Computational protocol: Low acclimation capacity of narrow‐ranging thermal specialists exposes susceptibility to global climate change

Similar protocols

Protocol publication

[…] To determine whether there are differences in SMR between wide and narrow‐ranging species when acclimated at different temperatures, we ran a phylogenetically controlled Markov chain Monte Carlo generalized linear mixed model (MCMCglmm) with repeated measures (Hadfield, ). MCMCglmm uses a Bayesian approach to fit general linear models and includes the phylogenetic variance–covariance matrix as a random effect in the regression model, allowing for any genetic influences in the data to be accounted for.All modeling was conducted in R ver. 3.1.2 (R Core Team, ) using packages “ape” (Paradis, Claude, & Strimmer, ) and “MCMCglmm” (Hadfield, ); see Appendix for R code. Fixed factors to examine the influence on VO2 included: test temperature (5, 15, 25°C), acclimation temperature (14 and 22°C), and range size (latitudinal extent, as well as species grouped as wide vs. narrow‐ranging). To assign species as wide or narrow‐ranging, a natural break was found between species with latitudinal extents greater than or less than five degrees of latitude. Sex and genus were also included as covariates in the model. As Desmognathus salamanders are often associated with streams and seeps, they could be better buffered from temperature extremes than more terrestrial Plethodon salamanders. This could have a potential influence on salamander physiological tolerances and SMR. The covariate “genus” therefore divides Plethodon from Desmognathus to help to control for differences in thermal habitat associated with each group.As individual salamanders were used in multiple trials, an additional random effect was included in the model to account for repeated measures of individuals. The initial model included several interactions (test temperature × acclimation temperature, test temperature × range size, acclimation temperature × range size, and test temperature × acclimation temperature × range size), however, the three‐way interaction, as well as the two‐way interactions for test temperature × acclimation temperature and test temperature × range size and had p‐values >.3 and were removed from the model. Our priors took the form of: prior <‐list(G = list(G1 = list(V = diag(2), nu = 2, alpha.mu = c(0,0), alpha.V = diag(2) × 1,000)), R = list(V = diag(1), nu = 0.002)), where a 2 × 2 covariance matrix is being estimated for the random effects (G) and a scalar variance for the residuals (R). Using trace plots, we observed the distribution of samples to remain stationary over time, therefore giving us confidence that our posterior is a good approximation of the true distribution. For optimal outcomes, we ran the analysis for 300,000 iterations, with 25,000 samples of burn‐in, and sampling every 1,000th generation.Further, additional MCMCglmm analyses were performed where data were grouped separately by range size into wide and narrow‐ranging species. This grouping enabled examination of the affect of acclimation temperature on SMR within each group (wide and narrow‐ranging) and for each test temperature (5, 15, 25°C). Acclimation (14 and 22°C) was included as a fixed factor, while sex remained as a covariate. T‐tests were also performed on each species for each test and acclimation temperature, to determine which species showed evidence of thermal acclimation or metabolic depression. [...] Remaining statistical analyses do not have the phylogeny incorporated into the model as in earlier MCMCglmm analyses. As phylogenetic non‐independence may influence results in comparative analyses of multiple species, we need to test for the phylogenetic influence of measured traits (e.g., acclimation ability, CTMax, thermal range, latitudinal extent). Lambda tests were employed using “Fit Continuous” model tests in the “geiger” package of R (v.3.0.2, R Development Core Team, ). In all cases, lambda was chosen as the best model with lambda scores for traits between 0.00 and 0.086. These low lambda values indicate that these particular traits have very little phylogenetic signal, enabling us to use the original data without further concern for phylogenetic influence. [...] To better assess the relationship between physiological tolerances and local thermal environments, we estimated the thermal range of localities for each species and performed linear regressions between acclimation ability versus environmental thermal range and acclimation ability versus latitudinal extent. Acclimation ability was defined as any positive increase in VO2 from lower to higher acclimation temperatures, using VO2 at 22°C acclimation minus VO2 at 14°C acclimation (Table ). Any increase in VO2 for individuals acclimated at a higher temperature indicates acclimation capacity, whereas any decrease was considered metabolic depression. Only SMR data at the 25°C test temperature were used to determine warm temperature acclimation ability, as this is where we find the greatest influence of temperature on acclimation.To approximate the annual thermal range of each locality sampled, thermal data (averages 1950–2000) were obtained from the Worldclim online database at 1 km2 resolution (Hijmans, Cameron, Parra, Jones, & Jarvis, ). The program DIVA‐GIS (Hijmans, Guarino, & Rojas, ) was used to georeference and map salamander localities. Data were then extracted for the bioclimatic variable Bio 7 (temperature annual range), which represents air temperature of the local area. Salamanders are found to conform quickly to the temperature of their environment and air temperature has been found to be a good proxy of operative conditions actually experienced by terrestrial salamanders (Lunghi et al., ). Genus (Plethodon vs. Desmognathus) was included as a covariate.Finally, intraspecific regressions were performed to test how acclimation ability is influenced by natural thermal regimes experienced by populations. Nine of 16 species had data for multiple localities across the geographic range and could be used to test for relationships between temperature range of the environment and acclimation ability. […]

Pipeline specifications

Software tools APE, GEIGER, PHYSIG, DIVA-GIS
Application Phylogenetics