The influence of proxy selection on global annual mean temperature reconstructions during the Common Era

The reconstruction of global annual mean temperatures made by the PAGES 2k Consortium in 2019 represents one of the most influential sequences of global climate variability over the Common Era. However, it is still not clear how the reconstruction can be influenced by the selection of reconstruction methods and the selection of proxy records with different temporal resolutions over different regions. We adopt a widely used Composite-Plus-Scale method to elucidate the effects of the selection of the proxy records on temperature reconstruction. To ensure the uniformity of data, different types of proxy records spanning the past ∼2000 years from the PAGES 2k proxy network were used to investigate the potential effects of proxy selection in hemispheric and global temperature reconstructions during the past two millennia. The long-term trends, spectral characteristics, and volcanic responses of the annual temperatures were studied based on the reconstructions. Our results reveal a significant cooling trend in the global annual mean temperature using both tree-ring and non-tree-ring records during the 1–1850 CE period, and show that the cooling exhibits a stronger trend in the Southern Hemisphere (SH) than that in the Northern Hemisphere (NH). Yet, the long-term trends vary according to different combinations of proxy records. Different reconstructions based on different types of proxies also exhibit different features in terms of volcanic responses and spectral properties. Tree-ring-based temperature reconstructions show stronger cooling responses to tropical volcanic eruptions, while non-tree-ring-based reconstructions suggest less robust volcanic responses, which may be related to dating uncertainties and low temporal resolution of the proxies. Tree-ring width records tend to preserve a substantial proportion of high-frequency (<200 years) variability, whereas non-tree-ring proxies tend to capture a larger fraction of low-frequency (>200 years) variations. Efforts are needed to reduce uncertainties of the temperature reconstruction over the Common Era associated with the insufficient spatiotemporal coverage of the current proxy network, especially for the first millennium and for the SH and tropics, also to develop statistical methods and to improve the signal strength and constrain uncertainties in existing proxy records.