Effects of Odor Flux and Pulse Rate on Chemosensory Tracking in Turbulent Odor Plumes by the Blue Crab, Callinectes sapidus

2004·5.2 MB·

by  Troy A Keller

Reference: Bull. 207: 44-55. (August 2004)
B/o/.
2004 Marine Biological Laboratory
Effects of Odor Flux and Pulse Rate on Chemosensory
Tracking in Turbulent Odor Plumes by the Blue
Crab, Callinectes sapidus
TROY A. KELLER1 AND MARC J. WEISSBURG:
School ofBiology. Georgia Institute of Technolo^v. 310 Ferst Dr.. Atlanta. Georgia 30332-0230
Abstract. The ability of animals to track through chem- extract information on distance and direction (Moore and
ical plumes is often related to properties ofevanescent odor Atema. 1991: Mafra-Neto and Carde. 1994: Vickers and
bursts and to small-scale mixing process that determine Baker. 1994). These plume features, variously termed odor
burst properties. However, odor plumes contain variation tilaments, pulses, or bursts, are the result of turbulent mix-
over a range of scales, and little is known about how ing processes that distribute chemicals within the plume.
variation in the properties ofthe odor signal on the scale of Accordingly, odor stimulation in plumes is evanescent: al-
one to several seconds affects foraging performance. We though turbulent mixing may result in temporal variation in
examined how flux and pulse rate interact to modulate the stimulus intensity over a variety of scales, odor filaments
search behavior ofblue crabs, Callinectes sapidus. locating typically last less than 1 s and commonly may be less than
odor sources in controlled flume flows. Experimental treat- 200-500 ms (Muriis et al.. 1992: Moore ct nl., 1994:
ments consisted of continuous plumes and plumes with Crimaldi and Koseff, 2001: Webster and Weissburg, 2001 ).
discrete odor pulses at intervals of 2.5 s and 4 s at two The time scale of odor fluctuations may not be the same
fluxes. Crabs experienced diminished search success and time scale that characterizes either the neural coding of
reduced search efficiency as flux decreased and the inter- chemical signals or the behavioral response, for several
pulse interval lengthened. There often were significant in- reasons. The physical and molecular events underlying the
teractions between flux and pulse length, and neither prop- transduction process set limits on neuronal response times.
erty completely determined search behavior. Thus, over the For example, sensory neurons in crustaceans and insects
time span of several seconds, the blue crab chemosensory cannot respond distinctly to odor pulses presented at fre-
system is not a simple flux detector. The sensitivity ofblue quencies exceeding 4 and 10 Hz. respectively (Kaissling et
crabs to inter-pulse intervals in the range ofseveral seconds til.. 1987; Gomez et al. 1994). Even when sensory systems
indicates that larger-scale mixing processes, which create may be able to encode information rapidly, averaging or
odor variation on comparable scales, may exert a significant integrating over longer periods may increase the signal-to-
impact on foraging success in nature. noise ratio and result in more accurate depiction ofstimulus
properties. Indeed, temporal averaging is necessary fornav-
Introduction igational strategies that rely on mean properties of odor
Many aquatic and terrestrial animals orient to fluid-borne plumes, since stochastic variation in signal intensity in
chemical plumes to locate predators, prey, mates, or dwell- turbulent plumes reduces the accuracy ofrapid assessments
ing sites. A common sensory strategy of animals in these of local mean concentration (Webster and Weissburg.
environments is to use tine-scale features ofodor plumes to 2001).
Crustaceans such as crabs and lobsters clearly rely, at
least in part, on brief odor pulses as they navigate through
Received 15 Januarx 20(14; accepted 17 May 2004. turbulent plumes (Moore and Atema. 1991; Weissburg and
' Current address: DepartmentofWater Resources. St. Johns RiverWater Ziminer-Faust. 1994; Mead, 2002). However, animals may
Ma"naTgoemewnthoDmistriccto.rrP.eOs.poBndoexnc1e429.shPoalualtdka. bFeL 32a1d7d8r.essed. E-mail: alter behavior in response to larger-scale flow structures if
[email protected] they integrate information over time scales (i.e.. seconds)
44
(Ill MOSlNSOin IK \( KIM, l\ I'll SI I) I'l I Ml S
corresponding to the period over which these structures 1988). was 3.1 mm s '. calculated using the Law-of-the-
cause signal variation. Unfortunately, there is little informa- Wall (Weissburg and Zimmer-Faust, 1993; Keller el nl..
tion on how blue crabs orother crustaceans respond to odor 2003). This value conforms well to expectations for turbu-
signal variation over time scales of one to several seconds. lence in open channel flows. The Roughness Reynolds
Such information is essential for a meaningful analysis of number (Re*), which measures the penetration of turbulent
plume tracking in nature, where large-scale flow structures eddies into the boundary layer, was 2.65. which suggests
(i.e.. eddies) may interact with small-scale turbulence to smooth flow (Denny. 1C)SS). These hydrodynamic condi-
produce odor fluctuations on time scales ranging from mil- tions during behavior trials are well within the range re-
liseconds to minutes (Atema, 1995; Weissburg. 2000; Web- ported forblue crab habitats in the field (i:i>., Zimmer-Faust
ster and Weissburg. 2001). Thus, this study examined the et nl., 1995). Trials look place between about 1400 and
navigation of blue crabs (Callinectes siipitins) in turbulent 1700. Light levels were lowered during trials to minimi/e
flow in response to artificially pulsed odor plumes. Pulse visual cues during navigation and to reproduce the low light
rate and flux were varied independently to examine whether levels corresponding to early morning and evening periods
crab responses to pulsed plumes varied with the general when foraging activity peaks in the field (Clarke et at..
