Understanding Last Millennium Climate Change: Reconstructions, Testing climate model response to volcanic forcing, and Evaluating efficacy of reconstruction methods

A reconstruction of summer temperature from pollen records in central North America is presented covering approximately the last 1000 years. A relatively warm Medieval period (~1100-1500) and cool Little Ice Age (~1500-1900, coolest during the early 19th century) are reconstructed using a generalized linear model employing Bayesian inference. Plausible modern circulation analogs are considered, showing pressure composites that may have been more prevalent during the earlier compared to the later period. These analog patterns are consistent with independent reconstructions of positive NAO/AMO-like conditions during Medieval times, both of which are similar to circulation patterns simulated by GCMs using generalized sea surface temperatures consistent with La Niña-like conditions in the tropical Indo-Pacific region. The climatic patterns that generally lead to warmer summers in the study region are additionally associated with anomalous high-pressure ridging and large-scale drought, which is also consistent with drought reconstructions for North America during Medieval times. New tree ring-based reconstructions of the surface temperature field over the past 500 years in western North America and the adjacent Pacific Ocean are also presented. A key feature of these reconstructions occurs in the early 19th century, which is the coldest period across the region during this time, similar to the center-continent pollen-based record. The spatially-explicit field reconstructions are used to drive analysis of regional responses to volcanic forcing, providing a basis for examining agreement between the reconstructed response patterns and the corresponding responses in GCM simulations. The ultimate goal of comparison is to identify regionally best-performing GCM output to use as forcing input in regional impacts models. Initial application of the regional post-volcanic fingerprint to parallel analyses done with members of the NCAR CCSM suite of models suggests that significant improvement in regional fidelity has resulted from continued model development. Additional examination using a new millennium-length CCSM integration and adding European post-volcanic field reconstructions yields a more mixed picture. Finally, a rigorous experimental evaluation of the efficacy of climate field reconstruction (CFR) methods is presented, derived from the western North American temperature reconstructions. This evaluation compares the fidelity of CFRs based on real proxy predictors to those obtained by using non-informative predictors. The non-informative proxies are designed to have the same autocorrelation structure as the real proxy data, but contain no climatic information. Large ensembles of reconstructions are generated in both cases, providing estimated Monte Carlo distributions of reconstruction skill. The skill metric distributions of the real proxy-based CFRs indicate good reconstruction quality and clearly (and almost entirely) separate from the poor skill distributions generated using the non-informative proxies, in contrast to a recent similar study that suggests proxy-based reconstructions have little efficacy, but which did not evaluate CFR methods.