Quantifying long-term historical climate is fundamental to understandingrecent climate change. Most instrumentally recorded climate data are onlyavailable for the past 200 years, so proxy observations from natural archivesare often considered. We describe a model-based approach to reconstructingclimate defined in terms of raw tree-ring measurement data that simultaneouslyaccounts for non-climatic and climatic variability. In this approach we specifya joint model for the tree-ring data and climate variable that we fit usingBayesian inference. We consider a range of prior densities and compare themodeling approach to current methodology using an example case of Scots pinefrom Tornetrask, Sweden to reconstruct growing season temperature. We describehow current approaches translate into particular model assumptions. We explorehow changes to various components in the model-based approach affect theresulting reconstruction. We show that minor changes in model specification canhave little effect on model fit but lead to large changes in the predictions.In particular, the periods of relatively warmer and cooler temperatures arerobust between models, but the magnitude of the resulting temperatures arehighly model dependent. Such sensitivity may not be apparent with traditionalapproaches because the underlying statistical model is often hidden or poorlydescribed.
展开▼
