Skip to main content

Data from: Does migration promote or inhibit diversification? A case study involving the dominant radiation of temperate Southern Hemisphere freshwater fishes

Cite this dataset

Burridge, Christopher; Waters, Jonathan M. (2020). Data from: Does migration promote or inhibit diversification? A case study involving the dominant radiation of temperate Southern Hemisphere freshwater fishes [Dataset]. Dryad.


While theory predicts that dispersal is a pivotal influence on speciation and extinction rates, it can have contradictory effects on each, such that empirical studies are needed to quantify its role. In many studies, dispersal reduction appears to promote diversification, although some analyses of migratory species suggest otherwise. Here we test for a relationship between migratory status and diversification rate within the dominant radiation of temperate Southern Hemisphere freshwater fishes, the Galaxiidae. We reconstructed a molecular phylogeny for the group, comprising >95% of extant lineages, and applied State-dependent Speciation Extinction models to estimate speciation, extinction, and diversification rates. In contrast to some previous studies, we revealed higher diversification rates in non-migratory lineages. The reduced gene flow experienced by non-migratory galaxiids appears to have increased diversification under conditions of allopatry or local adaptation. Migratory galaxiid lineages, by contrast, are genetically homogeneous within landmasses, but may also be rarely able to colonise other landmasses in the temperate Southern Hemisphere. The life history state exhibiting higher diversification likely reflects the spatial context of the study system relative to species dispersal abilities, by means of the intermediate dispersal model; the accurate quantification of dispersal abilities will aid in our understanding of these likely interactions.


Phylogenetic analyses incorporated as many galaxiid species as possible. While studies of diversification rates can accommodate missing taxa by incorporating their proportion during analysis, they assume the missing taxa are randomly distributed across the tree (Beaulieu and O’Meara 2016). However, the previously most comprehensive phylogenetic analysis of Galaxiidae (Burridge et al. 2012) lacks representation within two groups of morphologically cryptic species. Therefore, these groups were incorporated into the existing DNA dataset (599 bp lrRNA, 1154 bp cyt b, 1695 bp RAG-1, 1083 bp S7), and re-analysed to construct a time-calibrated phylogeny, following the same approach as previously (Burridge et al. 2012). Firstly, the South African Galaxias zebratus clearly represents a group of morphologically cryptic species (Waters and Cambray 1997; Chakona et al. 2013). Cytochrome b data were added to represent the ‘nebula’, ‘Heuningnes’, ‘Klein’, ‘Joubertina’, ‘Riviersonderend’, ‘Goukou’, ‘Breede’, ‘rectognathus’, ‘slender’ and ‘mollis’ putative species (Waters and Cambray 1997; Chakona et al. 2013). Another distinct lineage previously identified from the Olifants River is also retained (Waters and Cambray 1997). Secondly, the Australian mountain galaxias complex (Galaxias olidus s.l.) is now recognised as a radiation comprising at least 15 species (Adams et al. 2014; Raadik 2014). While not all species within the mountain galaxias complex are reciprocally monophyletic for DNA variation (Raadik 2014), we chose cyt b sequences that represent a lineage distinct for each species. While the molecular phylogeny reconstructed for these recently-diverged mountain galaxias may not represent those of the actual species, all species exhibit the same life history and the chronology of their molecular divergence is very shallow. Following Adams et al. (2014) and Raadik (2014), the previously analysed data ascribed to G. olidus now corresponds to G. oliros. In addition to data from these two recently resolved complexes, cyt b, lrRNA and S7 data from a new species Galaxiella toourtkoourt (Coleman et al. 2015) were also included. It has become apparent that previously analysed sequence data considered to represent Aplochiton taeniatus actually represents A. zebra, and therefore new cyt b data have been substituted. Homologous data from the recently resurrected A. marinus were not available at the time of analysis, representing the only excluded galaxiid species. All new data analysed relative to Burridge et al. (2012) are summarised in Table S1. The proportion of missing species was set at 0.07 for diadromous species (1/14), and 0.03 (2/68) for non-diadromous species (estimating that two species may still not be recognised within G. zebratus s.l., G. olidus s.l., or elsewhere).

