A weighted-averaging regression and calibration model for inferring lakewater salinity using chrysophycean stomatocysts from lakes in western Canada

Abstract

Chrysophycean stomatocysts were identified and enumerated from the surface sediments of 60 lakes located on the Interior Plateau of British Columbia. The lakes span a salinity gradient from freshwater to hypersaline (0.0–92.4 g L−1), with the majority being fresh to hyposaline. One hundred and ten stomatocyst morphotypes, almost all of which were previously described, were identified from the lake sediments. The first axis of direct gradient analysis, which was highly significant, was essentially a salinity axis (i.e. [Ca], [Mg], [K], [Na], [SO4], [DIC], and [Cl]). Most cysts were found to have fairly broad tolerances, with the narrowest tolerances occurring among morphotypes with the lowest salinity optima. Weighted-averaging regression and calibration techniques were used to develop an inference model to measure the relationship between measured average lakewater salinity and stomatocyst inferred salinity (apparent r2=0.80). Simple weighted-averaging produced a model with a lower bootstrapped RMSE of prediction than weighted-averaging with tolerance downweighting. These data indicate that chrysophyte stomatocysts are useful quantitative indicators of past lakewater salinity (in the freshwater to hyposaline range) in B.C. lakes, and can be used to strengthen the interpretations from diatom-inference models already developed from the same region.