Refining the statistical parameters for constructing tree-ring chronologies using short-lived species: Alder (Alnus glutinosa Gaertn)

The dendrochronological potential of short-lived species has had varying degrees of success in the past. Where there has been a level of success with short sequenced assemblages, the focus has been on visual comparisons, based on the occurrence of signature rings. Of vital importance to alder’s ability to be cross-correlated is that it produces a significant amount of distinguishable signature rings. Between 2012 and 2013, a large artificial island (crannog) of medieval date was excavated at Drumclay, County Fermanagh, Northern Ireland, and revealed a site of significant longevity, dating from the 9th century AD to the Post-Medieval period. This excavation exposed a vast number of well-preserved waterlogged archaeological features, resulting in the retention of over 9,000 individual wood samples. Oak timbers were used scarcely in the construction of the crannog, with the dominant wood species identified during excavation being alder. While the oak timbers have proved successful in providing spot dates and indicating phases of activity, the full chronological potential of the wood assemblage lies in the ring patterns of the principal species, alder, particularly with respect to understanding construction phases and site evolution. Previous failures to build chronologies using alder have been attributed to the short-lived and site specific nature of the species. Here, we test whether the measurement of large numbers of samples from a single context within a single site overcomes the limitations posed by alder. We measured the ring-widths of 90 alder samples from archaeological features within the crannog’s infrastructure to test if a robust context chronology could be built. The average ring sequence length ranged from 30 to 60 rings, with one timber extending to 108 rings. Visual correlations were used to aid ring pattern matching in conjunction with statistical correlation. We used radiocarbon wiggle-matching to test the robustness of our constructed chronology and to anchor it to an absolute timescale. Our results to date show that problems of autocorrelation can arise when long alder sequences (>100 rings) are used in conjunction with short sequences (30 to 60 rings). Establishing a rigorous protocol for sample selection has enabled us to develop a more statistically refined methodology that has produced t-values as high as 8.3. We show that in order to construct the best possible alder chronology, multiple ring patterns need to be examined from each context. We recommend examining short-lived assemblages on their own merits; the best approach in these cases is not to look for the longest sequences but instead to focus on the those from the mean sequence range.