level ofstimulus intensity. The results indicate that flux and 1999).
pulse rate interact to determine crab foraging behavior, Chemical stimuli consisted of a solution prepared by
ASW
suggesting that small-scale turbulent features are not the soaking whole, intact (previously frozen) shrimp in
only determinants of olfactory-mediated foraging ability. drawn from the flume. Shrimp were soaked in ASW for 1 h
at a concentration of either 3.5 or 7.0 gm 1 '. and the
Methods solutions were prepared immediately before behavioral tri-
als. This odorant solution was released parallel to the flow
Animals 2.5 cm above the bed from a 4.7-mm-diameter brass nozzle
Male and female blue crabs (Cullinectes su/iuliis Rath- with a fairing to minimize the flow perturbation.
bmuenn, S1u8p9p6l)iewse),refrcoolmlehcatbeidt,atussiandgjabcaeinttedtotraDpisck(sGounlfBSapyecii-n ASDWualsoiunrpcuetsainndtotthheeondoozrzalnetascocluotmimono.daTtheed uasneodoofrltewsos
Panacea, Florida(3000' N, 8422' W). Crabs were shipped separate flows allowed us to alternate between odorless and
to Atlanta, Georgia, kept in communal tanks filled with odor-containing solutions to introduce odor pulses into the
artificial seawater(ASW; Instant Ocean, 33 ppt, 20 C), and flume channel. The two flow streams converged at the
tested within 20 d of collection. Animals were maintained horizontal arm of the L-shaped nozzle (about 2 in. long),
on a cycle of 12 h light and 12 h dark, and fed freshly and each was equipped with an in-line flowmeter sothat the
thawed shrimp ad libitum every other day. We withheld velocities could be independently controlled. Each flow
food from the crabs about 12 h before testing to ensure that stream passed through athree-way solenoid valve, and both
they were not satiated and to standardize the time of last valves were controlled by a valve driver such that one valve
access to food. was on (i.e., diverting the flow into the nozzle) when the
other was off (diverting the flow into a waste reservoir). A
Theflow and stimulus environment pulse generator(AM Systems) provided the timing signal to
the valvedrivertocontrol theon-and-offtimes. Airpressure
We characterized blue crab search behavior and hydro- imparted the necessary force to drive the fluid through the
dynamics in an indoor, recirculating flume ( 12.5-m long X system.
0.75-m wide; total capacity about 3200 gal, or roughly Experiments were run using three patterns of stimulus
12.1 13 1) lined with sand to provide a natural substrate. The release for each ofthe two odorant concentrations. In addi-
working section where foraging trials took place begins tion to a continuous odor plume, we also generated pulsed
about 10 m downstream ofthe flume entrance. This facility plumes of two types. In the first condition. 2.5 s ofodorant
has been described previously, and results in reproducible, release alternated with 2.5 s ofodorless ASW. The second
realistic, and well-defined flow environments (Keller et al, condition consisted of alternating 1-s odor pulses and 4-s
2003). Average flow velocity was maintained at 4.9 cm intervals ofodorless ASW. Initial flow visualization exper-
s~' 0.08 (mean SD) with a water depth of 23.0 cm iments suggested that it was difficult to produce coherent
0.348 (mean SD). The vertical velocity profile of our puKcs using on-or-off times of less than 1 s. The flow rate
flows showed a good fit to the Law-of-the-Wall equation for the odorless ASW was set to 60 ml min ' to ensure
(Keller el al.. 2003). which specifies the expected relation- cleanly separated odor pulses. The flow rate forthe odorant
ship between flow velocity and height above the substrate in solution was systematically adjusted to equalize the volume
an equilibrium boundary layer (Weissburg, 2000). At this of odorant solution released in each of the three plume
flow speed, the boundary layershearvelocity, *, a measure types. Release rate for the continuous case was 6 ml
of stress due to turbulent velocity fluctuations (Denny. min '. which corresponded to isokinetic release, as flow
46 T. A. KELLER AND M. J. WEISSBURG
velocity 2.5 cm above the bed is approximately 0.5 cm s ' Behavioral experiments
(Keller el <;/.. 2003). The 2.5-s and 1-s odor pulses were The ability of blue crabs to locate the odor source by
released at rates of 12 and 30 ml min"', respectively. Since navigating through the chemical plumes was tested in the
experiments took place in the same water flow speed, a flume, using methods similar to those previously described
sensor ofconstant area would experience the same number (Keller et ai, 2003). Briefly, blue crabs were carefully
of stimulus molecules delivered to its surface over the 5-s moved to the flume and placed in a flow-through acrylic
cycle length for plumes created with solutions of a given plastic box (27.2-cm long. 19.5-cm wide. 16.5-cm high).
odorant concentration. Our manipulations using different Animals were acclimated in the box for 13 min prior to the
m
pulse lengths and stimulus concentrations therefore de- release of odorant solution from a source located 1.5
couple the temporal pattern of stimulation from rate of directly upstream ofthe centerofthe acrylic box. The front
odorant delivery (flux) to crab chemosensors. door was raised 2 min later. Trials lasted for a maximum of
15 min andwere terminatedifthe animal successfully found
the source and attempted to grab it, or moved either up-
Flow visualization stream of the source or downstream of the acrylic box.
