MULTIVARIABLE ENVIRONMENTAL INFERENCES BASED ON DIATOM ASSEMBLAGES FROM SUDBURY (CANADA) LAKES

1. Diatoms were identified and enumerated from a surface sediment calibration set of seventy-two Sudbury lakes, and the relationships between species composition and limnological variables were examined using canonical correspondence analysis (CCA), a multivariate technique based on a species-packing model assuming unimodal species distributions along environmental gradients. 2. Weighted-averaging regression and calibration were then used to develop predictive models for inferring lakewater pH, total [Al], [Ni], [Ca], and conductivity. 3. The applicability of the diatom inference models was further supported by similarities noted between recent diatom-inferred values to actual lakewater measurements made in the 1980s on Chiniguchi Lake.