Effect of temperature on the growth, survival, development and foraging behaviour of Sardina pilchardus larvae

MARINE ECOLOGY PROGRESS SERIES Mar Ecol Prog Ser Vol. 559: 131–145, 2016 doi: 10.3354/meps11881 Published November 9 INTRODUCTION Temperature influences metabolism and growth in poikilotherms (Clarke & Johnston 1999) and is considered a key factor shaping the ecophysiology and life history strategy of fish (Pörtner & Peck 2010). Temperature plays a particularly important role during the early life of marine fish by influencing rates of metabolism, growth and mortality of marine fish larvae (Houde 1989, Blaxter 1991). The ability of a species, a population or an individual to persist over a range of temperatures is limited by the thermal adap- © The authors 2016. Open Access under Creative Commons by Attribution Licence. Use, distribution and reproduction are un - restricted. Authors and original publication must be credited. Publisher: Inter-Research · www.int-res.com *Corresponding author: gar[email protected] Effect of temperature on the growth, survival, development and foraging behaviour of Sardina pilchardus larvae S. Garrido1,2,*, A. Cristóvão2, C. Caldeira2, R. Ben-Hamadou3, N. Baylina4, H. Batista4, E. Saiz5, M. A. Peck6, P. Ré2, A. M. P. Santos1,7 1Instituto Português do Mar e da Atmosfera, Av. Brasília s/n, 1449-006 Lisboa, Portugal 2Marine and Environmental Sciences Centre, Faculdade de Ciências, Universidade de Lisboa, Campo Grande, 1749-016 Lisboa, Portugal 3Department of Biological and Environmental Sciences, College of Arts and Sciences, Qatar University, PO Box 2713, Doha, Qatar 4Oceanário de Lisboa, Esplanada D Carlos I, 1900-005 Lisbon, Portugal 5Institut de Ciències del Mar – CSIC, Ps. Marítim de la Barceloneta 37-49, 08003 Barcelona, Spain 6Institute for Hydrobiology and Fisheries Science, University of Hamburg, Olbersweg 24, 22767 Hamburg, Germany 7CCMAR, Centro de Ciencias do Mar, Universidade do Algarve, Campus de Gambelas, 8005-139 Faro, Portugal ABSTRACT: The effect of water temperature on the growth, survival, development and foraging behaviour of European sardine Sardina pilchardus larvae was examined in the laboratory. First, the capability of early sardine larvae to cope with starvation was assessed at temperatures from 10 to 22°C. Second, we examined under ad libitum feeding conditions and across the range of temperatures experienced by sardines during spawning along the Atlanto-Iberian coast (13−17°C) the ontogenetic changes in growth, survival and foraging behaviour of sardine larvae. Unfed larvae had similar maximum survival times (11−12 d post hatching, dph) from 13 to 15°C, but the survival time was significantly shorter at the coldest and warmest temperatures tested. The survival of exogenously feeding larvae increased with temperature, but younger endogenously feeding larvae had higher survival at colder temperatures. The cumulative mortality after 25 dph, however, was similar at the 3 temperatures. Not only larval growth rate increased with increasing temperature, but ontogenetic development also occurred sooner and at smaller sizes. Notochord flexion, which is a developmental milestone for fish, occurred 10 d earlier at 17 rather than at 13°C. The time spent swimming and the foraging behaviour (orientations to prey, feeding strikes and successful capture) significantly increased throughout the ontogeny and with temperature. This study highlights how even modest changes in spawning temperature can lead to large changes in the survival and growth of larval sardine. This study also reveals some of the mechanisms whereby inter-annual and seasonal variability in temperature can have significant ecological impacts at the population level. KEY WORDS: Sardina pilchardus · Growth rate · Pelagic fish · Mortality · Foraging behaviour O PEN PEN A CCESS CCESS Mar Ecol Prog Ser 559: 131–145, 2016 tive capacity of biochemical, cellular and organismal processes which constrains the geographic distribution of stocks and influences inter-annual variability in survival (Jordaan & Kling 2003) and also by the severity of climate impacts (Rijnsdorp et al. 2009). Even within the limits of thermal tolerance, subtle changes in temperature experienced by early larvae cannot only influence survival by altering physiological rates and behaviour but can also influence the performance of survivors later in life (Johnston et al. 2001, Koumoundouros et al. 2009, Moyano et al. 2016). Discriminating the relative contribution of genetic versus environmental factors to the variation in larval performance is necessary if we hope to provide robust predictions of the effects of climate change on fish stocks (Jordaan & Kling 2003, Dahlke et al. 2016). The European sardine Sardina pilchardus is distributed along the eastern North Atlantic coast, from Iceland and the North Sea to Senegal and the Mediterranean Sea (Whitehead et al. 1985). One of the main areas of sardine spawning in the northeast At - lantic is the Iberian Peninsula, where sardine larvae often dominate the ichthyoplankton community (Garrido et al. 2009). Due to persistently low recruitment since 2006, the size of the Iberian sardine stock has substantially decreased, reaching historical minimum values in 2012−2013 (ICES 2014). The de crease in recruitment of European sardines is thought to be directly related to increasing water temperature ob - served in recent years in the Western Iberian Up - welling Ecosystem (ranging from 0.02 to 0.03°C yr−1 since 1985) (Relvas et al. 2009). Field surveys in that region report the highest occurrence of spawning at 14 to 15°C with little or no spawning at water temp - eratures colder than 12°C and warmer than 16°C (Stratoudakis et al. 2007, Peck et al. 2013). The sizeat-maturation is generally smaller in sardines from warmer waters and, likely for this reason, length-atmaturation has decreased and the spawning period has increased in western Iberian waters during the last 2 decades (Silva et al. 2006). In order to understand the modifications in sardine spawning and development, it is essential to know the underlying physiological and behavioural responses of individual larvae at different temperatures. Laboratory studies can serve to simplify some of the interactions of inherently complex natural systems by removing factors that confound temperature effects, such as food limitation (Jordaan & Kling 2003). To our best knowledge, there are no previous laboratory studies examining ontogenetic changes and/or the effect of temperature on survival, growth and behaviour of exogenously feeding European sardine, with all previous studies restricted to either embryos or unfed larvae (Blaxter 1969, Miranda et al. 1990, Bernal et al. 2008). The present study investigated the effect of temperature on the growth, survival and foraging behaviour of European sardine larvae. First, the capacity of larvae to withstand starvation as a function of temperature was assessed. Second, the effect of temperature on growth, mortality and foraging behaviour was studied for larvae reared with excess food across at a range of temperatures encompassing the vast majority (95%) of those associated with sardine spawning in Atlanto-Iberian waters (Coombs et al. 2006). This study examines the effect of temperature on growth, mortality and foraging behaviour of European sardine larvae under controlled laboratory conditions, and our results are discussed in relation to patterns derived from the field. MATERIALS AND METHODS Larval rearing and growth experiments Sardine larvae were hatched from eggs spawned by broodstock European sardines Sardina pilchardus maintained in a 15000 l cylindrical tank at the Oceanário de Lisboa. Fish were originally captured in 2009 and 2010 by purse seine in coastal waters off Peniche (western Portugal). Adult fish started spawning naturally after adjusting the temperature (15°C) to natural conditions and increasing the light regime (16 h light:8 h dark) to the maximum number of daylight hours occurring in western Iberia. The eggs were collected daily in the morning from the broodstock