Assessing the potential for developing internal diatom-based transfer functions for Lake Baikal

We explore possible quantitative relationships between diatom species and environmental data for Lake Baikal using multivariate techniques. Our approach differs from published studies in other regions (on training sets and transfer functions) because (1) although only one lake is examined, we use the internal lake gradients rather than gradients among lakes and (2) the majority of the dominant diatom taxa are endemic. Canonical correspondence analysis on 93 surface sediment diatom assemblages reveals that major taxa show distinct relationships with measured environmental variables. Five significant variables were identified: snow thickness on ice (which accumulates on the frozen lake surface between January and April/May), water depth, suspended matter, annual solar radiation, and mean July temperature of surface waters. The strongest relationship is with snow thickness, which influences light levels in the water below the ice. A diatom-based inference model has been developed to predict snow thickness on the ice using weighted averaging (WA) and WA with tolerance down-weighting (WA _tol) models. Results suggest that Lake Baikal diatom assemblages in surface sediments can be used to develop a robust model for estimating snow thickness across frozen Lake Baikal (r ² _jack = 0.607, RMSEP = 0.138 log cm). This model has potential applications for recent paleoclimate studies in this region of continental Eurasia.