A comparison of relative abundance versus class data in diatom-based quantitative reconstructions

Abstract

Relative species abundances are the most frequently applied data type used for modern or paleolimnological diatom studies. In contrast, plant ecologists save time by commonly using ordinal scale data (class data), where the abundance of a species is estimated using dominance classes, instead of relative abundance data. This study compares the performance of models based on ordinal diatom species class data (class 1: sporadic (<0–1%) up to class 6: dominant (>60%)) with similar model types based on relative abundance data for different regional training sets and sediment cores. First, relative diatom abundances were converted into ordinal classes. Species response to total phosphorous (TP) was modelled using both types of data - relative abundance and ordinal class data. Secondly, TP was reconstructed for six sediment cores from North-East Germany, Switzerland, and Denmark using WA and WA-PLS based on both types of data. Thirdly, 20 lake sediment surface samples with known relative diatom abundances and known water TP concentrations were recounted using an ordinal data scale to create an independent test set. No significant differences were found between relative abundance and class data for (1) explained species variance, (2) reconstructed TP values, and (3) inferred TP values of the 20 recounted samples. This approach demonstrates that past TP concentrations may also be reliably reconstructed using class data instead of relative diatom abundances. Thus, by using class data lake managers may not only obtain more long-term records past water quality, but this approach is also quicker and therefore more cost effective. Moreover, the findings of this study may also advance the use of automatic diatom identification with digital image recognition, as we demonstrate that not every damaged diatom valve needs to be identified.