Fluctuating salinity during development impacts fish life histories
Description
To test the effect of developmental environment and age, we ran models with environment (freshwater, stable-saline, fluctuating-saline), age (young, old) and their interaction as fixed factors. Age was excluded if a trait was only measured at one age. If there was a significant interaction, we examined the environmental effect separately at each age. If the interaction was non-significant, it was removed from the model to estimate the main effects. Given a significant environment effect, we ran post-hoc Tukey’s tests to test for pairwise differences. Each trait in each sex was analysed separately. Brood ID (for all models) and individual ID (for male reproductive traits) were included as separate random factors. We first ran linear mixed models (LMMs) and checked whether the model residuals fulfilled the model assumptions. If not, we either log-transformed the data or ran generalized linear mixed models (GLMMs) with an appropriate error distribution. We included the same variables in zero-inflated models as those in main models. We ran overdispersion tests (DHARMa package) to ensure that the variance was not greater than the mean for GLMMs. For immunity, we included standardised body length as a covariate. For log-transformed gut length, we accounted for its allometric relationship with body size by log-transforming body length and including it as a covariate (standardised). For relative telomere length, we ran an additional LMM including age since birth (standardised) as a covariate to test how chronological age affect telomere shortening. Growth We ran separate LMMs for growth in juveniles, adult females and adult males. We considered body length as the response variable, and treated environment as a fixed factor, and age (in weeks) as a fixed covariate. We treated brood ID (for both juveniles and adults) and individual ID (for adults) as random factors to control for repeated measures from the same family and/or fish. We included the environment-by-age interaction, and age^2 to account for non-linear growth. If the interaction was significant, we ran individual LMMs (and Tukey’s tests) at each age to determine when body size diverged between environments. Mortality Adult mortality was evaluated separately for each sex using Cox regression models (coxme package) with environment as the fixed factor and brood ID as a random factor. Surviving fish were treated as right-censored data. Sex differences We calculated the standardised mean difference (effect size = Cohen’s d) for the effect of salinity (stable-saline relative to freshwater) and the effect of fluctuations (fluctuating relative to stable-saline) for traits in both sexes. We further tested the difference in effect sizes between females and males using Welch’s t-test (α = 0.05, two-tailed) to assess sex-specific responses.