Presence-absence of marine macrozoobenthos does not generally predict abundance and biomass

Many monitoring programmes of species abundance and biomass increasingly face financial pressures. Occupancy is often easier and cheaper to measure than abundance or biomass. We, therefore, explored whether measuring occupancy is a viable alternative to measuring abundance and biomass. Abundance- or biomass-occupancy relationships were studied for sixteen macrozoobenthos species collected across the entire Dutch Wadden Sea in eight consecutive summers. Because the form and strength of these relationships are scale-dependent, the analysis was completed at different spatiotemporal scales. Large differences in intercept and slope of abundance- or biomass-occupancy relationships were found. Abundance, not biomass, was generally positively correlated with occupancy. Only at the largest scale, seven species showed reasonably strong abundance-occupancy relationships with large coefficients of determination and small differences in observed and predicted values (RMSE). Otherwise, and at all the other scales, intraspecific abundance and biomass relationships were poor. Our results showed that there is no generic relationship between a species’ abundance or biomass and its occupancy. We discuss how ecological differences between species could cause such large variation in these relationships. Future technologies might allow estimating a species’ abundance or biomass directly from eDNA sampling data, but for now, we need to rely on traditional sampling technology.

Fig. S1 Regional intraspecific temporal relationships for bivalve species.   . S1 Regional intraspecific temporal relationships for the bivalves species that were not plotted in Fig. 2. Abundance-occupancy relationships are shown in the left column, and biomass-occupancy relationships are shown in the right column. Each row represents one species. Each data point represents a yearly measurement of either a species' abundance (m -2 ) or biomass (g m -2 ), and occupancy (fraction of sampling stations occupied) in the entire Dutch Wadden Sea. We modelled the log10 of abundance or biomass as a function of the logit of occupancy (solid line). To assess the strength of relationships, each panel shows the coefficient of determination (R 2 ) and back-transformed Root Mean Squared Error (RMSEbt, %). Non-significance of linear models is indicated by dashed lines. Points are labelled with the last two digits of the sampling years.

Fig. S2
Regional intraspecific temporal relationships for the polychaete species that were not plotted in Fig. 3. Abundance-occupancy relationships are shown in the left column, and biomass-occupancy relationships are shown in the right column. Each row represents one species. Each data point represents a yearly measurement of either a species' abundance (m -2 ) or biomass (g m -2 ), and occupancy (fraction of sampling stations occupied) in the entire Dutch Wadden Sea. We modelled the log10 of abundance or biomass as a function of the logit of occupancy (solid line). To assess the strength of relationships, each panel shows the coefficient of determination (R 2 , proportion) and back-transformed Root Mean Squared Error (RMSEbt, %). Non-significance of linear models is indicated by dashed lines. Points are labelled with the last two digits of the sampling years.

Fig. S3
Local intraspecific temporal relationships. Abundance-occupancy relationships are shown in the left column, and biomass-occupancy relationships are shown in the right column. Each row of two panels represents one species. The lines represent tidal basins and are based on fitting linear regressions to eight data points of yearly measurements (2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015) of abundance (n m -2 ) or biomass (g m -2 ), and occupancy (fraction of sampling stations). To assess the variability and strength of relationships, each panel shows the mean coefficient of determination (R 2 ) and back-transformed Root Mean Squared Error (RMSEbt, %) with their standard deviation between brackets. Tidal basins are numbered (as shown in Fig. 1A) from west (dark red) to east (blue). Note that abundance-occupancy relationships were not estimated if a species was not observed in more than two out of eight years (see methods).

Fig. S4
Spatial intraspecific relationships. Abundance-occupancy relationships are shown in the left column, and biomass-occupancy relationships are shown in the right column. Each row of two panels represents one species. The lines represent years and are based on fitting linear regressions to ten data points (one for each tidal basins, Fig 1A) of abundance (m -2 ) or biomass (g m -2 ), and occupancy (fraction of sampling stations occupied). To assess the variability and strength of relationships, each panel shows the average coefficient of determination (R 2 , proportion) and back-transformed Root Mean Squared Error (RMSEbt, %). The values in brackets represents standard deviations. Years are coloured from 2008 (dark red) to 2015 (blue).

Fig. S5
The difference in strength of intraspecific relationships according to the scale of the analyses. Abundance-occupancy relationships are shown in the left column, and biomassoccupancy relationships are shown in the right column. Each row represents a spatiotemporal comparison of relationships: comparing the temporal relationship within tidal basins and that of the entire Dutch Wadden Sea (small vs regional), comparing spatial relationships with temporal relationships for the entire Dutch Wadden Sea (spatial vs regional), and comparing spatial relationships with temporal relationships within tidal basins (spatial vs local). For each species and comparison, we calculated the difference (in percentage points) for the backtransformed Root Mean Squared Error (RMSEbt) and coefficient of determination (R 2 ) by subtracting the statistics of the first mentioned scale from the second. For example, positive values in the first row of panels indicate that the local temporal relationship were stronger than the regional temporal relationships. For comparisons of temporal relationships within tidal basins, and spatial relationships between tidal basins, we used median values of R 2 and RMSEbt. To guide the eye, we indicate the bivariate distribution of change (ellipse) with its centroid (cross), and indicate an increase in the strength of a relationship (shaded area). To indicate the absolute strength of relationships, we present filled symbols if the relationship has an RMSEbt below and R 2 above 50%. Different plotting symbols represent different taxonomic groups (see legend). # fit non-linear smoother smoother <-loess(log_afdm ~ log_L, span=span, data=d2, control = loess.control(surface = "direct")) #calculate residual flesh mass R<-d2$rel_afdm <-resid(smoother) ## identify outliers #hist(R, n=30) resid.q <-quantile(R,prob=c(0.25,0.75)) #calaculate quantiles iqr <-diff(resid.q) #calculate Inter Quartile Range limits <-resid.q + criterium*iqr*c(-1,1) #abline(v=limits, col=2, lty=2) score <-(pmin((R-limits[1])/iqr,0) + pmax((R -limits[2])/iqr,0)) d2$outlier<-abs(score)>0