Introduction
Processes responsible for population structuring are key to understanding population dynamics and speciation. Isolation occurs on a continuum from longer, multi-generational timescales, whereby populations genetically diverge through selection and genetic drift, to much shorter timescales (even a single generation), when immigration is rare enough to produce distinct ecologically closed (i.e. self-recruiting) populations, that may still maintain genetic connectivity over the long term (Dunning, Danielson, & Pulliam, 1992; Kekkonen et al., 2011; Slobodkin, 1980; Jones et al., 2009). Furthermore, genetic shifts may occur through strong selection acting over relatively short timescales (Carroll, Hendry, Reznick, & Fox, 2007; Fussmann, Loreau, & Abrams, 2007), despite populations maintaining some genetic connectivity (Moody et al., 2015). As a result, it is often difficult to determine the degree and timescale at which population connectivity occurs (Fussmann et al., 2007; Levin, 1992). Therefore, evaluating both short- and longer-term population structuring is key in determining both ecological and evolutionary dynamics. To determine why population structuring occurs and differs on various spatial and temporal scales, the mechanisms driving population structuring must also be examined (Levin, 1992).
Behaviour and life history traits can interact with landscape and generate context-dependent patterns of isolation, and therefore population structure, across various temporal scales. Dispersive or migratory behaviours (Cowen, Lwiza, Sponaugle, Paris, & Olson, 2000; Dingle & Drake, 2007) tend to maintain open populations, where individuals are often exchanged. (Pinsky et al., 2017). If migratory pathways fragment and behavioural mechanisms favour local retention as opposed to connectivity, isolation may occur causing open populations to become closed (self-recruiting) (Booth, Montgomery, & Prodöhl, 2009; Hughes, Schmidt, Macdonald, Huey, & Crook, 2014). Understanding behavioural mechanisms is essential in instances where structured populations form despite no apparent barriers to genetic exchange or structuring mechanisms (Levin, 1992).
Amphidromous fishes provide an ideal system to investigate behavioural mechanisms leading to open and closed population structuring across timescales due to the combination of stream resident adults and a potentially dispersive marine pelagic larval phase (Augspurger, Warburton, & Closs, 2017; McDowall, 2007). Amphidromous fishes spawn in freshwater. After hatching, larvae drift downstream to a marine pelagic environment, develop for a period of typically 3–6 months, then return to freshwater fluvial environments as juveniles where they remain for life. Many such species are thought to disperse during the larval phase and maintain open populations across broad geographic distributions (McDowall, 2003, 2007). In contrast, other species appear to resist dispersal, resulting in relatively closed populations across short (Hogan, Blum, Gilliam, Bickford, & McIntyre, 2014; Sorensen & Hobson, 2005; Warburton, Jarvis, & Closs, 2018) and even long timescales (Hughes et al., 2014) despite no obvious barriers preventing connectivity.
Amphidromous fishes also often form landlocked populations, the life-history of which is nearly identical to their diadromous counterparts, with fluvial juvenile and adult forms and pelagic larvae. In this case, however, larval development occurs in a lake rather than the marine environment (Augspurger et al., 2017). This landlocking provides further potential for closed and open population structuring across timescales, as lakes can provide an opportunity for the isolation and genetic divergence of populations (Gouskov & Vorburger, 2016; King, Young, Waters, & Wallis, 2003). Closed landlocked populations diverge further genetically, subsequently radiating into non-migratory forms and species complexes (Allibone & Wallis, 1993; Allibone et al., 1996; Burridge, McDowall, Craw, Wilson, & Waters, 2012; Goto, Yokoyama, & Sideleva, 2015; Yamasaki, Nishida, Suzuki, Mukai, & Watanabe, 2015). In other cases, landlocked populations maintain connectivity with diadromous populations across longer-term timescales (Goto & Arai, 2003; Hicks et al., 2017). Behavioural mechanisms during the larval pelagic phase, such as orienting into current, may play a role in determining connectivity across both short- and long timescales.
Galaxias brevipinnis is a facultatively amphidromous fish distributed throughout New Zealand, with great capacity for forming both diadromous and landlocked populations (McDowall, 1990). Landlocked populations of G. brevipinnis are potentially isolated from diadromous populations as their larvae develop in lakes, but do not appear to drift downstream out of them (Hicks et al., 2017). This may be due to their strong rheotactic behaviour after hatching, which may limit dispersal from their pelagic developing environment (Hicks, 2012). Further, diadromous adults are rarely found upstream of lakes, despite the absence of any obvious in-stream barriers blocking access, possibly due to the lack of a rheotactic cue allowing juveniles to navigate through large pelagic environments (Hicks, 2012; Jarvis & Closs, 2019). Thus, populations of G. brevipinnis may potentially exhibit context-dependent degrees of population isolation and genetic structuring, creating an ideal opportunity to examine the importance and interaction of behaviour and landscape on population connectivity across temporal and spatial scales.
Here we investigate patterns of hierarchical population structuring inG. brevipinnis across spatial and temporal scales, and the possible role of behaviour in generating population structuring. We use genetic analyses to determine population structure across long multi-generational timescales, analysis of otolith trace element signatures to evaluate population structure over short timescales (the processes sustaining a population over a single generation), and larval trawling to determine larval distribution as a result of behaviour. We hypothesized that population structuring would interact with landscape, and generated a number of hypotheses in this context. Over long timescales, we predicted that: (1) coastal populations would tend to show high genetic homogeneity, indicating open populations like those of other amphidromous fishes, while (2) landlocked populations would show closed populations and some genetic divergence from other landlocked populations and coastal sites. Over short timescales, we hypothesized that: (3) landlocked and coastal stream populations would develop within their respective systems (e.g. lake or ocean), and (4) otolith trace element signatures would reflect catchment level meta-populations within lakes and along coastlines, forming semi-closed (self-recruiting) populations on short-term timescales that are isolated to varying degrees despite some level of genetic exchange across longer-timescales. Finally, we hypothesized that: if otolith clustering occurred, (5) larval distribution would show higher larval densities in river plumes.