Limiting motorboat noise on coral reefs boosts fish reproductive success

Permits and ethics approval

Animal collections and all experimental procedures were conducted with ethical approval from the University of Exeter (2013/247), James Cook University (A2361) and Lizard Island Research Station, permission from the Great Barrier Reef Marine Park Authority (GBRMPA) (G17-39752.1) and under licence from the Australian Government Department of Fisheries (170251).

Study species

The spiny chromis (Acanthochromis polyacanthus) is a damselfish that exhibits bi-parental care of eggs and juveniles at nests within shallow reef habitat in the tropical Western Pacific³⁹ (Fig. 4). Spiny chromis are planktivores on the Great Barrier Reef, importing nutrients from the plankton to the reef⁴⁰. As such, their preferred habitat is the reef edge (within ~7 m). Most pairs raise one clutch per season—in our study, second clutches occurred in only 7% of the population—so we tracked first clutches from adult pairs. Adults enhance their reproductive success by fanning eggs to oxygenate them, and chasing away potential predators from eggs and juveniles^16,41. Adults also allow their offspring to eat some of their body mucus in a behaviour known as ‘glancing’ that has potential nutritional and/or immunological benefits⁴².

Field study

We conducted the field study at Lizard Island Research Station (LIRS) (14° 40′ S, 145° 28′ E), Great Barrier Reef, Australia over an entire spiny chromis breeding season (90 days from 23 October 2017 to 20 January 2018).

Sites and nests

We selected six coral reef edge sites (113–218 m in length) within the lagoon on the south of Lizard Island (Fig. 5). Water temperature ranged from ~26 °C at the start to ~29 °C at the end of the season. The reefs were composed mainly of a mixture of live and dead coral. Prevailing currents were wind-driven from south to north. Three of the sites were exposed to ‘busy boating’ while boating was limited at the other three. Treatments were allocated to sites partly randomly and partly allowing for ease and safety of motorboat access. Nest positions within sites did not differ between treatments in: depth of bottom next to reef (1–5 m at mid-tide with a tidal range of ~2 m, N = 67 measurements at a range of tidal heights, mean ± SE depth = 2.3 ± 0.1 m; LMM, sound treatment: Χ²₁ = 0.82, p = 0.365, random effect of site: variance = 0.14, standard deviation=0.38); nest height above the sand (range = 0–4 m, mean ± SE = 0.6 ± 0.1 m; sound treatment: Χ²₁ = 0.52, p = 0.470, site: variance=0.33, standard deviation = 0.57); or distance from the edge of the reef (range = 0–6.6 m, mean ± SE = 1.2 ± 0.2 m; t-test: t_33,32 = 1.61, p = 0.112; a linear model did not fit the data). A small number of broods (mean ± SE per site = 4.7 ± 1.0) were already present at each site at the start of the season; spiny chromis occasionally breed outside the main breeding season. These were marked and ignored for the purpose of the experiment. The mean ± SD number of nesting pairs identified at each site at the start of the season was 9.5 ± 2.0 for limited-boating sites and 10.7 ± 3.2 for busy-boating sites. Nesting pairs did not always form clear, stable pairs with obvious territories and so pairs without broods were not tracked through the experiment. The mean ± SD total number of adults at each site at the start of the experiment was 185 ± 92 for limited-boating sites and 119 ± 20 for busy-boating sites.

Motorboat traffic exposure and protection

There is a navigable channel through the lagoon where the experiment was conducted. Fishing boats, tourist boats and research station boats pass through the channel, but the main source of traffic is research station boats. We chose six sites along the navigable route and randomly allocated these to treatments. Following random allocation, two sites were switched for safety reasons for motorboat drivers (Fig. 5). We experimentally elevated motorboat noise at three of the six sites (busy-boating treatment) to mimic typical traffic around a port, harbour or regularly visited reef. At these sites, we drove eight different 5 m aluminium motorboats with 40 hp Suzuki four-stroke outboard engines repeatedly along the length of the site within 10–30 m of the edge of the reef. Busy-boating sites received an average of 180 motorboat passes each day during 3–6 ‘exposure periods’ lasting 15–20 min each; this totalled 1.25–1.5 h per day of traffic noise at each busy-boating site. The other three sites were protected from motorboat traffic (limited-boating treatment), by marking these reefs on the research station map as areas to avoid by at least 100 m and monitoring activity in the lagoon daily. When experimenters needed to access protected sites, speed was reduced to that where no wake was created (roughly ¼ throttle) within 100 m and boats were anchored 20 m from the reef. See Supplementary Information for further details of motorboat traffic exposure and protection.

