Key message We found that indiscriminately using tree-ring MXD data with inhomogeneous temporal distribution from different elevations might cause biased chronologies. A mean-adjusting method was developed to overcome this bias.
Abstract
Here we analyse maximum latewood density (MXD) of Pinus sylvestris L. (Scots pine), from deadwood (dry) and subfossil wood (from lakes) collected along an elevation gradient in and close to the central Scandinavian Mountains, in the province of Jämtland, Sweden. Focusing on two common time periods (900–1150 CE and 1300–1550 CE), the mean absolute MXD of deadwood samples showed an inverse relationship with elevation, i.e. the absolute MXD decreases with elevation. However, the MXD values of the subfossil samples did not show such a consistent relationship with elevation. It was also noted that the differences in mean absolute MXD values among sites of different elevations in a given period were larger than among sites of similar elevation between the two time periods, where the former was assumed to be warmer than the latter. Using a theoretical model and a real example, it was shown that indiscriminately using MXD data with inhomogeneous temporal distribution from different elevations may cause biased chronologies, which can have significant effects on subsequent interpretations of past climate variability. A mean-adjusting method was developed to overcome this bias, and its usefulness was demonstrated by comparing two chronologies built on mean-adjusted and unadjusted MXD samples. It was concluded that unadjusted samples from different elevations with inhomogeneous temporal distribution can distort the long-term trend in a final chronology, while this bias can be alleviated if mean-adjusted samples are used.