Behavioural and oceanographic isolation of an island-based jellyfish (Copula sivickisi, Class Cubozoa) population

Mapping the population

The geographic extent of the Copula sivickisi population inhabiting the east coast of Magnetic Island was mapped in the 2017 medusae season (September to November) with underwater Jellyfish Camera units (JCams; Fig. 1b;⁹). The photopositive C. sivickisi medusae were attracted to the light on each JCam and recorded by the adjacent camera in 30-min deployments. Following established methods for managing potential repeat counts²⁵, the abundance of medusae was measured by counting the maximum number of medusae in any single frame of video during the deployment period (Nmax).

The east coast is comprised of numerous adjacent bays which each contain shallow fringing reefs. The distribution of C. sivickisi medusae has been found to overlap with the islands reefal habitat bands which are abundant in Sargassum sp. algae and coral⁹. JCams were, therefore, deployed at sites on reefal habitat in the east coast bays: Picnic Bay (6 sites), Geoffrey Bay (2 sites), Alma Bay (6 sites), Alma north (6 sites), Arthur Bay (6 sites) and Florence Bay (6 sites). Further, Middle Reef lies approximately 3 km to the south west of Magnetic Island, between the island and the mainland and could potentially act as a bridge between the Magnetic Island C. sivickisi population and any mainland populations. JCams were also deployed at 6 sites on Middle Reef to explore this possibility. The coordinates of the sites were pre-determined from Google Earth satellite images to guide the night sampling. Darkened patches in the images were assumed to correspond to moderate to high reefal habitat availability (i.e., > 33% substrate coverage by Sargassum sp. and coral) following⁹, and the presence of reefal habitat at the sites was later confirmed from the JCam footage. Each bay/reef, excluding Geoffrey Bay, was sampled two times over six non-consecutive nights from the 25th of September to the 30th of October. Two sites within Geoffrey Bay, randomly placed near the C. sivickisi hotspot identified in⁹, were sampled on each of the 6 trips to confirm the presence of C. sivickisi medusae throughout the sampling period. Geoffrey Bay was sampled at a lower spatial resolution compared to the other bays/reefs (2 sites compared to 6 sites) because we could be confident in detecting medusae in Geoffrey Bay given the information derived from the comprehensive sampling of the bay in previous medusae seasons (presented in⁹). Further, a higher temporal resolution (6 trips to Geoffrey Bay compared to two for the other bays/reefs) was required to verify that C. sivickisi were present during the entire sampling period, and this had associated logistical constraints such as the weather.

Biophysical modelling

Hydrodynamic description

The two-dimensional version of the Second-generation Louvain-la-Neuve Ice-ocean Model [SLIM; 26] was used to model the currents off the eastern coast of Magnetic Island and in the surrounding region. A detailed description of SLIM and the specifics of its application in this study is presented in the supplementary information. Magnetic Island lies in the central area of the Great Barrier Reef (GBR). The currents in the GBR system are largely driven by the jets of the South Equatorial Current (SEC) which flow westward across the Coral Sea and collide with the outer reefs of the GBR²⁷. In the central GBR, the North Caledonian Jet (NCJ) from the SEC generally diverges around the Queensland Plateau before meeting the outer reefs²⁷. In addition to these regional scale forcings, the waters within the GBR system are shallow (mostly < 60 m) so winds can force currents at a local scale²⁸. The model domain extended westward into the Coral Sea (Fig. S1a). The northern boundary of the model was approximately set in the middle of the Queensland Plateau, above the latitude where the southern divergence of the NCJ typically meets the GBR. The southern extent of the domain was far to the south of Magnetic Island to avoid confounding errors from the open boundary forcing. The unstructured SLIM grid was made coarser in open water and finer near coasts and over reefs. This reduced the number of elements required to effectively capture both the regional scale currents and the small-scale currents near complex bathymetry where there was horizontal current shear. The sea surface elevation simulated in SLIM was validated against a local tide gauge and the simulated currents were validated against current meters at 4 sites at or near Magnetic Island (Figs. S1b, S3, Table S1). The hydrodynamic fields were simulated every 3 min, and saved every 15 min. This high temporal resolution was used to match the fine spatial resolution of the SLIM grid to ensure that complex coastal and reefal hydrodynamic features were effectively simulated. The dispersion of medusae was simulated by coupling the hydrodynamic output with models of C. sivickisi medusae behaviour.

