藤壶幼虫中游动附着过程和水流场检测方案(CCD相机)

：@PLOSoNE ：@PLOSoneBarnacle Settlement in Turbulent Flow OPENACCESS Citation： Larsson Al， Granhag LM， Jonsson PR(2016) Instantaneous Flow Structures andOpportunities for Larval Settlement： Barnacle LarvaeSwim to Settle. PLoS ONE 11(7)： e0158957.doi：10.1371/journal.pone.0158957 Editor： Stuart Humphries， University of Lincoln，UNITED KINGDOM Received： November 18，2015 Accepted： June 24， 2016 Published： July 27，2016 Copyright： @ 2016 Larsson et al. This is an openaccess article distributed under the terms of theCreative Commons Attribution License， which permitsunrestricted use， distribution， and reproduction in anymedium， provided the original author and source arecredited. Data Availability Statement： All relevant data arewithin the paper and its Supporting Information files. Funding： This work was supported by FORMAS(http：//www.formas.se) through contracts 215-2012-1134 to AlL and 209-2008-1115 to PRJ. The SwedishResearch Council (http：//www.vr.se) providedadditional support to PRJ (contract 621-2011-3600).The work was performed within the Centre for MarineChemical Ecology (http：//cemace.science.gu.se) andthe Linnaeus Centre for Marine Evolutionary Biology(http：//cemeb.science.gu.se) at the University ofGothenburg. The funders had no role in study design， Instantaneous Flow Structures andOpportunities for Larval Settlement： BarnacleLarvae Swim to Settle Ann l. Larsson*， Lena M.Granhag"，Per R. Jonsson Department of Marine Sciences， University of Gothenburg， Tjarno， Stromstad， Sweden a Current address： Department of Shipping and Marine Technology， Chalmers University of Technology，Gothenburg， Sweden * Ann.Larsson@marine.gu.se Abstract Water flow affects settlement of marine larvae on several scales. At the smallest scale localflow regime may control the probability of adhesion to the substrate. Our aim was to mecha-nistically understand the transition from suspended to attached larvae in turbulent flow.Recently it was proposed that opportunities for larval settlement in turbulent boundary lay-ers depend on time windows with suitable instantaneous flow properties. In flume flow wecharacterized the proportion of suitable time windows in a series of flow velocities withfocus on the near-bed flow. The change in the proportion of potential settling windows withincreasing free-stream velocities was compared to the proportion of temporary attachmentof barnacle cypris larvae at different flow velocities. We found large instantaneous flow vari-ations in the near-bed flow where cyprid attachment took place. The probability of temporaryattachment in cyprids declined with local flow speed and this response was compatible witha settling window lasting at least 0.1 s with a maximum local flow speed of 1.9-2.4 cms .Cyprids swam against the near-bed flow (negative rheotaxis) and the swimming speed (1.8cm s) was close to the critical speed that permitted temporary attachment. We concludethat temporary attachment in barnacle cyprids requires upstream swimming to maintain afixed position relative to the substrate for at least 0.1 s. This behaviour may explain the abil-ity of barnacles to recruit to high-flow environments and give cyprids flexibility in the pre-set-tlement choice of substrates based on flow regime. Introduction In the marine environment， transfer of propagules from pelagic to benthic habitats is a crucialstep in many life cycles. The rates at which spores or larvae encounter and adhere to a substratemay strongly affect settlement patterns [1-4] and will influence population dynamics in sessileor sedentary organisms [5-7]. Coastal circulation is important for the transport of propagulesto potential settlement sites [8]， and on larger scales settlement patterns of sessile marine inver-tebrates are often linked to larval supply [9-12]. However， independent of larval supply， data collection and analysis， decision to publish， orpreparation of the manuscript. Competing Interests： The authors have declaredthat no competing interests exist. variation in local hydrodynamics between adjacent bays or sites within bays have recently beenfound to generate differences in settlement rates [3，13，14]. Small-scale hydrodynamics at thebed affect larval settlement patterns both through passive and active processes. The deliveryrate of larvae to the substrate is mainly a passive process governed by turbulent mixing [15，16]although behavioural control of larval vertical velocity may modify the probability of bedencounter [17-20]. After contact， flow-induced forces may prevent propagules from attachingto the substrate [21-22]. Active behavioural response to flow at the substrate may also affectfinal settlement depending on choice of micro-flow environments and through decisions toleave the surface returning into the water column [1，4，23，24]. Most marine propagules settle within a more or less developed turbulent boundary layer[25]. Even if the mean flow velocity in a turbulent boundary layer increases predictably withthe distance from the substrate there are large and chaotic fluctuations in time. Depending on，e.g. bed topography， waves and wind the turbulence intensity may differ greatly for a givenmean flow velocity. Larvae or spores approaching the substrate interact with the instantaneouslocal flow and an improved mechanistic understanding of the settling process requires detailedinformation about the distribution of flow velocities on temporal and spatial scales relevant tocontact， adhesion and behavioural responses [26-28]. Turbulent flow near the substrate is verydynamic， and high-velocity vortices known as sweeps will reach all the way down to a fractionof a millimetre away from the substrate with velocities in the same order as the mean free-stream velocity [29，30]. Consequently， the near-bed flow is expected to vary in time and space，potentially affecting spore and larval settlement. Crimaldi et al. [26] first suggested that larvaemaking contact with a substrate need some minimum time interval below a critical local stressto allow for attachment. The frequency of these so called settling windows in a particular flowregime will then determine the probability of successful larval attachment， a concept that hasbeen further explored by Reidenbach et al. [27] and Koehl et al. [28] for various habitats andflow conditions. Possible mechanisms defining the settling window are the flow-induced forcesat adhesion and the time required to make a secure attachment [15，31]. A larva or spore attach-ing to the bed is subjected to drag and lift forces induced by the flow. To attach and remainattached the adhesion strength of the propagule needs to resist the induced hydrodynamicforces [15]. Cypris-larvae of rocky-bottom intertidal barnacles make a strong temporary attachmentwith the antennular discs [32，33] (Fig 1). This temporary attachment is achieved by a combina-tion of cuticular villi covering the antennular disc and secretion of a viscoelastic adhesive [34].In temporary attachment the cyprid may walk with its antennules to explore the substrate forsuitable settlement sites， or the cyprid may decide to leave an undesirable substrate [35，36]. Itis， however， unclear