Principal component analysis applied to dendrochronology

This work uses samples of the species Ocotea porosa (Nees & Mart) Barroso (Imbuia), collected in the city of General Carneiro, Southeast region of the State of Paraná (26o24'01 25"S 51o24'03 91"W), Brazil, to generate average chronology (GC index) of this region. The objective of this article is to remove the natural growth trends of trees using a tool that is still little explored for this purpose, Principal Component Analysis (PCA). In each tree sample, the width of each growth ring was measured, obtaining a time series (1 ring per year). The samples were selected using Cluster Analysis, which classifies samples based on their similarities. Once the Principal Components (PCs) were obtained, the dendrochronological series were reconstructed without the first PC. This methodology is an estimate of the trend that best represents the natural growth of all trees on the site. The arithmetic mean of the series without the 1st PC is the GC index. It was found that PCA has three benefits: fast data processing, preservation of low-frequency signals and, when integrated with a powerful tool, the Alternated Least Squares (ALS) method, missing data estimation.