20TH-CENTURY SALINITY AND WATER-LEVEL FLUCTUATIONS IN DEVILS LAKE, NORTH-DAKOTA - TEST OF A DIATOM-BASED TRANSFER-FUNCTION

Diatom surface-sediment assemblages and associated water-chemistry data are used to develop a transfer function for the quantitative reconstruction of salinity change in the northern Great Plains of North America. The transfer function, derived from canonical correspondence analysis (CCA) of the surface-sample data, is applied to a core from Devils Lake in sediments encompassing the last century. The accuracy of the calibration equation is tested by comparing the diatom-inferred salinity with documented salinity and water-level data recorded for the lake over the last 100 yr. The diatombased reconstruction shows good agreement with the historical data, particularly in freshwater and at low to moderate salinities. Discrepancies between the measured and diatom-inferred salinity during high-salinity periods may reflect inadequacies with the inference method itself or more likely problems in the sedimentary record of the lake related to poor diatom preservation, sediment mixing and reworking, and dating inaccuracies. This diatom-based inference method for salinity reconstruction represents a powerful tool for paleohydrologic and paleoclimatic reconstruction in semiarid and arid regions.