how the ability for temporary attachment is controlled by flow in theboundary layer [33，37]. In a field experiment， cyprids of Balanus (Amphibalanus) improvisusDarwin 1854 showed reduced settlement and recruitment with increasing local flow speedalthough modelled larval contact rate increased with flow velocity [3]. The negative correlationbetween local flow speed and recruitment suggested a regulating mechanism operating imme-diately after initial contact but before temporary attachment. The potential to remain in con-tact with a substrate in flow includes both passive and active mechanisms： 1) adhesion tosurfaces， e.g. hydrophobic interactions [38]， 2) gravity may prevent propagules from becomingresuspended on horizontal surfaces， 3) mechanical interactions with a substrate using protrud-ing spines and setae [15]， 4) larvae may behave actively to make contact or maintain a positionat the substrate through directional swimming. To go from contact to temporary attachment acyprid must orient the body to a position that makes temporary attachment with the antennu-lar discs possible， and probably some time interval of low flow velocity relative the substrate isrequired. Crisp noted that barnacle cyprids swam upstream when close to the substrate July 27，2016 Fig 1. Cyprid of Balanus improvisus barnacle. The cyprid is the settling stage of the barnacle here makingcontact with the substrate before temporarily attaching with its antennular discs. Photo credit Kent Berntsson. doi：10.1371/journal.pone.0158957.g001 (negative rheotaxis) [37]. He hypothesized that this was an adaptation to reduce the relativevelocity to the substrate to allow temporary attachment in flow. Previous studies ofexpected larval settling probability as a function ofthe frequency of set-tling windows have been performed under various flow conditions [26-28]. Reidenbach et al[27] used values for larval adhesive strength to predict settlement probabilities of nudibranchlarvae on coral reefs. However， empirical data of how settling windows correspond to actuallarval attachment have never been collected. In this study we specifically studied how tempo-.rary attachment in barnacle larvae is attained and how attachment is controlled by boundarylayer flow conditions. We first characterized the turbulent flow structures in the near-bed layerat the height of settling larvae. We did this by estimating the proportion of time in lull periodsas a function of both critical bed shear stress and critical local velocity allowing temporaryattachment. We also calculated the hydrodynamic forces imposed on cyprids at the bed andcompared those to previously reported data of forces required to dislodge cyprids. The propor-tion of time for flow events acting as settlement windows and calculations of induced forceswere then compared to empirical data on attachment probability and swimming behaviour forB. improvisus cyprids released in flume flow. Finally， we tested which properties of flow and thelength of settling window that best predicted temporary attachment of cypris larvae. We alsodiscuss the possible role of active behaviour to attain attachment. Materials and Methods Turbulent flow structures in the boundary layer Near-bed flow in turbulent boundary layers was studied in a recirculating flume (7 m long， 0.5m wide； see description in Jonsson & Johansson [39]) designed to produce a fully developedturbulent boundary layer at the working section situated 5 m from the entrance. The flume wasfilled with seawater (salinity 33±1%o，18±0.5℃) to a depth of 10 cm and turbulence was trig-gered by a 5 mm bar fixed at the floor shortly after the flume entrance. The flume flow wascharacterized by mean and instantaneous flow parameters measured using acoustic doppler July 27，2016 Table 1. Characteristics of flume flow. z=30 mm z=10mm U..(ms') SD u (ms) SD u*(ms) 0.05 0.006 0.045 0.007 0.0021 0.10 0.010 0.093 0.011 0.0047 0.15 0.016 0.136 0.016 0.0064 0.20 0.020 0.178 0.020 0.0083 Stream-wise mean velocities (u) and standard deviations (SD) are given for 2 heights (Z) above the flumefloor. Measurements 30 mm above the flume floor represent approximate free stream velocities (U.). Datawere collected using ADV. Shear velocity (u*) is between 4-5% of the free stream velocity. doi：10.1371/journal.pone.0158957.t001 velocimetry (ADV)， and particle image velocimetry (PIV). ADV (20 Hz，Nortek AS) was usedto record approximate free-stream velocity at z=30 mm (z is height above bed)； the meanvelocity at this height is at least 90% of the free-stream velocity [40]. ADV measurements werefurther used to estimate shear velocity (u*) from the correlation between the horizontal andvertical fluctuating velocity components (uw) in the lower part of the turbulent logarithmiclayer (z=10 mm)， also known as the constant stress layer， according to Schlichting [41]： The flow characteristics are summarized in Table 1. Finally， we used ADV to check for pos-sible cross-stream differences in flow speed. Within the cross-section used for measurements(minimum 7 cm from the walls)， the mean and SD of flow speed differed less than 5% at allfree-stream velocities. A main objective was to characterize the instantaneous flow structures at the height of cyprislarvae interacting with the bed. This required high-frequency flow measurements very close tothe bed (z~0.5 mm) of sufficient duration to allow statistical analysis of flow variability. Char-acterization of turbulent flow structures was performed at free-stream velocities (U)of 0.05，0.1， 0.15 and 0.2 ms (±0.005 ms) over a smooth flume floor. Velocities at the innermost 5mm above the bed were measured with PIV； recordings were made in two replicate series of 2minutes for each free-stream velocity. The water was seeded with 3-um TiO2 tracing particlesand the flow was illuminated from above with a double pulsed Nd：YAGlaser (Litron， 30 mJ at532 nm). The beam was expanded by a combination of lenses into a 1 mm thick vertical lightsheet situated 7 cm from the transparent sidewall of the flume. The PIV-camera (LaVisionImager Pro X， 1600x1200 pixels) with a Nikon 50-mm Nikkor lens， an extension tube and a532 nm bandpass filter was run in the double frame mode. Recordings covering an area of20.8×15.6 mm were done at 14.77 Hz (maximum for the laser at the time) and images wereanalysed with the DaVis 7.2 software (LaVision) using cross correlation. Correlation calcula-tions were made on a 5x5 mm area at the bed using the multipass option with 50% overlappinginterrogation windows. The part closest to the boundary was analysed using the interrogationwindow size 32x32 pixels followed by 16x16 pixels and for the upper part 64x64 followed by32×32 was used when necessary for correct vector calculations. To increase the accuracy closeto the boundary and because the flow close to the boundary is forced in the horizontal direc-tion， horizontal， elliptical 4：1 Gaussian weighing function interrogation windows were applied.The window size of 16×16 pixels together with the 50% overlap resulted in a resolution of 10vectors per mm close to the bed. For calculations of instantaneous flow velocities in thestream-wise (u) and vertical (w) direction at the height of a settling cyprid (z=0.5 mm)， an July 27，2016 average of 10 vectors (corresponding to a width of 1 mm) was taken from each frame. Integraltime scales (based on autocorrelation analysis) for the local flow at z=0.5 mm were 2.1， 0.73，0.43 and 0.33 s for U.。 