Trials were also terminated ifthe animal failed to leave the
We performed a series of qualitative flow visualizations box within 5 min. Release rate and stimulus concentration
to roughly characterize the plume conditions. Our methods were varied daily, with three orfourtrials on any given day.
were designed only to ascertain that pulsing the source Consequently, no more than 1 1 ofthe stimulus solution was
results in discrete odor signals that propagate downstream, released into the Hume, which minimized accumulation of
and are similar to qualitative methods employed in other shrimp metabolites. Flume waterwas filtered through 5-jU.m
psutuldsieespr(oe.pge.r,tiBeeslarnegqeuriraensdmWeitlhliosd,s,19s9u6)c;h aasfulpllaannaalrysliasseor-f apabnsiocrubliantgemfeidltiears,,aancdtivUaVtedstcearrilbiozne,rsntitwroateo-ratnhrdeephtoismpehsatpee-r
induced fluorescence (Webster and Weissburg. 2001). that week for a period of 12-18 h.
Crab behavior was recorded on videotape using a low-
go well beyond that needed here. Initial visualizations sug- CCD camera mounted about 2 m above the
gweesrteedditshrautptceodhearsentthepyulmsoesvegdendeorawtnesdtraetatmhebuntozmzalientoariifniecde lwtiaogirhnkti-insgnegntswsiteoicvtreieodn-loifghtth-eemfiltutmien.g dAiowdaetserptoigwhetrebdacbkypaackwactocnh-
cohesion for at least 100 cm, setting an approximate (and battery was attached to the carapace of each blue crab
pwreobpahboltyogcroanspehrevdattihveep)lbuomuendaasriyt efvorolpvueldsefocrohreoruegnhcley.1T5hucsm, btieofnoarle sbtaetheavoiforaalnimeaxlpseritmhaetntsfa.ileWdetoasfsianydedthethesoumrocteivbay-
beginning at a distance of about 85 cm downstream of the placing them in a small tank (28 cm dia. containing 5 1 of
tnoazkzelne.wiPtihcatudriegsitaolfctahemeprlaummoeunastevdie0.w5edmfarboomveabthoevemiwdedrlee ffaliulmeed wtoatreers)paonnddotfofethriisngfotohdemwiathsiinng5lemsihnriwmep.reBldueesicgrnaabtsedthaast
ofthe working section, yielding a visualized areaofapprox- unresponsiveandwereomittedfrom subsequentanalysis. Sim-
imately 15 X 10 cm (1 X w). A 0.lf/r fluorescein solution ilar to observations by Weissburg and Zimmer-Faust (1993).
prepared in flume water was illuminated with a 2-cm-thick the percentage of unresponsive crabs ranged from about 1 l'7r
horizontal light sheet produced by a slide projector and to 19%, with no systematic variation across treatments. Each
focused on the plume, and photographed against a black crab was tested only once, for a total ofover 150 trials (Table
sand background. The three release conditions (continuous. 1 ) used for behavioral and kinematic analysis.
2.5-s pulse, and 1-s pulse) were replicated using the same
delivery and flow parameters as in the behavioral trials. Table 1
The digital images were converted to high-resolution
(1000 X 1500 pixel)jpeg files, then converted to grayscale Numberofsuccessfulandunsuccessfulsearches as afunction ofpulse
and inverted using Adobe Photoshop. We selected arandom length andsource conccntnition
subsection away from the plume (free of visible odor fila- Pulse length (seconds)
ments) in each image and used this to determine the back-
ground pixel intensity, which then was subtracted fromeach Flu\*
image. Images were analyzed using the public domain NIH
Image program (ver. 4.02: available at http://rsb.info.nih.
gov/nih-image/) to produce a frequency histogram of pixel
intensity values for each image. An average histogram for
each plume condition was constructed from rive images,
each taken from a separate visualization.
CHEMOSENSORY TRACKING IN PULSED PLUMES 47
The A.v coordinates ofthe centroid ofeach light-emitting interspersed with clean water. The continuous plume was
diode were determined using Motion Analysis software (30 more homogeneous than either of the t\\o pulsed cases,
Hz; ver. 3.1: Motion Analysis Corp.. Santa Rasa. CA), although it still exhibited considerable tine-scale structuie
smoothed using a moving average algorithm (window that is ageneral hallmark ofplumes created by sources with
size = 3 frames), and extracted to produce a 5-Hz time slow release rates relative to the ambient flow (Webster and
series. The tracks of successful and unsuccessful searches Weissburg, 2001). The concentration records along the
were used to calculate a variety of kinematic parameters of plume centerline reinforce the perceptions gained from ex-
foraging crabs, including walking speed, turning angle, and amining the images. The continuous plume showed some
net-to-gross displacement ratio (NGDR. a measure of path variation but had relatively long stretches with moderate
tortuosity; Weissburg and Zimmer-Faust, 1993). dye concentration. In contrast, dye concentration in the
pulsed plumes was either extremely high or undetectable.
Statistical analysis Peak-to-peak distances along the plume centerline ranged
between 8 and 15 cm in pulsed plumes, which at a relative
Data on crab foraging success in various plume condi- flow velocity of 5 cm s l, produced peak-to-peak intervals
tions were analyzed using log-likelihood methods for fre- ofapproximately 1.6-3 s. The intermittency factor, defined
quency analysis (Sokal and Rohlf. 1981). These methods as the total proportion ofthe field of view above the detec-
use maximum likelihood techniques to establish the signif- tion limit (i.e.. with a pixel intensity greaterthan 0; Chatwin
icance of treatment variables (and their interactions) for and Sullivan. 1989), was 0.557 0.23. 0.391 0.19. and
determining the frequency ofobservations in each table cell 0.139 0.24 (mean std err), for continuous, 2.5-s, and
by attempting to express the logarithm of expected cell 1-spulses, respectively, indicating more signal variability in
frequencies as a linear function ofexperimental parameters. pulsed plumes as a result of highly coherent dye patches.
