Discussion
Our results show various levels of population structuring across a range of spatial and temporal scales. Genetic structuring of lake- and ocean-developing G. brevipinnis populations indicates that lake populations are generally closed and therefore isolated from downstream coastal populations over long timescales. Collectively, lake tributary populations exhibit greater genetic structuring than coastal populations, but comparatively little genetic structuring is evident within individual lakes or along the coast, consistent with more open populations dynamics and within-system exchange of individuals. On shorter timescales, within-system otolith trace element signatures align with genetic results, suggesting that no migration occurs between lake tributary populations and coastal stream populations. Otolith trace element signatures also indicate further structuring on shorter timescales, with a semi-closed meta-population structuring of individual catchments or regions within lake and coastal systems. Larval plume residence is likely a simple consequence of rheotaxis (Hicks, 2012), a behavioural mechanism that may generate population isolation and structuring across both short and long timescales. Due to its ‘leaky’ nature, plume residence may result in populations being closed across short timescales and driven by self-recruitment, but not sufficient to prevent limited dispersal and genetic exchange among rivers, leading to open population dynamics on a longer timescale.
Long-term population structure
When compared to coastal stream populations in close proximity, lake tributaries have higher levels of structuring, indicating that each lake creates a closed population subject to genetic drift. In contrast, genetic structuring is relatively homogeneous within each lake and coastal region, indicating that the within-system meta-populations evident via otolith microchemistry results are not as isolated, and therefore relatively open, over longer timescales. Patterns of closed populations in lakes persist even where lakes are in close proximity to, and have open access to, the marine environment. Thus, lakes appear to act as barriers to upstream migration, despite presenting no obvious physical barrier to movement (see Gouskov & Vorburger, 2016). Mechanisms potentially leading to this isolation are unclear, but could be related to simple behaviours such as preference for current and upstream orientation (Hicks, 2012). By preferring to orient into the current, lake-developing larvae would seem likely to be retained in their natal river plume, as suggested by our larval trawling results, and adults and juveniles from downstream diadromous populations are unlikely to move upstream through a lake due to the absence of directional cues provided by water currents (Concepcion & Nelson, 1999), thus preventing exchange. Within lake and within-coast comparisons suggest relatively low genetic structuring, implying that catchment level meta-populations are open over long timescales, aligning with other studies on dispersive larvae which find connectivity at broad spatial and temporal scales (McDowall, 2003; McDowall, 2007).
Short-term population structure
Otolith trace-element signatures show both distinct system level population structuring, aligning with genetic results, and within system meta-population structuring, thus detecting a level of fine-scale relatively closed populations not detected by genetic methods. Distinct otolith trace element signatures formed between lake tributaries and coastal streams are not unexpected in our case due to impermeable hydro-electric dams downstream of lakes Wanaka and Wakatipu, as well as the distances involved. Further, a lack of mixing has previously been found in Lakes Moeraki and Paringa, which both have unimpeded access to the coast (Hicks et al., 2017), suggesting that connectivity is even more unlikely when a physical barrier is present. However, distinct trace element signatures between Lakes Wanaka and Wakatipu suggests retention within each lake, and therefore a closed population, despite the absence of any impermeable physical barriers that would prevent movement between the two. These closed populations must persist long term, as the same pattern can be seen in the genetic results. In contrast, both marine and freshwater populations seen in our study also formed otolith trace element clusters when grouped by tributary or region, indicating fine scale, catchment-level, semi-closed population structuring within coast or lake, similar to other studies (Hogan et al., 2014; Smith & Kwak, 2014; Warburton et al., 2018). Even the less distinct clusters (e.g. in L. Wakatipu) still had much higher reclassification success than would be expected if recruitment was largely from unstructured pools (e.g. Hickford & Schiel, 2016), suggesting at least some degree of larval philopatry (Warburton et al., 2018). While genetic population structuring could occur through mechanisms such as reproductive isolation via altered spawning time, we observed catchment and lake level structuring in the otolith trace element signatures of a single cohort. These results suggest behavioural mechanisms are restricting widespread larval dispersal from their natal streams, which would also result in retention within lakes.
