HIGH-RESOLUTION PALEOCLIMATIC RECORDS FOR THE LAST MILLENNIUM - INTERPRETATION, INTEGRATION AND COMPARISON WITH GENERAL-CIRCULATION MODEL CONTROL-RUN TEMPERATURES

Palaeoclimatology provides our only means of assessing climatic variat ions before the beginning of instrumental records. The various proxy v ariables used, however, have a number of limitations which must be ade quately addressed and understood. Besides their obvious spatial and se asonal limitations, different proxies are also potentially limited in their ability to represent climatic variations over a range of differe nt timescales. Simple correlations with instrumental data over the per iod since AD 1881 give some guide to which are the better proxies, ind icating that coral- and ice-core-based reconstructions are poorer than tree-ring and historical ones. However, the quality of many proxy tim e series can deteriorate during earlier times. Suggestions are made fo r assessing pl oxy quality over longer periods than the last century b y intercomparing neighbouring proxies and, by comparisons with less te mporally resolved proxies such as borehole temperatures. We have avera ged 17 temperature reconstructions (representing various seasons oi th e year), all extending back at least to the mid-seventeenth century. t o form two annually resolved hemispheric series (NH10 and SH7). Over t he 1901-91 period, NH10 has 36% variance in common with average NH sum mer (June to August) temperatures and 70% on decadal timescales. SH7 h as 16% variance in common with average SH summer (December to February ) temperatures and 49% on decadal timescales. markedly poorer than the reconstructed NH series. The coldest year of the millennium over the NH is AD 1601, the coldest decade 1691-1700 and the seventeenth is the coldest century. A Principal Components Analysis (PCA) is performed o n yearly values for the 17 reconstructions over the period 40 1660-197 0. The correlation between PCI and NH10 is 0.92, even though PCI expla ins only 13.6% of the total variance of all 17 series. Similar PCA is performed on thousand-year long General Circulation Model (GCM) data f rom the Geophysical Fluid Dynamics Laboratory (GFDL) and the Hadley Ce ntre (HADCM2). sampling these for the same locations and seasons as th e proxy data. For GFDL, the correlation between its PCI and its NH10 i s 0.89, while for HADCM2 the PCs group markedly differently. Cross-spe ctral analyses are performed on the proxy data and the GFDL model data at two different frequency bands (0.02 and 0.03 cycles per year). Bot h analyses suggest that there is no large-scale coherency in the serie s on these timescales. This implies that if the proxy data are meaning ful, it should be relatively straightforward to detect a coherent near -global anthropogenic signal in surface temperature data.