Evaluating diatom‐derived Holocene pH reconstructions for Arctic lakes using an expanded 171‐lake training set

Inference models from diatoms preserved in lake sediments can be used to reconstruct long‐term pH changes to better understand the process of lake ontogeny. An expanded diatom training set was developed using taxonomically harmonized modern assemblages in surface sediments of 171 lakes spanning a variety of geological and climatic settings across the Canadian Arctic. Lake‐water pH emerged as a significant variable and the most influential in structuring diatom assemblages. The resulting two‐component weighted‐averaging partial least squares pH inference model performs strongly, even after identifying effects of spatial autocorrelation at distances <20 km. The model was then applied to three dated Holocene diatom stratigraphies from Arctic regions of contrasting bedrock geology and buffering capacity, and the significance of the pH reconstructions was assessed. At Lake CF3 in a poorly buffered catchment, a gradual but significant pH decline begins 5000 years after lake inception, coincident with regional Late Holocene cooling. Reconstructions for two well‐buffered, more alkaline sites were not significant, probably due to poor analogues and other factors driving changes in diatom assemblages. Due to sparse soil and vegetation in these and other Arctic basins, bedrock composition is the most important regulator of Holocene pH, and only in poorly buffered lakes does pH primarily represent a climate signal.