of 0.05， 0.1， 0.15 and 0.2 ms， respectively. Average boundary layerprofiles at each free-stream velocity were calculated from the whole 5×5 mm area and the linearpart of the boundary layer was determined by eye. For the resulting velocity series (2 min each， 14.77 Hz) representing stream-wise and verticalvelocity fluctuations 0.5 mm above the bed， we quantified the proportion of time occupied bylulls exceeding some critical time interval for increasing tolerable local stress [26]. Crimaldiet al. [26] defined lulls in terms of the time below some critical Reynolds stress which was usedas a proxy for the total hydrodynamic forces a settling larva could tolerate to attain successfulattachment. Since larvae in all free-stream velocities attached within the linear part of theboundary layer (see Results)， we applied the total instantaneous bed shear stress for settlementprediction used by Reidenbach et al. [27]： where t is the total bed shear stress (sum of viscous and turbulent stress components)， u is thedynamic viscosity of seawater， Ou/Oz is the linear gradient in the flow between u at z = 0.5 mmand the flume floor where u=0cms ，and -pu’w’represents the Reynolds stress where p is thedensity of seawater and u'and ware the instantaneous fluctuations of velocity in the stream-wise and vertical directions， respectively. Total bed shear stress or Reynolds stress may be asuitable proxy for hydrodynamic forcing when the settling mechanism is unknown. In thepresent paper， however， we also test a specific hypothesis about the settling mechanism involv-ing rheotactic swimming to reduce the relative speed with the substrate. Consequently， besidesbed shear stress， we chose to assess instantaneous local velocity as the critical flow propertylimiting temporary attachment of cypris larvae. The critical time interval is defined as the minimal time a larva needs (ter)，below a certainlocal velocity or stress， for successful attachment to the substrate (Fig 2). Each time intervalbelow this critical local velocity or stress， which is equal to or longer than ter is a settling win-dow (t，). Since the available time during a critical time interval to start and complete the set-tling event is only tw-ten the proportion of time L(Utcr) for some free-stream velocity (U.) Fig 2. Schematic drawing of the interaction between flow velocity fluctuations and the opportunitiesfor larval attachment. A"stress lull" or"settling window" is defined as the minimum time interval below somecritical flow velocity required for successful attachment. According to this hypothesis the cyprid can only usethe proportion of time in settling windows for attachment. doi：10.1371/journal.pone.0158957.g002 available for attachment during a total time interval T for a given ter is： where N is the total number of settling windows and t>tr. We examined the proportion oftime available for attachment for critical time intervals of 0.14， 0.47， 1.02 and 2.03 s (recordingswere made at 14.77 Hz). The two replicate data series for each of the four free-stream velocitiesshowed very similar results and data are given as mean values. Measurements of local velocities 0.5 mm above the bed were also used to calculate thehydrodynamic forces encountered by cypris larvae attaching to the substrate. The aim was tocompare flow-induced forces with previously recorded data of attachment strength in cyprids.For force calculations the cyprids were assumed having a fixed position relative to the substrate(attached) and drag and lift forces were calculated as： where Fp=drag force， Cp=drag coefficient， p=fluid density， Ap=frontal area， U=fluidvelocity experienced by the organism， Fz =lift force， Cz = lift coefficient and Ap= planformarea. The frontal and planform areas of the cyprid were calculated from shape and dimensionsgiven in Larsson et al. [42] and for Ap the cyprid was assumed oriented facing the flow. Theresulting Ap and Ap from those calculations were 0.05 mm and 0.1 mm ， respectively. Sincethe Reynolds number (Re) for a settling cyprid was well below 1000 (1-100 range) we used theequation for the drag coefficient of a sphere from White [43]： Following Reidenbach et al. [27] who calculated the hydrodynamic forces on settling nudi-branch larvae， a constant C， value of 0.2 was used. This value of Cwas based on reanalysis offorce data [44] on sediment grains sitting on surfaces from Chepil [45]. Studies of larval adhesion and swimming Predictions of the probability of attachment from the analysis of turbulent fluctuations in thenear-bed flow were tested by studying barnacle cyprids in flume flow. Cyprids of Balanusimprovisus were reared in the laboratory according to Berntsson et al. [46]. The larvae wereaged by storing them in 14℃ for 3-5 days before experiments were performed. The workingsection was fitted with a 50 cm wide and 40 cm long transparent Plexiglas panel covered on theunderside with a thin white， opaque plastic coating that was illuminated from below. The pur-pose of the illumination was to attract the larvae to the flume floor and to facilitate the detec-tion of attached larvae. Cyprids were added to the bed immediately upstream of the workingsection c. 15 cm from the wall through a glass pipette with an exit diameter of 1 mm. The larvaewere added at the flume floor with the pipette pointing horizontally and downstream at 45°with the mainstream direction， which resulted in minimum flow interactions around thepipette exit. Between 20 and 40 larvae were added to the floor during about half a minute. Thetip of the pipette was filmed to count the number of added cyprids. The number of larvaefound on the 40 cm long illuminated flume floor (working section) 1 min after each release wasrecorded. The studies were performed at U.。of 0.05， 0.1， 0.15 and 0.2 ms (±0.005 ms). July 27，2016 Two different batches of cyprids were used and for each batch four experiments at each velocitywere performed in random order. In total 210-250 larvae were added per flow speed tested. AtU.。=0.05 ms the local velocity at the bed was so low that the larvae were only slowly driftingin the near-bed flow and could reside in the working section for 1 min without attaching withthe antennules. The proportion of larvae found at the bed (93%) thus represented all but thosedeliberately leaving the working section. Since our hypothesis only includes larvae thatattached with their antennules， the larval data at U.。