Triplefins (family Tripterygiidae), particularly Helcogrammoides chilensis, are small-bodied, cryptobenthic fish characterized by a bipartite life cycle. They spawn benthic eggs and show parental care up until a planktonic larva with pigmented eyes, small yolk sacs and open mouth hatches (Ruck 1973, 1980). Their pelagic larval duration is approximately 2 to 4 months (Mansur et al. 2014), after which they settle as juveniles in the rocky exposed shores (McDermontt and Shima 2006).

Helcogrammoides chilensis inhabits the coast of Chile from 20°18′S to 36°45′S in the southern Pacific, in shallow waters near the coast (Williams and Springer 2001, Cancino et al. 2010). It has carnivorous habits, with a diet mostly composed of copepods and amphipods (Muñoz and Ojeda 1998). Its pelagic larval stages are found throughout the year in nearshore waters (Hernández-Miranda et al. 2003, Landaeta et al. 2009, 2015), with maximum abundances between October and April (Palacios-Fuentes et al. 2014). H. chilensis larvae have a long pelagic larval duration (78-113 days; Mansur et al. 2014) given their slow growth rates (~0.15 mm day^-1; Palacios-Fuentes et al. 2014).

H. chilensis can be considered a model species due to its life history traits, which are similar to those of other cryptobenthic species that inhabit rocky or reef shores in temperate waters (e.g. benthic eggs, parental care, carnivorous, pelagic larvae and slow growth) (Caie and Shima 2019). Recently, Tripterygiids have been considered part of a group of cryptobenthic reef fishes (species shorter than 5 cm in length) which fuel reef trophodynamics, owing to their unique larval dynamics (high abundance, rapid growth and extreme mortality) (Brandl et al. 2019).

The subtidal rocky reefs of the semi-protected area of Montemar, Valparaíso Bay, central Chile (32°57′S, 77°33′W) (Fig. 1A), are characterized by their shallow depth (~1 m during neap tide, ~3 m during spring tides) and moderate swell. These reefs are subject to seasonal and diurnal oceanographic changes generated by events such as coastal storms that occur in winter, sea breeze in summer and daily tidal variation. The region is dominated by a semi-diurnal tidal regime that explains 98% of the sea level variability (Díaz-Astudillo et al. 2017). Sea level in the very nearshore is dominated by both tides and surge waves, but there is no information about sea level variability in intertidal or subtidal pools of central Chile. The seafloor is mostly lithic, composed of coarse sand and eroded shells. These reefs are the habitat of small and medium-size cryptobenthic fish species (Stepien 1990), housing large abundances of H. chilensis which settle preferentially on lithic substratum (Pérez-Matus et al. 2016).

During the late austral summer and early autumn, larval fish aggregations from intertidal rocky reefs nearby Montemar (Fig. 1A) were video-recorded (1080 p resolution) with a GoPro Hero3+ Black Edition camera (2014, 32 recordings from 19 February to 16 April) and a GoPro Hero4+ Silver Edition camera (2016, 29 recordings from 3 January to 15 March). The larvae were recorded by three observers (PPF, MFL, MAR) each day between 10:00 and 14:00 h, a similar time frame to that used in previous studies (e.g. Leis et al. 2014) and suitable considering our study’s aim to determine inter-daily behavioural responses.

Larval fish aggregations were recorded as long as they could be followed underwater and at an appropriate distance to avoid major disturbances in fish behaviour (~20 cm). In situ observations of Helcogrammoides larvae, juveniles and adults show that the presence of the diver does not generate significant differences in fish pool abundances (Javier Polanco, pers. comm.). Recordings were not available for weekends or days of extremely high tides or dangerous surges (2014: 3 days; 2016: 6 days).

Haphazard individuals were captured periodically with hand nets (10×13 cm, 2 mm mesh opening), then sacrificed in 1 mL of benzocaine BZ-20 (20 mg per 100 mL) dissolved in 1 L of seawater and immediately preserved in 96% ethanol. Then individuals were observed and photographed with a camera Moticam 2500 (5 Mpx) attached to an Olympus ZS-61 stereomicroscope (Olympus Corporation, Shinjukuku, Tokyo, Japan), and later identified as the triplefin Helcogrammoides chilensis (Cancino, 1960) according to Pérez (1979).

