Archer fish jumping prey capture

12 Pages·2017·2.74 MB·English

©2017.PublishedbyTheCompanyofBiologistsLtd|JournalofExperimentalBiology(2017)220,1411-1422doi:10.1242/jeb.145623
RESEARCHARTICLE
Archer fish jumping prey capture: kinematics and hydrodynamics
AnnaM.Shih*,LeahMendelson*andAlexandraH.Techet‡
ABSTRACT freshwater hatchetfish (Gasteropelecidae) jump upwards in a
ballistic motion, aided by their pectoral fins in exiting the water
Smallscalearcherfish,Toxotesmicrolepis,arebestknownforspitting
(Saideletal.,2004;Wiest,1995).TheTrinidadianguppy(Poecilia
jetsofwatertocaptureprey,butalsohuntbyjumpingoutofthewater
reticulata) jumps spontaneously for population dispersal in
toheightsofupto2.5bodylengths.Inthisstudy,high-speedimaging
waterfall-laden environments (Soares and Bierman, 2013). The
and particle image velocimetry were used to characterize the
guppy’s jumping behavior includes a preparatory phase of slow
kinematics and hydrodynamics of this jumping behavior. Jumping
backward swimming using the pectoral fins, followed by forward
usedasetofkinematicsdistinctfromthoseofin-waterfeedingstrikes
accelerationwithkinematicssimilartoC-startmaneuversandburst
andwassegmentedintothreephases:(1)hoveringtosightpreyat
swimming.
thesurface,(2)rapidupwardthrustproductionand(3)glidingtothe
Jumping to feed requires more precision than jumping for
prey once out of the water. The number of propulsive tail strokes
migrationorescape(e.g.Matthes,1977;Lowryetal.,2005;Pronko
positivelycorrelatedwiththeheightofthebait,asdidthepeakbody
etal.,2013).Forfeedingstrikesingeneral,Weihs(1973)surmised
velocity observed during a jump. During the gliding stage, the fish
that when a fish was sufficiently motivated by hunger, efficiency
traveled ballistically; the kinetic energy when the fish left thewater
wassecondarytoshort-termenergyuse.Ultimately,forlong-term
balancedwiththechangeinpotentialenergyfromwaterexittothe
survivalpurposes,theoverallenergeticcostoffeedingbyjumping
maximum jump height. The ballistic estimate of the mechanical
shouldnotexceedtheenergeticgainprovidedbythetargetedprey.
energy required to jump was comparable with the estimated
TheAfricantetrasBrycinusnurseandAlestesbaremozecanjump
mechanical energy requirements of spitting a jet with sufficient
upto1mintotheairtodislodgeseedsfromriceplantsandtheneat
momentum to down prey and subsequently pursuing the prey in
theseedsaftertheyhavefallenintothewater(Matthes,1977).The
water. Particle image velocimetry showed that, in addition to the
mangrove rivulus (Kryptolebias marmoratus) uses multiple
caudal fin, the wakes of the anal, pectoral and dorsal fins were of
kinematic modes, including jumping S-type ‘launches’, to travel
nontrivialstrength,especiallyattheonsetofthrustproduction.During
upbankstofeedonland(Pronkoetal.,2013).Lowryetal.(2005)
jump initiation, these fins were used to produce as much vertical
revealedsilverarawana(Osteoglossumbicirrhosum)jumpusingS-
accelerationaspossiblegiventhespatialconstraintofstartingdirectly
atthewater’ssurfacetoaim. starts similar to those executed by ambush predators (e.g. Webb,
1984;PorterandMotta,2000)afteraperiodofburstswimming.The
KEYWORDS:Preycapture,Jumping,Archerfish,Rapid arawana’s kinematics for aerial feeding were faster and included
maneuvering,Particleimagevelocimetry largeramplitudebodymotionsthanduringin-waterfeeding.
As emphasized by Lowry et al. (2005), having more than one
INTRODUCTION prey capture mode expands an organism’s ‘ecological niche’ and
Jumping by any organism requires substantial energy and precise givesita‘competitiveadvantage’inhuntingandforaging.Archer
muscularcoordination.Aquaticjumpersinparticularmustproduce fish(genusToxotes)arefoundinmangroveswamps,rivermouths
thrustinmannerscompatiblewiththetransitioninfluidmediaand andupstreambrackishandfreshwaterregions(Lüling,1963;Allen,
drop in density (and thus force-producing ability) between water 1978). Toxotes spp. eat insects or small aquatic animals that live
andair.Organismsranginginsizefromlargemarinemammalsand near the water’s surface. Archer fish employ several hunting
sharks(e.g.Brunnschweiler,2005;Davenport,1990;Hesteretal., strategies:spittingawaterjetatanaerialtargetsuchthatitfallsinto
1963; Hui, 1989) to small copepods (Gemmell et al., 2012) have thewatertobeeaten,rapidlylunginginthewaterforfallenpreyor
developedaquaticjumpingstrategiescompatiblewiththeirsizeand jumpingtocaptureaerialprey(BekoffandDoor,1976).Thiswide
y
survivalgoals(e.g.preycapture,escape,matingormigration). rangeofforagingbehaviorsmakesthearcherfishamodelpredator g
o
Among fishes specifically there is remarkable diversity in and a unique fish for further hydrodynamic and biological
l
o
jumping strategies. In some fish species, including trout and investigation. i
B
salmon, jumping is an oft-observed migratory behavior (e.g. Previous studies of archer fish hunting have used spitting,
l
Kondratieff and Myrick, 2006; Lauritzen et al., 2005, 2010), includingcontrolofjethydrodynamicsandaim,asanindicatorof a
t
executedfromdepthinplungepoolsatthebaseofwaterfalls.For the fish’s cognitive capabilities (e.g. Verwey, 1928; Rossel et al., n
e
predatorevasion, African butterfly fish (Pantodon buchholzi) and 2002;TimmermansandSouren,2004).Schlegeletal.(2006)found m
thatarcherfishfirelargermassesofwaterastheirpreygrowsinsize i
r
(and thus attachment strength to leaves and branches) to avoid e
p
ExperimentalHydrodynamicsLaboratory,DepartmentofMechanicalEngineering, expendingunnecessaryenergyonsmaller,weakerprey.Vailatietal. x
MassachusettsInstituteofTechnology,Cambridge,MA02139,USA. E
(2012)proposedthatarcherfishmodulatetheflowrateofthejetover
*Theseauthorscontributedequallytothiswork f
time to maximize force on prey. Experiments have also suggested o
‡Authorforcorrespondence([email protected]) thatjetsarefocusedforspecificpreyheights(GerullisandSchuster, al
2014) and hunting fish compensate for their distance to the prey n
A.H.T.,0000-0003-3223-7400 r
(Burnette and Ashley-Ross, 2015). This control over spitting is u
o
Received16August2016;Accepted27January2017 facilitatedbysophisticatedeyesight.Despiterefractionoflightatthe J
1411
RESEARCHARTICLE JournalofExperimentalBiology(2017)220,1411-1422doi:10.1242/jeb.145623
surfaceandgravityeffectsonthewaterjet,spittingstillhitspreywith kinematics and ballistic energetics of the jump. Particle image
impressive accuracy (Dill, 1977; Timmermans and Vossen, 2000; velocimetry(PIV)wasusedtounderstandtherolesofthefish’sfins
Schuster et al., 2004). Aiming studies have shown Toxotes can during upward propulsion. Combined, these data expand
account for viewpoint dependency (Schuster et al., 2004) and knowledge of a relatively unmentioned aspect of the archer fish’s
connectapparentsizeofthebaitwithrelativepositiontothetarget lifestyle:aerialpreycapture.
