Physical and chemical predictors of diatom dissolution in freshwater and saline lake sediments in North America and West Greenland

Diatom dissolution in surface sediment samples from two regional lake datasets in the Northern Great Plains (NGP; n 5 64) and West Greenland (n 5 40) is assessed using a morphological approach categorizing valves during routine diatom analysis. Two dissolution indices are derived to parameterize diatom dissolution, and, when compared between two analysts in a blind test, show good correspondence and are closely correlated to diatom fragmentation. We explore the relationships between hydrochemical and physical lake parameters (including meromixis) on dissolution within both lake regions using multivariate methods and modeled with logistic regression. Salinity is the sole significant predictor of dissolution in West Greenland but salinity, carbonate concentration ([CO3 22 ]) and meromixis are significant predictors in the NGP. Limnological parameters explain 40–59% of variation in dissolution in both regions for both dissolution indices. The dissolution index methodology is applied to a short sediment sequence from Devils Lake (North Dakota), where diatom-inferred salinity inferences can be compared with a historical record of salinity fluctuations over the 20th century. Absolute errors in paleosalinity estimates are strongly correlated with diatom dissolution, with salinity overestimated in 8 out of 11 poorly preserved samples. Preservation does appear to constrain the reliability of the inferred paleosalinity at this site and may also affect the quality of diatom-based paleoenvironmental inferences elsewhere (including estimates of biogenic silica), where preservation state is often not explicitly considered. The value of lake sediments as paleoenvironmental archives is widely recognized for studies of long-term environmental change (e.g., Smol 2002; Fritz 2003). In arid and semiarid regions of the United States, Africa, Europe, and West Greenland, diatom models have been developed for quantitative reconstructions of salinity as a direct proxy for effective moisture (e.g., Fritz 1990; Gasse et al. 1995; Ryves et al. 2002) and applied to sedimentary fossil assemblages to infer the nature and variability of climate during the Holocene (Verschuren et al. 2000; Laird et al. 2003; McGowan et al. 2003). Poor preservation of diatoms as a result of silica dissolution and valve fragmentation occurs in both freshwater and saline systems, although the latter are particularly susceptible (Barker 1992; Gasse et al. 1997; Ryves et al. 2003). Recent work on both freshwater and saline lake assemblages has shown that effects are differential between species (Barker et al. 1994; Ryves et al. 2001; Battarbee et al. 2005) and can lead to unpredictable error in quantitative reconstructions (Barker 1992). While dissolution can result in the complete destruction of the diatom silica record in sedimentary sequences, partial diatom dissolution can bias assemblages to more resistant taxa with profound, but often overlooked, implications for reconstructing environmental and ecological change. Dissolution and diagenesis may also cause significant distortion to isotopic proxy records being developed from diatom silica. Understanding the information loss associated with poor microfossil preservation is fundamental to assessing the quality of such paleoenvironmental inferences. Acknowledgments We thank Emily Bradshaw, Robert Hecky, and two reviewers for many constructive comments on the manuscript. We dedicate this paper to the memory of our diatomist colleagues, John Kingston and Platt Bradbury, both of whom contributed to the early development of our work on the NGP lakes. Part of this work was carried out within NERC grant GR9/02033 to D.B.R. and R.W.B., and additionally during an NERC studentship to D.B.R. (GT4/90/ALS/28). Limnol. Oceanogr., 51(3), 2006, 1355–1368 E 2006, by the American Society of Limnology and Oceanography, Inc.