The length of each video segment in which larvae appeared was determined to estimate sampling effort in seconds (s). Each video was examined to count the number of fish larvae in the aggregations per frame every 10 s (higher frequency than proposed by Hoare et al. 2004 to obtain a finer temporal scale). The final abundance was considered as the maximum observed value. A second revision was conducted to establish the time (in seconds) that the aggregations of fish larvae spent in schooling (i.e. when individuals were grouped, swimming polarized and coordinated in the same direction), shoaling (i.e. when individuals were grouped and stayed together as a whole, but with each larva swimming in its own direction), or solitary behaviour (i.e. when a single larva was found per frame) (Fig. 1B-D). This time was then standardized using the sampling effort (s) to obtain comparable measurements. If there was more than one video segment containing larvae per day, the information was pooled together so that each day contained only one data value. Days without larval aggregations were not considered for further analysis.

Meteorological and oceanographic data acquisition

Wind speed and direction were measured every 5 minutes from 6 January to 20 April 2014 and from 3 January to 15 March 2016 with a Davis Vantage Pro2 meteorological station installed ~20 m away from the subtidal pool of the study (Fig. 1). Wind speeds under 1 m s^–1 were assigned by the station as “calm winds”. To estimate the along-shore and cross-shore wind stress (τ), the wind vector was separated into its orthogonal components u (east-west) and v (north-south). A constant drag coefficient (C_D) of 1.3 10^–3 and a constant air density (ρ_A) of 1.2 kg m^–3 were used, following Shaffer et al. (1997). The kinetic energy dissipation rate (ε, W m^-3) was derived from wind speed as a proxy of wind-induced turbulence following the model developed by MacKenzie and Leggett (1993), using Z=2 m (mean water depth of the aggregations), C_D=1.3 10^–3, ρ_A=1.2 kg m^–3, a seawater density of ρ_w=1025 kg m^–3 and a von Karmann constant of 0.4. Wind stress components and wind-induced turbulence were then plotted through time to detect any temporal patterns that may influence the behaviour of larval fish.

In 2014 water temperature (°C) was measured directly from the intertidal pools during each sampling day using a Hanna HI9828 multiparameter probe. In 2016 a HOBO data logger anchored 50 m away from the pool was used to determine the water temperature at 3 m depth during the sampling period (Fig. 1A).

To explore relations between the observed behaviours and environmental variability, a principal component analysis (PCA, using correlation matrix) was performed using a biological matrix containing the proportion of time spent in each behaviour per date and an environmental matrix with four explanatory variables: water temperature, u- and v- wind stress and wind-induced turbulence. Spearman correlations were performed between larval behaviour proportions and the four environmental variables. Prior to analysis, proportional data were transformed using an arcsine function. Differences in environmental variables between each year were evaluated using Mann-Whitney U tests. Statistical analyses were performed using the Past 4.0 software (Hammer et al. 2001).

In 2014, larval fish observations varied between 1 and 148 individuals (mean 37.82±41.09) (Fig. 2A), with two main peaks of +60 individuals, the first on 28 February and the second during the first fortnight of April. In 2016, abundances ranged between 1 and 70 individuals (28.52±15.69) (Fig. 2B), with only one peak which started in mid-January and lasted until early February. Larval abundances were more variable in 2014 than in 2016 (Levene test, F=1.63, P<0.01).

All types of behaviour were observed. In 2014 (Fig. 2C), schooling behaviour was dominant from 19 February to 11 March, later changing into a solitary condition until 26 March, followed by predominate shoaling behaviour until 16 April. In contrast, in 2016 (Fig. 2D) a dominant schooling period was observed between 5 January and 10 February, later shifting into solitary behaviour until 9 March.

In both years, the majority of individuals observed in solitary behaviour were large (~3.7 cm) and had highly pigmented bodies, corresponding to juveniles of H. chilensis.

