Warming-induced positive age trends challenge MXD detrending

Raw measurements of maximum latewood density (MXD) typically decline with cambial age. This small but persistent trend may diminish or even reverse in living trees under global warming. We here present such a dataset from the Harz mountains in northern Germany that is characterized by an overall positive age trend retained in 92 MXD series of differently old Picea abies trees spanning the period from 1809-2020 CE. The positive age trend creates a new situation in which the identification of a temperature signal based on deviations from a horizontal line (e.g. by disabling positive slopes in ARSTAN; Cook, 1985) results in an underestimation of anthropogenic warming, because the residual between the horizontal line and presumed negative age trend remains unconsidered. We approach this problem by testing a range of detrending methods including classic curve fits, signal free (SF), and regional curve standardization (RCS). While none of the methods produces a chronology that perfectly depicts anthropogenic warming, the smallest offset is found when applying RCS to a trimmed MXD dataset. However, the selection of detrending methods based on the fit with instrumental target data, without prior understanding of their suitability, reduces the independency of the proxy record and causes statistical overfitting of the calibration model. Awareness of this problem is important in a rapidly warming world, in which positive age trends in MXD and other temperature-sensitive proxies are likely to increase.