IMAGe SeminarForward and inverse modeling the nonlinear relationship between
University of Arizona
Natural proxy records are generally the result of lossy, nonlinear interactions between multivariate climate and the biological or chemical recording process of the specific proxy archive. The width of tree rings depends on the relative availability of temperature and moisture resources integrated over time. Yet regression-based climate reconstructions typically treat tree-ring width as information-preserving and linearly related to either temperature or moisture-related quantities.
I present a simple yet biologically-motivated nonlinear forward model for tree-ring width as a function of monthly temperature and precipitation inputs. I validate the model, called VS-Lite, by comparing simulations of a network of ring-width chronologies across the continental United States with observed chronologies at the same locations. I will also present a Bayesian hierarchical modeling solution to the inverse problem of reconstructing climate from ring width data. VS-Lite is used to describe the climate-proxy relationship within the Bayesian hierarchy. Like the real-world ring width recording process, VS-Lite is not a one-to-one mapping between the space of climate histories and the space of modeled ring width series. The resulting nonidentifiability in our inverse statistical model is partly overcome when VS-Lite is combined with prior climatic information encoded in the process-level of the Bayesian hierarchy. The rest is translated into a rigorous and realistic representation of uncertainty in the resulting estimates of climate given the objective limits of real-world proxy data.
Friday, August 24, 2012
Mesa Lab Damon Room
Refreshments at 9:45
Lecture at 10:00 am