(Schuster et al., 2006). Archer fish are also capable of pattern
recognition,most dramaticallyexhibited byNewportetal.(2016), MATERIALSANDMETHODS
whosuccessfullytrainedarcherfishtospitatprintedhumanfaces. Fish
Successful prey capture following spitting requires advanced Ten smallscale archer fish [Toxotes microlepis (Günther 1860)]
rapid maneuvering capabilities (Wöhl and Schuster, 2007; wereimportedfromalocalaquariumstoreandhousedina75%full
Krupczynski and Schuster, 2013; Reinel and Schuster, 2014, 55 gallon (122×33×51cm) brackish water aquarium (25–28°C,
2016).Forrapidmaneuveringinwater,archerfishexecuteC-starts, salinity 9.5mScm–1, 9 h:15h light:dark cycle). Because the fish
alsoknownasescaperesponseswhenusedforpredatorevasion,in werenotbredincaptivity,agesandsexeswereunknown.Fishwere
which the body bends into a ‘C’ shape before straightening and fedfreeze-driedbrineshrimporbloodwormsatthewater’ssurface
accelerating (e.g. Domenici and Blake,1997). Wöhl and Schuster daily.Fishweretrainedtojumppriortoexperimentsbyreplacing
(2007) found no difference between the maximum speeds of dailysurfacefeedingwithfoodsuspendedfromastringabovethe
predictivefeedingstrikesandescaperesponsesresultingfrombeing tank.Duringtraining,foodwastemporarilyremovedifthefishspat
startled. C-starts were reported to reach peak linear speeds above instead of jumping. Training was performed at least twiceweekly
20bodylengthss–1andaccelerationsupto12timesthatofgravity foratminimum1monthpriortotesting.Fivefishjumpedforgreater
(Wöhl and Schuster, 2007). Reinel and Schuster (2014) further than 15 trials under the kinematic experimental conditions
foundthatarcherfishcancontrolspeedfollowingaC-startwithout described in the next section. All experiment protocols were
performinganyadditionaltailbeatstoaccelerate. approvedbytheMassachusettsInstituteofTechnologyCommittee
Cognitively,thepreycapturebehaviorsexhibitedbyarcherfish onAnimalCare(protocolno.0709-077-12).
are influenced by learning and competition. Rossel et al. (2002) Fish lengths and caudal fin surface areas were obtained by
found when prey were dislodged, archer fish used the initial digitally photographing each fish in a narrow aquarium with a
trajectorytopredictthedirection,speedanddistanceatwhichthe gridded background to provide scale and distortion correction.
prey would land. Fish would then turn accordingly to pursue Images were taken with a Canon EOS 20D digital SLR camera
regardlessofwhethertheyweretheshooter.Archerfishhuntingis equippedwithan18–55mmzoomlens.Cameracalibrationswere
influenced by competition, and fish kept in schools have been verified by photographing an object of known length in the same
observed to shoot morewhen competing for resources (Goldstein tank. Standard length (6.8–11.1cm range) was used as the
and Hall, 1990; Davis and Dill, 2012). Davis and Dill (2012) characteristic body length (BL) scale to normalize data. The
observed a high degree of kleptoparasitism in captive archer fish, caudalfinsurfacearea(1.7–5.3cm2)wasproportionaltothelength.
withfishstealingpreyshotdownbyanotherfish43.6%ofthetime Fishmass(7.9–28.3grange)wasobtainedbyweighingabeakerof
incaptivegroupsofthreetosevenfishandanincreasingprobability waterbeforeandafteraddingthefish.Physicaldimensionsofeach
oftheftthemoreshotswererequiredtobringdownaparticularprey. specimen are presented in Table 1, along with the number of
Additionally, Davis and Dill (2012) found that the frequency of kinematic trials for each fish, the maximum height and the prey
jumping was greater in larger groups of fish: jumping constituted capture success rate, which was determined as the percentage of
29%ofthetotalpreycapturebehaviorsobservedinagroupofseven trialsinwhichthemouthclosedonthebait.Failedjumpswithbait
fishcomparedwith17%ofbehaviorsinagroupofthreefish.Based heights below the maximum successful jump height of each fish
on field recordings, Rischawy et al. (2015) suggested that were considered for analysis, but failed jumps with bait heights
competition was an evolutionary rationale forarcher fish spitting, abovethemaximumsuccessfuljumpheightwerenot.Becausethe
visionandpredictivestarts,astheirhabitatswerealsoinhabitedby focusofthisstudywasontheheightattained,successfulandfailed
other species of surface-feeding fish capable of fast responses to capturekinematicswereanalyzedtogether.
hydrodynamicstimulisuchaspreyhittingthewater’ssurface.
While jumping may compromise stealth or be less successful Kinematicanalysis
with moving targets, it presents a self-sufficient mode of prey Kinematicimagingwasperformedinthehometank.Customplastic
y
capturewherethejumpingfishisthemostlikelyonetocapturethe dividerswereusedtoseparateonefishforstudywhilepermittingit g
o
prey.Thisstudybuildsonpreviousknowledgeofarcherfishaiming toswimfreelyandallowingwaterflowthroughtheentiretank.A l
o
and maneuvering to provide detailed characterization of the dried Gammarus shrimp (Tetra, Blacksburg, VA, USA) was i
B
jumping specializations in these fish. Study of this behavior will suspended on a wire above the water’s surface for a range of
l
a
ultimatelyenableassessmentoftheroleofjumpingasacompetitive heights from 0.25 to 2.5BL to elicit jumping maneuvers. Bait
t
foraging strategy. High-speed video was used to analyze the positionwithinthetanksufficientlypreventedhydrodynamicwall n
e
m
i
r
Table1.Standardlength,tailareaandmassofeachfish e
p
Specimenno. Length(cm) Tailarea(cm2) Mass(g) Jumps(allviews) Jumps(tailkinematicviews) Max.height[cm(BL)] Successrate(%) x
E
1 11.1 5.3 28.3 16 10 11.0(1.0) 94 f
o
2 9.8 5.1 23.6 24 17 14.7(1.5) 92
l
3 7.0 2.2 7.9 18 8 13.7(2.0) 94 a
n
4 6.8 1.7 10.2 16 8 15.5(2.3) 87 r
5 7.6 2.6 9.3 24 11 20.0(2.6) 71 u
o
Thetotalnumberofjumps,maximumobservedheightandsuccessrateofeachfisharealsoprovided. J
1412
RESEARCHARTICLE JournalofExperimentalBiology(2017)220,1411-1422doi:10.1242/jeb.145623
effects on the jump flow field. Jumps were recorded at 700– differences in stroke amplitude (in BL) and duration (in s).
