Inferring lake depth using diatom assemblages in the shallow, seasonally variable lakes of the Nebraska Sand Hills (USA): Calibration, validation, and application of a 69-lake training set

摘要

The Nebraska Sand Hills are a distinctive eco-region in the semi-arid Great Plains of the western United States. The water table underlying the Sand Hills is part of the High Plains/ Ogallala aquifer, an important water resource for the central Great Plains. Lake levels are affected directly by fluctuations in the water table, which is recharged primarily by local precipitation and responds quickly to climatically induced changes in regional water balance. Instrumental records are available for only 50–100 years, and paleolimnological data provide important insights into the extremes and variability in moisture balance over longer time scales. A set of 69 lakes from across Nebraska was used to establish a statistical relationship between diatom community composition and water depth. This relationship was then used to develop a diatombased inference model for water depth using weighted averaging regression and calibration techniques. Development of the inference model was complicated by strong intra-seasonal variability in water depth and the linkages between depth and other limnologic characteristics, including alkalinity, water clarity and nutrient concentrations. Analysis of historical diatom communities from eight lakes allowed for the reconstruction of lake-level fluctuations over the past several thousand years. Comparisons of the more recent portion of these reconstructions with the instrumental Palmer Drought Severity Index (PDSI) showed that sediment records may not faithfully reflect short-term fluctuations in water level, except where sedimentation rates are very high. However, large and persistent changes in moisture availability were discernible even in longer, low-resolution records. Thus, diatoms are a useful addition to the tools available for understanding past drought in the central Great Plains, especially when trajectories of change are constrained by data from multiple sites or other proxies.