为了解过去2000年北半球温度变化特征, 本文基于16条长度超过2000年的树木年轮年表, 利用最优信号提取法重建了过去2000年北半球年平均温度序列。结果显示, 过去2000年存在1个极端低温年和4个极端高温年、50个低温年和53个高温年, 以及2个冷期和3个暖期。其中, 第一个冷期大约在500~550A.D., 这期间出现了8个低温年, 第二个冷期是小冰期(1450~1850A.D.), 出现了38个低温年和1个极端低温年。3个暖期除发生了17个高温年的中世纪暖期(900~1250A.D.)外, 1000A.D.前还存在一个大约是在400~500A.D.的暖期, 该时期发生了1个极端高温年和11个高温年。第三个暖期是受人类影响显著的20世纪暖期, 该时期未出现极端低温年和低温年, 而出现了2个极端高温年和21个高温年, 这意味着在全球持续增温的情景下, 历史上曾发生过的低温年和极端低温年再次出现的概率很小。对重建温度变化进行归因分析, 发现大西洋多年代际涛动(Atlantic Multi-decadal Oscillation, 简称AMO)、太阳活动和火山活动均对北半球的温度变化有贡献。年平均AMO指数可以解释36% 的年代际以上尺度的温度变率, 而太阳活动的贡献仅限于与北半球的温度序列具有类似周期信号, 其相关性并未通过显著性检验, 这表明AMO可能是北半球过去2000年年平均温度变化的主要驱动因子。但由于AMO指数重建时包括了与本文资料的地理位置相同的资料, 那么资料本身的相似性可能会影响AMO和北半球温度的关系, 未来需要利用海洋沉积等其他类型的地质记录重建AMO指数, 以进一步对比检验AMO的贡献。重建结果与全球34个强火山事件对比发现, 火山爆发后的第一年北半球降温最为明显, 这种降温效应在第四年完全消失, 说明火山爆发对北半球年平均温度的影响主要体现在年际尺度上, 这与已有研究结果一致。
In order to further understand the past temperature variation over Common Era in the Northern Hemisphere(NH), we reconstructed the NH annual temperature variability over Common Era using the optimal information extraction(OIE)method, based on 16 tree-ring chronologies longer than 2000 years.The proxy dataset is composed of 15 tree-ring width chronologies and one tree-ring density chronology covering a large area extending across the Northern America, Europe, and Asia. All chronologies are strongly correlated with the multi-proxy reconstructed annual NH temperature anomalies over the past millennium at the 95% confidence level. The OIE method utilized the ensemble LOCal regression method to improve the ability to retain the low frequency signal, and used the correlation coefficient between the calibrated sequence and the local climate factor in the proxy record location to measure regional differences response to the climate signal. This method belongs to the Composite-plus-scale method classification.The results show that there are one extreme low temperature event and four extreme high temperature events, 50 low temperature events and 53 high temperature events, two cold periods and three warm periods over the past 2000 years. Among them, 8 low temperature events occurred in the first cold period(ca.500~550A.D.), one extreme low temperature event and 38 low temperature events appeared during the Little Ice Age(1450~1850A.D.). In addition, there are 17 high temperature events in the Medieval Warm Period(900~1250A.D.). Before 1000A.D., there is a warm period in ca.400~500A.D., during which existed one extreme high temperature event and 11 high temperature events. The third warm period is the 20th century, during which extreme low temperature and low temperature events did not appear but there have been two extreme high temperature events and 21 high temperature events. This implies that the extreme low and low temperature events over a hemispheric scale could not likely appear under further global warming scenario.Based on the attribution analysis of the reconstructed temperature variability, the forcings, including the Atlantic Multi-decadal Oscillation(AMO), solar activity and volcanic activity all could have affected the NH annual temperature anomalies variability over the past 2000 years. The AMO explains more than 36% NH temperature variability over multi-decadal time scale indicating that the AMO is the main driving factor of the NH annual temperature variations over the past 2000 years. Note that the tree-ring chronologies used in this reconstruction and in AMO index reconstruction may derive from the same positions. This similarity would cause the uncertainty in the influence of AMO on the NH temperature. This needs to be verified using the new AMO index based on the other type proxy records, e.g. coral record. Solar activity could have weak impact on the NH temperature variability on a centurial time scale. Compared with 34 stronger global volcanic events, the NH temperature obviously drops in the first year after the eruption. This cooling effect completely disappeared in the fourth year. This implies that the impact of the volcanic activity on NH temperature decrease is transient and takes effect only on the inter-annual time scale, which is consistent with other results.