C. sivickisi medusae behaviour

Two candidate models (base and dependent; Table S2) of the behaviour of C. sivickisi medusae were developed from the results of Schlaefer et al.⁹ and from data sourced from the literature^12,13,22. In both models, the behaviour of medusae changed with position in relation to reefal habitat (on/off) and with time of day (day/night; Table 1). Medusae ‘on habitat’ interacted with the habitat, while medusae ‘off habitat’ had no habitat associated behavioural cues. The C. sivickisi medusae were programmed to be nocturnal, i.e., they were inactive during the day and active at night. The positions of the model medusae were re-assigned every 3 min.

The two candidate models only differed by the on habitat, night time behaviour. In the base model, medusae on habitat at night were made to swim toward the habitat midline regardless of the current speed. A current speed dependent attachment behaviour was added to the dependent model. A dependent model medusa on habitat at night would only swim if the current speed at its position was less than a set cut off. If the current speed equalled or exceeded the cut off, the medusa would attach itself to the habitat, thereby becoming immovable.

Role of behaviour in retention

Biophysical modelling was used to investigate the role of medusae behaviour in maintaining their observed distribution in Nelly Bay and Geoffrey Bay of Magnetic Island, where they were predominantly found on shallow bands of fringing reef habitat (see⁹). Base and dependent model runs were performed. When the medusae swam in both models, their speed was set to their calculated critical swim speed (U_crit; 4.9 cm s⁻¹;⁹) i.e., the maximum swim speed C. sivickisi medusae can maintain for an extended period of time²⁹. The current speed cut off for the dependent model medusae was set to 6 cm s⁻¹ (⁹, Table S2). Medusae were additionally modelled as passive in a control scenario to determine the level of retention without behaviour.

C. sivickisi medusae were released from bands of reefal habitat in Nelly Bay and Geoffrey Bay (15 locations spaced at 200 m intervals, Fig. S1c) on a date at the end of September, near the start of the 2016 C. sivickisi season (Table S3). The runs ended after 30 days, approximately the length of one lunar cycle, allowing for the assessment of retention under spring and neap tides. No mortality was included. Five replicate runs were performed per candidate model/parameterisation. The number of medusae remaining in the bays was counted through time, with the outer limit of the Nelly and Geoffrey Bay catch zones set to the deepest cross shelf and furthest longshore boundaries of the bays reefal habitat bands (Fig. S1c). The catch zones represented the limits of the distribution of C. sivickisi medusae in the bays as measured by⁹.

Role of behaviour in retention sensitivity analysis

A sensitivity analysis was performed to examine how changing the dependent behavioural model parameterisation effected the simulated retention of C. sivickisi medusae on reefal habitat. Specifically, the retention of C. sivickisi medusae on reefal habitat in Nelly Bay and Geoffrey Bay, Magnetic Island, was re-simulated with the dependent model populated with different swim speeds and a range of current speed cut off values. The other model specifications were kept constant and matched the role of behaviour in retention analysis. Two medusae swim speeds were tested. In one set of runs the swim speed was set to U_crit⁹, as in the main analysis. There was considerable variability in the U_crit for C. sivickisi medusae calculated by⁹, U_crit = 4.9 cm s⁻¹ ± 4.4 standard deviation. Therefore, U_crit was halved in the second set of runs to more conservatively model the swimming capabilities of C. sivickisi medusae. Halving U_crit gave an estimate of the medusae’s sustainable swim speed (i.e., the maximum speed they could sustain without signs of fatigue³⁰; U_sust = 2.45 cm s⁻¹). Five different attachment to habitat current speed cut offs (3, 4.5, 6, 7.5 and 9 cm s⁻¹) were modelled to cover and extend the range of current speeds at which C. sivickisi medusae attached to the tank in the swim trials conducted by⁹ (Table S2). The two swim speeds were each combined with all cut off’s of greater speed. The cut-off speeds had to be greater otherwise the model medusae would never be active in currents faster than they could swim against, so they would physically never be advected from the virtual reefal habitat. Eight dependent model parameterisations were, therefore, tested in the sensitivity analysis (U_crit with cut offs 6, 7.5 and 9 cm s⁻¹, and U_sust with all cut offs). Note, the parameterisation with the swim speed set to U_crit and the cut off set to 6 cm s⁻¹ was the dependent model parameterisation presented in the main role of behaviour in retention analysis. The number of C. sivickisi medusae in Nelly Bay and Geoffrey Bay was counted through time as in the primary role of behaviour in retention analysis. The range of current speeds with the potential to expatriate medusae from the virtual habitat was calculated as the speed difference between the medusae swim speed and the set current speed cut-off at which medusae attached to the habitat. The effect of the size of the expatriating current range (explanatory variable) on the percentage of model medusae remaining in the bays at the end of the 30 day model runs (response variable) was tested with a regression analysis (python 3.7.4, statsmodels 0.11.0, Fig. S6).