=0.05 ms’were excluded from furtheranalysis. At U.of0.1 msor more， unattached larvae were transported with the near-bedflow. Most larvae reacted by swimming against the flow trying to attach and those not succeed-ing were swept off the working section with no chance of returning. Proportions of attachedcyprids were tested in a 2-factor ANOVA with free-stream velocity as a fixed factor and batchas a random factor. In the analysis a type 1 error (a) rate of 0.05 was used. Data were tested forhomogeneity of variances using Cochran’s C test and no transformation was needed. A post-hoc test was performed using Student-Newman-Keuls (SNK) procedure. The swimming speed of cyprids was estimated by releasing larvae in the water column inthe flume 5 cm above the illuminated flume floor section. The flume was operating at a lowspeed since this seemed to trigger swimming towards the attractive illuminated floor betterthan still water. When larvae started to swim towards the bottom， we used a stopwatch to mea-sure the time it took to reach it. Only larvae that started swimming very soon after release andswam persistently without pausing were included. Our impression was that larvae were verymotivated and swam vigorously to reach the bottom. The swimming speed was measured for16 larvae from 2 different larval batches. The sinking speed of Balanus improvisus cyprids islow (0.0013 ms， unpublished) and is only marginally affecting swimming speed measure-ments performed in the vertical mode. The importance of the observed rheotactic swimming response for temporary attachmentwas quantified in a separate study. Cyprids were added to the bed using the technique with theglass pipette described above. In a free-stream velocity of 0.15 ms， we observed the larvalbehaviour immediately preceding successful temporary attachment. A short distance down-stream from where larvae were released， a 17 mm section of the flume floor was video-recordedthrough the flume wall. Larvae were released in 3 series and a total of 29 cyprids attached to therecorded section of the floor. The behaviour preceding temporary attachment was divided into2 categories； 1) swimming against the current， 2) swimming/drifting downstream or any otherbehaviour. The result was analysed with a single-variable frequency test (chi-square). Comparison between flow analysis of settling windows and empiricallarval attachment data When comparing the predicted proportions of settling windows from the flow analysis withempirical data of larval attachment， the attachment proportions were either used directly ornormalized to flow speed.Larval attachment was measured over 40 cm at all free-stream veloci-ties whereas the probability distributions of settling windows are time based. With increasingspeed， the time period available for larvae to attach at the working section before being sweptoff was probably shorter， although this effect was partly reduced by larvae swimming againstthe flow. To set an upper bound to this bias we converted the length where attachment waspossible to a time period possible for attachment by dividing the length of the attachment plateby the average local flow speed at z =0.5 at the three free-stream velocities. The measured pro-portions of cyprid attachment were then multiplied by the ratio of the time available for attach-ment at the free-stream velocity of0.1 msand the time available at the free-stream velocitythe attachment data was recorded for. Hence the attachment data for 0.1 msremained July 27， 2016 unchanged whereas the attachment data for 0.15 and 0.20 mswere increased， compensatingfor the shorter time available for attachment. To find the best fit between the empirically determined proportion of larval attachment andthe predicted opportunities for attachment from suitable instantaneous flow events， the follow-ing procedure was applied. We expected that the attachment probability (Ap) is a function ofthe proportional time available for attachment during lull periods (L(U.o，tcr) in Eq 3) withsome proportionality constant (f) representing the attachment propensity independent of Laccording to： Since we have no information on the value of f， we used the relative change in successfulattachment of cypris larvae in the different free stream velocities for comparison with predictedopportunities from calculated flow events. The relative changes in attachment of cypridsamong free-stream velocities were compared to the relative changes in attachment opportuni-ties among free-stream velocities predicted by the flow events. The ratios between the observedattachment proportion (Ao) at U.。 =0.15 and 0.10 ms，U.。=0.20 and 0.10 ms，and U.=0.20 and 0.15 mswere calculated. From the probability distributions of proportional timeavailable in lull periods (proportional to Ae)， we calculated similar ratios between the free-stream velocity treatments for each potential critical velocity (0.001-0.1 ms ) or critical shearstress (0.001-0.3 Pa)， and for each of the critical time intervals for settlement (0.14， 0.47， 1.02and 2.03 s). Note that attachment is only expected in a time interval longer than the criticaltime and below the critical stress or velocity.The critical velocity (or stress) where the Ap-based ratio best agreed with the Ao-based ratio was noted. Thus 3 critical velocities or stresseswere obtained (one for each ratio tested) for each critical time interval. The best fit between theempirically measured attachment and the predicted opportunities from flow events was con-sidered to be the critical time interval where the variation (SD) among the 3 critical velocity orstress values was lowest since a low variation indicates that a similar critical velocity or stresswas predicted for all three free-stream velocities. We also wanted to assess if critical velocity or critical shear stress best explained ourobserved attachment data. Since velocity and stress have different units and this will affect themagnitude of the variation and make a direct comparison difficult， the data were Z-trans-formed： where X，is each individual value of critical velocity or stress and X and SD are the average criti-cal velocity or stress and standard deviation respectively across all critical time intervals. Thestandard deviations for the Z-transformed critical velocities and stresses were then used for thecomparison of best fit. Results Near-bed flow characteristics The average flow velocity (from PIV analysis) in the inner part of the boundary layer could beseparated into a logarithmic layer and a linear layer close to the bed (Fig3). Due to the high res-olution of the measurements (10 data points mm )， the height of the linear part could be thor-oughly determined. The height of the linear part decreased with increasing free-stream velocityand was about 1.6 mm at U.。=5 cms ， 1.1 mm at U.。=10cms ，0.8 mm at U.。=15 cms，and 0.6 mm at U.。