using 500 µm mesh egg collector bags placed in the skimmers of the tank. At that time, most eggs were found halfway through egg development (Stage V, Gamulin & Hure 1955). A total of 2000 viable eggs was placed into 5 l glass beakers containing gently aerated water from the broodstock tank. Over the next 1 to 2 h, the temperature was slowly adjusted to 17, 15 or 13°C. Finally, eggs were gently transferred to 30 l cylindrical tanks filled with seawater of salinity 35 at their rearing temperature. Eggs were incubated using a 16 h light:8 h dark light regime and gentle aeration to mix the water and maintain a high oxygen concentration. Surface light levels were kept at 55 to 58 µmol s−1 m−2 (Philips Master TL-D Super 80 58W fluorescent lamp). Growth experiments consisted of transferring batches of ca. 2000 sardine eggs to the abovementioned 30 l tanks and rearing the embryos and larvae either unfed or fed a mixed diet of dinoflagel132 Garrido et al.: Effect of temperature on sardine larval growth lates Gymnodinium sp., rotifers Brachionus sp. and nauplii, copepodites and adults of the copepod Para - cartia grani. Details of the feeding conditions, which translate into saturated feeding and growth rates of sardine larvae, can be found in Silva et al. (2014) and Caldeira et al. (2014). Each morning, after checking the food concentration in the tanks, 20 to 30% of the tank water was renewed by syphoning the bottom of the tank, and new food was added to obtain the preestablished food concentration. Every other day pH, ammonia and oxygen were measured (Hanna Instruments 9828) to confirm that high water quality was maintained throughout the experiment. In the starvation experiment, 5 different temperatures were tested (10, 13, 15, 17, and 22°C ± 0.4°C). Experiments using fed larvae were conducted at 13, 15 and 17°C, which correspond to the coldest, peak and warmest temperatures, respectively, measured during peak spawning by sardines in Iberian waters (Coombs et al. 2006). Throughout the experiment, groups of larvae were randomly sampled from experimental tanks and rapidly anesthetized with MS-222 to determine their size (total length, TL) under a stereomicroscope (±0.05 mm accuracy) and posteriorly preserved in formaldehyde. When embryos were found to be in an advanced developmental stage (Stage X; Gamulin & Hure 1955), tanks were inspected frequently to determine the time when 50% of the larvae hatched. Larval size was de termined at the time when ∼50% of the larvae had hatched. Given that at the temperatures tested the hatching period can last up to 10 h (Bernal et al. 2008), larval size-at-hatch would be temperaturedependently overestimated, and for that reason larval size at the day of hatching was not formally compared between temperatures. Larvae were sampled from each tank each day until death in the case of unfed larvae. In the case of fed larvae, samples were taken until 50 d post-hatching (dph) at 15 and 17°C and at 25 dph at 13°C, where technical problems did not allow measurements beyond that age due to accidental loss of larvae. Mean larval TL for fed larvae was determined from measurements made on 20 (at hatching), ≥10 (3 to 15 dph) and 5−10 (≥16 dph) individuals. Measurements of sardine larvae were made at 0, 3, and 5 dph and then every 5 d until the end of the experiment. Every day, the bottom of each tank was syphoned, and dead larvae were counted to determine age-specific survival rates. Sardine larvae preserved in formaldehyde were digitally photographed under a stereoscope (Leica S8 APO, zoom 8:1 with a Canon EOS SLR550 camera), and the development of important ontogenetic traits of fish larvae, such as fin formation and notochord flexion, were registered for larvae reared at each of the 3 temperatures. Foraging behaviour The behaviour of fed larvae was observed at 3 temperatures (13, 15 and 17°C). Every day and for each temperature treatment, randomly selected larvae (≥20 observations) were individually followed and their behaviour registered for a 1 min interval (focalanimal