A weighted—averaging regression and calibration model for inferring total phosphorus concentration from diatoms in British Columbia (Canada) lakes

Abstract

SUMMARY1. The relationship between surficial sediment diatom taxa (Bacillariophyceae) and measured limnological variables in forty‐six British Columbia lakes was explored using canonical correspondence analysis (CCA). Lake‐water total phosphorus concentration (TP), maximum lake depth, conductivity, and calcium concentration each accounted for independent and statistically significant directions of variation in the distribution of diatom taxa.2. Weighted‐averaging (WA) models were developed to infer lake‐water TP from the relative abundances of 131 diatom taxa in the surficial sediments of thirty‐seven lakes. WA regression and calibration with classical deshrinking provided the best model for TP reconstructions.3. Our quantitative inference model has two major advantages over existing multiple linear‐regression models: (i) inferences are based on the responses of individual taxa to TP, and do not involve grouping the taxa into a small number of ecological categories; and (ii) the model assumes that diatoms respond to TP in a unimodal, rather than a linear, fashion.4. The WA model can now be used to infer past lake‐water TP, within the range 5–28νgr1−1, from diatoms preserved in the sediments of British Columbia lakes. The model can provide quantitative estimates of the onset, rate, and magnitude of lake eutrophication in response to natural processes and human disturbances.