Separating Common Signal From Proxy Noise in Tree Rings

Abstract

AbstractTree rings are the most widely‐used proxy records for reconstructing Common Era temperatures. Tree‐ring records correlate strongly with temperature on an interannual basis, but studies have found discrepancies between tree rings and climate models on longer timescales, indicating that low‐frequency noise could be prevalent in these archives. Using a large network of temperature‐sensitive tree‐ring records, we partition timeseries variance into a common (i.e., “signal”) and non‐climatic (i.e., “noise”) component using a frequency‐resolved signal‐to‐noise ratio (SNR) analysis. We find that the availability of stored resources from prior years (i.e., biological “memory”) dampens the climate signal at high‐frequencies, and that independent noise reduces the SNR on long timescales. We also find that well‐replicated, millennial‐length records had the strongest common signal across centuries. Our work suggests that low‐frequency noise models are appropriate for use in pseudoproxy experiments, and speaks to the continued value of high‐quality data development as a top priority in dendroclimatology.