900 frames s–1 (0.0011–0.0014s between frames) using an IDT Comparisons were made using specimens and stroke numbers
Motion Pro X3 high-speed camera (1280×1024 pixels, 900– with n≥5 runs with visible tail trajectories. Specific factors
1008μs exposure time). A Nikon Nikkor 50mm lens (aperture considered were the specimen, the timing of a stroke in a jump
f/4–5.6) yielded fields of view between 12×15cm and 27×34cm. sequence(e.g.firststrokeversusthirdstroke)andtheoveralljump
Thescenewas back-illuminatedbyabankof diffused fluorescent height.
lights.
Kinematic image analysis was performed using MATLAB Flowvisualization
R2014a (MathWorks, Natick, MA, USA). A total of 98 jumping Wakefeatures generated byajumping archer fish werevisualized
sequences were analyzed (16–24 per specimen). To synchronize usingahigh-speed,near-infraredimplementationof2DPIV(Raffel
timescales,t=0swasdefinedaswhenthefirstjumpingstrokewas etal.,1998).A10gallontank(51×25×31cm)filledhalfwaywith
initiatedbythefish’scaudalfin.Positionsofthesnout,caudalfin, water from the home tank was seeded with neutrally buoyant,
bait and free surface were digitized using the custom MATLAB polyamidparticles(averagediameter50μm,density1.03gcm–3).
softwarepackageDLTdv5(Hedrick,2008).Thesnoutwastracked IlluminationwasprovidedbyaLasirisMagnumdiodecontinuous-
automatically,whileall otherpointsweredigitized manually.The wave laser with a maximum output of 2Wat 810nm. Integrated
vertical position of the snout was smoothed using a quintic laser optics produced a 10 deg fan light sheet. The laser was
smoothingsplinetoensureacontinuoussecondderivative(Walker, mountedbelowthetankwiththelightsheetparalleltothefrontwall.
1998). Jump height (h ) was defined as the maximum snout Particles were imaged with an IDT Y-3 camera (900 frames s–1;
max
heightabovethefreesurface.Upwardvelocity(v)andacceleration 1260×1024 pixels) equipped with a Nikon Nikkor 50mm lens
(a) were calculated by differentiating the spline fit of the snout (12×10cmfieldofview).
position. Body midline traces were determined in MATLAB by PIV was performed at bait heights of 0.5 and 1.0BL. Bait was
digitizingapproximately10pointsfromthesnouttothecaudalfin positioned between the front tank wall and the laser sheet.
along the fish’s centerline and fitting a cubic spline between the Positioning the bait 10cm or less from the wall ensured that the
points. The midline splinewas evaluated at every pixel along the fish jumped with its ventral side (as opposed to dorsal) facing the
lengthofthefishandconvertedtophysicalunitsforvisualization. camera.AsetupdiagramisprovidedinFig.S1.PIVanalysisfocused
Snout position was tracked to determine vertical trajectories ontheinitialmotionsofthecaudal,pectoraloranalfins,asalignment
becauseithadhighcontrastwiththeimagebackgroundandcould betweenthefishandthelightsheetwasnotguaranteedatlatertimes
be easily detected forautomatic tracking in theimage data. Snout inajump.DLTdv5wasusedtomanuallydigitizethetrajectoryofthe
positionalsoservedasabetterindicatorofaimthancenterofmass in-plane fin (caudal or anal) in the PIV images to determine the
(COM)positionbecausethegoalofajumpwasforthefish’smouth kinematicscorrespondingtotheobservedwakestructures.
to reach the bait. The gravitational potential energy (E =mgΔh), PIV images were processed with LaVision DaVis 7 software
P
kinetic energy (E =0.5mv2) and total energy (E= E +E ) were (LaVision, Göttingen, Germany) using a multi-pass cross-
K P K
calculatedovertimeforeachrunbyassumingtheCOMwasafixed correlation algorithm (32×32 pixel windows; 50% overlap),
distancebelowthesnoutandfollowedasimilarverticaltrajectory. yielding approximately 45 vectors per body length. Velocity
In the energy calculations, which treated the fish as a ballistic fields were post-processed in DaVis with median filtering,
projectile, m was the overall mass of the fish, g=−9.8ms–2 was iterative interpolation to fill empty vectors and a 3×3 averaging
gravitational acceleration and Δh was the change from initial to filter.Vorticity(ω)wascomputedfromthevelocityfieldsusingthe
maximum snout height. Given the nature of the jump kinematics, MATLABfunctioncurl.Tomeasurethestrengthofcoherentwake
approximating the COM position from the snout position vortices,circulationwascalculatednumericallyas:
contributed minimal error to the energy estimates. As shown in ð
X
Manodvtihee1re,twheassnnoouttsriegmniafiicnaendtabimodeydbaetnthdeinbgaiutnftoirltahfteejrubmapitdcuaprattuioren. G¼ ~v(cid:2)^ndA¼ vi;jdA; ð1Þ
A i;j
Direct tracking of a point at the stretched-straight COM in five
validation runs (one per specimen) had a root mean squared error where^nisthenormalvectorofthePIVmeasurementplane,iand
(RMSE) of 2–4% and a maximum of 8% disagreement with the j are indices in the x and y directions, respectively, and δA=
estimate of COM position as a constant offset from the snout. (16pixels)2=0.0225cm2istheenclosedareabetweengridpoints.
y
Similarly,TytellandLauder(2008)foundsmalldisplacementofthe A threshold of 25–45% of the local maximum vorticity in each g
o
COMofabluegillsunfishduringtheextremebendingofaC-start vortexcorewasusedtodeterminetheintegrationregionoverwhich l
o
maneuver,andXiongandLauder(2014)estimatedatmaximum5% Eqn1wasapplied(EppsandTechet,2007).Thevorticitythreshold i
B
errorinCOMtrackingduringforwardswimmingbyassumingthe for each case was chosen as the lowest threshold at which
l
COMwasafixedpointonthefishbody. close-proximity wake features did not merge. This calculation a
t
The digitized caudal fin trajectory was used to count peak-to- underestimatedthecirculationbecausevorticitybelowthethreshold n
e
peak propulsive tail strokes. Strokes that ended or initiated at the was excluded from summation. For a given vortex core, the m
bdoudraytiomnisdlwineerewmereeacsounresdidfeorerdruhnaslfinstrwokheics.hStthreokkeinaemmpaltiitcudiemsaagneds circulationΓthr,calculatedwithathresholdjj~v~vminjj,wasrelatedtothe eri
provided a ventral view of the body and complete tail kinematics truecirculationΓas: max xp
(cid:2) (cid:3) E
couldbeseenwithoutocclusionfromtheanalfin.Thepeak-to-peak j~v j f
amplitude of each tail strokewas measured by plotting the lateral G ¼ 1(cid:3) min G: ð2Þ o
thr j~v j
displacementofthetailfromthebodyaxisovertime;durationswere max al
n
measured as the interval between successive peaks in the ThecorrectionfactorinEqn2assumedaGaussiandistributionof
r
displacement curve. Mann–Whitney U-tests (MATLAB function vorticity(Speddingetal.,2003)andwasusedtoestimatethetrue u
o
ranksum) with P<0.05 considered significant were used to assess strengthofeachvortexfromthethresholdedvalue,thusfacilitating J
1413
RESEARCHARTICLE JournalofExperimentalBiology(2017)220,1411-1422doi:10.1242/jeb.145623
comparison of circulations calculated with different thresholds. thejumpingbehaviorwere:(1)hovering,(2)thrustproductionand
Circulationsweresmoothedovertimeusingalocalaveragingfilter (3)gliding.
withaneighborhoodoffivetimesteps(0.0056s). Beforeeachjump,thesnoutwaspositionedatthesurfacetosight
Theorientationofpropulsivejetsinthewakewasmeasuredasthe prey(Fig.1A,t=0.0ms).Alternatingmotionsbytheleftandright
mediandirectionofthetop80%ofvelocityvectorsinthevicinityof pectoral fins and an oscillatory wave in the caudal fin provided
avelocitymaximum.Anglesweremeasuredrelativetoverticaland stabilizingforcesthatallowedthebodytoremainfixedinspace.The
wereusedtoassessjetcontributionstoupwardversuslateralforces. vertical position of the snout measured from the surface during
The in-plane wake kinetic energy was calculated at each time hoveringwas−0.01±0.06BL(mean±s.d.).Startingsnoutposition
stepas: didnotvarysignificantlybyfish(Mann–WhitneyU-test,P<0.05).