technique, Martin & Bateson 1993). During observations, water mixing was stopped by removing the aeration so that individual larvae could be followed in the absence of turbulence which may affect larval behaviour. The time spent swimming by the larvae during the 1 min observation period (locomotory model action pattern), as an indicator of foraging effort, was logged. The foraging MAPs recorded, adapted from Barlow (1968), included the occurrence of orientations (also termed fixations or s-shape positions when detecting a prey) and the frequency of lunges or attacks on prey; given that not all orientations resulted in successful attacks, the percentage of complete feeding sequences with respect to the total number of orientations was also estimated. At 15 and 17°C, behavioural observations were made until the larvae were 50 and 25 dph at 13°C. Data analysis For fed larvae from 3 to 25 dph at each of the 3 temp erature treatments, growth rate was assessed from the exponential fit of TL (in mm) and age data using the equation: (1) where L0is length at 3 dph, kis the instantaneous growth rate and tis age in dph. In the case of larvae reared until 50 dph at 15 and 17°C, data were ad - justed from 3 to 50 dph to a Laird-Gompertz growth curve by an iterative nonlinear regression routine: (2) where A0is the growth rate at time 0, and αis the rate of exponential decay. To compare the relationship between size and age between larvae reared at different temperatures from age 3 to age 25 dph (common range in ages between the 3 temperatures), first generalized linear models (GLMs) with an identity link were used to as - Lkt TL e 0 = L At TL e 0 1e 0 = () () α−−α 133 Mar Ecol Prog Ser 559: 131–145, 2016 134 sess if the interactions of age and temperature were significant. When the interaction term was significant, pairwise comparisons of slopes were conducted using an ANCOVA model. When the differences between the slopes were significant, the regression was repeated excluding the interaction term. When the differences between the slopes were not significant, a new regression model excluding the inter - action term was fitted, and both regression models (with and without the interaction model) were compared using ANOVA, and then the most parsimonious model was selected. Growth rates of 3 to 50 dph larvae reared at 15 and 17°C and fitted to LairdGompertz growth curves were compared by using the method to compare non-linear models described in Chen et al. (1992). Cumulative survival rates of sardine larvae reared with excess food were calculated from Day 0 to 25 dph. Comparison of the survival rates of larvae reared with different temperatures was conducted using the Logrank (Mantel cox) z-test. Cumulative survival rates were considered different when p < 0.05. GLMs were used to test the significance of the interaction between age and temperature on the time spent swimming, the number of orientations, number of attacks on prey and the percentage of successful attacks on prey for pre-flexion sardine larvae (≤25 dph for larvae reared at 13 and 15°C and ≤20 dph for larvae reared at 17°C). When the interaction term was significant, pairwise comparisons of slopes were conducted using an ANCOVA model. When the differences between the slopes were significant, the re - gressions were repeated excluding the interaction term. When the differences between the slopes were not significant, a new regression model excluding the interaction term was fitted and both regression models (with and without the interaction model) were compared using ANOVA, and then the most parsimonious model was selected. Foraging efficiency (FE) was estimated as (3) where Kis the growth rate of sardine larvae and C is the capture rate (here estimated as the number of attacks on prey) at each of the 3 temperatures. Both K and Cwere expressed in carbon units (µg C) by using available conversions of larvae length to carbon content and prey mean length of prey (Paracartia grani) to carbon content (Caldeira et al. 2014). Statistical ana lyses were performed using