Uncertainties in climate reconstructions from ice core isotopic data

Ice cores provide information on temporal variability in stable water isotopes, and often the goal is to utilize these isotopes to reconstruct climate variability over a larger region. However, it is important to quantify the uncertainties associated with the assumption that a given ice core is representative of that larger region. Stacking multiple ice core records can help minimize the uncertainties, but how many cores are necessary? A suite of fifteen firn cores were drilled within a 17,389 km^2 area in the central West Antarctic Ice Sheet, and analyzed for δ18O and δD. Here, we assess how many cores are actually necessary to confidently represent the region of interest, assuming the full stack of all fifteen cores is representative of the regional average isotopic variability. We statistically compare records of varying numbers of cores to the stacked record of all cores, removing two cores step-wise from the isotopic stack. As might be expected, a greater number of cores included in the stacked record yield a tighter spread in correlation coefficients with the full stacked record. However, lower numbers of cores included in the stacked average, while they have a larger spread in correlation coefficients, contain many combinations of cores with extremely high correlations with the full stack. Indeed, as few as four stacked cores can correlate as highly with the stack of fifteen cores as combinations of ten to twelve cores. Surprisingly, the spatial distribution of the chosen cores to include in the stacked average is not necessarily a good predictor for how well they will correlate with the regional average. One goal of this study is to determine the minimum number of cores necessary to capture the regional average isotopic variability, and a method for selecting the optimal sites to drill the cores. The results of this study will also be used to quantify the uncertainties associated with climate reconstructions from isotopic data. In addition, we will test the methods developed here with suites of ice cores from other regions to determine how applicable the methods are for other areas of the Antarctic Ice Sheet.