Improved EEMD-based standardization method for developing long tree-ring chronologies

Xianliang Zhang,Xiaobo Liu,Zhenju Chen,Junxia Li

doi:10.1007/s11676-019-01002-y

Xianliang Zhang, Xiaobo Liu + Show 2 more

Open Access

https://doi.org/10.1007/s11676-019-01002-y

Copy DOI

Abstract

Long tree-ring chronologies can be developed by overlapping data from living trees with data from fossil trees through cross-dating. However, low-frequency climate signals are lost when standardizing tree-ring series due to the “segment length curse”. To alleviate the segment length curse and thus improve the standardization method for developing long tree-ring chronologies, here we first calculated a mean value for all the tree ring series by overlapping all of the tree ring series. The growth trend of the mean tree ring width (i.e., cumulated average growth trend of all the series) was determined using ensemble empirical mode decomposition. Then the chronology was developed by dividing the mean value by the growth trend of the mean value. Our improved method alleviated the problem of trend distortion. Long-term signals were better preserved using the improved method than in previous detrending methods. The chronologies developed using the improved method were better correlated with climate than those developed using conservative methods. The improved standardization method alleviates trend distortion and retains more of the low-frequency climate signals.

Highlights

Millennia-long tree-ring chronologies can be developed by overlapping tree ring series from living trees with dead wood, construction wood or sub-fossil material through cross-dating, extending the records of living trees back in time
Climatic variations or cycles that occur at wavelengths longer than the length of tree ring series are not captured by standardized chronologies because they are identified as part of the biologically based growth trend (Granger 1966; Cook et al 1995)
The results showed that there was a loss of long-term signals with the Months pre-ensemble empirical mode decomposition (EEMD)

Summary

Introduction

Millennia-long tree-ring chronologies can be developed by overlapping tree ring series from living trees with dead wood, construction wood or sub-fossil material through cross-dating, extending the records of living trees back in time. The methods used to remove growth trends from raw measurements of tree rings are referred to as standardization methods and classified as determination or data adaptive methods In determination methods, such as the linear regression line and modified negative exponential curve methods (Fritts et al 1969), a previously defined mathematical model is used to describe the radial growth of trees. Data adaptive methods, such as the smoothing spline method (Cook and Peters 1981), fit the behavior of the observed tree ring width. Long-term climate signals are lost when developing millennia-long tree ring chronologies using these standardized methods as a result of the segment length problem (Cook et al 1995)

Objectives

Methods

Conclusion