A universal error source in past climate estimates derived from tree rings

Abstract. Recently it has been shown that climate estimates derived from tree rings often tend to show erroneous long-term oscillations, i.e. there are spectral biases at low frequencies. The result is independent of parameter studied (precipitation or temperature) or measured proxy (tree ring widths or maximum latewood densities). In order to find reasons for such universal errors, a new reconstruction method is introduced where no age dependence of the tree rings is determined. The aim, however, is not to generate better reconstructions but to study error variances of long-term oscillations. It is shown that paucities and data gaps due to missing trees increase the risk for erroneous low-frequency variability. A general approximate formula is introduced in order to estimate the presence of such a risk. A case study using Torneträsk data from Northern Sweden illustrates how longer periods with missing trees cause paucities and gaps leading to erroneous climatic oscillations. Systematic underestimation of the temperature around AD 1600 and after 1950 (“divergence”) is in the study case explained by such data gaps and paucities.