Fig. 3

Quantitative relationships between trawl biomass and eDNA copy numbers of Atlantic cod. Panel (a) shows regional rank correlation. Regions of highest biomass were ranked in descending order and plotted against regions of highest eDNA copy numbers (R² = 0.64, P = 0.0102). The blue solid line denotes 1:1 relationship. In panels (b–d) the solid blue line denotes a type I (Ordinary Least Squares; OLS) regression model and the red line a type II (Ranged Major Axis; RMA) regression model. Shaded grey area is the 95% confidence interval calculated from type I regression. Type I regression assumes the biomass determined from the trawl survey as an independent and true reference value and thus regression error is associated only with eDNA concentrations. Type II regression takes into consideration that both trawl determined biomass and eDNA concentrations are field variables each containing variance (Legendre et al. 2018)⁴⁴. Panel (b) shows the correlation between regional sums of Atlantic cod biomass obtained from trawl surveys (kg) and eDNA quantities (copies). Type I OLS model (y = 33,888 × − 25,416; r² = 0.76; p = 0.003). Type II RMA model (y = 37,709 × – 34,009; r² = 0.79, p = 0.003). Error bars on y-axis are propagated errors for region sums determined from the analysis of technical replicates and represent ± 1 sd. Panel (c) shows the correlation between CPUE for trawl survey (kg h⁻¹) and sampling effort normalised eDNA quantities (copies L⁻¹). Type I OLS model (y = 8487 × −948; r² = 0.66; p = 0.008). Type II RMA model (y = 10016 × – 3540; r² = 0.71, p = 0.008). Error bars on y-axis are propagated errors for copies L⁻¹ determined from the analysis of technical replicates and represent ± 1 sd. Panel (d) shows the correlation between CPUE for trawl survey (kg + 1 h⁻¹) and eDNA concentrations (copies L⁻¹) for individual stations. Type I OLS model (y = 6596 × + 1890; r² = 0.15; p = 0.01). Type II RMA model (y = 12910 × – 7998; r² = 0.18, p = 0.01).