摘要
Currently available proxies were studied as networks for building reconstruction models of the Atlantic Multidecadal Oscillation (AMO). Only proxies that would double the current record length (backwards in time from AD 1564) were included. We present two proxy networks and corresponding reconstruction (transfer) models, one for tree-growth based proxies only and another for multiproxies. Both of them show a useful match in timing as well as amplitude with the AMO. These model structures demonstrated reasonable model performance (overall r<sup>2</sup> = 0.45 - 0.36). The time stability of proxy-AMO relationships was also validated. The new models produced acceptable results in cross-calibration-verification (reduction of error and coefficient of efficiency statistics in 1856-1921 and 1922-1990 vary between 0.41 and 0.21). The spatial distribution of these data series indicate that proxies respond to an AMO-like climatic oscillation over much of the Northern Hemisphere.
Currently available proxies were studied as networks for building reconstruction models of the Atlantic Multidecadal Oscillation (AMO). Only proxies that would double the current record length (backwards in time from AD 1564) were included. We present two proxy networks and corresponding reconstruction (transfer) models, one for tree-growth based proxies only and another for multiproxies. Both of them show a useful match in timing as well as amplitude with the AMO. These model structures demonstrated reasonable model performance (overall r<sup>2</sup> = 0.45 - 0.36). The time stability of proxy-AMO relationships was also validated. The new models produced acceptable results in cross-calibration-verification (reduction of error and coefficient of efficiency statistics in 1856-1921 and 1922-1990 vary between 0.41 and 0.21). The spatial distribution of these data series indicate that proxies respond to an AMO-like climatic oscillation over much of the Northern Hemisphere.
