Proportion relative bias for the total abundance estimator (), as estimated from different models (indicated on rows), and as a function of different transect configurations (columns). (Each value represents median proportional bias of n= 100 simulation replicates, where the posterior predictive mean of is used as a point estimator. Models could include landscape-level covariates (collectively referred to as ‘covs’), geographical stratum (‘stratum’), a measure of sampling intensity (intended as a potential fix for preferential sampling; ‘samp.dens’) and spatially autocorrelated ‘RE’. Different flight configurations are displayed in figure 4; the first arabic symbol in the flight name relates to the number of flights (4, 8 or 12 flights), while the second reflects the amount of effort devoted to nearshore surveys (1=no coastal surveys, 2=2 coastal surveys, 3=4 coastal surveys, 4=6 coastal surveys). Blank entries indicate cases where estimates were numerically unstable or models were otherwise overparametrized.)