Acoustic recordings and analysis

We made acoustic recordings of both pressure and particle motion at three locations within each site using an accelerometer with integrated hydrophone (M20-040 manufactured and calibrated by Geospectrum Techologies Inc. Dartmouth, Canada; sensitivity follows a curve from 0 to 5000 Hz) and a digital recorder (Zoom F4, Zoom Corporation, Tokyo, Japan; calibrated using pure sine waves measured with an oscilloscope). See Fig. 6 and Supplementary Information for further details of acoustic recordings, analysis and results.

Breeding and reproductive success

Each site was checked by snorkellers every other day to monitor breeding by spiny chromis pairs. Nests with new hatchlings were marked with flagging tape and continued to be checked every other day throughout the season to monitor reproductive success. The day in the season that broods hatched was used to test for an effect of motorboat exposure on the timing of breeding using a linear mixed-effects model (fitted in R) with site as a random effect. The proportion of nests retaining offspring at the end of the season in the two treatments was compared using a Chi-squared test.

Brood size

We counted the number of offspring in broods within four days after hatching at a subset of 59 nests; 32 in the three busy-boating sites (N = 9, 11, 12) and 27 in the three limited-boating sites (N = 5, 10, 12). Average clutch size at hatching in the wild was 126 ± 16 (mean ± SE). Some of these nests were part of the predator presence observations (details below), some were part of the size monitoring (details below) and some were independent. We counted the number of offspring in three photos and used the highest number for analyses. We tested for an effect of motorboat treatment on brood size at hatching using a Welch’s t-test.

Predator presence around nests

We determined baseline predatory threat (counts of heterospecific piscivores, potential predators of juveniles) using video camera (GoPro 5) deployments. Thirty-three nests (not studied for size) were videoed once or twice between 1 and 11 days post-hatching; 18 in the three busy-boating sites (N = 7, 6, 5) and 15 in the three limited-boating sites (N = 6, 5, 4). A total of 55 videos were analysed. A camera stand was placed at each nest on the first or second day post-hatching and remained in place 2–3 m from the nest. For each survey, after GoPro cameras were attached to camera stands, several minutes settling time was allowed (mean ± sd: 827 ± 35 s), followed by a 30-s recording of predator presence in the absence of any motorboats. All nests that were videoed were at least 10 m away from one another (parents spend most of their time within 2 m of the nest). Videos were randomly named and analysed by KEC without sound (to remain ‘blind’ to treatment) using BORIS 7.6.1⁴³. Spiny chromis offspring are small and vulnerable to any piscivore on the reef and parents defend their offspring by chasing potential predators. The number of heterospecific piscivores at each nest was surveyed (conspecifics are known to cannibalise offspring, but this is very rare¹⁶). A negative binomial regression parameterised such that the variance is a quadratic function of the mean was fitted using glmmTMB in R with piscivore counts as the response variable, motorboat treatment and days into the season as potentially interacting fixed factors, and nest and site as random effects.

Juvenile survival

It was not possible to observe the eggs as they were laid in caves, so we studied juveniles from when they could be observed above the substrate (shortly after hatching – this species completes the larval stage inside the egg and hatches at the juvenile stage¹⁶). We also counted the number of surviving offspring at the subset of 59 nests every 4–8 days. When all juveniles from a nest could be captured in a frame, we used three photos per time point and used the highest number. As juveniles aged, they used more space and could not be captured in a single photo; then, they were counted by snorkellers with experience in fish surveys (SLN and IKD). The reliability of snorkellers’ counts was tested against one another and did not differ when there were <20 juveniles and were accurate to the nearest 5 when there were >20 juveniles based on counts of 43 nests. Usually, however, when there were >20 juveniles at a nest, they were at an earlier developmental stage and counts could be taken from photos. Survival of juveniles was recorded as number of days from hatching until they were no longer seen at the nest. Where all offspring from a nest were apparently lost to predation (mean survival time = 21 days), the nest continued to be monitored every other day for the remainder of the season to ensure offspring had not temporarily disappeared and to check for second clutches. A Cox proportional-hazards survival model was fitted in R with motorboat treatment and initial hatching count as fixed effects, and nest and site as random effects. We discounted three nests where counts increased due to assumed experimenter error or immigration. The package Coxme was used in R to test for the interaction between treatment and start count (hatch day was not included in the model as there was no indication of an effect of treatment or site on hatch day), with nest and site as random effects. The package coxph was used to create Fig. 1B, which does not account for the random effects, but is used for illustrative purposes.

