Implications for meta-analysis of scale dependence in biodiversity change Seminar

Empirical studies of biodiversity have well-recognised sensitivities to spatial and temporal scale. Meta-analyses typically collate studies that vary widely in replication and scale of observed or estimated numbers or densities of species. Any scale dependence in the study-level biodiversity metrics then influences estimates of meta-effect size and confidence interval, with an as yet unquantified potential for bias. Here we sample from simulated communities of old-growth and secondary forests to evaluate sources of inaccuracy in study-estimates and meta-estimates of effect sizes, caused by within-study plot and sample sizes, biodiversity measure, and choice of effect-size metric. We find multiplicative influences of all these parameters on the magnitude and direction of change in effect-size accuracy with precision (1/variance, the study-level weighting). Across scales, species richness with log response ratio yields better accuracy than species density and Hedges’ g, which dangerously combines higher precision with persistent bias. Using empirical datasets, we demonstrate the challenge of detecting scale-dependence, which is governed by the weighting scheme and the effect-size metric, due to covariation between replication, variance and plot size. We propose solutions to reducing bias, urge  empirical studies to publish raw data to allow meta-analysts to evaluate covariation, and generally recommend against using Hedges’ g in biodiversity meta-analyses.

Speaker Information

Dr Becks Spake, Postdoc in Applied Biogeography, School of Geography and Environmental Science, University of Southampton