Causes of regional climate variability over the last millennium

During the last millennium, natural variations of regional to large-scale temperature were driven by varying orbital, volcanic and solar forcing. Reconstructions of surface temperature derived from natural proxy archives extend our records beyond the instrumental period and give us information about the extent of past variability and its causes. However, limited proxy data availability and quality, and systematic biases in the data reduce confidence in such reconstructions. Climate models serve as an additional independent source of information; however, they are only approximations of the real-world processes and rely on records of natural and anthropogenic drivers as input, which again are subject to uncertainty. Discrepancies between proxy reconstructions and model simulations suggest problems with either source of information, or both. We find that biological memory effects in tree-rings inflate low frequency temperature variability and suppress high frequency variability in pseudo proxy experiments, decreasing the cooling response to volcanic eruptions while prolonging its recovery time, and increasing the difference between the Medieval Climate Anomaly and the Little Ice Age. Due to the high seasonal dependency of millennial temperature trends caused by orbital forcing, and a seasonal delay between insolation and climate response, we can explain differing trends in individual proxy records. We find that long-term trends in simulations and proxy reconstructions agree better when accounting for proxy specific seasonal sensitivity. Lastly, we represent the effect of forcing uncertainty in a large ensemble of forced responses with an impulse response model. We find robustly that high solar variability is less consistent with proxy reconstructions than low solar variability, even when accounting for volcanic uncertainty. Uncertainties in the volcanic forcing record may explain certain discrepancies between the proxy and model volcanic response, but are disguised by internal variability. We conclude that it is necessary to account for these uncertainties when comparing model and proxy data, to ensure like-for-like comparison, but that carefully selected and evaluated proxy data is an excellent data source for reconstructing and explaining variability of regional to large-scale temperature changes during the last millennium