Tree-ring width and density data around the Northern Hemisphere: Part 2, spatio-temporal variability and associated climate patterns

Patterns of summer temperature over the Northern Hemisphere, obtained from a calibration of a treering network, are presented for every year from 1600 to 1877. The network of tree-ring density chronologies is shown to exhibit spatially coherent modes of variability. These modes closely match summer half-year tempera ture variations, in terms of similar spatial patterns and similar temporal evolution during the instrumental period. They can, therefore, be considered to be proxies for the temperature patterns, and time series for the eight most dominant patterns are presented back to the late seventeenth century. The first pattern represents spatially coherent warming or cooling and it appears to respond to climate forcings, especially volcanic erup tions. Most other patterns appear to be related to atmospheric pressure anomalies and they can be partially explained by heat advection associated with anomalous atmospheric circulation. This provides the potential for reconstructing past variations in atmospheric circulation for the summer half-year. To investigate this poten tial, modes of summer-pressure variability are defined, and an attempt is made to reconstruct them using principal components regression. Poor verification statistics and high sensitivity to the design of the regression procedure provide little confidence in the reconstructions presented, which are regarded as being preliminary only. A repeat study using instrumental temperature predictors shows that the poor performance is attributable mainly to the weakness of the relationship between air temperature over land and atmospheric circulation during summer: though a relationship exists, it is not strong enough to yield reliable regression models when only a relatively short overlap period (55 years in this study) exists for calibration and verification. Further attempts to reconstruct large-scale atmospheric circulation patterns that include precipitation-sensitive networks of tree-ring data are likely to produce improved results.