Species-specific effects on the oxygen isotope ratio of tree-ring cellulose

The mechanistic model of tree-ring cellulose oxygen isotope ratios (δ 18Ocell) provides a promising tool for paleo-environmental and eco-physiological studies. While the current model captures major fractionation effects leading to δ18Ocell, the full potential of the model is yet to be realized due to the existence of some weakness of the model. My dissertation was designed to address several weaknesses.;In an analysis of large-scale photosynthesis-weighted canopy temperatures (Tcanδ) pattern based on multiple datasets of tree-wood cellulose isotope ratios (δ18Owc), I confirmed the previous finding of a boreal-to-subtropical convergence of Tcanδ to around 20 °C; meanwhile, I show that the Tcanδ convergence pattern cannot be extended to the tropical regions, where Tcanδ averaged at around 26 °C, and was thus higher than the boreal-to-subtropical mean. These results caution against the common practice of taking Tcanδ to be the same as growing season air temperature in parameterizing the tree-ring model, in particular for boreal trees.;Measurements of leaf water isotope enrichment (Δ18O L) were made on multiple tree species growing either in a common garden or field environment. Using the data collected, I explored the nature of effective pathlength (L) of the Péclet effect, — a key component of the tree-ring isotope model. I revealed the common presence of an inverse relationship between L and transpiration rate (E), both at the within- and across-species levels. Such a L-E covariance pattern must be incorporated into the tree-ring model for improving its predictive power. The current tree-ring model does not explicitly capture the species effect. To build this knowledge gap, I conducted a field study to investigate the mechanisms behind the variability in late-wood δ18Ocell (δ18O lc) among three co-occurring tree species in a temperate forest. The results highlighted source water δ18O (δ 18Osw) as a key factor in explaining δ18O lc difference between black and chestnut oak. Through model-measurement comparison, I further show that the Péclet effect is only relevant to δ18Ocell of the oaks but not to pitch pine. Such a Péclet effect related difference needs to be invoked to explain the observed more enriched δ18Olc signals in pitch pine relative to the oaks. signals in pitch pine relative to the oaks.