Matlab 8.1 (R2013a) from Mathworks (mortality data) and the open source software R version 2.9.2 of R Development Core Team (growth and behavioural data). RESULTS Larval growth under starvation Sardine larvae hatched sooner at higher temperatures, and the time between egg collection to 50% hatch ranged from ca. 72 to 30 h at incubation temperatures of 10 and 22°C, respectively. The mean (±SD) TL at the time of 50% hatch was 2.9 ± 0.3, 3.6 ± 0.3, 3.8 ± 0.7, 4.3 ± 0.3 mm and 3.8 ± 0.2 mm at 10, 13, 15, 17 and 22°C, respectively. Time until the yolksac absorption and the larvae started exogenous feeding was 2.5 d at 22°C and 4.5 d at 10°C. The growth of endogenously feeding (yolk sac) larvae increased with increasing temperature except at the lowest (10°C) and highest (22°C) temperatures. At 10°C, larval length-at-age remained fairly constant until death and only increased from 0 to 3 dph at 22°C, sharply decreasing afterwards. At the other 3 temperatures tested, larval size increased until 6− 7 dph and then either abruptly dropped (17°C) or steadily declined (13° and 15°C) (Fig. 1). Maximum duration of survival of unfed larvae was 6, 12, 11, 12 and 7 dph at 10, 13, 15, 17 and 22°C, respectively (Fig. 1). Growth and survival of feeding larvae The batches of sardine larvae used in feeding trials hatched approximately 60, 48 and 30 h after the em - FE 100 /KC=× Fig. 1. Mean size-at-age (total length ± SE, mm) versus age (d post-hatching, dph) of unfed European sardine larvae at 5 dif - ferent temperatures. Black lines represent the interval at which the yolk sac was exhausted and eyes were pigmented. Mean sizeat-age at age 0 corresponds to the mean size when 50% of the larvae have hatched in the tank, not the precise size-at-hatch bryos were collected from the adult spawning tank, at 13, 15 and 17°C, respectively. There was a significant increase in size (TL) with age with increasing temperature (Fig. 2, Tables 1–3). The GLM model showed a significant effect of temperature on growth for larvae growing from 3 to 25 dph at 13, 15 and 17°C (Akaike’s information criterion [AIC] = 2694.8; p < 0.0001). The exponential increase in larval size from 3 to 25 dph (slope) was similar for larvae reared at 17 and 15°C and higher than that of larvae reared at 13°C (Tables 1 & 2). The ANOVA model comparGarrido et al.: Effect of temperature on sardine larval growth 135 Fig. 2. Total length (mm) versus age (d post-hatching, dph) of fed European sardine larvae at 3 different temperatures: (a) 13°C, (b) 15°C, (c) 17°C. Exponential growth curve fitted for the period from 3 to 25 dph for all temperatures (dashed line). Grey area represents the 95% confidence interval of the Laird-Gompertz curve fitted to the period from 3 to 50 dph for larvae reared at 15 and 17°C Period Temp. Regression No. of Confidence interval Goodness of fit (°C) equation ind. (95% confidence bounds) Adjusted R2 One-tailed probSlope Intercept ability (p-value) 3−25 dph 13 257 0.0231−0.0286 4.44−4.77 0.58 <0.001 15 291 0.0395−0.0437 4.16±4.43 0.84 <0.001 17 292 0.0412−0.0453 4.28−4.55 0.85 <0.001 4.12e0.0258age y= 4.12e0.0416age y= 4.32e0.0433age y= Table 2. Exponential regression of the relationship between larval size (total length, mm) and age (d post hatching, dph) for Sardina pilchardus larvae reared at 13, 15 and 17°C from 3 to 25 dph Temp. df p-value F ρ R2 (°C) Intercept Slope 13 and 15 1 <0.001 0.001 941 <0.001 0.81 13 and 17 1 0.023 <0.001 1033 <0.001 0.82 15 and 17 1 0.005 0.714 1710 <0.001 0.88 Table 1. Results of ANCOVA models analysing the relationship between larval size (total length, mm) and age (d post hatching, dph) for Sardina pilchardus larvae reared from 3 to 25 dph at 13,15 and 17°C ing the regressions with and without the interaction term of age and temperature for larvae reared at 15 and 17°C from 3 to 25 dph showed that removing the interaction did not significantly affect the fit of the model (F = 0.13; p = 0.71), suggesting that exposure temperature did not significantly alter the growth rate of larvae for this age interval. The