ð The body was inclined at an angle below the surface (see also
X
1 1
EK;wake ¼ A2rjVj2dA¼2r i;j jVi;jj2dA; ð3Þ Fthieg.bSo1d)ya(nFdigth.e1Aca,ut=da0l–f5in.6wmass)d.eAfltetchteedstlaartteorafltlhyetothwraursdtpornoedsuicdteioonf
stage, the pectoral and pelvic fins abducted (Fig. 1A, t=5.6–
where |V| is the in-plane velocity magnitude and ρ is the water 16.8ms) and remained abducted until exiting the water.
density(1.00gcm–3forexperimenttemperatureandsalinity).The Simultaneously, the fish initiated a series of propulsive tailbeats
control volume for kinetic energy calculations was identified with an oscillatory wave traveling along the body (Fig. 1B). The
manually to include all wakestructures surrounding thefish body snout trajectory was predominantly vertical. As more of the body
while limiting the inclusion of ambient flow structures and noise. left the water, the motion envelope of the tail decreased in
For consistent filtering between kinematic and hydrodynamic amplitude.Thethrustproductionphasewasconsideredoverwhen
energy data, the in-plane kinetic energy measurement was thefishceasedactivetailbeatsorcompletelyleftthewater(Fig.1A,
smoothedusingaquinticspline. t=56ms). The gliding phase was when the body accelerated only
due to gravity or changes in posture (Fig. 1A, t=56.0–140.0ms).
RESULTS This phase was considered over when the fish reached maximum
Kinematics height.Themouthclosedonthebaitduringtheglidingphaseofa
Jumpingkinematicsfollowedthethreestageframeworkforin-water successfuljump.
fast starts (Weihs, 1973; Domenici and Blake, 1997), with Bait height (hbait) and jump height (hmax) were strongly
preparatory, propulsive and variable stages. The three phases of correlated (Fig. 2A). The maximum jump height observed was
A B
2.5 201.6
190.4
0 ms 5.6 ms 11.2 ms 16.8 ms 22.4 ms 28.0 ms 33.6 ms 179.2
2
168.0
156.8
1.5 145.6
134.4
123.2
1
112.0
L)
B 100.8
39.2 ms 44.8 ms 50.4 ms 56 ms 84 ms 112 ms 140 ms y (
0.5 89.6
78.4
y
g
67.2 o
0 l
56.0 o
i
B
44.8
l
a
–0.5 33.6 nt
e
22.4 m
11.2 ri
–1 e
0 p
–0.4–0.2 0 0.2 t (ms) Ex
x (BL) f
o
Fig.1.RepresentativejumpingimagesequenceandcorrespondingbodymidlinetracesofToxotesmicrolepis.(A)Imagesequenceofspecimen5 l
a
jumping2.3bodylengths(BL)abovethesurface.Theaimingstageendedattimet=0ms.Framesfromt=5.6to56.0msshowedaseriesofthrust-producing n
tailbeats.Abductionofthepectoralfinswasalsoprominentfromt=5.6to11.2ms.Fort>56.0ms,thefishwasabovethewaterglidingtowardthebait.Movie1 r
u
showsvideoofthisjumpingsequence.(B)BodymidlinetracesofthejumpseeninAshowtravelingwavesalongthefishspine.Thesnoutpositionisindicatedby o
the‘o’markersandthetailpositionisdenotedbythe‘+’markers.Thebaitheightisdenotedbythedottedline. J
1414
RESEARCHARTICLE JournalofExperimentalBiology(2017)220,1411-1422doi:10.1242/jeb.145623
A B
3 0.8
hmaxBL–1=1.06hbaitBL–1+0.13 (hmax–hbait)BL–1=0.06hbaitBL–1+0.13
–1L] 2.5 r2=0.95 P<0.001 –1L] 0.6 Pr2==00..00171
B B
max 2 )ait
height [ 1.5 hh–maxb 0.4
ut h ot [(
no 1 ho
x. s ers 0.2
a v
M 0.5 Linear fit O
1:1 fit
0 0
0 1 2 3 0 1 2 3
Bait height [h BL–1] Bait height [h BL–1]
bait bait
Archer 1
C D
3 0.8
hmaxBL–1=0.46Vm2ax(gBL)–1+0.23 (hmax–hbait)BL–1=0.03Amaxg–1+0.07 Archer 2
2.5 r2=0.59 P<0.001 r2=0.16 Archer 3
–1] –1BL] 0.6 P<0.001 Archer 4
BLmax 2 h–)bait Archer 5
h x
height [ 1.5 hot [(ma 0.4
mp 1 sho Successful bait capture
Ju ver 0.2
0.5 O
0 0 Failed bait capture
0 1 2 3 4 0 2 4 6 8 10
Peak velocity [V2 (gBL)–1] Peak acceleration [A g–1]
max max
Fig.2.Relationshipsbetweenbaitandjumpheight,baitheightandovershoot,peakvelocityandjumpheight,andpeakaccelerationandovershoot.All
quantitiesarenon-dimensionalized.Openmarkerscorrespondtotrialswherethefishdoesnotcapturethebaitandclosedmarkerscorrespondtotrialswith
successfulbaitcapture.(A)Maximumsnoutheightwassignificantlycorrelatedwithbaitheightwithaslopegreaterthana1:1relationship.(B)Overshoot,defined
asjumpheightabovethebaitheight,showedweakpositivecorrelationwithbaitheightbutwasgreatestatintermediatebaitheights.(C)Maximumvelocity,non-
dimensionalizedbyhydrodynamicFroudenumber½V2 ðgBLÞ(cid:3)1(cid:4),correlatedstronglywithjumpheight.(D)Maximumacceleration,normalizedbygravity,showed
max
apositivecorrelationwithovershoot.
approximately 2.5BL. Jump heights above 2BL were realized in accelerations were correlated with greater overshoot. Peak
three specimens (specimens 3–5), while the two fish of greater acceleration varied among individual fish: two of the fish
length and mass (specimens 1 and 2) had maximum observed (specimens 3 and 4) typically achieved higher accelerations than
jump heights of 1.0BL and 1.5BL, respectively. The minimum theothersacrossalljumpheights(Fig.S2B).
