Sampling of fish assemblages
The fish community within each site was surveyed using Baited Remote Underwater Video surveys (BRUVS). Three replicate BRUVS were deployed at each site at least 200m apart and within 5-10mins of each other, with each survey lasting 30mins. Sampling was done once a year, at a similar time of the year (Austral winter), over a 17-year period between 2002 and 2018, except for 2012, 2016 and 2017 when no sampling was done. BRUVS units consisted of video cameras attached to a metal frame and bait pole with a bait bag containing ~800g of pilchard (Sardinops neopilchardus ) 1.5m in front of the cameras. The maximum number of individuals of a species in a single frame (MaxN) was recorded for each site and year using SeaGIS Event Measure software (seagis.com.au) as a conservative measure of relative abundance, as this method eliminates the chances of recounting the same fish. All fish within the field of view within 3m of the camera were identified to species level where possible.
Each fish species identified in the BRUVS was then classified into one of four trophic guilds: planktivore, herbivore, invertivore or piscivore, based on the predominant food source for adult individuals of the species as reported in the literature e.g. (Allen 1999; Malcolmet al. 2007) and/ or on Fishbase (Froese & Pauly 2000). Using these same sources, species were also classified into ecoregion (tropical or temperate) based on distribution data (see full list of species and assigned trophic guild and distribution in Supplementary Table 2). For species that show a strong distributional crossover, ecoregion was allocated based on whether the core range extends furthest towards either the tropics or in a poleward direction.