regression of larval size depending of larval age and temperature without an interaction term for larvae reared at 15 and 17°C from 3 to 25 dph showed that the intercept was significantly different for larvae reared at the 2 temperatures (F = 24.77; p < 0.0001); therefore, sizeat-age was significantly higher for larvae reared at 17°C. Growth rates of larvae reared from 3 to 50 dph at 15 and 17°C were significantly different (F = 5.76; p > 0.0001), and larval growth rate was higher for larvae reared at the higher temperature (Fig. 2, Table 3). Cumulative survival rates were significantly different between 13 and 15°C (z = 4.75, p < 0.001), be - tween 13 and 17°C (z = 5.26, p < 0.001) and between 15 and 17°C (z = 2.04, p = 0.041 < 0.05) (Fig. 3). Mortality during the endogenous feeding stage increased with increasing temperature, but when exogenous feeding was the only feeding mode used by the larvae (>5 dph), daily mortality was significantly higher at lower temperatures. From 12 dph (for 15 and 17°C) and 15 dph (for 13°C) onwards, daily mortality was extremely low when compared to the first days after hatch (Fig. 3). The development was faster at warmer temperatures. In general, fin formation occurred earlier and at smaller sizes with increasing temperature (Table 4). At 10 dph, the pectoral and caudal fins were first ob - served at 10 dph for larvae reared at 13 and 15°C and at 5 dph for larvae reared at 17°C. The beginning of notochord flexion was observed at 10, 15 and 20 dph, for larvae reared at 13, 15 and 17°C, respectively, and the complete flexion occurred at 25 and 20 dph for larvae reared at 15 and 17°C, respectively. Not only did notochord flexion occur earlier in time for higher temperatures, it also occurred at smaller sizes. At 13°C, notochord flexion was not complete by the end of the observations (at 25 dph). Development of the caudal fin was complete at 40 dph (16.2 mm TL) and 30 dph (13.7 mm TL) for larvae reared at 15 and 17°C, respectively. Locomotory behaviour and foraging of well-fed larvae At 13, 15 and 17°C, 1 to 2 dph sardine larvae spent most of their time inactive, at the surface or near the bottom of the tank. Older larvae started swimming horizontally and vertically, spending most of the time in the upper half of the tank until 25 dph and near the bottom half of the tank at ages >25 dph. Prior to an age of 25 dph, larvae paused frequently to search for prey while from approximately 25 dph on wards, larvae Mar Ecol Prog Ser 559: 131–145, 2016 136 Period Temp. (°C) Regression equation No. of ind. Confidence interval (95% confidence bounds) L0 α r 3−50 dph 15 386 3.52−4.02 0.064−0.086 0.035−0.047 17 403 3.30−4.03 0.074−0.101 0.044−0.058 TL=3.68e 0.074 0.041 1e0.041t () () −− TL 3.86e 0.081 0.048 1e0.048t = () () −− Table 3. Relationship between larval size (total length, mm) and age (d post hatching, dph) for Sardina pilchardus larvae reared at 15 and 17°C from 3 to 50 dph adjusted to a Laird-Gompertz growth curve. t: age in dph; L0: length at 3 dph; αand r: coefficient of the Laird-Gompertz growth curve (Eq. 2) Fig. 3. Cumulative survival rate of Sardina pilchardus larvae reared with high concentration diet and under 3 different temperatures using the Kaplan-Meier method. Survival data not shown after 15 d post-hatching (dph) as subsequent changes were negligible (<3%). Red arrows represent the time period when the mouth opens (MO) and exogenous feeding begins started searching for prey while swimming. Time spent swimming by pre-flexion sardine larvae in - creased significantly with age, and its interaction with temperature was also significant (GLM AIC = 6995; p < 0.0001) (Tables 5 & 6, Fig. 4). The rate of increase in time spent swimming was higher for larvae reared at 17°C when compared to those reared at 13 and 15°C. The ANOVA testing for differences between the regression models with and without the interaction term of temperature and age for larvae reared at 13 and 15°C were not significant (F = 0.846; p = 0.357), suggesting that the rate of increase in