Unravelling heterogeneity of soft bottom communities in littoral zone using cluster analysis: Methodical recommendations

• Published in: Journal of sea research 2019-04, Vol.146, p.46-54
• Main Authors: Filippova, Nadezhda A., Kozin, Vitaly V., Gerasimova, Alexandra V., Maximovich, Nikolay V.
• Format: Article
• Language: English
• Subjects: Benthic communities; Benthos; Biota; Cluster analysis; Communities; Community structure; Heterogeneity; Intertidal environment; Intertidal zone; Littoral environments; Littoral zone; Long-term changes; Macrobenthos; Methods; Procedures; Reliability analysis; Sampling design; Similarity; Soft bottom communities; Structures; The White Sea
• ISSN: 1385-1101; 1873-1414
• DOI: 10.1016/j.seares.2019.01.009

Study of spatial-temporal heterogeneity of the marine benthos is often done using the classification methods, cluster analysis in particular. To date, numerous procedures for determining the number of groups, present in a data set, have been proposed. As a rule, these methods are based on averaged characteristics of the structure of benthic communities, not taking into account the variability of biota abundance or biomass among individual samples. But the number of replicates can strongly affect similarity measures, especially ones, that have often been favored in field studies. The main aim of the present research was to study the effect of different number of replicates on the results of cluster analysis of the benthic community structures. Objects of investigations were the typical soft-bottom associations in the intertidal zone of the White Sea, occupying 9 locations and studied since 2008. The repeated comparisons of community structures described by different number of replicates (10 to 1) in each location were performed. The study showed that the weaker the differences between compared communities, the greater amount of samples were needed to obtain a reliable result. When comparing community structures from different locations at the same year as few as 5 samples were sufficient to obtain relatively reliable community descriptions. However, if the task was analysis of long-term changes in the community structure in the same site, the number of samples should be increased to 11. The reliability of the cluster analysis results depended on 1) similarity level, at which all stations merged into a single group, 2) similarity level, at which clusters were identified, and 3) ratio of these similarity levels. •Cluster analysis results depends on the sampling design.•The weaker the differences between compared communities, the more samples are needed.•Rarely can reliable community comparison be made with less than five samples.•Revealing inter annual changes in community structure requires >10 samples.