=20 cms. The dimensionless height z*(z u/v； v is kinematic viscosity) July 27，2016 Fig 3. Mean flow velocities in the near-bed region measured by Particle Image Velocimetry (PIV).Velocity gradients are shown for free-stream velocities (U..) of 0.05， 0.1， 0.15 and 0.2 ms. The height of thelinear part of each velocity profile is indicated by a fitted straight line. doi：10.1371/journal.pone.0158957.g003 was between 3.2 and 4.9， which agrees well with a typical z* ofabout 5 [47]. The results suggestthat the cyprids (0.5 mm long) making contact with the bed were within the linear part of thelayer at all studied free-stream velocities. Note that although the average velocity increases line-arly with height from the substratum where larvae reside at the bed， the velocity fluctuationscan be substantial (Fig 4). Flume studies of larval attachment success and swimming Observations of larvae showed that they swam actively against the current and towards the bot-tom to remain in contact with the bed. We conclude that larvae at U =0.1 and 0.15 msprobably were in contact with the flume floor most of the time whereas at 0.2 ms they mayhave been temporarily resuspended from the bed. When attempting to attach， the studied lar-vae kept both antennules extended in front of their body. The temporary attachment of cyprids Fig 4. Stream-wise velocity fluctuations at z=0.5 mm above the flume floor. Velocity fluctuations weremeasured at 15 Hz in free stream velocities (U.) of 0.05-0.2msusing PIV. doi：10.1371/journal.pone.0158957.g004 Fig 5. Flow-dependent proportion of temporary attachment in cyprids. Bars represent the proportion(mean±SE，n=4) of attachment for 20-40 cyprids released in a flume flow at free-stream velocities of 0.1，0.15 and 0.2 ms. Lowercase letters denote significant differences among factor levels (P<0.05， SNK post-hoc). The labels show the mean local velocity at the height where cyprids interact with the bed (z=0.5 mm).Data shown are original attachment proportions not normalized to flow speed (see Materials and Methods)and include two larval batches. doi：10.1371/journal.pone.0158957.g005 released at the bed decreased sharply with increasing local velocity at z=0.5 mm (2-factorANOVA， F2，2=1032，P=0.001； Fig 5). The difference was significant among all velocities(SNK post-hoc， P0.05). There was no effect of larval batch and no interaction between larvalbatch and velocity. The swimming speed of cyprids was estimated to 0.018±0.004 ms(mean ±SD). In the detailed study of larval attachment behaviour， swimming against the cur-rent was of significant importance to attachment success (x=21.6， df=1， P<0.001). Of the29 cyprids recorded while temporary attaching to the flume floor， 27 (93%) swam against thecurrent immediately preceding attachment. The other 2 larvae managed to attach while over-turning head first putting the antennular discs on the floor downstream oftheir body. Velocity and stress fluctuations and predicted opportunities for larvalattachment Following the analytical procedure in Crimaldi et al. [26] we produced cumulative frequencydistributions for a range of critical instantaneous local velocities (Fig 6a) and instantaneous bedshear stresses (Fig 6b) for various critical time periods required for attachment. These attach-ment probability distributions were compared to empirical data of cyprid attachment in threefree-stream velocities. We used the minimum of standard deviation (on Z-transformed data) forthe three critical velocities or stresses derived for each critical time interval to find the best fitbetween the empirical frequencies of larval attachment and the predicted opportunities forattachment from suitable instantaneous flow events. The SD generally decreases with decreasingthreshold time needed for attachment indicating cyprids can utilize very short stress lulls (~0.1s) for attachment (Table 2). The empirical larval attachment data fits better to the distributionsof instantaneous velocities than to the distributions of instantaneous bed shear stresses indicat-ing that instantaneous local velocity is the flow property better predicting temporary attachmentof cypris larvae compared to total bed shear stresses (Table 2). The best fit (minimum SD)between the original larval attachment data and the frequency of suitable flow events is found July 27，2016 Critical velocity (m s-1) Critical stress (Pa) Fig 6. Cumulative probabilities for time in settlement windows. Probabilities displayed as a function of a) tolerable threshold localvelocity or b) bed shear stress and minimum time required for temporary attachment. The vertical lines represent the best fit betweenempirical settlement proportions to the cumulative probability curves for critical velocity and critical stress respectively. The solid lines showbest fit based on original attachment data and the dashed lines represent the best fit using attachment proportions normalised to flow speed.See text for more details. doi：10.1371/journal.pone.0158957.g006 for a threshold local flow velocity of 0.019±0.001 ms (±SE) for a minimum duration of 0.14s. If the comparison instead is done using data of temporary attachment normalised to flowspeed (see Materials and Methods)， a local flow velocity of 0.024±0.000 msfor 0.14 s resultsin the best fit. The vertical lines in the panels of Fig 6 indicate the threshold local flow velocities(a) and the threshold bed shear stresses (b) that best fitted the cyprid attachment data. Hydrodynamic forces on larvae We calculated the resulting hydrodynamic force (vector sum of drag and lift) on attached lar-vae at 0.024 ms， which was the most likely threshold velocity for larval attachment July 27，2016 Table 2. Fit of larval attachment data to critical flow properties. Attachment data Critical probability distribution Threshold attachment time (s) 0.14 0.47 1.02 2.03 Original Velocity Mean 0.019 0.025 0.031 0.042 ZSD 0.17 0.28 0.53 0.95 Bed shear stress Mean 0.053 0.069 0.079 0.096 ZSD 0.39 0.70 0.83 1.07 Normalised Velocity Mean 0.024 0.030 0.038 0.048 Z SD 0.054 0.14 0.47 0.79 Bed shear stress Mean 0.062 0.076 0.090 0.12 ZSD 0.37 0.54 0.49 0.63 Average critical velocities (m s"l) and bed shear stresses (Pa) derived from comparison between empirically determined larval attachment and predictedopportunities for attachment from suitable instantaneous flow events. The SD-values are from Z-transformed data to facilitate comparison of the magnitudeof variation between the different units. The best fit (minimum SD) is given in bold for original attachment data and attachment data normalised to flow speedrespectively. See text for more details. doi：10.1371/journal.pone.0158957.t002 (normalised data) and at 0.116 ms， which was the highest instantaneous flow velocity atz=0.5 mm obtained during our measurements (at U。=20 cms). The resulting force at aninstantaneous velocity of 0.024 m swas 9.3×10-8Nand at 0.116 ms7.7×10-N.To com-pare these results with cyprid attachment strength we used data produced by Eckman et al.[33] for Balanus amphiterite cyprids， which are of equal size as B. improvisus cyprids. Themean instantaneous force resulting in cyprid dislodgement was 6.0×10-N. Hence the forceimposed on B. improvisus cyprids in this study at the threshold velocity allowing attachment istwo orders of magnitude lower than the average force needed to dislodge temporary attachedcyprids in the study by Eckman et al. [33]. The force imposed by the highest instantaneouslocal velocity measured at U.。