHTPAPD]2.0.CO%3B2http://dx.doi.org/10.1111/j.2007.0906-7590.05236.x

Brady & Targett: Movement of fish around hypoxia

Kneib RT (2000) Salt marsh ecoscapes and production trans-fers by estuarine nekton in the southeastern U.S. In: Weinstein MP, Kreeger DA (eds) Concepts and contro-versies in tidal marsh ecology. Kluwer Academic Pub-lishers, Dordrecht

Kokko H, Lopez-Sepulcre A (2006) From individual disper-sal to species ranges: perspectives for a changing world.Science 313: 789−791

Lankford TE Jr, Targett TE (1997) Selective predation byjuvenile weakfish: post-consumptive constraints on ener -gy maximization and growth. Ecology 78: 1049−1061

Macpherson E, Duarte CM (1991) Bathymetric trends indemersal fish size: is there a general relationship? MarEcol Prog Ser 71: 103−112

Manderson JP, Pessutti J, Hilbert JG, Juanes F (2004) Shal-low water predation risk for a juvenile flatfish (winterflounder; Pseudopleuronectes americanus, Walbaum) ina northwest Atlantic estuary. J Exp Mar Biol Ecol 304: 137−157

McNatt RA, Rice JA (2004) Hypoxia-induced growth ratereduction in two juvenile estuary-dependent fishes.J Exp Mar Biol Ecol 311: 147−156

Meng L, Powell JC, Taplin B (2001) Using winter floundergrowth rates to assess habitat quality across an anthro-pogenic gradient in Narragansett Bay, Rhode Island.Estuaries 24: 576−584

Meng L, Cicchetti G, Chintala M (2004) Nekton habitatquality at shallow water sites in two Rhode Island coastalsystems. Estuaries 27: 740−751

Miller MH (2010) Movement of summer flounder (Para -lichthys dentatus) in relation to hypoxia in an estuarinetributary. MS thesis, University of Delaware, Lewes, DE

Minello TJ, Able KW, Weinstein MP, Hays CG (2003) Saltmarshes as nurseries for nekton: testing hypotheses ondensity, growth and survival through meta-analysis. MarEcol Prog Ser 246: 39−59

Pan W (2001a) Akaike’s information criterion in generalizedestimating equations. Biometrics 57: 120−125

Pan W (2001b) Model selection in estimating equations. Bio-metrics 57: 529−534

Pan W, Connett JE (2002) Selecting the working correlationstructure in generalized estimating equations with appli-cation to the lung health study. Statist Sinica 12: 475−490

Paterson AW, Whitfield AK (2000) Do shallow-water habitatsfunction as refugia for juvenile fishes? Estuar Coast ShelfSci 51: 359−364

Pirotta E, Matthiopoulos J, MacKenzie M, Scott-Hayward L,Rendell L (2011) Modelling sperm whale habitat prefer-ence: a novel approach combining transect and followdata. Mar Ecol Prog Ser 436: 257−272

Powers SP, Peterson CH, Christian RR, Sullivan E, PowersMJ, Bishop MJ, Buzzelli CP (2005) Effects of eutrophica-tion on bottom habitat and prey resources of demersalfishes. Mar Ecol Prog Ser 302: 233−243

Railsback SF, Harvey BC (2002) Analysis of habitat-selec-tion rules using an individual-based model. Ecology 83: 1817−1830

Reckhow KH, Chapra SC (1999) Modeling excessive nutri-ent loading in the environment. Environ Pollut 100: 197−207

Ross SW (2003) The relative value of different estuarinenursery areas in North Carolina for transient juvenilemarine fishes. Fish Bull 101: 384−404

Rountree RA, Able KW (1993) Diel variation in decapodcrustacean and fish assemblages in New Jersey polyha-line marsh creeks. Estuar Coast Shelf Sci 37: 181−201

Rountree RA, Able KW (1997) Nocturnal fish use of New Jersey marsh creek and adjacent bay shoal habitats.Estuar Coast Shelf Sci 44: 703−711

Rypel AL, Layman CA, Arrington DA (2007) Water depthmodifies relative predation risk for a motile fish taxon inBahamian tidal creeks. Estuaries Coasts 30: 518−525

Sackett DK, Able KW, Grothues TM (2007) Dynamics ofsummer flounder, Paralichthys dentatus, seasonal migra-tions based on ultrasonic telemetry. Estuar Coast ShelfSci 74: 119−130

Shimps EL, Rice JA, Osborne JA (2005) Hypoxia tolerance intwo juvenile estuary-dependent fishes. J Exp Mar BiolEcol 325: 146−162

Shirey C (2005−2007) Annual summaries of fish kill investi-gations. Delaware DNREC, Division of Fish and Wildlife,Fisheries Program, Dover, DE

Skilleter GA, Peterson CH (1994) Control of foraging behav-ior of individuals within an ecosystem context: the clamMacoma balthica and interactions between competitionand siphon cropping. Oecologia 100: 268−278

Stierhoff KL, Targett TE, Miller K (2006) Ecophysiologicalresponses of juvenile summer and winter flounder tohypoxia: experimental and modeling analyses of effectson estuarine nursery quality. Mar Ecol Prog Ser 325: 255−266