jump height observed in any of the trials was approximately Fishexecutedmorepeak-to-peaktailstrokes(n )tojumphigher
TB y
0.5BL, though bait was positioned as low as 0.25BL. The (Fig. 3A). Tail strokes were not uniform; amplitudes (A ) varied g
relationship between bait height and jump height was also across specimens and stroke number (Fig. 3B–D, Fig.TBS3). For lo
o
quantified in terms of the overshoot (hmax−hbait). Overshoot was specimens2and3,thethirdstrokewassignificantlynarrowerthan Bi
weakly correlated with bait height (Fig. 2B). The median the first two tail strokes. In the case of specimen 5, there was a
l
a
overshoot across all fish and jump heights was 0.16BL. significantdecreaseinamplitudebetweenstrokes5and6andearlier
t
Variation in overshoot range between specimens was observed tail strokes (Fig. 3B–D). Stroke duration differed between n
e
(Fig. S2A): specimens 1 and 2 had lower median and maximum specimens,withspecimens1and2havingslowertailstrokesthan m
overshoot than specimens 3–5. Undershoot, with a jump height specimens 3–5, possibly because of their larger caudal fins. ri
e
less than the bait height, was observed during early specimen Durations were similar between all strokes by a single specimen
p
training but not during kinematic experiments. (Fig.S3). x
E
Maximum body velocities (V ) during each run ranged from Time profiles of position, velocity, acceleration and energies
max f
0.6to1.7ms–1withpeakaccelerations(A )from10to103ms–2. (Fig.4)highlightedthreekeytimesinthejump:t ,whenthetail o
max glide
The hydrodynamic Froude number [V2 ðgBLÞ(cid:3)1] was derived leftthewatercompletely;t ,whenthesnoutreachedthebaitheight; al
max bait n
from peak velocity to normalize for size effects and correlated andt ,whenthesnoutreachedmaximumheight.Peakvelocityand
max r
strongly with jump height (Fig. 2C). Acceleration, normalized by thusballistickineticenergyalwaysoccurredwhilethetailwasstillin u
o
gravity, was strongly correlated with overshoot (Fig. 2D): higher the water and before the fish reached its prey. Acceleration was J
1415
RESEARCHARTICLE JournalofExperimentalBiology(2017)220,1411-1422doi:10.1242/jeb.145623
Archer 2 Fig.3.Tailkinematicsoverincreasingjumpheight
3 0.4
A B andoverthecourseofasinglejump.(A)Thenumberof
1] a peak-to-peaktailstrokesexecutedbythefishincreased
–BLmax2.52 AAArrrccchhheeerrr 123 −1BL] 0.3 a Bl(iPnLe<)a0wr.0liyth0w1s)ittr.ho(kBjue–mnDpu)mhVebaiegrirhafttoio(rinnspBoefLca)imwmeiptnhlistsu2idg(enBif()i,nc3oar(nmCt)caaolinzrrededl5abt(iDoy)n.
height [ 1.5 AArrcchheerr 45 Ade [TB b Sinpseucffiimcieenntsn1uamnbder4owferurensexschlouwdeindgbfuelclatauislekinoefmanaticsfor
h u 0.2 greaterthantwotailstrokes.Boxplotsshowthemedian,
out 1 mplit upperandlowerquartilesandwhiskerstowithinthree
n interquartilerangesoftheupperandlowerquartiles.
s A
Max. 0.5 r2=0.80 P<0.001 0.1 Cleitrtcelressadbeonvoetebooxuetlsiedrsenboetyeosntdattihsitsicraallnygdei.stLinocwtegrrcoauspes
h BL–1=0.37n –0.19 withineachspecimen(P<0.05)usingMann–Whitney
0 max TB
U-tests.Datawerecomparedwithinbutnotbetween
0 2 4 6 1 2 3
specimens.Additionaltailkinematicstatisticsareshown
Number of tail strokes Stroke no. inFig.S3.
Archer 3 Archer 5
0.4 0.4
C D
a b
A−1BL]TB 0.3 c A−1BL]TB0.3 a,b a a,b a,b,c c,d
mplitude [ 0.2 mplitude [ 0.2 d
A A
0.1 0.1
1 2 3 1 2 3 4 5 6
Stroke no. Stroke no.
greatestatjumponset.Afterthefish completelyleftthewaterand compared with the changes in velocity, acceleration and energy
again after bait capture, continued tail oscillations (see Movie 1) duringthrustproduction.Despitefluctuations,themeanacceleration
causedchangesinbodyposturereflectedinthevelocity,acceleration between t and t was −11.6ms–2, close to gravitational.
glide bait
and energy traces; the magnitude of these fluctuations was small Kineticenergyreachedzeroatthemaximumheight.Asidefromthe
Thrust prod. Glide Tail exits (tglide) Reaches bait (tbait) Max. height (tmax)
Displacement (BL) Position (m) Velocity (m s–1) Acceleration (m s–2) Energy (mJ)
0.18 50
1 Gravity 25 Kinetic
2.5 2 0.16 1.6 Mean glide Potential
3 40 Total
45 0.14 1.4 20
2 30
1.2
0.12
1 20
0.1 15 y
1.5 g
0.08 0.8 10 o
l
1 0.06 0.6 0 10 Bio
0.4 l
a
0.04
t
0.5 0.2 –10 5 n
0.02 e
0 m
–20
0 0 –0.2 0 eri
0 0.1 0.2 0 0.1 0.2 0 0.1 0.2 0 0.1 0.2 0 0.1 0.2 p
t (s) Ex
f
Fig.4.Displacementsforalltrialsandposition,velocity,accelerationsandenergiesforthe2.3BLjumpfromFig.1.(A)Representativeverticaltrajectory o
fromspecimen5(thickblackline)comparedwithdisplacementtrajectories(relativetoinitialposition)forallspecimens(numericlegend)andjumpheights. l
a
(B)Verticalpositionrelativetothesurfaceincreasedduringthrustproductionandglidestages.(C)Velocityreachedamaximumshortlybeforethetailleftthe n
waterandthethrustproductionstageended.(D)Accelerationwasgreatestatjumponset.Theblackdashedlineshowsgravitationalaccelerationandthe r
u
greydottedlineshowsthemeanaccelerationduringtheglidestage.(D)Kinetic,potentialandtotalenergy.Kineticenergywasgreatestimmediatelybeforethetail o
leftthewater.Thetotalenergyincreasesduringthrustproductionandplateausduringtheglidestage. J
1416
RESEARCHARTICLE JournalofExperimentalBiology(2017)220,1411-1422doi:10.1242/jeb.145623
1.5 viewer’s left side advected out of the measurement plane by
Archer 1 t=0.038s, while the pair on the viewer’s right side remained in-
Archer 2
Archer 3 plane. The anal fin also shed a vortex pair with a predominantly
Archer 4 lateraljet(76–80degfromvertical).Thesamewakestructureswere
–1g) EAKrc=hEePr 5 observedforPIVperformedonspecimen4atthesamejumpheight
k (seeFig.S4).
