Methods to merge overlapping tree-ring isotope series to generate multi-centennial chronologies

Stable isotopes ratios in tree-rings are increasingly used to assess long-term environmental variation. As with conventional tree-ring investigations, crossdating overlapping sequences of wood material allows chronologies longer than the lifespans of individual trees to be developed. We measured data from four overlapping cohorts of recent (AD 1650–2004) and historic (AD 1100–1198, AD 1150–1325, AD 1256–1700) wood to yield millennial-length δ13C, δ​D and δ18O chronologies. However, we found that offsets in absolute isotopic values between overlapping segments challenge the preservation of low frequency climate signals. In this paper, we describe the development of these records, signal strength assessments, and present the problems associated with overcoming these offsets and preserving low-frequency variance. We then propose and evaluate four different solutions to link the cohorts. These methods include various types of transformations of mean and variance; all present potential advantages and drawbacks. Inflated variance, loss of absolute isotope values, and the loss of low-frequency signals are major problems. Only one method is capable of fully preserving low-frequency trends, but this method may also risk inducing long-term artifacts. While the various solutions we propose may be useful to help overcome such biases, the challenges described here, if typical for other isotopic series, will best be solved by additionally modifying sampling and measurement procedures. We recommend avoiding pooling of tree rings in the overlap period and to increase the replication in order to extract the original signal and to maintain low-frequency signals.

► We present the first published methods to merge isotope chronologies from tree-rings.
► Five trees per chronology is not sufficient to merge overlapping isotope chronologies.
► No preservation of low frequency signals exceeding the cohort segment with our method.
► The proposed sampling strategies may preserve low-frequency signals in future studies.