Journal of Paleolimnology 14: 49--67. 1995. 49 © 1995 Kluwer Academic Publisher.f. Printed in Belgium. An expanded weighted-averaging model for inferring past total phosphorus concentrations from diatom assemblages in eutrophic British Columbia (Canada) lakes Euan D. Reavie', Roland I. Hall',2 & John P. Smol' 1 Paleoecological Environmental Assessment & Research lLJb (PEARL), Department of Biology, Queen j. University, Kingston, Ontario, Canada, K7L 3N6 2present address: Ontario Ministry of Environment & Energy, Science & Technology Branch, Aquatic Science Section, Bellwood Acres Rd., P. 0. Box 39, Dorset, Ontario, Canada, POA J EO Received 12 August 1994; accepted 19 December 1994 Key words: diatoms, eutrophication, lake management. paleolimnology, British Co lumbia, lakes. phosphorus, training sets Abstract Eighteen lakes were added to a published training set of 46 British Columbia (BC) lakes in order to expand the original range of total phosphorus (TP) concentrations. Canonical corresponden ce analysis (CCA) was used to analyze the relationship between diatom assemblages and environmental variables. Specific conductivity and {TP] each exp lained significant (PSO.05) directions of variance in the distribution of the diatoms. The relationship between diatom assemblages and [TP] was sufficiently strong to warrant the development of a (WA) regression and calibration model that can be used to inf er past trophic status from fossil diatom assemblages. The relationship between observed and inferred [TP] was not improved by the addition of more eutrophic lakes, however the [TP] range and the number of taxa used in the transfer function are now superior to the original model. Diatom species assemblages changed very little in lakes with TP concentrations greater than 85 j.Lg 1-1, so we document the development of a model co ntaining lakes with p,g 1- 1. The updated model uses 59 training lakes and covers a range of species optima from 6 to 41.9 J.Lg 1-1 TP, and a total of 150 diatom taxa. The updated inference model provided a more realistic reconstruction of the anthropogenic history of a highly eutrophic BC lake. The model can now be used to infer past nutrient conditions in other Be lakes in order to assess changes in trophic status. Introduction Water quality has declined in many western Canadi- an lakes as a result of anthropogenic nutrient loading. These lakes have often been subject to perturbations since European settlement began in the 1800s, and in some areas, Indian settlements may have affected water quality well before this time (Woolliams, 1979). However, water quality information measured before 1950 is scarce or lacking for most systems, and realistic estimates of previous ecological conditions are often only available in the form of proxy data from the sed- iment record. Paleolimnological approaches can vide information on the causes, timing and extent of ecological disturbances. Such data are often necessary for proper management decisions (Smol, 1992). More- over, with knowledge of the pre-disturbance state, istic restoration goals can be estab li shed. The siliceous remains (valves) of diatoms have been widely used as biomarkers to elucidate environ- mental conditions in lakes (e.g. Dixit et aI., 1992). Recently, ecologicall y relevant statistical methods have been developed for inferring environmental ditions from diatom assemblages. These methods are based on multivariate ordination and weighted aging rNA) regression (that effectively estimates the optima and tolerances of several species) and calibra- tion (Birks et at., 1990a). Ecological parameters of the