To examine the influence of pulse length, flux, and their However, pulsed plumes were more variable even when the
interaction on foraging success, we used LOGLIN, a log- effects ofsignal absence (no detectable dye) were removed.
linearanalysis procedure formultiway frequency tables that Frequency histograms of dye intensity that include only
is available in the SYSTAT software package, ver. 10.2 non-zero pixel values (Fig. 2) indicated that low dye con-
(SYSTAT Software Inc., Richmond, CA). centrations occurred often in pulsed plumes. In contrast,
Analysis ofvariance (ANOVA) was used to examine the continuous plumes had a more even distribution ofintensity
influence of pulse length, flux, and their interaction in values, a lower frequency of dilute dye, and more frequent
determining kinematic parameters for animals that were cases ofintermediate concentration. Thus, continuous cases
clearly engaged in olfactory navigation (i.e., successful represented fairly constant stimulation, whereas pulsed
searchers). Calculation of kinematic variables (speed, plumes represented odor environments in which odor stim-
NGDR. turning angle, andtime spentmotionless) wasbased ulation was highly intermittent, on the scale of one to
on the entire path of a given successful or unsuccessful several seconds. Plumes with the shortest pulses were the
forager. The Bonferroni method was used to establish sig- most intermittent in space, and therefore would be expected
nificance values for multiple post hoc tests (Sokal and to show the greatest temporal variability as the plume is
Rohlf, 1981) to permit comparisons across the three pulse advected downstream.
treatmentsateach flux rate. Weperformedasimilaranalysis
ANOVA
on unsuccessful searchers, but the suggested that
Behavior
plume properties had no impact on behavior (see Results).
Finally, we used multiple analysis ofvariance (MANOVA) Success rate. Flux and release properties affected the
to compare overall kinematic behavior of unsuccessful and ability of foraging crabs to navigate through the plume to
successful foragers. In this analysis, datawerepooledacross the source (Table 1 ). Groups of crabs challenged with
pulse and flux conditions for each state of search success various plume types located the source at rates from 67% to
(see Results). 28% depending on the combination ofpulse length and flux.
In general, a larger flux rate (i.e., plumes produced with the
Results more concentrated stimulus solution) significantly increased
the success rate (log-likelihood x2 '= 21.35. P < 0.01. df =
Flow visualization 6)relativetothe lowerflux plumes. Pulsedplumes were less
Qualitative flow visualization revealed that our system easily tracked by foraging crabs at both high and low fluxes;
cgeonnetriantueoduspuclassee,d palnudmdeosctuhmaetndtieffdersemdalslubsdtiafnfteiraelnlcyesfrionmsptahe- hinoswiegvneifri,catnhte(elfofegc-tliokfelpiuhlosoedorn su=cce1s3.s8r7a,tePs w=as0.m0a5r4g.indafll=y
tial variability between plumes generated with 1-s versus 7). The decrement in foraging success when navigating
2.5-s pulses (Figs. 1,2). Plumes created with both pulse through pulsed plumes was greater at the higher flux, but
times were composed of highly intermittent odor bursts the interaction between pulse properties and flux was not
48 T. A. KELLER AND M. J. WEISSBURG
= 0.56 + 0.23
Continuous Plume 160 I
200 400 600 800 1000 1200 1400 1600
Along-streamdistance(pixels)
Pulsed Plume - 2.5 s
=0.39 + 0.19
I
_
400 600 800 1000 1200 1600
Along-streamdistance(pixels)
Pulsed Plume - 1 s 180 = 0.14 + 0.24
160 1
140
120
>>
1 100
1
"-c 80
o>
-" *
> 60
40
20
200 400 600 800 1000 1200 1400 1600
Along-streamdistance(pixels)
Figure 1. Characteristics forcontinuous and pulsed odorplumes: a side-bv-side comparison ofa represen-
tative plume image (left panel) and the distribution of pixel intensity values (right panel) along the plume
centerlme(solidline),foreachofthethreeplumeconditions. Pixelintensityvaluesarederivedfromthedigitally
processedimages, inwhichhighervaluesindicategreaterdyeconcentration.The intermittence,/, definedasthe
proportionofpixelsintheimageabovethedetectionthreshold(i.e.. withapixelintensitygreaterthan0;Chatwin
and Sullivan. 1989). is given above each graph ofpixel intensity. The figure gives the mean /( SEM) based
on five images, foreach plume type.
CHEMOSHNSORY TRACKING IN PULSED PLUMES
100
-|
c:
0o)
CD
Q_
0)
E
60 -
OZ5
50 -
40
20 25 30
Pixel intensity bin
Figure 1. Distribution of pixel intensity values in continuous and pulsed plumes: the average cumulative
frequencydistribution( 1 stderr)asafunctionofpixelintensitybinforeachplumetype.Eachbincorresponds
to eight pixel intensity values, except for the first bin where the pixel value (no detectable dye) has been
omitted. The average frequencies are based on a sample size offive images for each plume type.
significant (pulse *concentration interaction x1 = 13.25, 26.77, F2^ = 17.00. F2 ,3 == 21.66. respectively: P
P = 0.066, df =7). 0.001 for all cases). Crabs locomoted extremely quickly
Search kinematics. Tracking behavior ofcrabs was qual- undermaximally stimulatory conditions (continuous plumes
itatively similar to that described in a variety of other at high fluxes), displaying mean movement rates ofapprox-
studies on chemosensory navigation in aquatic crustaceans imately 10 cm s" ', which exceeds the mean speed of
(Moore et ai, 1991; Weissburg and Zimmer-Faust, 1994; unsuccessful foragers (Fig. 3). Crabs tracking plumes in
Mead, 2002). Animals moved towardthe source, sometimes most other conditions moved at rates of 4-6 cm s ,
directly, but at othertimes using circuitous routes involving somewhat less than those of unsuccessful foragers. More
cross-stream tacks or meanders. Animals occasionally consistent odor stimulation produced higher mean speeds
stopped foroneto several seconds enroute tothe source, but only at high fluxes, whereas mean speed was similar in
long periods of motionlessness were rare; crabs generally pulsed versus continuous plumes at low fluxes.
