Proxy record selection to refine Holocene temperature reconstruction in Europe

Holocene climate reconstructions are characterized by rather smooth temperature trends owing to the low temporal resolution, dating uncertainty, and non-climatic noise of underlying proxy records. In this study, we apply methods to refine such reconstructions in Europe by calibrating a network of 126 low-resolution Holocene proxies against a continental-scale (smoothed) high-resolution temperature reconstruction over the past 2000 yr. This approach differentiates between 35 records that correlate with the shorter continental reconstruction, and 91 records that do not correlate or have other issues such as abrupt variance, temperature level, or temporal resolution changes. The separation into ‘calibrating’ and ‘non-calibrating’ proxies has limited statistical skill, however. This is because the smoothed record autocorrelations exceed r = 0.99 and constrain the degrees of freedom of any correlation-based analysis. The selection based on fit with a common target naturally increases covariance among the records, from an inter-series correlation (Rbar) of 0.00 for all 126 proxies, to Rbar = 0.33 for the 35 calibrating proxies over the Common Era (CE). Covariance in the latter group also reached Rbar = 0.30 during the period Before the Common Era (BCE), compared to Rbar = 0.04 among the non-calibrating proxies. This increase in covariance prior to the CE, while not statistically significant, indicates that the selection process based on fit with a much shorter, 2000 yr target may have improved the temperature reconstruction over the Holocene.