Bias Correction of Paleoclimatic Reconstructions: A New Look at 1,200+Years of Upper Colorado River Flow

摘要

Bias correction, while widely used with climate-model output, is not typically applied to paleoclimatic reconstructions. While many reconstruction models have low average error, they still may contain bias, especially in the tails of distributions. Bias correction, used cautiously, can be a valuable procedure that alters interpretations of past events. Analyzing the iconic tree-ring reconstruction of Upper Colorado River flow, we find that its probability distribution is markedly different from that of observed flow. Using quantile mapping to bias correct the reconstruction, we analyze the full reconstructed record and two events in particular: the 1100s megadrought and the early 1600s pluvial. Overall, bias correction made the 1100s megadrought, the largest in the 1,200+year record, even more extreme. After bias correction, the early 1600s pluvial marginally exceeds the early twentieth century pluvial in magnitude but not in duration. Overall, bias correction should be considered whenever paleoclimatic reconstructions are compared directly to observations.