An Approach for Quantifying and Correcting Sample Size-Related Bias in Population Estimates of Climate-Tree Growth Relationships

摘要

Dendroecology is based on the estimation of target population climate sensitivity from a finite number of trees (N). Recent studies showed a sample size-related bias in the estimations of climate-tree growth relationships, decreasing sample size leading to a weakening of the bootstrapped correlation coefficients. The present analysis points out that the bias equals the squared root of the expressed population signal of the growth chronology built from N trees and then proposes a correction factor to accurately estimate the population sensitivity to climate. The interests, limits, and implications of this correction are illustrated from 504 individual growth chronologies of silver fir (Abies alba [Mill.]) sampled in the Jura Mountains (France) along an altitudinal gradient of increasing climate forcing. This data set was split into three groups of 168 trees (low, medium, and high elevation). Our results show that the signal common to all trees strengthened with increasing climate forcing and that the accuracy of the correction slightly decreased with both decreasing sample size and climate forcing. Corrected bootstrapped correlation coefficients still underestimated the strength of the population climate-tree growth relationships when less than 10 trees were used at low elevation against 4 at high altitude.