J 1
(P Energy transfer between the archer fish and the water was
E
d assessed quantitatively through the circulation of individual wake
ze vorticesandthetotalin-planekineticenergyovertime(Fig.8).The
mali in-plane kinetic energy was a lower bound on energyexpenditure
or because dissipation and fluxes out of the measurement plane
n 0.5
s- reduced the cumulative strength of early features at later times.
s
Ma Analysisconsideredthewakestrengthsoftheanalandpectoralfins
r2=0.72 fora0.5BLjump(Fig.7)andthecaudalfinduring1.0BL(Fig.6)
P<0.001 and 0.5BL (Fig. S5) jumps, all by specimen 5. Kinematic
parametersforthejumpsareprovidedinTableS1.
0
The in-plane kinetic energy for the 1.0BL jump reached a
0 0.5 1 1.5
maximum at t=0.030s, which corresponded to when three tail
Mass-normalized E (J kg–1)
K stroke vortices and one dorsal fin vortex were seen in the wake
Fig.5.Energycomparisonduringtheglidestage.(A)Kinetic(E )and (Fig. 8A). The in-plane kinetic energy of the 0.5BL case was
K
potentialenergy(E )balanceduringtheglidestageforjumpheightsabove greatestatt=0.037s,whichcorrespondedtothethirdandfinaltail
P
1BL.Kineticenergywascalculatedattglide;potentialenergywasmeasuredas strokebeforepreycapture.Althoughkinematicdatashowedthatthe
thechangeinheightbetweentglideandtmax.Thedottedlinedenotes accelerationwasgreatestatjumponset,themechanicalenergyinput
conservationofkineticandpotentialenergy.Energieswerescaledbythe
duringajumpwasnotasinstantaneous.
massofthefishforcomparisonacrossspecimens.
PIV performed looking at the anal and pectoral fins had two
energypeaks:oneatt=0.020s,whenthefirstjumpingmotionsby
aforementioned posture oscillations, the total ballistic energy theanalfinandpectoralfinswerecompleted,andonethatremained
plateauedduringtheglidestage. relativelyconstantfromt=0.040stot=0.050s,afterthesecondanal
Energy balance was assessed during the glide stage for jumps fin stroke had been completed. By t=0.060s, the in-plane kinetic
greater than 1.0BL. During thrust production and for jumps in energy had dropped considerably, as the pectoral fin and anal fin
which the body was partially submerged at bait capture, vortices had advected out of the measurement plane. The
hydrodynamic energy and the kinetic energies of individual fins contribution of the pectoral fins to the total energy in this case
would need to be considered along with the ballistic energies wasalsounderestimatedbecauseonlyhalfofthepectoralfinwake
measured in this study. Maximum ballistic kinetic energies wasin-plane.Theinitialslopes(t<0.015s)ofthetwo0.5BLenergy
calculatedduringeachjumprangedfrom2.6to25.3mJ.Changes casesweresimilardespitemeasurementsbeingfocusedondifferent
ingravitationalpotentialenergyrangedfrom2.8to33.7mJ.During fins.
theglidestage,thekineticenergywhenthecaudalfindepartedthe Thetimerequiredforthecirculationofanyindividualfinvortex
wateratt balancedthechangeingravitationalpotentialenergy toreachpeakstrengthwasmuchshorterthanthetimescaleforthe
glide
fromt tot (Fig.5). kinetic energy to reach a maximum. No single kinematic stroke
glide max
contributed a majority of the kinetic energy to the wake.
Hydrodynamics Circulations for the caudal fin cases (Fig. 8C,D) had additional
PIVrevealedthewakestructuresproducedbythefishduringthrust maximaafterinitialshedding,likelybecauseof3Dmotionaffecting
production.Thecaudalfinandbodywakeduringthefirstthreetail vortex alignment with the light sheet or interaction with out-of-
strokes of a 1.0BL jump consisted of coherent vortex structures plane flow features. For the 1.0BL jump, the circulation of the
shedbythecaudalfinandjetsinitiatedbythebodywavebetween second tail stroke (Fig. 6, t=0.017–0.023s) was the largest. This
the pectoral fins and the caudal peduncle (Fig. 6). The caudal fin strokehasthelargestamplitudeandshortestdurationofanystroke
y
wake resembled the classic reverse Kármán street of forward fish inthesequence(TableS1). g
o
locomotion with a single vortex core shed per peak-to-peak tail l
o
stroke. The first two tail strokes produced a strong vortex pair DISCUSSION i
B
betweent=0.007and0.023s.Att=0.030s,multiplepositivevortex Useofoscillatorykinematicsforjumping
l
a
cores were observed: one from the caudal fin and one from the Archer fish jump utilizing traveling body wave kinematics
t
dorsalfin,whichrawimagesshowedwasalsopresentinthelight (Fig. 1B), an acceleration behavior with little kinematic similarity n
e
sheet at this time (see Movie 2). A patch of negative vorticity to their in-water C-starts. Velocities (0.6–1.7ms–1) and m
betweenthetwocaudalfinvorticesappearedtopinchofffromthe accelerations (10–103ms–2) measured in the present study were ri
e
negativevortexcoreofthesecondtailstroke. comparablewith those reported by Wöhl and Schuster (2007) for
p
The appearance of wake structures from fins other than the tail C-startmaneuversbyToxotesjaculatrix,acloselyrelatedspeciesof x
E
(Fig.6,t=0.030s)andsimultaneousmotionofmedianandpaired similar morphology and size, in aquatic prey capture (0.45–
f
fins at the onset of a jump (Fig. 1A) motivated PIV focused on 2.10ms–1, 17–118ms–2) and escape scenarios (0.39–1.68ms–1, o
locationsbeyondthecaudalfin.Thewakesofthepectoralandanal 20–90ms–2). If similar speeds and accelerations are achieved by al
n
fins for a 0.5BL jump are seen in Fig. 7. Propulsive jets of two different sets of kinematics, the question is raised of why
r
downward orientation (±15–18 deg from vertical) appeared near oscillatoryaccelerationkinematicsarepreferableforjumping,while u
o
bothpectoralfinsatt=0.010s.Thepectoralfin vortexpaironthe C-type maneuvers are used for in-water feeding strikes. Precision J
1417
RESEARCHARTICLE JournalofExperimentalBiology(2017)220,1411-1422doi:10.1242/jeb.145623
ω (s–1)
150
90
30
–30
–90
–150
57 deg
Caudal
(stroke 1)
Caudal (hover)
1 cm
0.5 ms–1 t=–0.004 s t=0.001 s t=0.007 s
Dorsal
Caudal 46 deg 37 deg Caudal
(stroke 2) (stroke 3)
–53 deg
–36 deg
–17 deg
t=0.017 s t=0.023 s t=0.030 s
Fig.6.Timeseriesofparticleimagevelocimetry(PIV)imagesfora1.0BLjumpbyspecimen5.Brightlocationsonthebodymaskshowthepositionof
thelasersheet.Thepectoral,pelvicandanalfinswerelocatedinfrontofthelightsheet.Thegreyboxshowsthelocationofthefreesurface.Wake
structuresarelabeledbythedirectionoftheirpropulsivejetsandthefinthatcreatedthem.Jetanglesarerelativetovertical.Thecaudalfinshedavortexduring
eachtailstrokeandpropulsivejetswereinitiatedalongthebodyintheregionbetweenthepectoralfinsandthecaudalpeduncle.Avortexwakestructure
fromthedorsalfinwasalsopresentinthePIVplaneatt=0.030s(seealsoMovie2).