moved consistently upstream until they reached the nozzle Low rates of movement in particular stimulus-release
and grabbed it with their chelae. Release conditions signif- conditions were associated with increases in motionless
icantly affected crab foraging behavior (see below). In con- periods (Fig. 4). Animals in high-flux plumes rarely paused
trast, the behavior ofunsuccessful foragers was remarkably during odor tracking and even at the shortest pulse length
constant and lacked detectable differences in any kinematic stopped fortotal periods ofless than 5 s. Motionless periods
parameter (speed, NGDR, stop time, body angle) as a func- exceeded 5 s in all low-flux plumes, and crabs remained
tion ofpulse length or flux. In fact, the majority ofP values stationary for over 20 s in plumes with short pulse lengths
for these tests were greater than 0.5, and only one was less (long inter-pulse intervals). Flux, pulse length, and their
than 0.2. Thus, we combined the paths of all unsuccessful interaction all had significant effects on the duration of
foragers for presentation and analysis. motionless periods (F, 53 == 12.86. P < 0.001: F253 =
Behavior ofsuccessful searchers. ANOVA revealed that 23.12, P < 0.001: /s53 == 4.76. P < 0.05. respectheh i.
flux, pulse length, and their interaction all significantly The trend for increased motionlessness with decreasing
affected tracking speed for successful searchers (F, 53 = pulse length was nonsignificant at high Htix rates, although
50 T. A. KELLER AND M. J. WEISSBURG
18- higher than for any other group. The upper and lower
16- quartiles and 10th to 90th percentage limits span a smaller
range than in successful searchers operating in any single
14 -
plume type, and are much smallerthan the variation encom-
passed by successful searchers across the range ofdifferent
wE 10 plume types. Mean and median stop times for unsuccessful
foragers are lower than for successful foragers in nearly all
0) plume types, and display a very low range as well. These
C(QO!). Eli patterns are somewhat surprising if the behavior of unsuc-
cessful animals reflects a failed search attempt. The signif-
icant behavioral responses toplumeconditions in successful
foragers argue for similar behavioral differences among
unsuccessful foragers if they are attempting, albeit poorly,
us 1s 2.5s 1s 2.5s to locate the source. Thus, one might reasonably expect a
High Low greater degree of variation among the pooled group of
Treatment unsuccessful searchers compared to that of successful for-
agers in any given condition.
Figure 3. Walking speed statistics for all unsuccessful foragers (US) The MANOVA examining the difference in kinematics
accronendattifenodurowusisutchcpelosudsmoferuasln,tffosororalguehtriisg,ohn-sifnloufxpl7uaamnneddsl3ow.wi5-tfhglusx1h-srtirmaepnadtmeI2n".t51s-,sr(eips.eup.le.scetpsi,lveulmayen)sd. bhiegthwleyensignalilficsaunctce(/s="s_f,.u,l,, =and134u0n.s9u7c,cePss<fu0l.0s0e1a)r,cihnedriscatwiansg
The gray boxenclosesthe 25thand 75th percentiles. errorbarsabove and that movements differed according to whether animals
below the box enclose the 90th and 10th percentiles. circles represent found the source. Testing the pooled successful forager
pthoeinbtsoxouitsstihdeemoefdtihaen,90atnhdatnhde 1w0htihtepedracsenhteidlelsi.nethies tbhleacmkeasonl.idLilnineeswaibtohvien treatment against unsuccessful foragers is a conservative
the box plots join pulse treatments not significantly different from each test for overall differences between animals that locate, or
other at P < 0.05 using Bonferroni adjustedpast hoc tests. Sample sizes fail to locate, the source. In general, pooling across treat-
(numbers ofpaths) for these groups are given in Table 1. ments that display significant differences (i.e., successful
foragers) is ill-advised since it compounds error variance
post hoc tests detected significant differences among some
treatment groups. The 1-s pulse length induced significantly 60 -i
greater motionlessness in low-flux plumes relative to the
other two pulse treatments.
Path linearity varied significantly as a function of pulse
lwleeannsggttnhhoi(nFs-ti,egrn,_ai,cfti=icoann5t.(9eF7f2.fecPt=o<f3.f05l.7u0,x1)Pitsa<enldf0.(t0Fh52e),^fallu=xth5or.au9tg7eh.-ptPuhlesr<ee 0E) 30
0.01). The overall effect of these factors was uniformly QO. 20-
moderate path linearity at low fluxes, and moderate to high CO = T
path linearity in high fluxes, with paths displaying signifi-
cantly more meander at the shortest pulse length (Fig. 5). -L.
Behavior of successful versus unsuccessful searchers.