=0.2 ms， is still one order of magnitude lower than the forcerequired for dislodgement. Discussion It still remains a major challenge to resolve the relative importance of larval supply， contactrate with the substrate and active habitat choice in determining settlement patterns. Contactwith the substrate will strongly depend on coastal circulation， turbulent mixing and gravity.Moreover， behavioural responses may modify contact rate through responses in the water col-umn [17-20] and possibly also close to the substrate [37]. A crucial step is to after initial con-tact maintain a position at the substrate long enough to ensure a more secure temporary orpermanent attachment [3]. The least understood step is how the transition between initial con-tact and a secure attachment is acquired and how it is affected by flow regime， surface proper-ties and larval behaviour. In the present study we found a decline in temporary attachment of barnacle larvae at meanfree-stream flow velocities between 0.1 to 0.2 ms. The gradual decline without a suddenthreshold suggests that cypris larvae either show large individual variability in their ability tomake an attachment or that they respond to non-mean flow properties. Crimaldi et al. [26]suggested that larval attachment in a turbulent boundary layer depends on the presence of suf-ficiently long lull events below some critical stress. Thus， the frequency of time in these settlingwindows in a given flow regime should determine the probability of successful attachment.The distribution of suitable instantaneous flow events is strongly affected by unidirectionalflow speed， wave exposure and presence of microhabitats [26-28]. The instantaneous flow July 27，2016 structures in Crimaldi et al. [26] were measured in the turbulent logarithmic part of the bound-ary layer， but in accordance with Reidenbach et al. [27]， our measurements clearly illustratethat also within the linear part of the boundary layer， where larval attachment usually takesplace， flow velocities in the stream-wise direction vary substantially (Fig 4). We conclude fromthe fit to observed larval attachment probabilities in Fig 6 that a fixed response to instantaneousflow velocity， in this case lulls below 0.019-0.024 ms， can explain the gradual decline inattachment probability with increasing flow speeds. Furthermore we can conclude that instan-taneous velocity is a better predictor of cyprid attachment probability compared to instanta-neous bed shear stress (Table 2). This could also be true for other larvae that are strongswimmers whereas instantaneous stress may be more important for larvae incapable of rheo-taxis. We also calculated the drag and lift forces on larvae and found that the resulting hydro-dynamic force imposed on cyprids at the critical local flow velocity allowing attachment is twoorders of magnitudes lower than the hydrodynamic force needed to dislodge cyprids duringtemporary attachment [33]. This indicates that attachment strength is not an important factorlimiting settlement of cyprids in turbulent flow. Therefore models of settlement probabilitybased on measurements of larval attachment strength， e.g. [27]， may overestimate settlementprobability in cases where the crucial limiting step precedes temporary or permanentattachment. Instead a limiting step can be the transition between initial contact and temporary (or per-manent) attachment and this study illuminates the importance of investigating near-surfacelarval behaviour and larval attachment mechanisms when predicting larval settling probabili-ties from flow properties. For barnacle cyprids， we here show that this transition is likely gov-erned by rheotactic swimming allowing the cyprid to remain in contact with the substrate untila stress lull can be utilized for temporary attachment. In this study we repeatedly observed howcyprids of B. improvisus swam upstream when close to the bed， a behaviour that was foundvery important for accomplishing temporary attachment (>90%of observed attachments wereimmediately preceded by upstream swimming). Interestingly， the predicted critical local flowvelocity allowing temporary attachment is close to the swimming speed， 0.018 m s， of B.improvisus cyprids. Although larvae swam vigorously in our swimming speed measurements，the measured speed is probably somewhat lower than the absolute maximum larvae can gener-ate when encountering very challenging conditions. Crisp also observed that cyprids of the bar-nacle Semibalanus balanoides showed strong negative rheotaxis [37]. Crisp suggested thatcyprids in this way may adjust their swimming speed to the ambient flow to facilitate settle-ment [48]. Our results support this hypothesis and based on our findings we propose a scenariowhere the cyprid upon contact with the substrate gains a reference to exercise negative rheotac-tic swimming. This reduces the relative speed to the substrate and while orienting the body to asuitable position the cyprid can make a strong temporary attachment with its antennular discs.This attachment is likely facilitated by the highly flexible joints between segments of the anten-nules [49]. According to our analysis the whole sequence can be completed in only 0.1 s allow-ing settlement also in challenging flow environments. The maximum swimming velocity thatcyprids may orient in an upstream direction in response to substrate contact may thus deter-mine at what flow speed barnacles can settle. In a downwelling flume， DiBacco et al.[50] foundthat the upward swimming velocity of S. balanoides increased with velocity of the downwellingflow and could be as high as 0.072 msfor short time periods. Crisp [37] measured swimmingspeeds for S. balanoides and Balanus crenatus that were more than twice as fast as the speedswe measured for B. improvisus. An additional aspect is that larger larvae will experience fasterlocal flow speeds because of the boundary layer. Cyprids of S. balanoides and B. crenatus areabout twice the size of B. improvisus cyprids so in a linear velocity gradient their swimmingspeeds should allow settlement at similar free-stream velocities. Crisp [37] also measured the July 27，2016 Many rocky-bottom intertidal barnacle species inhabit natural environments where ambi-ent flow velocities due to wave action or tidal currents often greatly exceed the flow velocitiesinvestigated in the present study. With increasing flow velocity， the frequency of lull periods inthe near bed flow with speeds lower than the cyprid swimming speed decreases and eventuallybecomes almost nonexistent. Hence our results suggest that cyprids in order to attach in high-flow environments might rely on the periods when waves and tides are turning. It may furtherbe speculated that fast swimming bursts against the boundary-layer flow is an adaptation toallow cyprids to actively select relatively high-flow environments. It also seems likely thatcyprids have the ability to fine-tune their flow exposure during temporary attachment bysearching for optimum sites among surface roughness elements [51]， or by actively leaving asubstrate exposed to excessive flow speeds； cyprids settling in high flow speeds may sufferreduced feeding rates as post-metamorphic juveniles [24]. Few marine larvae rival barnaclecyprids in terms of swimming speed. Ascidian tadpole larvae may reach above 0.01 m s[52]and could exploit a similar settlement behaviour as cyprids， although no information is yetavailable on a possible rheotactic behaviour. Also species with slow-swimming larvae can colo-nize high-flow environments， e.g. hydroids and bryozoa [53]. Little is known