alonedoesnotrationalizethisdifferenceinkinematics;archerfish concluded that the fish do not use their mouths to aim during
performC-startswithsufficientcontroloverspeedandorientation spitting but angle their bodies instead. Because the mouth is not
topursuefallenpreybycontrollingthebodybendingspeed(Wöhl usedtoaim,itisthereforefeasiblethataimingmechanismsarenot
andSchuster,2007;ReinelandSchuster,2014). spitting-specific and can be applied to other behaviors. Jumping
Posture and direction of motion provide some rationale for and predictive feeding strikes are both initiated at the surface,
this kinematic dissimilarity. Timmermans and Souren (2004) allowing the fish to accuratelysight prey located above thewater.
y
g
o
l
o
18 deg –15 deg Bi
l
a
t
n
e
ω (s–1) –11 deg –10 deg m
150 i
r
90 e
p
30 x
E
1 cm –30 80 deg 76 deg f
o
–90
0.5 ms–1 t=0.010 s –150 t=0.022 s t=0.038 s nal
r
u
Fig.7.PIVtimeseriesfora0.5BLjumpbyspecimen5withthelightsheetinitiallyalignedwiththeanalandpectoralfins.Propulsivejetsandvortices o
wereproducedbythepectoralandanalfinsduringjumponset.ThepelvicfinwasinfrontofthePIVplane. J
1418
RESEARCHARTICLE JournalofExperimentalBiology(2017)220,1411-1422doi:10.1242/jeb.145623
A B
2.5 Caudal Caudal Anal & pectoral 100 Right Left Caudal
1) (1.0 BL) (0.5 BL) (0.5 BL) pectoral pectoral Anal (stroke 3–4)
–
m
mJ c 2 –1s) 50
nergy ( 1.5 2n (cm 0
e o
etic 1 ulati
n c
e ki 0.5 Cir –50
k
a
W
0 –100
0 0.01 0.02 0.03 0.04 0.05 0.06 0.01 0.02 0.03 0.04 0.05 0.06
150 C Caudal Caudal Caudal 150 D Caudal Caudal Caudal
(hover) (stroke 1–2) (stroke 3) Dorsal (hover) (stroke 1–2) (stroke 3)
100 100
2–1m s) 50 2–1 s)m 50
Circulation (c –500 Circulation (c –500
–100 –100
–150 –150
–0.01 0 0.01 0.02 0.03 0.04 0.05 –0.01 0 0.01 0.02 0.03 0.04 0.05
t (s) t (s)
Fig.8.QuantitativeanalysisofPIVdataforthreejumpingcasesbyspecimen5.(A)In-planekineticenergyforthePIVrunsshowninFigs.6,7andS5.
(B)Circulationsofvorticescreatedbytheanalandpectoralfinsfora0.5BLjump(Fig.7).(C)Circulationsfromthreecaudaltailstrokesandthedorsalfinvortexfor
a1BLjump(Fig.6).(D)Circulationsfromthreecaudalfinstrokesofa0.5BLjump(Fig.S5).InB–D,markersareplottedateveryotherPIVtimestepforclarity
(timebetweenmarkers0.004s).
During a C-start, the archer fish bodyorientation transitions from specimens3–5.Themaximumovershootsforspecimens3–5were
the snout angled upward (for spitting oraiming) to level with the observed at intermediate bait heights (1–2BL). In this range, the
surface (Wöhl and Schuster, 2007), whereas during a jump the initialaccelerationwasinsufficienttoreachthebaitbutadditional
bodyremainspointedvertically.Bodyposturealsovariedbetween tailstrokesmayhaveprovidedmorethrustthanwasneeded.Atbait
aerial and in-water feeding in the silver arawana. During aerial heights greater than 2BL, specimens overshot less because they
feeding, the arawana was angled upwards, compared with level were incapable of jumping higher. The two fish with the highest
with the surface during aquatic feeding (Lowry et al., 2005). median and maximum overshoots (specimens 3 and 4) typically
However, unlike the archer fish, S-type kinematics were seen at exhibitedhigherFroude-scaledvelocitiesthantheothers(Fig.2C).
both postures. Acceleration, which was also greatest in these two fish (Fig. S2),
Despitethepresenceofsimilaraimingmechanismsforjumping wassignificantlycorrelatedwithovershoot(Fig.2D).Specimens3
and spitting, the behaviors had different overshoot patterns. and4alsoexhibitedhighinitialstrokeamplitudesandshortinitial
Timmermans and Vossen (2000) found that spitting archer fish stroke durations relative to other specimens (Fig. S3); the rapid
overshotandundershotincomparablefrequency,andthatspecific acceleration of the tail necessary to achieve these kinematics
aimingpatternsvariedsubstantiallybyindividual.Whileindividual correspondstoalargeinitialthrustforceandlikelylessprecisejump
variationinaimwasalsoobservedinthepresentstudy,undershoot control.
y
wasneverobservedduringkinematicexperiments, andspecimens Therequirementofproducingthrustevenafterpartialwaterexit g
3–5hadamuchhigherrangeofovershootsthanspecimens1and2 also influences jumping kinematics. Gazzola et al. (2012) lo
o
(Fig. S2).Thefailuremodes of spitting and jumping help explain concluded that C-starts were the optimal kinematics for traveling i
B
this discrepancy. For spitting, overshoot, undershoot and poor distancesbecausealargevolumeoffluidmassaccelerateswiththe
l
a
horizontal aim all result in failed prey capture. In contrast, during bodytocreateapowerfuladdedmassforce.Ifhalfofthebodyis
t
jumpingonlyundershootorpooraimresultsinfailure,andpreycan airborne, the added mass force of the C-start is reduced to utilize n
e
still be captured even with substantial overshoot. For jumping, a onlythesubmergedvolume.SubstantialbodybendingforaC-type m
small amount of overshoot also ensures that the mouth closes acceleration maneuver would also put the fish into an unstable ri
e
securely on the prey. In this study, the median snout overshoot posture with the COM offset from the vertical trajectory and the
p
(0.16BL), the constant term (0.13BL) in the linear relationship snoutnolongeraimeddirectlyatthebait.Oscillatorykinematicsare x
E
betweenbaitandjumpheight(Fig.2A)andthesnoutlengthofthe moreadaptableasthefishleavesthewater.Forspecimens2,3and
f
fish(0.15–0.17BL)wereallsimilarinlength.Theseresultssuggest 5,asthefishaccelerated,thetailstrokesdecreasedinamplitudewith o
thattheremaybeanadvantagetothesnouttravelingapproximately nosignificantchangeinduration(Fig.3B–D,Fig.S2G–I).Atjump al
n
itslengthbeyondthebait. onset, PIV showed jets formed along the body ahead of the tail
r
Overshootwasonlyweaklycorrelatedwithbaitheight(Fig.2B), (Fig.6).Asthebodyleftthewater,therewaslesspropulsivebenefit u
o
butbaitheighthadinfluenceontheaimingbehaviorsobservedin to a full body wave, as flow could only be generated by the J
1419
RESEARCHARTICLE JournalofExperimentalBiology(2017)220,1411-1422doi:10.1242/jeb.145623
submerged portion. Smaller tail strokes with a reduced motion thrust production stage and was not dominated by the initial
envelopecouldcontributethrustwhileminimizingwastedenergyor acceleration. Energetic wakes were also formed by the anal and
instability above water. The initial tail stroke and simultaneous pectoralfinsinadditiontothecaudalfin.
