There are no whole truths in meta-analyses: all their truths are half truths

In a recent letter, Thomsen & Wernberg (2015) rean-alyzed data compiled for our recent paper (Lyonset al., 2014). In that paper, we examined the effectsof macroalgal blooms and macroalgal mats on sevenimportant measures of community structure and eco-system functioning and explored several ecologicaland methodological factors that might explain someof the variation in the observed effects. Thomsen &Wernberg (2015) re-analyzed two small subsets of the data, focusing on experimental studies examining effects of blooms/mats on invertebrate abundance.Their analyses revealed two interesting patterns.First, they showed that macroalgal blooms reducedthe abundance of communities that Thomsen andWernberg categorized as ‘mainly infauna’, whileincreasing the abundance of communities categorized as ‘mainly epifauna’. Second, they showed that theimpacts of macroalgal blooms on ‘mainly infauna’communities increased with algal density in experiments that included multiple levels of algal density.These findings, as well as the conclusions that Thomsen & Wernberg (2015) draw from them, are largely consistent with our own expectations and interpretations. However, we also feel that some caution is required when interpreting the results of their analyses.

Analyses of sublittoral macrobenthic community change in a marine nature reserve using similarity profiles (SIMPROF).

Sublittoral macrobenthic communities in the Skomer Marine Nature Reserve (SMNR), Pembrokeshire, Wales, were sampled at 10 stations in 1993, 1996, 1998, 2003, 2007 and 2009 using a Day grab and a 0.5 mm mesh. The time series is analysed using Similarities Profiles (SIMPROF) tests and associated methods. Q-mode analysis using clustering with Type 1 SIMPROF addresses multivariate structure among samples, showing that there is clear structure associated with differences among years. Inverse (r-mode) analysis using Type 2 SIMPROF decisively rejects a hypothesis that species are not associated with each other. Clustering of the variables (species) with Type 3 SIMPROF identifies groups of species which covary coherently through the time-series. The time-series is characterised by a dramatic decline in abundances and diversity between the 1993 and 1996 surveys. By 1998 there had been a shift in community composition from the 1993 situation, with different species dominating. Communities had recovered in terms of abundance and species richness, but different species dominated the community. No single factor could be identified which unequivocally explained the dramatic changes observed in the SMNR. Possible causes were the effects of dispersed oil and dispersants from the Sea Empress oil spill in February 1996 and the cessation of dredge-spoil disposal off St Anne’s Head in 1995, but the most likely cause was severe weather. With many species, and a demonstrable recovery from an impact, communities within the SMNR appear to be diverse and resilient. If attributable to natural storms, the changes observed here indicate that natural variability may be much more important than is generally taken into account in the design of monitoring programmes.