Stierhoff KL, Targett TE, Power JH (2009a) Hypoxia inducedgrowth limitation of juvenile fishes in an estuarine nurs-ery: assessment of small-scale temporal dynamics usingRNA: DNA. Can J Fish Aquat Sci 66: 1033−1047

Stierhoff KL, Tyler RM, Targett TE (2009b) Hypoxia toler-ance of juvenile weakfish (Cynoscion regalis): laboratoryassessment of growth and behavioral avoidance re -sponses. J Exp Mar Biol Ecol 381: S173−S179

Szedlmayer ST, Able KW (1993) Ultrasonic telemetry of age-0summer flounder, Paralichthys dentatus, movements in asouthern New Jersey estuary. Copeia 1993: 728−736

Tuzzolino DM (2008) Examining the prey resource value ofdiel-cycling hypoxia impacted benthic habitats to juve-nile summer flounder (Paralichthys dentatus) and weak-fish (Cynoscion regalis) in an estuarine tributary. MS the-sis, University of Delaware, Lewes, DE

Tyler JA, Rose KA (1997) Effects of individual habitat selec-tion in a heterogeneous environment on fish cohort sur-vivorship: a modelling analysis. J Anim Ecol 66: 122−136

Tyler RM, Targett TE (2007) Juvenile weakfish Cynoscionregalis distribution in relation to diel-cycling dissolvedoxygen in an estuarine tributary. Mar Ecol Prog Ser 333: 257−269

Tyler RM, Brady DC, Targett TE (2009) Temporal and spatialdynamics of diel-cycling hypoxia in estuarine tributaries.Estuaries Coasts 32: 123−145

Wannamaker CM, Rice JA (2000) Effects of hypoxia onmovements and behavior of selected estuarine organ-isms from the southeastern United States. J Exp Mar BiolEcol 249: 145−163

Weinstein MP (1979) Shallow marsh habitats as primarynurseries for fishes and shellfish, Cape Fear River, NorthCarolina. Fish Bull 77: 339−357

Wildhaber ML, Lamberson PJ (2004) Importance of the habi-tat choice behavior assumed when modeling the effectsof food and temperature on fish populations. Ecol Model175: 395−409

Yamashita Y, Tominaga O, Takami H, Yamada H (2003)Comparison of growth, feeding and cortisol level inPlatichthys bicoloratus juveniles between estuarine andnearshore nursery grounds. J Fish Biol 63: 617−630

219

Editorial responsibility: Stylianos Somarakis, Heraklion, Greece

Submitted: September 13, 2012; Accepted: July 1, 2013Proofs received from author(s): September 23, 2013

http://dx.doi.org/10.1046/j.1095-8649.2003.00175.xhttp://dx.doi.org/10.1016/j.ecolmodel.2003.08.022http://dx.doi.org/10.1016/S0022-0981(00)00160-Xhttp://dx.doi.org/10.1007/s12237-008-9108-xhttp://dx.doi.org/10.3354/meps333257http://dx.doi.org/10.2307/5970http://dx.doi.org/10.1126/science.291.5502.297http://dx.doi.org/10.2307/1447234http://dx.doi.org/10.1016/j.jembe.2009.07.020http://dx.doi.org/10.1139/F09-066http://dx.doi.org/10.3354/meps325255http://dx.doi.org/10.1007/BF00316954http://dx.doi.org/10.1016/j.jembe.2005.04.026http://dx.doi.org/10.1016/j.ecss.2007.03.027http://dx.doi.org/10.1007/BF03036517http://dx.doi.org/10.1006/ecss.1996.0134http://dx.doi.org/10.1006/ecss.1993.1050http://dx.doi.org/10.1016/S0269-7491(99)00092-5http://dx.doi.org/10.3354/meps302233http://dx.doi.org/10.3354/meps09236http://dx.doi.org/10.1006/ecss.2000.0640http://dx.doi.org/10.1111/j.0006-341X.2001.00529.xhttp://dx.doi.org/10.1111/j.0006-341X.2001.00120.xhttp://dx.doi.org/10.3354/meps246039http://dx.doi.org/10.1007/BF02907657http://dx.doi.org/10.2307/1353258http://dx.doi.org/10.1016/j.jembe.2004.05.006http://dx.doi.org/10.1016/j.jembe.2003.12.004http://dx.doi.org/10.3354/meps071103http://dx.doi.org/10.1890/0012-9658(1997)078[1049%3ASPBJWP]2.0.CO%3B2http://dx.doi.org/10.1126/science.1128566

cite10: cite12: cite14: cite23: cite16: cite25: cite18: cite32: cite27: cite4: cite34: cite43: cite36: cite8: cite38: cite52: cite50: cite54: cite47: cite61: cite56: cite49: cite45: cite29: cite65: cite5: cite9: cite11: cite13: cite20: cite15: cite24: cite17: cite31: cite26: cite40: cite19: cite28: cite6: cite35: cite37: cite51: cite33: cite53: cite60: cite44: cite55: cite48: cite62: cite64: cite57: cite66: cite3: cite7: