A new approach to correct the overestimated persistence in tree-ring width based precipitation reconstructions

Abstract

Tree-ring width is one of the most widely used proxy in paleoclimatological studies. Due to various environmental and biological processes, however, the associated reconstructions often suffer from overestimated low-frequency variability. In this study, a new correction approach is proposed using fractional integral techniques that corrects for the overestimated long-term persistence in tree-ring width based hydroclimatic reconstructions. Assuming the high frequency interannual climate variability is well recorded by tree rings, the new approach is able to (i) extract the associated short-term forcing signals of various climate conditions from the reconstructions, and (ii) simulate the long-term impacts of these short-term forcings by setting a proper fractional integral order in the fractional integral statistical model (FISM). In this way, the overestimated long-term persistence, as well as the associated low-frequency variability in tree-ring width based reconstructions can be corrected. We apply this approach to a recently published dataset of precipitation field reconstructions over China covering the past half millennium and removed the redundant, non-precipitation related long-term persistence. Compared to the original reconstruction with multi-century long-term dry conditions in western China, the corrected reconstruction considerably shortened the wet/dry periods to decadal scales. In view of the widespread non-climatic/mixed-climatic signals in tree-ring widths, this new approach may serve as a useful post-processing method to reconsider previous reconstructions. It may even be combined with the current detrending approaches by upgrading the pre-whitening methods.