swimming with pre-flexion larvae age was similar for larvae reared at these 2 temperatures. However, the intercept was significantly different between the two (F = 393; p < 0.0001; R2 = 0.39); therefore, the time spent swimming at age was higher at 15°C. At the start of exogenous feeding (3 to 5 dph), foraging activity was low; the (mean ± SD) frequency of orientation was 0.19 ± 0.54, 0.79 ± 1.36, and 1.75 ± 1.99 prey min−1 for larvae reared at 13, 15 and 17°C, respectively. The frequency of attacks (lunges) for first-feeding larvae was 0.09 ± 0.31, 0.40 ± 0.95 and 0.81 ± 1.24 attacks min−1 at 13, 15 and 17°C, respectively (Fig. 4). Foraging activity significantly in - creased with increasing age for pre-flexion larvae, and temperature significantly affected the slope of the increase in the frequency of larval orientations or fixations to prey (AIC = 8573; p < 0.0001), attacks on prey (AIC = 6995; p < 0.0001) and percentage of successful attacks on prey (AIC = 16776; p < 0.0001). The slope of the relationship between these 3 foraging behaviours with age increased with increasing temperature, with 13 < 15 < 17°C, except for the slope of the percentage of successful attacks to prey, which was significantly higher at 15 than at 17ºC (Tables 5 & 6). After flexion and dorsal fin began to develop (25 dph for 13 and 15°C, 20 dph for 17°C, Table 4), time spent swimming was significantly different according to the temperatures with which larvae were reared (ANOVA, F= 8.2; p = 0.0005) and was particularly high at 17°C and similar between 13 and 15°C (p-values of Tukey-test pairwise comparisons of 0.57, 0.001 and 0.02 for 13−15, 13−17 and 15−17°C, res pec - tively). Orientations towards prey as well as at tacks on prey were also significantly different be tween temperatures during the flexion stage (ANOVA tests: F= 12.1 and 4.6; p < 0.0001 and 0.012, for com - parisons of orientations and attacks, respectively). Orientations to prey at this stage were lower at 13°C and similar between 15 and 17°C (p-values of Tukeytest pairwise comparisons of 0.001, <0.0001 and 0.238 for 13−15, 13−17 and 15−17°C, respectively). Attacks on prey were also lower at 13°C and similar between 15 and 17°C (p-values of Tukey-test pairwise comparisons of 0.002, 0.015 and 0.990 for 13−15, 13−17 and 15−17°C, respectively). The successful attacks on prey were not significantly different between the 3 temperatures (ANOVA; F= 0.99; p = 0.376). Larvae reached 50% of their maximal value of time spent swimming at an age of 5.3 and 14.7 dph at 17 and 15°C, respectively (Fig. 4, Table 7). At approximately 15 and 20 dph for 17 and 15°C, respectively, sardine larvae spend all the observational time swimming, presenting a constant swimming behaviour. Garrido et al.: Effect of temperature on sardine larval growth 137 Temp. (°C) Age (dph) Mean size ± SD (mm) Hatching day, no pigmentation 13 0 3.6±0.64 15 0 3.8±0.27 17 0 4.1±0.34 Pigmentation, yolk sac, mouth open 13 4/5 5.5±0.63 15 3/4 5.4±0.41 17 3 5.0±0.49 Pectoral fins and beginning of caudal fin formation 13 10 5.5±0.60 15 10 5.9±0.60 17 5 5.5±0.41 Beginning of notochord flexion 13 20 8.1±1.32 15 15 7.7±1.29 17 10 6.3±0.92 Beginning of dorsal fin development 13 20 8.1±1.32 15 20 10.0±1.41 17 15 9.7±1.26 Notochord flexion complete 13 − − 15 25 13.1±1.52 17 20 10.8±1.25 Dorsal fin complete 13 − − 15 30 13.2±1.25 17 25 12.1±1.82 Beginning of anal fin development 13 − − 15 35 14.5±2.29 17 25 12.1±1.82 Caudal fin complete 13 − − 15 40 16.2±2.49 17 30 13.7±2.44 Table 4. Timing of the main morphological events occurring during larval development of sardine Sardina pilchardus at 13, 15 and 17°C. Pigmentation refers to the appearance of many small scattered melanophores in the dorsal region extending from the head to the tail, with one ventral caudal melanophore. Pectoral fins refer to the time when the insertion of the pectoral fins become oblique. Caudal fin complete refers to the time when the caudal fin has the complete number of fin rays. dph: d post hatching; −: no data Mar Ecol Prog Ser 559: 131–145, 