The molecular phylogeny of the Galaxiidae, with branches scaled proportionally to evolutionary time, was obtained following the approach of Burridge et al. (2012). The analysis used a relaxed molecular clock and included nine biogeographic calibration points within the family, together with nine fossil age calibration points from outgroup lineages. BEAST 2.6.2 (Bouckaert et al. 2019) post-burnin run exceeded 6.9x108 steps, with stationarity and convergence assessed using Tracer (Rambaut et al. 2018). 35,586 trees were retained for the construction of a maximum clade credibility tree using TreeAnnotater. This tree was well resolved at most nodes, with topological uncertainty generally restricted to shallow nodes that do not involve transitions in diadromy status. Outgroups and lineages that represent intraspecific galaxiid divergences were pruned such that each galaxiid species was represented by a single lineage, with two exceptions. First, the diadromous Galaxias brevipinnis is resolved as polyphyletic, with New Zealand and Australian clades exhibiting molecular divergence consistent with interspecific differentiation elsewhere in the family (Waters et al. 2010; Burridge et al. 2012). Secondly, Galaxias maculatus exhibits substantial molecular divergence among Australian, New Zealand, and South American populations, and arguably represents three species (Waters and Burridge 1999). Galaxias niger was excluded as this appears to be synonymous with Australian G. brevipinnis based on the depth of molecular divergence and morphological plasticity of the latter.

Taxa were coded as diadromous or non-diadromous based on knowledge derived from published accounts of species biology (largely summarised by Augspurger et al. 2017). Observations of migratory status of Aplochiton spp. indicate that populations of A. taeniatus might variously be freshwater-resident or diadromous (catadromous or amphidromous), and some landlocked populations are known also for A. zebra (Alò et al. 2019). The widespread distribution of the latter (including the Falkland Islands / Islas Malvinas), however, implies a recent diadromous history, and therefore this species was considered to possess both life histories. Semi-anadromy is observed in Lovettia (Schmidt et al. 2014), the sister-lineage of Aplochiton. The majority of diadromous species also exhibit landlocked populations (Aplochiton spp., G. brevipinnis-Australia and New Zealand, Galaxias maculatus, Galaxias truttaceus, Galaxias argenteus, Galaxias fasciatus), and were treated in two ways. Firstly, they were coded as diadromous, because their landlocked populations are likely recent derivatives of diadromous populations, which better reflects their state through evolutionary time (Ovenden et al. 1993; Waters et al. 2010). Secondly, we also tested models under which species could occupy multiple states simultaneously (using GeoHiSSE, below).

Analyses were conducted under the Hidden-state Speciation Extinction (HiSSE) framework (Beaulieu and O’Meara 2016) to test the null hypothesis of no relationship between diversification rate and migratory status. This approach avoids potential errors associated with the use of “equal rate” null hypotheses (Rabosky and Goldberg 2015), and also accommodates shifts in diversification that could also be related to states that are unobserved or “hidden”, to reduce inferential errors with respect to migratory status (Beaulieu and O’Meara 2016). All analyses were performed in R using the hisse (Beaulieu and O’Meara 2016) and diversitree (FitzJohn 2012) packages. The parameters of the HiSSE model are rates of transition among character states (diadromy, hidden state: presence/absence), and species turnover (speciation + extinction) and extinction fraction (extinction/speciation). State transition rates (diadromous<->non-diadromous) were either equal or unequal. Simultaneous changes in migration status and the hidden state were excluded. The null model assumed that species turnover and extinction fraction did not vary with respect to migratory status, but did with respect to hidden state (“A” or “B”). The null was then compared to a model lacking hidden states, where species turnover and extinction fraction were allowed to vary with migratory status. Models were compared based on Akaike information criterion (AIC). Additional models were constructed to test whether the presence of hidden states within either diadromous or non-diadromous lineages could explain an apparent difference in diversification with respect to migratory status (Beaulieu and O’Meara 2016). A contribution of a nested hidden state to diversification might reveal that additional characteristics are also important in explaining diversification rates (Beaulieu and O’Meara 2016).

Models were also conducted under the GeoHiSSE framework (Caetano et al. 2018), originally developed for biogeographic analysis where species can occupy multiple areas, but here applied to accommodate those species exhibiting both diadromous and non-diadromous populations (see list above). Analyses analogous to HiSSE were performed using the hisse package. These analyses allowed transitions directly between diadromy and non-diadromy (no requirement of a polymorphic intermediate state; “include.jumps=TRUE”) and also from a dispersal-polymorphic state to either of these individual states (“separate.extirpation=TRUE”).


Royal Society of New Zealand, Award: UOO-0404