Fig. 3.

Summary of the effect of using cross-habitat weights (taxonomic weights different to the sample habitat). Light grey numbers show F-measure as a percentage. Bespoke weights (when taxonomic weights match sample habitat) are always superior. Occasionally (8 times) cross-habitat weights achieved higher accuracy than weights that were averaged across the 14 EMPO 3 habitats. Frequently the cross-habitat weights were better than uniform weights (which is current best practice, 109 times). Occasionally the uniform weights outperformed the cross-habitat weights (65 times). Source data are provided as a Source Data file