Juvenile size

We monitored juvenile size at a subset of 22 nests; 11 in the three busy-boating sites (N = 3, 4, 4) and 11 in the three limited-boating sites (N = 3, 4, 4). Up to 10 juveniles (depending on catch success) were caught by snorkellers or SCUBA divers using hand nets from each nest each week. A total of 275 juveniles between 1 and 53 days post-hatching were sampled. Juveniles were transported to the field station in bags of fresh seawater. We measured standard length either under the microscope at 10× magnification or with a Vernier caliper, depending on fish size. All nests within a site were sampled on the same day and each site was visited for juvenile size sampling each week. Data were log-transformed and analysed using a linear mixed-effects model (fitted in R), with age and motorboat treatment as fixed effects, and nest and site as random effects.

Laboratory study

We conducted the laboratory study in the Marine and Aquaculture Research Facilities Unit (MARFU) at James Cook University, Townsville, Australia from March to July 2018. Spiny chromis adults were caught with fine monofilament barrier nets and hand nets from the section of reef on the lagoon side of Palfrey Island within the lagoon around Lizard Island in the northern Great Barrier Reef (14° 41′ S, 145° 27′ E) during November 2016. All adults would have experienced equivalent prior noise exposure. The mean ± SE standard length of adults was 10.7 ± 0.1 cm. Spiny chromis were randomly allocated to treatments and were housed in 30 male–female pairs, and maintained at a mean ± SE temperature of 27.7 ± 0.1 °C in the presence of either a busy-boating treatment (playback of four of the five recorded motorboats in a pattern matching exposures in the field) or a no-boating treatment (playback of ambient reef sound). We kept most of each brood with the parents to measure survival (cannibalism can rarely occur in this species under stress¹⁶) and isolated 50 individuals per brood as a single group in a separate tank (where parents could not compete with offspring for food) with the same playback treatment to measure size. See Supplementary Information for further details of tank setup and conditions, acoustic exposure regime, playback construction, acoustic recording analysis and results for the tanks.

Breeding

We checked all adult tanks daily after lights were switched on, but before playbacks began, for the presence of a newly laid clutch. The number of days since the start of the treatment that clutches were laid was used to test for an effect of motorboat noise exposure on the timing of breeding using a two-sample Welch’s t-test.

Clutch size and brood size

We photographed newly laid clutches and estimated clutch size by counting the number of eggs in a square on an overlaid grid and counting the number of grid squares containing eggs. All adult tanks were checked daily after lights were switched on, but before playbacks began, for the presence of a newly hatched brood. Clutch size and brood size were compared between treatments using two-sample Welch’s t-tests.

Egg characteristics and embryonic development

We monitored clutch-level and individual-level egg and embryo characteristics at days 1 and 10 during the egg phase of the first clutch laid by each breeding pair. Measures taken were: (1) egg area, (2) yolk sac area, (3) dorsal spine length (day 10 embryos only), and (4) dry weight (at 10 days). Egg area, yolk sac area and spine length were obtained by measuring 10 randomly sampled individuals per clutch under a light microscope (Olympus SZXY). For dry weights, fish were dried in an oven for >24 h at 60 °C and weighed on a Mettler microbalance with ±0.001 mg accuracy. The number of days between laying and hatching was used as the embryonic developmental time. Linear mixed-effects models (fitted in R) were used, with clutch as a random effect and motorboat treatment as a fixed effect.

