While attempting to assemble a reference image set of Southern Ocean diatoms for training automatic classification algorithms, we encountered numerous specimens which we were unable to classify unequivocally into one of three highly similar Fragilariopsis species. Problems about the delimitation of these species were also raised at the last Polar Marine Diatom Workshop in 2015 in Salamanca. The present study originated from these two sources. Using semi-automated microscopy and image analyses, we assembled a set of 501 specimen images and accompanying morphometric data, and 12 members of the polar marine diatomist community contributed their identification of these specimens independently from each other. After comparing the identification results themselves, we used the morphometric features extracted in an attempt to clarify the nature of morphometric distinction of the three taxa in uni-and bivariate analyses, and performed multivariate classification experiments and tested their agreement with expert consensus opinion. Beyond the specific insights into morphometric distinction of the studied taxa, our study also highlights some of the more generic challenges and possibilities of research at the interface between automatic identification and traditional taxonomy.

The zip file contains information making all processing steps taken in the paper transparent from image analysis results to statistical analyses; in detail: Subfolder with images: - "SHERPA output": all original analysis output files from SHERPA R scripts: - 1-Fragilariopsis-merge-datafiles.R: data preparation/merging - 2-Fragilariopsis-features.R: custom features (heteropolarity etc.) - 3-Fragilariopsis-plots-analyses.R: data analyses and plots presented in the paper Data files: - Fragilariopsis-IDs-final-04.04.2017.xlsx: final table summarizing all identification results - Fragilariopsis-IDs-final-04.04.2017.csv: same, in csv format, for importing into R - Fragilariopsis-SHERPA-output.csv: output of SHERPA analysis of included specimen images - Frag-3spp-all.txt: information from the above files merged, plus 60 x (X,Y) coordinates of each valve outline and 14 x 4 elliptic fourier coefficients. The file was prepared from the above files and from data under "SHERPA output" with the R script Fragilariopsis-3-spp-merge-datafiles.R - Frag-3spp-all-Gabor-2.txt: the same as above, after addition of further feature values (heteropolarity, eccentricity of broadest position, stria orientation) Variables included in the data files are explained in the README.txt.

Supplement to: Beszteri, Bánk; Allen, Claire Susannah; Almandoz, Gastón Osvaldo; Armand, Leanne K; Bárcena, María Angeles; Cantzler, Hannelore; Crosta, Xavier; Esper, Oliver; Jordan, Richard William; Kauer, Gerhard; Klaas, Christine; Kloster, Michael; Leventer, Amy; Pike, Jennifer; Rigual-Hernandez, Andrés S (2018): Quantitative comparison of taxa and taxon concepts in the diatom genus Fragilariopsis: a case study on using slide scanning, multiexpert image annotation, and image analysis in taxonomy¹. Journal of Phycology, 54(5), 703-719