Otolith microchemistry techniques and analyses
To collect fish for otolith analyses, ten juveniles that had recently returned from their pelagic developing habitat to a fluvial environment were collected from each coastal stream and lake tributary via backpack electric fishing (Kainga EFM300; NIWA Instrument Systems), euthanized using an overdose of anesthetic, and preserved in 90% ethanol. Sagittal otoliths were then retrieved in the laboratory. Otolith surfaces were cleaned by sonication in ultra-pure water for 30 s, allowed to air dry, then mounted on a standard glass microscope slide covered with double sided sticky tape (Warburton, Reid, Stirling, & Closs, 2017). Larval otolith trace element signatures were obtained by depth profiling laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) as described in Warburton et al. (2017). Depth profiling was completed at the University of Otago Centre for Trace Element Analysis on an Agilent 7500cs ICP-MS coupled to a Resonetics (now ASI) 155 RESOlution M-50 laser ablation system powered by a Coherent 193 nm ArF excimer laser. During depth profiling, the laser was repeatedly fired at the same spot on the surface of the otolith, ‘drilling’ through the otolith and obtaining a complete life-history profile from otolith surface (point of capture) through to the core (hatch). Possible down-hole effects associated with depth profiling were minimized following the recommendations of Warburton et al. (2017). A 75 µm spot size was used at a firing rate of 10 Hz tuned to capture an 11-element suite (Ca, Ba, Sr, Rb, Mn, Li, Cu, Ni, B, Mg, Al) with a sample fluence of 2.5 ± 0.1 Jcm-2. Condensation and contamination were minimized by completing ablations in a chamber containing an atmosphere of pure helium gas. Possible machine drift was accounted for by taking a 20-s gas blank at the start of each ablation, and standards (NIST 610, NIST 612, MACS-3) were run bracketing every 10 samples. Standard values for NIST were obtained from Jochum et al. (2011) and the standard value for MACS-3 was obtained from Chen et al. (2011). The chance of hitting otolith cores was maximized by mounting slides in a sampling cell and using a 400x video microscope to view otoliths.
Data were processed using the Trace_Elements and Trace_Elements_IS data reduction schemes in IOLITE version 2.5 (Paton, Hellstrom, Paul, Woodhead, & Hergt, 2011), with 43Ca set as the internal standard, resulting in elemental data expressed as a molar ratio of mols element/mols Ca. NIST 610 was set as the calibration standard, while NIST 612 (Jochum et al., 2011) and MACS-3 (Chen et al., 2011) were used as reference materials. To interpret otolith signatures, the otolith core was identified by a spike in Mn (Ruttenberg et al., 2005), and the 7–10 µm of drilling depth before this spike was selected as the larval development period, as indicated by traces (see Supplementary material, Appendix 1, Fig. A1). Elemental concentrations within this period were averaged to use as a standardized, comparable larval development value. Otolith trace element signatures lacking Mn spikes were excluded from further analyses due to the likelihood of having missed the natal core region, and this resulted in final sample sizes of 47 (coastal), 44 (Wanaka), and 47 (Wakatipu). Marine versus lake development was determined by interpreting patterns in Ba and Sr. High Sr and low Ba are characteristic of marine development, and the opposite is generally true in freshwater (Campana, 1999; Warburton et al., 2017).
Statistical methods similar to Hogan et al. (2014) were used to determine population structuring. To determine whether a linear discriminant analysis (LDA) could be used for evidence of population structure in otolith signatures, boxplots for each element were used to determine which elements were most likely to show variation. A six-element suite (Sr, Ba, Rb, Mg, Cu, Ni) was selected based on these box plots and likelihood of element stability in the otolith (see Campana, 1999). An LDA was then run on these six elements in the R statistical environment using the MASS package (R Core Team, 2015; Ripley et al., 2011) to examine clustering of otolith micro-chemical signatures that would indicate discrete larval development pools. To determine the geographic scale on which population structuring occurred, an LDA was first run at the broad-scale system level (i.e. attempting to reclassify individuals only to the three broad-scale habitats of ‘Coast’, ‘Wanaka’, and ‘Wakatipu’). To determine to what extent, if any, populations are structured within-system, additional LDAs were then run at a finer scale, i.e. within each lake or coastal region by site. See Fig. 1 for sampling sites and a priori regional classification. Re-classification rate was determined using training and test data where half of the samples were used to determine clustering and the other half then used for re-classification.