Establishment of a 4650-year-long eigenvalue chronology based on tree-ring cores from Qilian junipers (Juniperus przewalskii Kom.) in Western China

Abstract

This paper describes a tree-ring width chronology that spans the past 4650 years, established using the recently developed eigenanalysis technique. The aim is to show whether this eigenanalysis method enables the extraction of long-term tree-growth variations that are due to climatic changes, from a large dataset comprising 1263 tree-ring width records sampled from the highlands of Western China. In order to exclude the so-called growth rate/life span association effect, tree-ring width records were sorted into six subsets, based on the life spans of the trees sampled: 200–400, 400–600, 600–800, 800–1000, 1000–1500 years old, and trees older than 1500 years. Some partial chronologies were created, by pairing the set of all tree samples (living, dead, archaeological remains) with the living trees belonging to each subset. We computed the contours of tree-growth variations (on both 100-year and longer time scales) for each subset, ending with six pairs of these partial subset chronologies. Two sums of all these partial chronologies thus yielded a record of precipitation variations over a period ranging from 2627 BCE up to 2012 CE. It was found that this record shows a high degree of similarity to the existing chronology produced using the regional curve standardization (RCS) method applied to the same dataset, indicating that the eigenvalue chronology is capable of faithfully extracting long-term climatic variations. This also confirms that the first eigenvector represents the growth pattern that is characteristic of each biologically unique tree as well as the micro-environment of each tree stand. The variations observed over the last millennium seem to be connected to a cycle of solar activity with a period of ∼200 years. However, a clear lack of correspondence between solar activity and tree growth prior to 1000 CE indicates that the recent consistency may be coincidental. We believe that the eigenanalysis technique is readily applicable to other kinds of tree-ring datasets from different parts of the world.