We analyzed surface-sediment samples collected along transects from three sub-basins of a relatively large (~115 ha), bathymetrically complex lake, in northwest Ontario, Canada, to assess the reproducibility of diatom species habitats and diversity along a water-depth gradient. Transects displayed different orientations with respect to prevailing wind direction and varied in complexity and degree of slope along the lake bottom. Each transect consisted of three replicate samples at a resolution of ~1 m water depth from ~1 to 30 m for the two deep-basin transects and from ~1 to 18 m in the shallower basin. Distinct diatom assemblages were identified in all transects: (1) a near-shore community composed largely of attached life-forms and some motile benthic taxa, (2) a mid-depth community composed largely of motile life-forms and other benthic taxa that are adapted to lower light conditions (e.g. Staurosirella pinnata), and (3) a deep-water community dominated by planktonic taxa. Species richness was highest in the benthic zones (<9 m), with greatest species evenness in the mid-depth zone (~3–9 m). Species richness and evenness were highly correlated across the three transects (r = 0.89–0.93, p < 0.01). Diatom-inferred depth models were developed from the individual transects to assess reproducibility and applicability for down-core analyses using modern analog (MAT) and weighted-averaging (WA-PLS) approaches. Coefficients of determination (r2) for these models ranged from 0.80 to 0.98, and RMSEP ranged from 1.2 to 4.2 m. The models developed from the transect with the highest resolution sampling, gentlest non-complex slope and shallowest maximum depth were the strongest (\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$ r_{\text{MAT}}^{2} = 0.97 $$\end{document}; \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$ r_{\text{WA - PLS}}^{2} = 0.98 $$\end{document}) and had the lowest RMSEP (MAT = 1.2 m, WA-PLS = 1.3 m). These inference models can be used to infer past fluctuations in the depth of the benthic/planktonic boundary from cores retrieved near this ecotone and provide a sensitive record of the past change in location of the benthic zone. These types of data can be used to assess past variability in droughts and lake levels to better plan for potential future extremes. Such records incorporate more realistic estimates of natural variability than the ~100-year instrumental records currently used by water resource managers.