The behaviorofunsuccessful searchers occasionally resem-
bled that of successful crab foragers when individual kine- us 1s 2.5s 1hs 2.5s
matic parameters were examined. For instance, mean speed High Low
or stop times for unsuccessful searchers are in the middle of Treatment
the range displayed by successful searchers. However, it is
intriguing to note that even when mean values are similar sucFciegssufruel4.foraSgteorpstiinmeplstuamteissticwsitfohra1l-lsuannsduc2c.e5s-ssfuplulfsoersa,gearnsd(UcSo)ntainnduofuosr
between these groups, crabs failing to locate the source plumes, for high-flux and low-flux treatments (i.e.. plumes created with
displayed more uniform behavior. For most kinematic pa- odorant solutions of7 and 3.5 g shrimp 1~'. respectively). The gray box
rameters, both the median and the range (expressed as enclosesthe25thand75thpercentiles.errorbarsaboveandbelowthe box
quartiles or percentage limits) indicate that crabs failing to enclosethe90thand 10thpercentiles.circlesrepresentpointsoutsideofthe
locate the source moved more uniformly than did successful 90thand 10thpercentiles.theblacksolidlinewithintheboxisthemedian,
andthewhitedashedlineisthemean.Linesabovetheboxplotsjoinpulse
searchers. Thisistypifiedby dataforpath linearity andbody treatments not significantly different from each other at P < 0.05 using
angle (Figs. 5. 6). and to a lesser extent by mean stop time Bonferroni adjusted post hoc tests. Sample sizes (numbers of paths) for
(Fig. 4). Median values for both NGDR and body angle are these groups are given in Table 1.
CHEMOSENSORY TRACKING IN PULSED PLUMES 51
1.2-t of particular filament features is still being debated (e.g.,
Webster el ai. 2001; Webster and Weissburg, 2001).
1.0 - The results presented here indicate that larger-scale odor
fluctuations ofone to several seconds alter the behavior ot
animals tracking chemical cues. Animals in plumes with
short pulses (long inter-pulse intervals) showed degraded
cQOc tracking performance. Similarly, field and laboratory studies
ofpheromone tracking in insects have indicated that larger-
scale variation in plume featuresdecreases the initiation and
efficiency ofsearch behavior(David ct ai. 1983; Baker and
Haynes, 1987; Zanen ct ul., 1994). In aquatic creatures as
well as in insects, it appears that understanding olfactory-
mediated navigation will require us to investigate physical
US 1s 2.5s 1s 2.5s processes that create long-term fluctuations in odorconcen-
High Low tration, in addition to examining fine-scale mixing.
Treatment Behaviorally relevant signalproperties ofpulsedplumes
Foraging crabs in pulsed plumes found the source less
unsFuicgcuersesfu5.l foNreatg-etros-g(rUoSs)sadnidspfloarcseumcecnetssrfautliofo(raNgGeDrsR)insptlautimsetiscswiftohr la-lsl frequently, reduced their walking speed, spent less time
and2.5-spulses,forhigh-fluxandlow-fluxtreatments(i.e., plumescreated walking, and tacked across-stream more extensively than
with odorant solutions of7 and 3.5 g shrimp P'. respectively). The gray did animals in continuously released plumes. Interestingly,
boxenclosesthe25thand75thpercentiles,errorbarsaboveandbelowthe decreased search success was not simply an extension of
boxenclose the 90th and 10th percentiles, circlesrepresentpointsoutside inefficient search. Unsuccessful searchers acted in a rather
omfedtihaen,90tahndanthde1w0htihtepedrcaesnhteidlesl,intehiesbtlhaeckmesaonl.idLliinneeswiatbhoivnetthheebbooxxipsltohtes uniform manner as a group, showing behaviors (very
joinpulsetreatmentsnotsignificantlydifferentfromeachotheratP<0.05 straight paths, high body angles) characteristic ofanimals in
using Bonferroni adjustedpost hoc tests, with NGDR arcsin-transformed the absence of stimulus sources (Weissburg and Zimmer-
prior to analysis. Sample sizes (numbers of paths) for these groups are Faust, 1993, 1994; Weissburg el al., 2003). This suggests
given inTable 1. that the apparent decrease in search success in pulsed con-
with variance due to treatment effects, and reduces the
power of a test to detect a difference. Angular data require
special test methods and could not be included in the
MANOVA analysis. However, a Raleigh test (specific for U)
aninfgiuclaanrtldyatlao)wererveaavlesratghaetbsoucdcyesasnfgullesfotrahgaenrscrdaibssplfaayielidngsigt-o Qg1 60 _L
find the source (F,.,13 = 13.09, P < 0.001 ). further empha-
sizing the differences between the behavior ofunsuccessful 5 40
and successful searchers. T3
20
Discussion
Turbulent fluid motion has been shown to have a dra-
matic effect on the behavior of aquatic and terrestrial che- us 1s 2.5s 1s 2.5s
mosensory foragers (Weissburg and Zimmer-Faust, 1993; High Low
Mafra-Neto and Carde, 1994; Vickers and Baker, 1994; Treatment
Mfionoe-rsecalaendmixGriinlgls,pro1c9e9s9s)e.sItnhvaetstpirgoadtuocrse vhaarvieatifooncuinseoddoorn- forFisugcucreess6f.ulBfoordaygearnsglinepstlautmisetsicswiftuhrall-lsuannsducc2.e5s-ssfuplulfsoersa,gefrosr ihtigShi-lalnud\
filament structure as a key feature mediating olfactory- and low-flux treatments (i.e., plumes created wifh odorant solution-, ul 7
based navigation. The peak odor intensity of filaments and and 3.5 g shrimp I"1, respectively). The gra\ hu\ encloses the 25th and
bpfielleuanmmeesntutgogdeuitsrsattesidoountr,ocermiosienduatliavmtaeeri(etsthlyeopoeaf)bi,lcirateynadttuorfersterqa(ucMekonocayrnehoaadvnoedr 7pd1e5a0rttschhheenptpdeierlrlceciensene,nttiitilshleeetssh,b.elemarccerikarocrnsl.oeblsaLiridsrneleapisbrneeoasvhweeicntntahenipdntohitbenhetehlsuob\wooxupttlhsiosetisdthehjueo\iomnfeepndtucihllaesones,'eXtalrttnehhdaetatm9hne0ednlthwshMainHn.t.hdei
Atema. 1991; Weissburg and Zimmer-Faust. 1994; Atema. significantlydifferentfrom eachotherby a Raleigh lest al /' < 0.05 using
1996). The importance ofsmall-scale mixing in modulating a Bonferroni adjustment. Sample sixes(numbersul paths)forthese groups
tracking performance is quite clear, even if the importance are given in Table 1.