about the mecha-nisms governing the transition between initial contact and secure attachment in most larvaebut it can be speculated that hydrophobic interactions， mucous strings and ciliated body sur-faces are involved [35，54] and subsequently many larvae can strengthen the attachment byadhesive glands， e.g. in bivalves and in bryozoa [55，56]. However， in contrast to barnaclecyprids， larvae adopting a strategy based on high surface adhesion will not have the same flexi-bility to select habitats based on flow speed. In conclusion， we here for the first time compare settlement predictions from flow analysiswith actual attachment of larvae in turbulent flow. Our study suggests that marine larvae lessthan one mm in size can exercise behavioural responses at the bed that enhance attachmentprobability. Directed swimming against the flow allows barnacle cyprids to utilize very shortperiods of velocity lulls in the turbulent flow for attachment. The relative importance of larvaldelivery rate to the substratum and larval behavioural responses before and after temporaryattachment on settlement patterns remains challenging to estimate. However， although highturbulence levels increase the rate of larval transport to the substrate， Crimaldi et al. [26] foundthat the decrease in attachment probability due to high turbulence had a larger effect on settle-ment success than the increase in larval transport caused by turbulent mixing. Hence the abilityof barnacle larvae to increase the attachment rate by behavioural actions at the substrate raisesthe possibility to control habitat choice and can modify relationships between larval supplyand recruitment. Supporting Information S1 Dataset. Underlying data. The raw data underlying the statistical analyses， tables and fig-ures in the present study. (XLSX) Acknowledgments We thank Martin Ogemark for rearing of barnacle larvae. July27，2016 Author Contributions Conceived and designed the experiments： AIL LMG PRJ. Performed the experiments： AILLMG. Analyzed the data： AIL PRJ. Contributed reagents/materials/analysis tools： AIL PRJ.Wrote the paper： AIL LMG PRJ. References ( 1. Mullineaux LS， Garland ED. Larval recruitment in res p onse to manipulated field flows. Mar Biol. 1993； 116：667-683. doi： 10.100 7 /BF00355484 ) ( 2. Judge ML， Craig SF. Positive flow dependence in the initial colonization of a fouling community： resu l ts f rom in situ water current manipulations. J Exp Mar Biol Ecol. 1997 ； 210：209-222.doi： 1 0 . 1 0 16 / S0022-0981 ( 96)0269 1 -3 ) 3. Jonsson PR， Berntsson KM， Larsson Al. Linking larval supply to recruitment： Flow-mediated control ofinitial adhesion of barnacle larvae. Ecology.2004；85：2850-2859. doi： 10.1890/03-0565 4. Dobretsov S， Wahl M. Larval recruitment of the blue mussel Mytilus edulis： the effect of flow and algae.JExp Mar Biol Ecol.2008；355：137-144.doi： 10.1016/j.jembe.2007.12.018 5. Caley MJ， Carr MH， Hixon MA， Hughes TP， Jones GP， Menge BA. Recruitment and the local dynamicsof open marine populations. Annu Rev Ecol Syst. 1996； 27：477-500. doi： 10.1146/annurev.ecolsys.27.1.477 ( 6. Shkedy Y，R o ughgarden J. Barnacle re c ruitment and population dynamics predicted from coastal upwelling.Oikos. 1997； 8 0：487-498. doi： 1 0 .2307/ 3 54662 2 ) ( 7. Knights AM， Walters K. R e cruit-recruit in t eractions density-dependent processes and population per-sistence in the eastern oyster Crassostrea virginica. Mar Ecol Prog Ser.2010； 404：79-90. doi： 1 0 .3354/m ep s08480 ) ( 8. Paris CB， Cherubin LM， Cowen RK. Surfing spinning or diving from reef to reef： effects on populationconnectivity. Mar Ecol Prog Ser. 2007 ； 347：285-300. doi： 10.3354/me p s 0 6985 ) 9. Gaines S， Brown S， Roughgarden J. Spatial variation in larval concentrations as a cause of spatial vari-ation in settlement for the barnacle Balanus glandula. Oecologia. 1985； 67：267-272.doi： 10.1007/BF00384297 10. Roughgarden J， Pennington JT， Stoner D， Alexander S， Miller K. Collisions of upwelling fronts with theintertidal zone： the cause of recruitment pulses in barnacle populations of central California. ActaOecol. 1991； 12：35-51. 11. Bertness MD， Gaines SD， Wahle RA. Wind-driven settlement patterns in the acorn barnacle Semibala-nus balanoides. Mar Ecol Prog Ser. 1996； 137：103-110. doi： 10.3354/meps137103 12. Hoffmann V， Pfaff MC， Branch GM. Spatio-temporal patterns of larval supply and settlement of intertidalinvertebrates reflect a combination of passive transport and larval behavior. J Exp Mar Biol Ecol.2012；418-419：83-90. doi： 10.1016/j.jembe.2012.03.008 13. Pineda J， Riebensahm D，Medeiros-Bergen D. Semibalanus balanoides in winter and spring： Larvalconcentration settlement and substrate occupancy. Mar Biol.2002； 140：789-800. doi： 10.1007/s00227-001-0751-Z 14. Jenkins SR. Larval habitat selection not larval supply determines settlement patterns and adult distribu-tion in two chthamalid barnacles.JAnim Ecol.2005； 74：893-904. doi： 10.1111/.1365-2656.2005.00985.x 15. Abelson A， Denny M. Settlement of marine organisms in flow. Annu Rev Ecol Syst. 1997； 28：317-339.doi： 10.1146/annurev.ecolsys.28.1.317 16. McNair JN， Newbold JD， Hart DD. Turbulent transport of suspended particles and dispersing benthicorganisms： how long to hit bottom? J Theor Biol. 1997； 188：29-52. doi： 10.1006/jtbi.1997.0453 17. Tamburri MN， Finelli CM， Wethey DS， Zimmer-Faust RK. Chemical induction of larval settlementbehaviour in flow.Biol Bull. 1996； 191：367-373. doi： 10.2307/1543009 18. Fuchs HL， Mullineaux LS， Solow AR. Sinking behavior of gastropod larvae (llyanassa obsoleta) in tur-bulence.Limnol Oceanogr.2004； 49：1937-1948.doi： 10.4319/lo.2004.49.6.1937 19. Hadfield MG， KoehlMAR. Rapid behavioral responses of an invertebrate larva to dissolved settlementcue. Biol Bull. 2004；207：28-43.doi： 10.2307/1543626 PMID： 15315941 20. Fuchs HL， Hunter EJ， Schmitt EL， Guazzo RA. Active downward propulsion by oyster larvae in turbu-lence.J Exp Biol. 2013；216：1458-1469. doi： 10.1242/jeb.079855 PMID： 23264488 July 27，2016 21. Pawlik JR， Butman CA. Settlement of a marine tube worm as a function of current velocity： Interactingeffects of hydrodynamics and behavior. Limnol Oceanogr.1993；38：1730-1740. doi： 10.4319/lo.1993.38.8.1730 22. Granhag LM， Larsson Al， Jonsson PR. Algal spore settlement and germling removal as a function offlow speed. Mar Ecol Prog Ser. 2007；344；63-69. 23. Mullineaux LS， Butman CA. Initial contact exploration and attachment of barnacle (Balanus-Amphitrite)cyprids settling in flow. Mar Biol. 1991； 110：93-103. doi： 10.1007/BF01313096 24. Larsson Al， Jonsson PR. Barnacle larvae actively select flow environment supporting post-settlementgrowth and survival. Ecology. 2006； 87：1960-1966. doi： 10.1890/0012-9658(2006)87[1960：BLASFE]2.0.CO；2 PMID： 16937634 ( 25. Dade WB. Near-bed turbulence and hydrodynamic control of diffusional mass transfer at the sea floor.Limnol Oceanogr. 1993； 38：52-69. doi： 1 0 . 4 3 1 9 /l o .19 9 3.38.1 . 0 05 2 ) ( 26. Crimaldi JP， Thompson JK， Rosman JH， Lowe RJ， Koseff JR. Hydrodynamics of larval settlement： Theinfluence of turbulent stress events at potential r ecruitment sites. Limnol Oceanogr. 2002； 47：1137- 1151. doi： 1 0 .4319 / lo.2002 . 