Patterns of diatom species distribution in relation to total N (TN), total
P (TP), and other environmental variables from riffle sites on 2 streams in
southern Ontario, Canada, were determined using canonical correspondence a
nalysis (CCA). Relationships with TN and TP were sufficiently strong to dev
elop weighted-averaging (WA) regression-calibration models for inferring st
ream water concentrations of these nutrients. The models were accurate with
in +/-2.4 mu g/L for TP (apparent r(2) = 0.52) and +/-2 mg/L for TN (appare
nt r(2) = 0.53). An evaluation of the goodness of fit of these models with
and without bootstrapping indicated that they performed better (bootstrappe
d r(2) = 0.44 for TP and bootstrapped r(2) = 0.42 for TN) than published TN
and TP inference models for which similar assessments were made. Based on
Organisation for Economic Cooperation and Development eutrophication rating
s, the TP model predicted 76% of the mesotrophic and 57% of the eutrophic s
amples correctly. The model correctly predicted only 20% and 33%, respectiv
ely of oligotrophic and hypereutrophic samples. WA inference models were im
proved when seasonal variation was removed by using mean summer water quali
ty and diatom data (apparent r(2) = 0.76 and bootstrapped r(2) = 0.61 for T
P; apparent r(2) = 0.82 and bootstrapped r(2) = 0.70 for TN). Overall, rye
conclude that epilithic diatoms can be related to TN and TF using these met
hods, and that WA inference models have utility for indicating eutrophicati
on in southern Ontario lowland streams.