In 2014 water temperatures decreased gradually over the course of the study period, fluctuating between a maximum of 17.12°C on 5 February and a minimum of 11.98°C on 3 March (mean 14.23±1.26°C) (Fig. 3A). Strong fluctuations of about 4°C over periods of less than three days during the first week of February, the first days of March and the second week of April can be associated with strong southwesterly wind events (Fig. 3A). Wind stress in the v component (maximum v-stress of 0.08 N m^–2, minimum of –0.11 N m^–2) showed lower variation than the u component (maximum u-stress of 0.02 N m^–2, minimum of –0.17 N m^–2), illustrating the high variability of cross-shelf winds and the lower kinetic energy of meridional winds. Westerly winds were more intense than easterly winds during summer. Turbulent dissipation rates showed a similar pattern to wind stress, fluctuating between 0 and 0.004 W m^-3 (mean of 7.9 10^–5±0.0002 W m^–3), with high turbulence events during the first week of February and the second week of April.

In 2016 water temperatures fluctuated between a minimum of 14°C on 22 February and a maximum of 18.9°C on 22 January, with a mean of 16.8±1.17°C. A gradual decrease of the temperature was observed, given that higher temperatures were observed during late January and lower ones in mid-March (Fig. 3B). Winds were predominantly from the southwest and highly variable in intensity, with temporary windows of up to 8 days of calm winds (<1 m s^-1) between events of intense winds. Similar values of wind stress were observed in both components, varying between –0.13 N m^–2 and 0.04 N m^–2 in the east-west component and between –0.13 N m^–2 and 0.06 N m^–2 in the north-south component (Fig. 3B). Two particularly strong southwest wind events were observed in mid-January and late February, causing abrupt decreases in water temperatures. Turbulent dissipation rates were generally low, except during windows of strong winds in mid-January, late February and late April. They varied between 0 and 0.008 W m^–3, with a mean of 7.5 10^–5±0.0002 W m^–3.

Water temperatures were significantly higher in 2016 than in 2014 (Mann-Whitney U=26, P<0.001), while no differences were found in u-wind stress (Mann-Whitney U=353, P=0.322) and v-wind stress (Mann-Whitney U=326, P=0.161) between the years. Nonetheless, wind-induced turbulence was significantly higher in 2014 (Mann-Whitney U=247.5, P=0.004).

The PCA for 2014 explained 65.6% of the variability in the first three axes (Fig. 4A), with the first accounting for 26.7% of the variance, which mainly showed a positive (negative) correlation between u-wind stress and wind-induced turbulence with solitary (shoaling) behaviour. The second axis (22.6% of the variance) illustrated a positive relationship between the water temperature and schooling behaviour, while the third axis explained 16.3% of the variance and showed a positive (negative) correlation between v-wind stress and solitary (schooling) behaviour (Table 1). In 2016, the first three axes of the PCA explained 83.5% of the variability (Fig. 4B). Axis 1 (38.4%) showed a positive correlation between wind stress (in both components) and wind-induced turbulence with solitary behaviour, and a negative correlation between water temperatures and solitary individuals. Axis 2 (28.6%) showed a positive (negative) relation between u- and v-wind stress and schooling (solitary) behaviour. Axis 3 (16.4%) showed a positive correlation between u-wind stress and shoaling behaviour (Table 1).

The Spearman correlations showed that in 2014 wind-induced turbulence showed a significant correlation with solitary (positive correlation, ρ=0.45, P<0.05) and shoaling (negative correlation, ρ=–0.44, P<0.01) behaviour, meaning that when turbulence was high the proportion of solitary fish increased while aggregations in the form of shoals decreased. In 2016 water temperatures were correlated negatively with solitary behaviour (ρ=–0.51, P<0.05) and positively with shoaling behaviour (ρ=0.43, P<0.05). Shoaling was also correlated negatively with wind stress in both the u- (ρ=–0.45, P<0.05) and v- (ρ=–0.41, P<0.05) components. (Table 2).