52 T. A. KELLER AND M. J. WEISSBURG
ditions reflects a failure to initiate search, as opposed to an the plume, suggesting that increased stopping behavior in
inability to locate the source once search has begun. Thus, pulsed plumes may improve the foraging success ofbenthic
blue crabs in pulsed plumes often elect not to search, and arthropods in intermittent-stimulus environments. At
they probably experience reduced search success when they present, we lack the data necessary to clarify the role of
do attempt to track to the source. stopping versus casting in other walking arthropods. Some
Our results suggest that the chemosensory system in blue walking insects increase the extent of cross-stream mean-
crabs is not a simple flux detector, at least over the time ders as plumes become more intermittent (Kanzaki et til.,
scale of several seconds that is an important level of tem- 1992). Such a casting strategy is based on an internally
poral variation in naturally occurring odor plumes (e.g., generated counter-turning scheme similar to the one that
Murlis, 1986; Finelli et til.. 1999). The performance ofblue produces casting of insects when they are aloft (Tobin,
crabs in different release conditions ought to be similar if 1981; Kanzaki et ai. 1992). Other walking insects (Willis
flux was the sole determinant of behavior, since plumes and Baker, 1987), as well as benthic crustaceans (Weiss-
created with the same stimulus concentration had equal flux burg, 2000), do not use a counter-turn generator, but their
rates over the 5-s cycle length. The deleterious effects of response to loss of an odor signal has not been examined.
short pulses suggest that the integration of incoming chem- Thus, the extent to which search strategies are molded by
ical signal strength requires periods substantially longer evolutionary constraints rather than by a common stimulus
than the integration time of chemosensory neurons, which environment remains unknown.
in lobsters can code stimulus intensity within several hun- One difference in the responses ofaquatic and terrestrial
dred milliseconds (Gomez and Atema, 1996). Integration arthropods is the maximum interval between odor pulses
may act to smooth out random fluctuations in signal inten- that still allows the animal to maintain smooth and rapid
sity, particularly when physiological response times are upstream progress toward the source. Blue crabs responded
shorter than the relevant scale of variation of the incoming negatively to inter-pulse intervals that exceed 2.5 s. In
stimulus. Indeed, it has been suggested that a major advan- contrast, inter-pulse intervals in the range of0.25 s to 1 s are
tage of olfactory transduction is that biochemical events at sufficient to disrupt orientation in moths flying through
the cellular level act as an integrator so that rapid binding pheromone plumes (Vickers and Baker, 1992; Mafra-Neto
and unbinding of stimulus molecules at the receptor does and Carde, 1995, 1998). Although the difficulty ofcreating
not result in spurious fluctuations in neuronal output (i.e., cleanly separated pulses at higher frequencies prevents ex-
action potentials) (Zufall el ai. 1494). amination of shorter inter-pulse intervals, it is unlikely that
Flux was not completely irrelevant to the tracking per- blue crabs respond significantly more quickly to signal
formance of the crabs, however. Lower flux did indeed offset than current observations suggest, given the relatively
produceeffects similartothose ofplume pulsing; animals in slow rate at which primary chemosensory afferents in crus-
the low-flux (low-concentration) plume walk more slowly, taceans operate. The responses of insect olfactory receptor
stop more, and find the source less often than crabs in the neurons to brief pulses remain distinct at frequencies of up
higher flux plume. Thus, flux can have a significant impact to at least 10 Hz (Rumbo and Kaissling, 1989; Vickers and
on chemosensory search when superimposed on changes in Baker, 1992). In contrast, lobster chemosensory neurons
the longer-term pattern of temporal variation in signal in- fuse signals when the frequencies exceed 4-6 Hz (Gomez
tensity. Unfortunately, few experiments have decoupled <7 ill.. 1994). The generally smaller turbulence intensity in
flux from pulse rate or inter-pulse interval, so it is currently aquatic versus terrestrial environments (Murlis, 1986;
unclear whether olfactory systems in other animals display Weissburg, 2000) results in larger mixing lengths. In turn,
similar properties. this larger mixing scale (and more moderate flow speeds)
Moths, another group of well-studied animals that em- translates into slower variation of odor signals in water-
ploy chemosensory tracking, respond to pulsed plumes sim- borne plumes, possibly resulting in longer processing times
ilarly to blue crabs. Increasing the interval between succes- in aquatic animals than in terrestrial ones. Interestingly, the
sive odor pulses decreases the initiation of search flight, inter-pulse intervals required to disrupt plume-following in
decreases flight speed, and increases the extent of cross- the oriental silk moth (Bimibyx mori) are about 2-3 s (Kan-
wind movement (i.e., cross-wind casting; Vickers and zaki ettil.. 1992). Silkmothswalktopheromone sourcesand
Baker, 1992; Mafra-Neto and Carde, 1995). Decreased are within the low-velocity region of a boundary layer
speed and increased casting may enhance the probability where odor signals are less disrupted by turbulent fluid
that a moth will reacquire the plume if it temporarily loses motion. Thus, odor signals may be more coherent to a
the signal (Mafra-Neto and Carde, 1995). Flying arthropods walking insect than to its flying brethren, resulting in a
must keep moving to stay aloft, which necessitates cross- reduced sensitivity to short periods of odor absence.
