Behavioural models

This study gives insight into the paradigm that larval fish can influence their dispersal patterns through horizontal and vertical swimming behaviours, by predicting how these patterns are affected using a biophysical dispersal model (BDM). Giving the larval fish the ability to move (vertical, horizontal, or both) drove changes in dispersal. The connectivity

patterns of the different behavioural models grouped into models that included or did not include orientated horizontal swimming (OHS). Swimming behaviour, both horizontal and vertical, generally allowed larval fish to settle closer to their natal region (increasing local retention), reduced the distance of dispersal (restricting the connectivity pathways), and increased their chances of settlement success. The strength of this behavioural effect on connectivity patterns was noticeable with orientated horizontal swimming (OHS) and

weakest with ontogenetic vertical migration (OVM). The compelling driver of the variation in connectivity patterns was the spawning location and presumably timing of spawning (not a factor in these permutations), which have been seen to be influential on predicted dispersal patterns in other modelling studies using similar approaches (Ayata et al. 2010; Treml et al. 2015; Kvile et al. 2018), and more influential than vertical migration behaviours (Puckett et al. 2014), although not in all systems (North et al. 2008). Driving these connectivity patterns are the hydrodynamics of the coast of New South Wales (NSW), dominated by a southward western boundary current (the East Australia Current) that separates off the coast

approximately at the location of regions 7 and 8 that generates mesoscale eddies and entrains ichthyoplankton (Suthers et al. 2011).

Orientated horizontal swimming (OHS) was the behaviour that affected the modelled connectivity patterns the greatest, although the size of this effect presumably depends on the spatiotemporal scale of the modelled domain. Larvae with OHS behaviour increased local retention to the natal region, increased self-recruitment and reduced the dispersal distance. Giving larvae the ability to sense, orient, and swim towards a reef, also substantially increased their chances of settlement success. The reduced dispersal distance resulted in lower

settlement richness, diversity, and connectance. These results are in agreement with

modelling studies from other oceanographic locations that have seen OHS increase the local retention and self-recruitment of larval fish at similar levels (Wolanski and Kingsford 2014) and improve the settlement success and reduce the dispersal distance (Staaterman et al. 2012). It must be borne in mind that the implementation of OHS in the BDM used in this study was closer to the implementation used by Staaterman et al. (2012) than Wolanski and Kingsford (2014). Aligned with the results of the models, it is predicted that larval fish that can sense and orientate towards a reef benefit with an increased likelihood of settlement success and the ability to directly influence the trajectory of dispersal. The effect of OHS on the connectivity

OHS produced homologous patterns of connectivity. Even subsequent models with

additional vertical behavioural traits only marginally altered the connectivity patterns. This strong influence of OHS behaviour on connectivity patterns, relative to the other behaviours (i.e. DVM and OVM), is likely the result of the directed nature of the swimming (towards viable settlement habitat) as much as the swimming speed itself. The ability to alter dispersal comes from larvae being able to swim in the direction of a reef if they can swim faster than the prevailing current. While the other migration behaviours provide larvae with mechanisms that alter dispersal, without OHS they can only indirectly influence their dispersal outcome. Although, it must be put into context with the influence of other behavioural traits (Cowen and Sponaugle 2009), e.g. another study comparing a wider range of behavioural traits found OHS to only have an intermediate effect compared to other traits such as mortality, PLD, and size of the settlement competency window (Treml et al. 2015).

Surprisingly, compared to the other behaviours, OVM did not have patterns that differed strongly from passive larvae and had limited influence on the connectivity metrics. Local retention and settlement success were not affected when adding OVM to a passive model, only the self-recruitment of a region was increased. Larvae in the model exhibit vertical behaviour more during the pelagic larval stage than OHS (larvae orientate and swim when a reef is sensed), therefore it was hypothesised that vertical migration would have a greater influence on the dispersal. Vertical migration occurs early in ontogeny and changing position in the water column allows larvae to be advected by currents with different velocities. In contrast, for horizontal swimming to influence connectivity patterns, the larvae have to be competent to settle, and within a sensory distance of the reef to orientate and swim towards the settlement habitat. In general, observed distributions tend to be smaller than estimates of BDMs using passive drift (Shanks 2009). Previous studies have found OVM to increase modelled local retention (Paris et al. 2007; Kough et al. 2016) and transform the predicted dispersal patterns compared to non-swimming (passive) larvae (Butler et al. 2011; Drake et al. 2013; Kough and Paris 2015), however, this is not universal (Zhang et al. 2015). Vertically migrating with ontogeny is considered a strong retention mechanism as moving downwards in the water column exposes larvae to weaker and directionally divergent currents (Paris and Cowen 2004), yet in this study the evidence for increased retention was weak. The results suggest the effect size of OVM on larval retention is dependent on the regional

and a limitation of the hydrodynamic model resolution in capturing the coastal dynamics. Local retention was highest when the EAC was closest to the coast and reduced dramatically at the separation zone. Non-swimming larvae that were moved by a vertical component (Passive+VA) had connectivity patterns that were more similar to models with OVM than the non-swimming passive larvae, suggesting the vertical component advected larvae downwards at a slow rate that closely resembled the restricted movement of the stage based ontogenetic vertical migration used. Diel vertical migration (DVM) had a stronger effect on the

connectivity patterns and connectivity metrics than OVM (where OVM used the stage-based method described above), suggesting the time the larvae spend at different depths in the water column influences local retention. Connectivity models with DVM, similar to OVM, have been seen to increase settlement success, local retention and decrease the dispersal distance (Aiken et al. 2011; Butler et al. 2011; Robins et al. 2013). The results of this study further confirm these previous findings.