2016 138 Developmental stage Relation Temp. (°C) a(±SE) b(±SE) p R2 Pre-flexion SWI vs. Age 13 0.029±0.002 0.15±0.024 <0.001 0.30 15 0.027±0.001 0.30±0.020 <0.001 0.37 17 0.038±0.01 0.31±0.001 <0.001 0.47 Orient. vs. Age 13 0.09±0.011 0.28±0.159 <0.001 0.11 15 0.28±0.017 −0.67±0.277 <0.001 0.27 17 0.50±0.034 −0.26±0.417 <0.001 0.32 Attacks vs. Age 13 0.02±0.005 0.12±0.069 <0.001 0.05 15 0.17±0.013 −0.77±0.207 <0.001 0.20 17 0.24±0.034 −0.29±0.417 <0.001 0.27 Success attacks vs. Age 13 0.58±0.235 14.15±3.217 0.01 0.01 15 1.73±0.013 10.15±0.207 <0.001 0.11 17 1.34±0.283 24.72±3.445 <0.001 0.05 Post-flexion SWI vs. Age 15 0.002±0.0004 0.88±0.015 <0.001 0.06 17 −0.0001±9 ×10−5 0.99±0.0032 0.196 0.001 Orient. vs. Age 15 0.35±0.031 −1.20±1.194 <0.001 0.16 17 0.08±0.038 12.66±1.319 0.033 0.01 Attacks vs. Age 15 0.36±0.028 −5.07±1.051 <0.001 0.204 17 0.12±0.033 6.63±1.154 <0.001 0.023 Success attacks vs. Age 15 1.30±0.124 14.83±4.682 <0.001 0.145 17 0.54±0.033 46.61±1.154 <0.001 0.03 Caudal fin formation SWI vs. Age 15 −0.0001±0.006 1.00±0.316 0.09 0.01 17 0.0002±0.012 0.98±0.513 0.302 0.0001 Orient. vs. Age 15 0.35±0.145 −1.67±6.637 0.014 0.021 17 0.001±0.079 16.14±3.160 0.982 −0.002 Attacks vs. Age 15 0.326±0.133 −3.67±6.115 0.015 0.020 17 −0.06±0.069 14.77±2.774 0.352 −0.003 Success attacks vs. Age 15 0.91±0.447 31.85±21.843 0.056 0.011 17 0.06±0.246 66.71±9.807 0.782 −0.002 Table 5. Results of the linear regression analyses (y = ax + b) of the effect of age (d post hatching, dph) on the time spent swimming (SWI,%), frequency of orientations (Orient.), frequency of attacks and frequency of successful attacks on prey of Sardina pilchardus larvae during 3 developmental stages: pre-flexion (from 3 to 25 dph at 13 and 15°C and from 3 to 20 at 17°C) and post-flexion (25 to 50 dph at 15°C and 20 to 50 dph at 17°C) and after caudal fin formation (40 dph at 15°C and 50 dph at 17°C) Developmental stage MAPS Temperatures (°C) df p-value F ρ R2 Intercept Slope Pre-flexion SWI 13 and 15 1 <0.001 0.358 262 <0.001 0.39 13 and 17 1 <0.001 <0.001 270 <0.001 0.44 15 and 17 1 <0.001 <0.001 270 <0.001 0.42 Orientations 13 and 15 1 0.004 <0.001 190 <0.001 0.31 13 and 17 1 0.193 <0.001 270 <0.001 0.44 15 and 17 1 0.390 <0.001 183 <0.001 0.33 Attacks 13 and 15 1 <0.001 <0.001 140 <0.001 0.25 13 and 17 1 0.047 <0.001 229 <0.001 0.41 15 and 17 1 0.130 0.007 115 <0.001 0.23 Success 13 and 15 1 0.366 <0.001 44 <0.001 0.09 13 and 17 1 0.026 0.042 34 <0.001 0.09 15 and 17 1 0.262 0.001 36 <0.001 0.09 Post-flexion SWI 15 and 17 1 <0.001 <0.001 26 <0.001 0.06 Orientations 15 and 17 1 <0.001 <0.001 59 <0.001 0.13 Attacks 15 and 17 1 <0.001 <0.001 70 <0.001 0.15 Success 15 and 17 1 <0.001 <0.001 42 <0.001 0.09 Caudal fin formation SWI 15 and 17 1 0.205 0.310 5 <0.001 0.02 Orientations 15 and 17 1 0.027 0.047 3 0.017 0.01 Attacks 15 and 17 1 0.010 0.014 3 0.033 0.01 Success 15 and 17 1 0.175 0.136 2 0.058 0.007 Table 6. Results of ANCOVA models analysing the relationship between foraging behaviour (time spent swimming [SWI,%], frequency of orientations, frequency of attacks and frequency of successful attacks on prey) and Age (d post hatching, dph) of Sardina pilchardus larvae during 3 developmental stages: pre-flexion (from 3 to 25 dph at 13 and 15°C and from 3 to 20 dph at 17°C) and post-flexion (25 to 50 dph at 15°C and 20 to 50 dph at 17°C) and after caudal fin formation (40 dph at 15°C and 50 dph at 17°C). MAPS: modal action patterns Garrido et al.: Effect of temperature on sardine larval growth 139 Fig. 4. Mean ± 0.95 confidence interval bars of the (a,b,c) time spent swimming, (d,e,f) frequency of fixations or orientations, (g,h,i) frequency of lunges during a 60 s period of observation throughout larval ontogeny of Sardina pilchardus and (j,k,l) percentage of successful attacks on prey of larvae reared with high prey concentrations and at 3 different temperatures: 13, 15 and 17°C. Red arrows represent the timing when notochord flexion is complete. Parameters of the linear and logistic regressions are provided in Tables 5 & 7