Computational protocol: Benthic meiofaunal community response to the cascading effects of herbivory within an algal halo system of the Great Barrier Reef

Similar protocols

Protocol publication

[…] Benthic community data matrices (Bray-Curtis) were square root transformed for increased homoscedasticity []. For multivariate analyses, algal canopy height and mean particle size were normalised using Z-score transformation to account for differences in sampling units []. To analyse benthic invertebrate community variation, a mixed-effects PERMANOVA was used with distance from reef and particle size included as fixed terms, and patch reef identity included as a random factor to account for the multiple samples taken along each transect. Permutations were set at 9999, and significant factors were identified through step-wise removal of nonsignificant terms. The PERMDISP (Permutational Analysis of Multivariate Dispersions) function was used to determine whether significant PERMANOVA p-values were a result of variance around or between means. Distance Based Linear Models (DISTLM) were then used to determine the contribution of environmental covariates algal canopy height and mean sediment particle size to the overall multivariate assemblage variation []. Similarity Percentages Analyses (SIMPER) using overall community abundances were used to highlight groups driving any dissimilarity between distances.Following initial analyses of the whole data set, and based on SIMPER percent contribution, benthic invertebrate assemblages were divided into their corresponding taxa: Polychaeta, Nematoda, Mollusca and Arthropoda. Individual groups were then analysed against distance from reef, algal canopy height, mean particle size and C:N ratio using linear mixed-effects models (LME) from the “lme()” function within “nlme” package in R [], with patch reef again treated as a random factor. Invertebrate abundance homogeneity of variance at each distance level was confirmed. A polynomial (quadratic) equation was applied to distance from patch reef to allow for curvilinear relationships. Due to co-linearity between distance from patch reef and algal canopy height, where a quadtratic distance term was found to be insignificant it was removed completely and linear models were re-run []. Variables that made significant contributions to taxa-specific patterns were identified through Akaike Information Criterion (AIC) and step-wise removal of nonsignificant terms. The “predictSE.lme()” function, within the “AICcmodavg” package was implemented to approximate 95% confidence intervals of model fixed effects using the delta method. Two core samples from separate patch reefs at distances 30 m and 22 m were not used in the analyses due to missing nutrient and sediment particle size data, respectively. Data were analysed using PRIMERv6 (Primer-E Ltd, Plymouth, UK), PERMANOVA+ (Permutational Multivariate Analysis of Variance) and R 3.1.0 []. […]

Pipeline specifications

Software tools lme4, nlme
Application Mathematical modeling