Monitoring Diatom Algae in Northeastern Minnesota

Abstract

Surface sediment samples from 37 Itasca region lakes were appended onto an existing Minnesota lakes diatom calibration set to create a new 145 lake diatom training set with better representation of lakes throughout Minnesota and better reconstruction performance for total phosphorus values above 60 μg/L. Relationships among 15 chemical, physical and spatial (i.e., region) variables and 170 diatom species distributions in the new 145 lake training set were explored using canonical correspondence analysis (CCA), a multivariate ordination technique. Environmental variables that independently explained a significant portion of variation in species distribution were identified using forward selection. Total phosphorus (TP) was one of the most explanatory variables in the refined calibration set, including pH, lake depth, color, chloride, Secchi depth, conductivity, mean lake depth, lake area, and ANC. Weighted averaging partial least squares regression and calibration was identified as the appropriate method for model evaluations. A transfer function for inferring TP was generated from three versions of the lake calibration set: 1) the full 145-lake calibration set; 2) a reduced, 129-lake set where lakes with pH ≥ 9.0 removed to better represent conditions in the Itasca region; and 3) a further-reduced model that eliminated five lakes with poor diatom-inferred (DI) total phosphorus concentrations. The 124-lake calibration set produced a total phosphorus transfer function covering a gradient spanning lakes from 5-364 μg/L TP, and strong predictive ability for TP (r2 = 0.87, r2 jack = 0.77, RMSEP = 0.210 log(μg/L+1)). The transfer function was applied to estimate historical epilimnetic total phosphorus from subfossil diatom assemblages in a sediment core recovered from Jessie Lake, Itasca region. Pre-settlement conditions in Jessie Lake were inferred to be eutrophic, with a temporary augmentation in nutrient load between ~1950 and ~1980.