Subsampling of large-volume samples in waterlogged sediments. A time-saving strategy or a source of error?

For the archaeobotanical analysis of waterlogged sediments, which are usually very rich in organic remains, subsampling is often necessary. Subsampling strategies for this kind of material are mostly based on non-empirical experience and have rarely been tested. In this paper, we compare results from small-volume subsamples with those from the large-volume residue of the master sample subsamples were drawn from. The extra-variability caused by lumpiness of the unprocessed waterlogged sediment was quantified in order to find out how much the necessary sample sizes need to increase for this type of sediment. We found that subsampling of unprocessed waterlogged material in general should only be done if it is indispensable, as it can introduce a bias. We propose methods to adapt sample sizes based on random sampling for unprocessed waterlogged sediment (where random sampling is impossible) so that the proportions or also the diversity of plant species can be estimated with sufficient precision in these sediments. However, it would be best to use an appropriate sieving method first and then subsample the processed residues of the material. We also think that it is important to clearly present the methods used for subsampling in publications and that more detailed tests about subsampling should be performed.