Measuring and controlling data quality in biological assemblage surveys with special reference to stream benthic macroinvertebrates.

Summary 1. Biological assemblage surveys primarily aim to characterise species composition and relative abundance at one or more spatial or temporal scales. Data interpretation and conclusions are dependent on how well samples characterise the assemblage of interest. 2. Conventional measures of data quality, e.g. standard deviations or coefficients of variation, were designed for single variable estimation, and they are either insufficient or invalid for assessing the quality of data describing entire assemblages. Similarity indices take species composition and relative abundance into account and may be used to effectively measure and control the quality of data used to characterise assemblage structure. 3. The average Jaccard coefficient (JC) calculated across multiple pairs of replicate samples, i.e. autosimilarity JC (AJC), is conceptually and numerically related to the average coefficient of variation in the densities of all species recorded, a measure of sampling precision, and to the proportion of total species richness sampled, a measure of sampling accuracy. 4. We explored how AJC can be used to assess the effect of different potential sources of error on the quality of assemblage survey data, including the sampling effort used both within regions and at individual sites, the individuals collecting samples, sub-sampling procedures, and consistency of taxon identification. 5. We found that the autosimilarity-based approach overcomes most weaknesses associated with conventional measures of data quality and can be used to effectively measure and control the quality of assemblage survey data. [ABSTRACT FROM AUTHOR]