Connectivity modelling for the Capricorn-Bunker Group

The selected model domain was a rectangle tilted –15° off vertical that encompassed the Capricorn Group of 17 reefs (Heron, Wistari, Sykes, One Tree, Irving, Polmaise, Masthead, Erskine, North, Tryon, North West, Broomfield, Wilson, Wreck, Lamont, Fitzroy and Llewellyn). The Bunker Group of 4 reefs (Lady Musgrave, Boult, Hoskyn and Fairfax Reefs) immediately to the south was not included in the model domain.
The hydrodynamic grid was set up in Relocatable Coast Ocean Model (RECOM), with the numerical modelling performed in Sparse Hydrodynamic Ocean Code (SHOC) within CSIRO’s Environmental Modelling Suite (EMS). The hydrodynamic model had a resolution of 305 m × 305 m. eReefs GBR1 output was used to provide boundary conditions. Model runs had a timestep of 12 minutes. Measured wind speed and direction were obtained from Australian Bureau of Meteorology weather stations in the Capricorn area. Hydrodynamic hindcasts were performed for 3 November 2021 – 29 January 2022 and for 31 October 2022 – 14 January 2023.
Lagrangian particle dispersion was performed on hydrodynamic model outputs using CONNIE, a software for hydrodynamic particle advection. Whilst CONNIE is widely known through its world wide web graphical user interface, we ran CONNIE as a MATLAB R2021a executable, called from within a Python wrapper that managed input variables and output files. All runs were performed on a high-performance computing cluster.
At our source reefs digitisation of polygons was based on areas in each source reef containing “coral/algae” in the reef benthic maps of the GBR Reef Explorer. Particle release dates from these polygons were based on the nights we observed spawning to occur during fieldwork in November 2021 and December 2022. For November 2022, we were not in the field to make observations, so we instead inferred probable spawning dates from spawning observations taken in the central GBR during that month. On each particle release date, we released 1000 particles per hour between 6 pm and 10:59 pm. To mimic the gradual sinking behaviour of coral larvae, particles remained at 0.25 m deep for 12 hours , then 2 m depth for 24 hours, before going to 4.5 m depth for the remainder of the dispersal period. Windage during the dispersal phase closest to the surface was set at 3 % of the wind field at 10 m height, and at 0 % during the deeper phases. Dispersal runs were continued until the date that field observations ended or the date hydrodynamic runs ended, whichever was sooner.
Natural mortality of coral larvae and competency (ontogenetic readiness to metamorphose into a benthic-attached spat) as a function of time following spawning were applied to particles using the piece-wise equations of Moneghetti et al. (2019, Ecology: 100:e02730), developed for Acropora tenuis.
Particle arrival was assessed at 305 x 305 m square sink polygons placed at each field sampling station. Particles were weighted according to the amount of time that they spent within the target region. Particles were also weighted by the relative size of their source polygon, scaled to between 0 and 1 using the size of the largest source polygon. Particle numbers and weights were then used to calculate the degree of connectivity between each source reef and sink polygon.

Sensitivity analyses were performed to measure the impact of varying a number of parameters and structural elements of the lagrangian particle tracking methods:
(i) the impact of using different equations for competency and mortality.
(ii) the impact of propagules spending no time at the surface following spawning or a full 24 hours at the surface following spawning and prior to sinking to the other depths.
(iii) a sensitivity test of windage by comparing runs with windage at 3 % to runs with windage at 0 %.
(iv) a sensitivity test of the impact of receiving polygon size and coverage