|
|
abstracts
A Bayesian approach to reconstructing climate fields from proxy data
Martin Tingley, Peter Huybers
We present a Bayesian model to assimilate incomplete (in space and time) instrumental and proxy data sets to estimate, with uncertainties, the time evolution of a climate field. The Bayesian model consists of a process level that describes the evolution of the true climate field as a multivariate AR(1) process with spatially correlated innovations; a data level that specifies the relationships between the measurements (proxies and instrumental) and the true field values; and a prior level that specifies diffuse prior distributions for all unknown parameters. Multiple draws from the posterior produce a spatially and temporally complete ensemble of field evolutions compatible with the data and the model assumptions. Probability distributions for various statistics can be estimated from this ensemble, from simple measures like the time series of spatial means to more exotic quantities like the probability that a given year featured the most extreme value of the climate field during the reconstruction.
We demonstrate the utility of this approach with two applications: 1) a 600 year surface temperature reconstruction for high Northern latitudes based on tree ring, ice core, and lake sediment core data, as well as the Climate Research Unit's compilation of instrumental observations; and 2) a 500 year drought reconstruction for the four corners region of the USA based on tree ring time series and the Palmer drought severity index over the 1990-2003 interval.
Martin Tingley, Department of Earth and Planetary Sciences, Harvard University,20 Oxford st, Cambridge, MA, 02138, United States, tingley@fas.harvard.edu
Peter Huybers, Harvard University, United States
Session: F2: Regional Climate Dynamics
Download Poster:
> PAGES_YSM09_Tingley.pdf
Download Talk:
> YSM09_OralB_Tingley.pdf
|
|
