Utilising historical tree-ring data for dendroclimatology: A case study from the Bavarian Forest, Germany

Abstract

Summary Although there are abundant historical tree-ring (TR) data available in Europe, these data have rarely been used for dendroclimatic reconstruction. Dendro-historical dating is possible because of a strong common signal in the TR data, which, if consistent over large areas, must represent a common climatic forcing upon growth. The most commonly cited limitations to the use of historical TR data for climate reconstructions are (1) mixed climate signals within the living TR data used to crossdate the historical TR data (e. g. European Oak) and (2) the unknown provenance and/or site ecology of the historical wood. In addition, as historical timbers were often taken from relatively young trees, they may have a different climate response from the mature living trees used for calibration studies. The short mean segment length of the historical timbers may also limit the potential to recover low frequency growth trends from these series. In this study, we address these issues using a well-replicated TR data-set comprised of 678 living and historical TR width series of Norway spruce (Picea abies (L.) Karst) sampled in the Lower Bavarian Forest region in south-east Germany. The historical TR series were collected from 40 buildings in towns along the Danube River. Local knowledge suggests that almost all wood used for construction prior to the 20th century had a local origin. The historical spruce TR series correlate with living spruce chronologies from sites located below 700 m, but do not correlate with spruce chronologies from above 1050m. Comparison of the statistical characteristics of chronologies from historical timbers (e. g. ring-width, mean sensitivity and correlation with high elevation spruce chronologies) with those from living trees at low elevation sites indicated that the timbers were derived from ecologically similar, low elevation sites. As RW series of these living-tree sites contain a strong spring/summer precipitation signal, it is suggested that the historical TR data are similarly climate-sensitive. Differences in age-related response to climate were addressed by comparing the response of TR chronologies composed of trees 110 years with spring/summer precipitation. Results from a moving window regression procedure indicate no significant difference in the climate response of these two chronologies. The segment length problem was examined by dividing the 678 TR series into 3 age groups and applying regional curve standardisation. We demonstrate that multi-centennial trends can be captured from these data even when sample lengths of