HR: 1340h
AN: PP43A-0613 [Abstracts]
TI: A Detailed and Comprehensive Reconstruction of the Northern Annular Mode and Pacific North America
Pattern Using Tree Rings
AU: * Wettstein, J J
EM: justinjw@atmos.washington.edu
AF: University of Washington
Department of Atmospheric Sciences, Box 351640, Seattle, WA 98195
United States
AU: Wallace, J M
EM: wallace@atmos.washington.edu
AF: University of Washington
Department of Atmospheric Sciences, Box 351640, Seattle, WA 98195
United States
AB:
The cross-validated skill of existing multi-proxy climate variability reconstructions is difficult to assess due to a variety
of factors. The use of statistical methods that cloud interpretation of the physical links between predictors and
predictand and the use of non-standard cross-validation techniques or intervals are common problems.
In this study, we generate reconstructions of the Northern Annular Mode and Pacific North America patterns using
extratropical Northern Hemisphere tree ring width site chronology time series. Four independent methods of predictor
selection are employed: 1) mean ring width time series, generated to represent particular species over subjectively chosen
regions where a consistent temperature or precipitation growth signal occurs, are used as predictors, 2) significant
principal components generated from the matrix of ring width data are used as predictors, 3) natural clusters are formed from
the matrix of ring width data and used as predictors and 4) individual site chronologies are used as predictors. Each
method has two variants, one based on multivariate linear regression and another based solely on the covariance between
individual predictors and the predictand. The calibrated and cross-validated skill of each reconstruction variant are
compared in a consistent manner.
Results indicate that essentially the same few records provide the bulk of the cross-validated skill in the Northern Annular
Mode reconstruction, regardless of the selection method employed. The same can loosely be said of the Pacific North America
reconstruction. Many of the candidate ring width site chronologies are not very useful for this particular effort because
they either are located in the wrong area or are obtained from a relatively insensitive species. The tree ring based
reconstructions are also frequency dependent and the high-frequency component is the most robust. In addition to creating
well-verified reconstructions, we are able to explore the utility of particular records and groups of records with this
relatively detailed and comprehensive approach.
UR: http://www.atmos.washington.edu/$\sim$justinjw/AGU2004
DE: 3344 Paleoclimatology
DE: 3309 Climatology (1620)
DE: 1620 Climate dynamics (3309)
SC: Paleoceanography and Paleoclimatology [PP]
MN: 2004 AGU Fall Meeting