Population structure

Additional modelling runs were performed to elucidate the spatial structure of the C. sivickisi population on Magnetic Island. The dependent model was used because the dependent model behavioural suite was more biologically realistic. This was affirmed by the results of the role of behaviour in retention analysis and the sensitivity analysis. The high retention of C. sivickisi medusae on the reefal habitat in Nelly Bay and Geoffrey Bay simulated in the dependent model runs matched the measured restriction of C. sivickisi to the bays narrow reefs⁹. Further, high within-bay retention was simulated under numerous dependent model setups, so the selection of any one of these parameterisations for use in the population structure analysis would give biologically realistic results (see results section ‘sensitivity analysis’).

The spatial and temporal extent of the modelling was scaled up for the population structure analysis, and this had associated computational costs. The dependent model was, therefore, parameterised with a medusae swim speed of U_crit, and a current speed cut off of 6 cm s⁻¹ for the population structure analysis. In addition to simulating biologically realistic retention, this parameterisation had relatively low between replicate variability. Consequently, fewer model medusae were needed to get consistency between the population structure replicates (see results section ‘population structure’), which is indicative of a robust outcome³¹.

The population structure modelling was informed by the JCam population mapping. Specifically, the mapping revealed that C. sivickisi medusae were present all along the east coast of Magnetic Island, from Picnic Bay to Florence Bay (see results section ‘Population extent’), and confirmed the presence of suitable habitat (shallow reefs with Sargassum sp. and coral) in all the sampled bays and Middle Reef. Model C. sivickisi medusae were, therefore, released across the extent of the mapped east coast distribution. Further, as part of assessing the potential for Middle Reef to act as a bridge between the island and mainland populations, medusae were also seeded from Middle Reef despite their absence from the reef in the JCam survey. Medusae were, therefore, released from 32 locations spaced every 500–600 m along the near continuous reefal habitat band which runs from Middle Reef to Florence Bay (Fig. S1d; Table S4). The movements of the model medusae were simulated over an entire medusae season (September to November 2017; Table S4). Mortality was included in the model as an exponential decay function to simulate the natural attrition of medusae (Fig. S5, Table S4). The curve approached but would never reach 100% mortality, and the small percentage of medusae remaining after 54 days (near maximum age) were killed (3.2% of released medusae). Five replicate model runs were performed.

A measure of relative connectivity was generated to assess the internal structure (i.e., levels of self-seeding and inter bay/reefal connectivity) of the Magnetic Island C. sivickisi population (aim 3). The ‘on habitat’ zones in the behavioural models were divided into 32 detection zones, around the 32 seed locations. The detection zones had an average area of 0.08 km² ± 4 × 10^–3 (range 0.04–0.13 km²). Notably, because the detection zones were approximately evenly distributed in space, larger bays/reefs had more detection zones than smaller bays/reefs. The instances of unique connections (from source seed location to sink detection zone) made by adult medusae within and between zones were counted throughout the season. Medusae were considered adults 25 days post release because C. sivickisi medusae > 5 mm in diameter (half their maximum size⁶), are generally sexually mature²⁰, and C. sivickisi medusae would take around 25 days to grow to 5 mm [unpublished data]. To generate the measure of relative connectivity, a log base 10 transformation was performed on the counts + 1 data, and the transformed data was scaled from 0 (no connections) to 1 (most connections).

The potential for connectivity between the island and mainland populations, and thereby the potential for Magnetic Island to represent an isolated stock (aims 4), was assessed by tracking the positions of all adult medusae lost from habitat at/near Magnetic Island through time. The combined trajectories showed the maximum extent of the emigration plume from the Magnetic Island population.

Ethics approval

This work was supported by the ARC Centre of Excellence for Coral Reef Studies and the Australian Lions Foundation. It was conducted in accordance with James Cook University’s Animal Ethics Committee’s policies, procedures and guidelines. No specific permissions were required.

Source: Ecology - nature.com