Figure 3
From: Defining thresholds of sustainable impact on benthic communities in relation to fishing disturbance
Comparison of the effect size of fishing intensity (FI) vs natural time variation on the abundance and biomass of epifaunal taxa (top row, a–e), the presence of epifaunal taxa (midle row, f–j) and the abundance and presence of infaunal taxa (bottom row, k–o). Some taxa have a different response in sand and gravel as indicated. For each row, the first panel (a,f,k) is the pre-fishing observed estimate ± standard deviation (density is given for data analyzed as presence/absence only); densities and biomasses are log-transformed in the top and bottom panels using log(x100), with original data in number or g/100 m2. The second panels (b,g,l) are the relative changes due to the interaction between FI and survey time and the third panels (c,h,m) the survey effect alone, both from the GLMM outputs using continuous FI in the predictor variables. The fourth panels (d,i,n) are the taxon-specific FI thresholds and the fifth panels (e,j,o) the relative change at those thresholds as estimated from the GLMMs outputs using the categorized FI. Note that for presence this is the change in odds (i.e probability of presence compared to probability of absence). Light grey = before to after fishing (March to May), dark grey = before to four months after fishing (March to September). Estimates shown in the second, third and fifth panels are the effect sizes, exp(\(\beta \)) from eq1, with 90% confidence interval (α = 0.1). Abundance was modeled with Poisson or negative binomial, biomass with gamma distributions and presence with binomial (log-log link) distributions GLMMs. The vertical dash lines are ±0.5. 1 means no change, ±0.5 means ±50% abundance, biomass or odds. Missing values in (e,j,o) are taxa for which the model failed to converge.