additional fin motions produced the greatest acceleration of the Size effects may play a role in motivating jumping. Larger prey
body, but the smaller accelerations of later tail strokes ultimately providemoreenergytothefishuponconsumption,butarealsomore
yielded the velocity and thus kinetic energy necessary to reach energeticallycostlytohuntusingspitting.Schlegeletal.(2006)found
higher prey heights. The correlation between acceleration and that because the adhesion strength of prey attached to leaves or
overshoot (Fig. 2D) suggests that using multiple propulsive tail branchesincreaseswithsize,alarger,higher-momentumjetandmore
strokesgivesthefishgreaterjumpcontrolandmayberesponsible shooting attempts were needed to capture larger prey by spitting.
forhighpreycapturesuccessrates(Table1).Whiletailstrokescan Jumpingenergeticcostsdependsolelyonthepreyheight,notonits
be considered a discrete unit of thrust production (Fig. 3A), size.Inthepresentstudy,twolargerarcherfish(specimens1and2)
kinematic differences between successive strokes (e.g. amplitude hadlowerpeakjumpheightthanthreesmallerspecimens(specimens
andtailspeed)mayberesponsibleforthehighestjumps,speedsor 3–5). One possible explanation is diminishing returns between the
overshoot.Thehighercirculationsinthecaudalfinwakeduringa energyrequiredtojumpandtheenergeticgainsofaparticularprey.A
1.0BL jump than a 0.5BL jump (Fig. 8) suggest this level of largersamplesizethanthatusedinthepresentstudyisneededtofully
controlmaybepossible,thoughstatisticaldataatbothjumpheights characterizesizeeffectsonjumpingbehavior.
areneededtoconfirm.
Theuniquevisioncapabilitiesofthearcherfish,whichcreatethe Finfunctionatjumponset
constraint of jumping from the surface, have facilitated the PIVshowedthatthewakesoftheinitialfinmotionsbythecaudal
development of a compatible jumping strategy. Archer fish fin, pectoral fins and anal fin were all of non-trivial strength. The
exhibit a unique set of behaviors among similarly sized fish: force produced by a fin is proportional to its size, speed and the
predominantly vertical jumping of multiple body lengths, circulation of its wake. Therefore, the comparable circulations
considerable accuracy (>70% for all specimens in this study) and (Fig. 8B–D) between the wakes of caudal, anal and pectoral fins
a stationary start at the surface. The Trinidadian guppy jumped suggest that despite the correlation between tail strokes and jump
verticallytocomparableheights(inbodylengths),butwithminimal height,thecaudalfinwasnottheonlyfincontributingtotheupward
aim. The guppy also started from a depth sufficient to execute thrust.Theenergymeasurements (Fig.8A) likewisesuggestthese
multiple C-start accelerations and transition to burst swimming finswerepartiallyresponsiblefortheenergeticcostsofforminga
before exiting the water (Soares and Bierman, 2013). The silver coherentwake.
arawanaalsoacceleratedbeforeaerialfeedingstrikesbyusingburst The functions of the anal and dorsal fins during locomotion
swimming followed by an S-start (Lowry et al., 2005). In both have been widely studied, with suggested roles including thrust
cases,fishwereabletobuildmomentumbytravelingsignificantly augmentationandstabilization(StandenandLauder,2005;Tytell,
fartherthanabodylengthinthewaterbeforeexiting.Considering 2006; Tytell and Lauder, 2008; Chadwell et al., 2012; Borazjani,
fish water exit maneuvers initiated from the surface, mangrove 2013). These fins were also found to be active during specialized
rivulusinitiatedlauncheswithS-typekinematics,buttookoffatan behaviors such as backwards swimming (Flammang and Lauder,
angle closer to 45 deg to maximize horizontal distance traveled 2016). In a rapid-acceleration context, Tytell and Lauder (2008)
ratherthanheight(Pronkoetal.,2013).Mangroverivulusalsoused found that the dorsal and anal fins contributed momentum to the
S-type kinematics during aimed prey capture pounces on a threeprimarypropulsivejetsgeneratedbythebodyandtailduring
simulated bank, but the body was level with the surface and a bluegill sunfish C-start. In the present study, the PIV
travelednofurtherthanapproximately0.5BL(Pronkoetal.,2013). measurements showed the orientation of the anal fin and caudal
finjetscoulddiffersubstantiallyduringjumping(76–80degfrom
Energeticrationaleforjumping verticalfortheanalfinversus36–53degforcaudalfinjets).This
The ballistic velocities from this study were used to compare the variation suggeststhattheanalfin maybeproducingindependent
mechanicalenergyrequirementsofspitting,jumpingandin-water wake structures. In measurement planes toward the front of the
pursuit. Using maximal values reported by Vailati et al. (2012) analfin,theanalfinwakeappearedasanisolatedvortexpairand
(velocity4.5ms–1,volume0.1cm3,height0.14m)andthedensity jet (Fig. 7), suggesting independent functionality. When the PIV
of freshwater (1gcm–3), an upper bound on the ballistic kinetic lightsheetwaspositionedtowardthebackofthefin(Fig.S5),the
y
energyofspittingwasestimatedasapproximately2mJpershot.A caudalfin’sthirdstrokewasobservedtopassthroughthewakeof g
o
rigid-bodyapproximationofthekineticenergyrequiredforin-water thefirsttwoanalfinstrokes,suggesting thattheaftportion ofthe l
o
pursuit was estimated using the velocities reported by Wöhl and anal fin also provides momentum that the caudal fin utilizes i
B
Schuster(2007)at2.5–40mJfora10gfish,similartothejumping duringalatertailstroke.Theforce-producingabilityofsecondary
l
kineticenergyestimatedinthepresentstudy(2.7–46.9mJ).These fins may be limited to key times in the jumping sequence; a
t
mechanical energy estimates suggest that in-water pursuit Borazjani (2013) found that anal and dorsal fins produced less n
e
dominates energy consumption during hunting by spitting. In than 5% of the thrust in a C-start except for one instance right m
competitive scenarios in which the fastest pursuit is required, before the start of the propulsive stage. The lateral orientation of ri
e
mechanical estimates suggest jumping may be of energetic parity the anal fin jet also suggests a stabilization role. The body and
p
with spitting, followed by in-water pursuit. However, these rigid- caudal fin produced lateral force in addition to upward thrust as x
E
bodymechanicalestimatesofprey-captureenergyrequirementsdo propulsivejetswereatanon-zeroanglerelativetovertical;lateral
f
not take into account the efficiency of each behavior, including forces produced bythe analfin could counteract caudal fin forces o
energy transferred from the fish to the water. A ballistic energy to keep the fish’s snout upright. al
n
balancewasonlyapplicableoncethefishcompletelyleftthewater PIV of the pectoral fins during their abduction at jump onset
r
(Fig. 5). The PIV kinetic energy time history (Fig. 8A) showed revealedjetsactingclosetovertically(10–18deg).Bymomentum u
o
energylostbythefishtowakeformationoccurredduringtheentire conservation, these wake structures indicate that the pectoral fins J
1420