Calibration of Littoral Diatoms to Water Chemistry in Standing Fresh Waters (Flanders, Lower Belgium): Inference Models for Historical Sediment Assemblages

Abstract

Relationships between littoral surface-sediment diatom assemblages and ambient limnological conditions were examined in 186 lentic fresh waters throughout lower Belgium (Flanders). Most of these waters were small, unstratified, alkaline and rich in nutrients. Using weighted-averaging techniques, robust and accurate transfer functions were developed for median pH-values ranging from 3.4 to 9.3 and dissolved inorganic carbon concentrations from <1.6 to 63 mg l−1 (jackknifed r 2≈ 0.87, RMSEP <10% of the observed range), while a less precise model was obtained for sodium (2–571 mg l−1; jackknifed r 2 0.69, RMSEP 9.9% of the range). Restricting the data set to circumneutral and alkaline sites (pH≥6.5) revealed the importance of additional variables, including calcium, silica, chemical oxygen demand and potential gross oxygen production (a proxy for metabolic activity and phytoplankton abundance). Calibration models for these variables were strong enough to be useful (jackknifed r 2 0.57–0.59, RMSEP 13.1–16.4% of the observed range), although estimations should not always be considered entirely independent. Except for the predominant pH gradient, removal of all taxa with a distribution unrelated to the variable of interest improved model performance. In general, such taxa were proportionally represented among taxa classified according to their principal habitat. Application of the present models to diatom assemblages of shallow-water sediments obtained from historical samples and, most importantly, herbarium-macrophyte specimens, will improve hindsight into regional freshwater conditions and add to base-line setting of ecological quality standards in a highly impacted region.