Larval behaviour as a mechanism for population structuring
Larval plume residence found in this study suggests a behavioural mechanism that isolates populations within different streams or regions, and leads to long term isolation of lake populations. Nearly all larvae found in the study were present in river plumes, suggesting substantial retention within each stream draining into a lake or the sea. This process should result in the formation of semi-closed meta-population within lakes and along coastlines, concordant with the patterns seen in otolith trace element signatures. Further, plume residence seemingly provides a mechanism by which larvae avoid drifting out of lakes (David et al., 2019; Hicks, 2012), isolating populations within lakes, even when close to the coast. The actual behaviour maintaining larvae within plumes is unclear, although the observation of strong rheotaxis even in newly hatched larvae (Hicks, 2012), could provide a simple explanation for the retention of larvae in river plumes. Similarly, the river plumes entering lakes studied here anecdotally appeared to have more zooplankton than other sites, suggesting food availability may also favour retention in these systems. River plumes represent dynamic environments and thus the degree of retention is likely to vary both spatially between river plumes, and temporally in response to changes in flow. Similarly rare dispersal events mediated by environmental conditions have been identified in other systems such as arid floodplains (e.g. Mossop et al., 2015), resulting in relatively low genetic structuring despite largely disconnected ecological populations. Whilst the majority of larvae collected were in river plumes, the collection of small numbers of larvae in offshore habitats suggests leakage from river plumes, likely providing a degree of population connectivity within lake and marine pelagic habitats, thus maintaining population and hence genetic connectivity resulting in open population signatures over longer timescales.
Evolutionary implications
Behaviours that retain larvae within river plumes and resist dispersal provide a simple mechanism that could interact with landscape to produce context dependent patterns of population connectivity and isolation. Collectively, our results form a case study describing how behaviour might interact with geologic events (Craw et al., 2016) and landscape to facilitate divergence of landlocked populations of an ancestral G. brevipinnis, that may have led to the subsequent radiation of the non-migratory G. vulgaris species-complex (Allibone & Wallis, 1993; Allibone et al., 1996; Burridge, McDowall, Craw, Wilson, & Waters, 2012; Waters et al., 2010). Our results confirm that lake-developing populations of G. brevipinnis are isolated and genetically divergent, despite the relative proximity and potential connectivity with other lake populations and even diadromous populations in some cases. Similar speciation patterns are repeatedly seen in other amphidromous fish which universally have a pelagic larval phase but isolated adult populations (Augspurger, et al., 2017), including members of the Cottidae (Dennenmoser, Rogers, & Vamosi, 2014; Goto et al., 2015), Eleotridae (Nordlie, 2012) and Gobiidae (Keith & Lord, 2011), which are likely influenced by similar interactions between landscape, geology and behaviour.
Our results also have implications for the management and conservation of amphidromous species, particularly those in which non-diadromous populations are readily formed. Our otolith microchemical results, indicating that short-term population-sustaining processes may occur at the scale of individual catchments or isolated geographical regions, add further weight to the notion that management may have to be undertaken at the local (Hicks et al. 2017; Warburton et al., 2018) rather than island or distribution wide (Cook, Bernays, Pringle, & Hughes, 2009) scale. Further, the translocation of individuals from landlocked populations of facultatively amphidromous species has been suggested as a potential management strategy to enhance typically more degraded, lowland diadromous populations (e.g. David, 2003). Our finding of high genetic differentiation in landlocked populations, together with previously identified ecological differences between landlocked and diadromous populations (Augspurger & Closs, 2019), suggest that such a strategy should be approached with caution. Indeed, the translocation of a facultatively amphidromous shrimp before high levels of cryptic diversity were recognized has been recognized as resulting in the rapid extinction of the local genotype (Cook et al. 2006; Hughes et al. 2003).