Determination of Similarity Cutoff Value for the Classification of Macrobenthic Communities in the Korean Coast

The selection of optimal cutoff points has been questioned in many cluster analyses. Here, we aimed to determine a percent similarity cutoff value that can be generally applied to macrobenthic community classification. Biological datasets were collected from the Chokchon tidal flat in Incheon, Korea, in July 1994, and from the subtidal, soft-bottom areas of the Korean coast of the Yellow Sea between 2006 and 2008. For the tidal flat, $${R}_{0}$$ , a similarity at zero distance apart (indicating an identical community), and $${R}_{0}/2$$ , a cutoff similarity indicating that half of the community is different, were estimated based on the linear relationship between similarity and inter-sample distance. The subtidal area $${R}_{0}$$ was estimated as a representative value of similarities among replicates within a station. The $${R}_{0}/2$$ for the tidal flat and subtidal areas were determined to be 29 and 28%, respectively. The validity of this classification method using the similarity cutoff values and similarity profile analysis (SIMPROF) was verified using independent macrobenthos and environmental datasets from 54 tidal flats. BIO-ENV analysis determined that mean grain size, sorting, and acid-volatile sulfide were selected as the best determinants of community distribution, and these were compared among biological groups derived from both classification methods. The former was found to have higher classification performance and ability to assemble homogeneous groups. We discussed the sources of variation in similarity and the necessity of allowing researchers to choose an appropriate one within a range. Considering the confidence intervals, we suggest a cutoff value range between 26 and 30%.