Reproducibility of coccolith morphometry:: Evaluation of spraying and smear slide preparation techniques

It is of great importance to assess the internal accuracy and reproducibility of coccolith morphometry, and to understand any systematic differences between various sample preparation techniques, so that data obtained with different methods can be adequately compared. Here, we performed a comparative study between two techniques regularly used to prepare nannofossil microscope slides, the standard smear slide and spraying methods. With each technique, ten replicate slides were prepared and morphometric measurements were carried out on the coccolith genus Calcidiscus as well as full assemblage counts to determine the reproducibility of coccolith size measurements and relative species abundances for each method. For either method, two significant sources of variance are related to the morphometric data, the variance between replicate slides and the variance within slides. Thus, in order to reduce the total variance of the estimate of mean coccolith size, it is beneficial to increase the number of replicate slides as well as the number of measurements on each slide. If the aim of the morphometric study is to address the mean size of a coccolith taxon or species complex (a group of closely related sub-species), one could rely on either preparation technique, since no significant difference in (log-normalised) mean size between the two tested methodologies was found. Likewise, the 10th percentiles are statistically equal for both methods. However, the 90th percentile values are significantly larger in the sprayed slides than in the smear slides, indicating that the spraying method may either favour larger coccoliths or better resolve the full extent of size variability present in the sample. Future tests are needed to investigate whether, and if so, how, size-fractionation may occur when using the spraying method. Nevertheless, the spraying method is preferred based on the reproducibility of proportion estimates since no significant difference between replicate samples was observed, in stark contrast to the smear slide series with 3-4 times higher variance. It appears that information gained from one sprayed slide requires counting at least three replicate smear slides. (c) 2005 Elsevier B.V. All rights reserved.