Concord and discord among Northern Hemisphere paleotemperature reconstructions from tree rings

We review the current generation of large-scale, millennial-length temperature reconstructions derived from tree rings and highlight areas of agreement and disagreement among these state-of-the-art paleotemperature estimates. Although thousands of tree ring-width chronologies are now available from temperate and boreal forest sites across the Northern Hemisphere, only a small fraction of those records are suited as proxies for surface temperature. Maximum latewood density is clearly a superior temperature proxy but is less available, with few densitometric records that are both long and up-to-date. Compared to previous efforts, the newest generation of tree-ring reconstructions correlate more strongly against hemispheric summer temperatures and show better performance in tracking decadal/multi-decadal variability and year-to-year fluctuations. They also fit the observed memory structure of instrumental temperatures more closely than their predecessors. These new estimates still show signs of the so-called ‘divergence problem’ (the apparent loss of temperature sensitivity under recent warming), but do not extend after 2004 and cannot be used to evaluate the impact of the past decade's warming on northern temperature-limited forests. We caution against averaging together the latest hemispheric-scale reconstructions because they have each been constructed to suit different purposes and share much of the same underlying tree-ring data, especially prior to CE 1500. Past temperatures are recorded more clearly in maximum latewood density than total ring-width, so we recommend the Northern Hemisphere densitometry network be modernized through a new round of field collections and observations.