`Electronic subsampling' of macrobenthic samples: how many individuals are needed for a valid assessment result?

We developed and tested an `electronic subsampling technique' with benthic invertebrate samples taken in three German stream types to investigate, how strongly the number of individuals analysed influences the results. For each of 152 samples (`reference samples') 100 subsamples of the sizes 100, 200, 300, 500 and 700 individuals were generated randomly. To evaluate subsample deviation from the reference sample 45 metrics were calculated. In general, the variability of metric results increases with decreasing subsample size. Individual metrics show different sensitivity to decreasing subsample size. Three of the metrics tested (German Saprobic Index, German Fauna Index and Ecological Quality Index using Macroinvertebrates) are part of the German AQEM assessment system, for which they are transferred into quality classes. More than 40% of the 100-individuals subsamples are classified into a different quality class compared to the reference samples, but less than 20% for 700-individual subsamples. A certainty >20% is obtained with a subsample size of 300 individuals in lowland streams, whereas 700 individuals are needed to achieve the same level of confidence in mountain streams. Metrics, which rely on absolute abundances or abundance classes (e.g. BMWP, number of taxa) show higher sensitivity to a changing number of individuals than metrics, which depend on relative abundances (e.g. [%] Lithal preferences, [%] of Gatherers/collectors). Thus, the reliability of the metrics is related to subsample size, stream type and metric type. [ABSTRACT FROM AUTHOR]