Parental care of embryos

We filmed parental activity (distance moved by both parents) and time spent fanning eggs at day 10 of the egg phase during periods of playback. Two cameras were used: a Logitech HD Webcam C615 camera positioned 45 cm above the tank looking down with the entire tank in the field of view (‘top camera’), and a GoPro HERO 5 positioned inside the tank in front and looking into the nest (‘side camera’). Following a 30-minute settling time, minimising the disturbance to the fish, baseline behaviour was observed for five minutes. Then, in ‘busy-boating’ tanks, motorboat noise was played for a further five minutes, while in ‘no-boating’ tanks, a different ambient track was played for five minutes.

Two key nest-caring parental behaviours were identified for analysis:

1. (1)

   Activity – the distance travelled by parents. Average distance travelled within each breeding pair was calculated from the total distance travelled by both the male and female. Activity was observed from the top camera. Distance was calibrated by measuring a known distance on the bottom of the tank, present in all videos. The distance travelled was calculated by marking the position of the fish every second using the manual tracking feature of ImageJ version 1.52d (https://imagej.nih.gov/ij/download.html).

2. (2)

   Fanning – the amount of time parents spent fanning the clutch of eggs. Fanning was observed from the side camera and analysed using Solomon Coder software (https://solomoncoder.com/download.php).

T-tests were used to test for effects of treatment on parental care behaviour.

Juvenile survival

We counted the number of juveniles that survived with their parents at day 21 post-hatching (maximum count from three photos for each tank). Survival was measured at day 21 because that was the mean survival time in the field; also, by day 42 post-hatching, most parents had produced a second clutch which confounded observations of survival. The fish removed at hatching were included in the final count by modelling their survival as equal to that of the rest of the brood. Survival was converted to a percentage from the number of eggs laid in the clutch. This measure of survival is conservative compared with that expected in the field because the only potential predators of the juveniles in the tanks were their parents. Percentage survival rates were non-normally distributed and so were compared between treatments using a Wilcoxon signed-ranks test.

Juvenile size

We measured the standard lengths (from photos using ImageJ) and dry weights of ten juveniles per clutch at day 21 post-hatching and of eight juveniles per clutch at day 42 post-hatching, following humane sacrifice; fish were dried in an oven for >24 h at 60 °C and weighed on a Mettler microbalance with 0.001 mg accuracy. We measured length and weight from juveniles that were isolated from the parents to avoid the possibility that the parents could compete with the juveniles for food. Dry weights at day 21 and 42 were compared between treatments using an LMM with clutch as a random effect.

General statistical approaches

For measures of parental care, or at the level of clutch, t-tests or Wilcoxon signed ranks tests were used. Where we measured several individuals from within multiple sites, clutches or broods, LMMs or GLMMs were used to control for the random effects of site, clutch or brood, provided models fit the data satisfactorily. Plots of residuals vs fitted values were examined to check model fit and where models did not fit the data, standard tests such as t-test/Wilcoxon signed ranks were used in their place. Any effects among nests (such as slight variations in water flow) were therefore controlled for by the LMM or GLMM statistical models. The variance attributed to, and standard deviation of the variance for, random effects are presented as part of the full output from models. We used the same approach for model selection as in¹³. To establish the best-fitting model, terms were eliminated one by one from a maximal model. Simplified models were compared with more complex ones using maximum likelihood ratio tests that employ chi-square statistics to establish whether a simpler model performed significantly worse at explaining the data than a more complex model. If the simpler model was not significantly worse when a term was removed, the simpler model was deemed better and thus the removed term was dropped. If the simpler model was significantly worse, the term was maintained in the model⁴⁴. The degrees of freedom from maximum likelihood tests presented in the Results of the main paper are the difference between the degrees of freedom of the simpler and the more complex models. All potential interactions of fixed effects were examined and are only presented where their exclusion from the model made the model significantly worse at explaining the data at the significance level p < 0.10. When an interaction was presented, the effect of the main effects was established by comparing a model with the main effect in question, although not its interaction, with a model that dropped that effect. The package ‘lme4’ was used for LMMs. Effect sizes and standard errors given are the results from the model summaries within the R code packages used to create models. Full model outputs are presented in Table S1.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Source: Ecology - nature.com