Extracting a common high frequency signal from Northern Quebec black spruce tree-rings with a Bayesian hierarchical model

• Published in: Climate of the past 2009-01, Vol.5 (4), p.607-613
• Main Authors: Boreux, J.-J., Naveau, P., Guin, O., Perreault, L., Bernier, J.
• Format: Article
• Language: English
• Subjects: Bayesian hierarchical model; Continental interfaces, environment; dendrochronology; Mathematics; Mathématiques; Ocean, Atmosphere; Physical, chemical, mathematical & earth Sciences; Physique, chimie, mathématiques & sciences de la terre; Sciences of the Universe; uncertainties
• ISSN: 1814-9332; 1814-9324; 1814-9332
• DOI: 10.5194/cp-5-607-2009

One basic premise of dendroclimatology is that tree rings can be viewed as climate proxies, i.e. rings are assumed to contain some hidden information about past climate. From a statistical perspective, this extraction problem can be understood as the search of a hidden variable which represents the common signal within a collection of tree-ring width series. Classical average-based techniques used in dendrochronology have been applied to estimate the mean behavior of this latent variable. Still, depending on tree species, regional factors and statistical methods, a precise quantification of uncertainties associated to the hidden variable distribution is difficult to assess. To model the error propagation throughout the extraction procedure, we propose and study a Bayesian hierarchical model that focuses on extracting an inter-annual high frequency signal. Our method is applied to black spruce (Picea mariana) tree-rings recorded in Northern Quebec and compared to a classical average-based techniques used by dendrochronologists (Cook and Kairiukstis, 1992).