NOAA/WDS Paleoclimatology - Western North America 500 Year Annual Temperature Reconstruction

STUDY NOTES: This article presents a tree ring-based reconstruction of western North America annual temperature over 1500-1980, with 5x5 degree grid cell coverage for the region bounded by 30°-55°N and 95°-130°W. A 1000-member probabilistic ensemble derived by bootstrap methods is employed for characterization of uncertainty. A key result based on the ensemble methodology is that the most recent decades are, at the level of the bi-decadal average, very likely the warmest across the region during the past 500 years. The reconstruction is used in a side-by-side comparison with a parallel reconstruction driven by non-informative predictors that have the same temporal autoregressive structure as the tree ring data. The proxy-based reconstruction clearly outperforms the non-informative predictor-based reconstruction across a suite of calibration and validation skill measures.
ABSTRACT SUPPLIED BY ORIGINATOR: The performance of climate field reconstruction (CFR) and index reconstruction methods is evaluated using proxy and non-informative predictor experiments. The skill of both reconstruction methods is determined using proxy data targeting the western region of North America. The results are compared to those targeting the same region, but derived from non-informative predictors comprising red-noise time series reflecting the full temporal autoregressive structure of the proxy network. All experiments are performed as probabilistic ensembles, providing estimated Monte Carlo distributions of reconstruction skill. Results demonstrate that the CFR skill distributions from proxy data are statistically distinct from and outperform the corresponding skill distributions generated from non-informative predictors; similar relative performance is demonstrated for the index reconstructions. In comparison to the CFR results using proxy information, the index reconstructions exhibit similar skill in calibration, but somewhat less skill in validation and a tendency to underestimate the amplitude of the validation period mean.