Diatom dissolution in saline lake sediments : an experimental study in the Great Plains of North America
Environmental reconstructions are limited by the quality of the original data from which they are derived. In situations where microfossils are poorly preserved problems may arise, both through taxonomic uncertainty and more subtly from the alteration of the death assemblage as a result of the differential robustness of species. Diatom dissolution tends to be a particular problem in saline lakes. Laboratory-based dissolution experiments on fresh, modern diatoms collected from lakes in North and South Dakota and Saskatchewan were carried out to establish the abundance and composition changes of assemblages as dissolution progressed. Analysis of experimental results demonstrates that species exhibit regular dissolution relationships and can be ranked according to susceptibility to dissolution. Changes in valve morphology for selected key taxa were categorised under scanning electron and light microscopy into 'dissolution stages'. These data provide the basis for developing dissolution indices for individual taxa and assemblages, which can be related to absolute abundance changes of diatom valves. Experimental data were applied to two separate weighted averaging (WA) transfer functions to predict measured dissolution parameters (such as dissolved silica) and to a salinity transfer function developed from the Northern Great Plains. In the former case, models incorporating dissolution stage counting were more accurate and robust (as validated by jackknifing). Species and samples were downweighted according to species robustness (dissolution rank) within the WA transfer function. Downweighting either, or both, species and samples in the transfer function algorithm lead to minor improvements in model performance in terms of both r2 and standard error (as RMSE), despite incomplete coverage of species. A short core from Spiritwood Lake, North Dakota, was used to test the differences variable weighting had on reconstructed salinity. Results suggest Spiritwood Lake is only responsive to more extreme climatic events, and has remained fresh (<0.5g/l TDS) or subsaline (0.5-<3g/l TDS) throughout the last 150 years. The approach of variable sample and species weighting to the rest of the NGP surface sediment assemblage training set may improve the model further, which could be tested at sites with an historical record of salinity.