4 7 . 4. 1 13 7 ) 27. Reidenbach MA， Koseff JR， Koehl MAR. Hydrodynamic forces on larvae affect their settlement on coralreefs in turbulent， wave-driven flow. Limnol Oceanogr. 2009； 54：318-330. doi： 10.4319/lo.2009.54.1.0318 28. Koehl MAR， Crimaldi JP， Dombroski DE. Wind chop and ship wakes determine hydrodynamic stresseson larvae settling on different microhabitats in fouling communities. Mar Ecol Prog Ser.2013； 479：47-62. doi： 10.3354/meps10193 29. Johansson AV，Alfredsson PH. On the structure of turbulent channel flow. J Fluid Mech. 1982； 122：295-314. doi：10.1017/S0022112082002225 30. Denny MW. Biology and Mechanics of the Wave-Swept Environment. Princeton： Princeton UniversityPress； 1988 31. Koehl MAR， Hadfield MG. Soluble settlement cue in slowly moving water within coral reefs induces lar-val adhesion to surfaces. J Mar Syst. 2004； 49：75-88.doi： 10.1016/j.jmarsys.2003.06.003 32. Walker G， Yule AB. Temporary adhesion of the barnacle cyprid： the existence of an antennular adhe-sive secretion. J Mar Biol Ass UK. 1984；64：679-686. doi： 10.1017/S0025315400030344 33. Eckman JE， Savidge WB， Gross TF. Relationship between duration of cyprid attachment and dragforces associated with detachment of Balanus amphitrite cyprids. Mar Biol. 1990；107：111-118. 34. Aldred N，Hoeg JT， Maruzzo D， Clare AS. Mechanics of the behaviours mediating barnacle cypridreversible adhesion. PLoS ONE 2013； 8(7)：e68085. 35. Berntsson KM， Jonsson PR， Larsson Al， Holdt S. Rejection of unsuitable substrata as a potential driverof aggregated settlement in the barnacle Balanus improvisus. Mar Ecol Prog Ser. 2004；275：199-210.doi： 10.3354/meps275199 36. Prendergast GS， Zurn CM， Bers AV， Head RM， Hansson LJ， Thomason JC.Field-based video obser-vations of wild barnacle cyprid behaviour in response to textural and chemical settlement cues. Biofoul-ing 2008； 24：449-459. doi： 10.1080/08927010802340135 PMID： 18696291 37. Crisp DJ. The behaviour of barnacle cyprids in relation to water movement over a surface. J Exp Biol.1955；32：569-590. 38. Mihm JW， Banta WC，Loeb GI. Effects of adsorbed organic and primary fouling films on bryozoan set-tlement. J Exp Mar Biol Ecol. 1981； 54： 167-179. doi： 10.1016/0022-0981(81)90142-8 39. Jonsson PR， Johansson M. Swimming behaviour patch exploitation and dispersal capacity of a marinebenthic ciliate in flume flow. J Exp Mar Biol Ecol. 1997； 215：135-153. doi： 10.1016/S0022-0981(97)00031-2 ( 40. Jonsson PR， van Duren LA， Amielh M， As m us R， Aspden RJ，Daunys D， et al. Making water flow：acomparison of the hydrodynamic characteristics of 12 d i fferent benthic biological flumes.Aquat Ecol.2006；40：409-438.doi： 1 0 . 1 007/ s 1 0 4 5 2-0 0 6-9049 - z ) ( 41. Schlichting H. Boundary-layer theory. 7th ed . New York： McGraw-Hill； 1979 ) ( 42. Larsson Al， Mattsson-Thorngren L， Granhag LM， Berglin M. Fouling release of barnacles from a boathull with comparison to laboratory data of attachment strength. J Exp Mar Biol Ecol.2010； 392：107-1 1 4. doi： 1 0 . 1016 / j .jembe . 2010.04.014 ) 43. White FM. Viscous fluid flow. New York： McGraw-Hill； 1974 ( 44. Wiberg PL， S m ith JD.Calculations of the critical shear stress for motion of uniform and heterogeneous sediments. Water Resour Res. 1987 ； 23：1471-1480. do i：1 0 . 1029/W R 023i008p0 1 47 1 ) ( 45. Chepil WS. The use of evenly spaced hemispheres to evaluate aerodynamic forces on soil surfaces.Eos Trans AGU. 1958；39：397-404. ) July 27，2016 46. Berntsson KM， Jonsson PR， Lejhall M， Gatenholm P. Analysis of behavioural rejection of micro-tex-tured surfaces and implications for recruitment by the barnacle Balanus improvisus. J Exp Mar BiolEcol.2000；251：59-83.doi： 10.1016/S0022-0981(00)00210-0 PMID： 10958901 47. Middleton GV， Southard JB. Mechanisms of sediment movement. Tulsa： Society of Economic Paleon-tologists and Mineralogists； 1984 48. Crisp DJ. Factors influencing the settlement of marine invertebratelarvae. In： Grant PT， Mackie AM，editors.Chemoreception in marine organisms， London： Academic Press； 1974. p. 177-265. 49. Lagersson NC， Hoeg JT. Settlement behavior and antennulary biomechanics in cypris larvae of Bala-nus amphitrite (Crustacea： Thecostraca： Cirripedia). Mar Biol.2002；141：513-526. doi： 10.1007/S00227-002-0854-1 50. DiBacco C， Fuchs HL， Pineda J， Helfrich K. Swimming behavior and velocities of barnacle cyprids in adownwelling flume. Mar. Ecol. Prog. Ser.2011； 433： 131-148. doi： 10.3354/meps09186 51. Wethey DS. Ranking of settlement cues by barnacle larvae： influence of surface contour. Bull Mar Sci.1986； 39：393-400. 52. Chia F-S， Buckland-Nicks J， Young CM. Locomotion of marine invertebrate larvae： a review. Can JZool. 1984；62：1205-1222. doi： 10.1139/z84-176 53. Qian PY， Rittschof D， Sreedhar B. Macrofouling in unidirectional flow： miniature pipes as experimentalmodels for studying the interaction of flow and surface characteristics on the attachment of barnacle，bryozoan and polychaete larvae. Mar Ecol Prog Ser.2000； 207：109-121. doi：10.3354/meps207109 54. Fenchel T， Ockelmann KW. Larva on a string. Ophelia.2002； 56：171-178. doi： 10.1080/00785236.2002.10409497 55. Gee JM， Knight-Jones EW. The morphology and larval behaviour of a new species of Spirorbis (serpuli-dae). J Mar Biol Ass UK. 1962； 42：641-54. doi： 10.1017/S0025315400054321 ( 56. Lynch WF. The behaviour and metamorphosis of the larva of Bugula neritina (Linnaeus)： Expe r imentalmodification of the length of the free-swimming period and the responses of the larvae to light and grav- ity. Biol Bull. 194 7 ； 9 2 ：115-150. doi ： 1 0 .2307 / 1538162 P MID： 2 0294096 ) July 27，2016 PLOS ONE|DOI：journal.pone. Water flow affects settlement of marine larvae on several scales. At the smallest scale localflow regime may control the probability of adhesion to the substrate. Our aim was to mechanisticallyunderstand the transition from suspended to attached larvae in turbulent flow.Recently it was proposed that opportunities for larval settlement in turbulent boundary layersdepend on time windows with suitable instantaneous flow properties. In flume flow wecharacterized the proportion of suitable time windows in a series of flow velocities withfocus on the near-bed flow. The change in the proportion of potential settling windows withincreasing free-stream velocities was compared to the proportion of temporary attachmentof barnacle cypris larvae at different flow velocities. We found large instantaneous flow variationsin the near-bed flow where cyprid attachment took place. The probability of temporaryattachment in cyprids declined with local flow speed and this response was compatible witha settling window lasting at least 0.1 s with a maximum local flow speed of 1.9–2.4 cm s-1.Cyprids swam against the near-bed flow (negative rheotaxis) and the swimming speed (1.8cm s-1) was close to the critical speed that permitted temporary attachment. We concludethat temporary attachment in barnacle cyprids requires upstream swimming to maintain afixed position relative to the substrate for at least 0.1 s. This behaviour may explain the abilityof barnacles to recruit to high-flow environments and give cyprids flexibility in the pre-settlementchoice of substrates based on flow regime.