wind flight to regain the plume signal. For an animal walk- The importance oflonger-term signal variation. Investi-
ing on the sea floor, remaining in the same spot is analogous gators have focused on variation in the fine-scale structure
to casting; a motionless animal has a chance to re-contact ofodor filaments as a kev feature mediating olfactory-based
CHEMOSENSORY TRACKING IN PULSED I'Ll'MHS 53
navigation. The results presented here indicate that signal ments (Manna and Insley. 1989: Finelli et ,il.. 1999. 2000:
variation on the order ofone to several seconds also affects Keller et ai, 2001 ). For instance, the slowest frequency for
both the success rate and the kinematics ofanimals tracking odor fluctuations in an estuurinc tidal channel is about 0.8
chemical cues. Thus, understanding olfactory-mediated H/. which probably reflects plume cross-stream meander
navigation will require us to investigate physical processes created by the largest eddies in ihe flow (Finelli et ni.
that create long-term fluctuations in odor concentration, in 1999).
addition to examining fine-scale mixing. Unfortunately, Many aquatic crustaceans forage must effectively in sp;i
controlled studies on the effects of such longer-term signal tially and temporally coherent plumes (Weissburg. 2000:
svcaarlieativoanriaatrieonr,arpearctoicmuplaarrleydintoaqiunavteisctighaabtiitoantss.onAffielwamelnatb-- mbuatxisemeizKeelltheer estp<e//e.d. 2o0f01na)v.iwghaitcihonmaatythreeflceocsttstorfatdeegcireesatsheadt
oratory and field studies (in addition to the study reported ability to operate in more complex stimulus environments
here) suggest that odor fluctuations over one to many sec- (Weissburg et al., 2002). Although turbulence may be the
onds significantly affect foraging success and behavior primary mechanism producing longer-term odor fluctua-
(Lapointe and Sainte-Marie, 1992; Keller et ai. 2001).
Whelks, for instance, are less likely to be caught in baited tions detrimental to olfactory foraging, potential prey may
traps when they are confronted with dramatic changes in exploit these signal characteristics to mask themselves from
current direction (Lapointe and Sainte-Marie, 1992) that their predators. This raises the possibility that predators and
likely cause long-term gaps in the odor signal orchanges in prey coupled to one another via waterborne signals may be
engaged in an arms race similar to that evolved in other
its intensity.
Turbulence that generates second-scale odor fluctuations systems (Vermeij, 1987); prey rendered apparent to their
is common and may be created by a variety ofmechanisms consumers viti transmission of fluid-borne chemicals may
that produce eddies of an intermediate size (e.g., 10-1000 modulate the release of these signals to evade predation.
cm), which are generally smaller than the largest scales and Bivalves, which are commonly consumed by crabs and
greater than the microscale flow structures (1-10 mm) that other olfactory predators, mimic quite well the general
are primarily responsible for the properties of individual properties ofthe signal source used here. Bivalvesdischarge
odor filaments. Eddies in this intermediate size range may attractive metabolites from a siphon, and in principle, can
either cause the entire plume to oscillate cross-stream or modulate the flow to release odor pulses over time scales
else repackage the odor into larger-scale structures, both of that degrade the tracking ability of their predators. In fact,
which will induce longer-term odor variation. The largest any animal capable of actively controlling the release of
turbulence scales are restricted by the largest flow dimen- chemical cues might profit from creating pulsed plumes
sion for instance, the water depth or channel width (Rob- rather than continuous ones. Prey species, including bi-
erts. 1990). Because energy present in the largest flow valves, use a variety of mechanisms to detect predators;
features is transferred to successively smaller scales (the once a predator is detected, the prey use a suite ofstrategies
Kolmogorov cascade), large turbulent eddies present in to decrease their apparency and vulnerability to their con-
open estuaries, embayments, etc., will invariably create sumers (Cote and Jelnikar. 1999; Leonard et al.. 1999:
eddies of a size sufficient to induce odor variation on the Nakaoka, 2000). It is not known whether anti-predation
scale of one to several seconds. Alternately, shallow or strategies include the modification of plume signal proper-
narrow tidal channels may initially generate eddies in the ties, though bivalve prey exposed to predators often cease
correct size range. Finally, objects in flow shed vortices ata pumping entirely (Irlandi and Peterson, 1991; Nakaoka,
cpirrecduilcatrabclyelifnrdeeqrueonfcyd.iamTehteerfdreqinueancfyl,ow/ opfrvoedluocceidtybUy ias 2000). Clearly, additional field and laboratory experiments
given by the Strouhal number, St. which is 0.2 over a large on both predators and potential prey are requiredtoevaluate
range of Reynolds numbers (White, 1991). Thus, 0.2 = the significance of longer-term variations in odor signals.
fD/U. so a 10-cm-diameter object will shed vortices at a
frequency of 1 Hzinaflowof50cm s~'. Given typical flow
velocities intidally dominated systems (ca. 1-100cm s~'), Acknowledgments
biological ornonbiological objects (worm tubes, protruding The authors thank Elizabeth Wood, Dan Sickbert. Chen
shells, etc.) may create turbulence at scales required to
disrupt olfactory foraging, at least in blue crabs. Chen, and Jacqueline Stewart for help on the behavioral
Time-series measurements ofodor plumes show fluctua- trials. This work was partially supported by grants from
tions on a variety of scales that reflect the processes de- DARPA/ONR (contract #NOO14-03-1-0366) and the Na-
scribed above. Low-frequency fluctuations (e.g., less than 1 tional Science Foundation (NSF-IBN #0321444) to MJW.
Hz) are often a dominant component of frequency spectra Two anonymous reviewers provided comments that in-
forodor fluctuations in both aquatic and terrestrial environ- creased the clarity and breadth of this work.