1National Institute of Polar Research, 10-3 Midoricho, Tachikawa, Tokyo, Japan
2Scripps Institution of Oceanography, University of California, San Diego, La Jolla, CA, USA
3Geological Survey of Denmark and Greenland, Copenhagen, Denmark
4Zhejiang University, Hangzhou, Zhejiang, China
5Meteorological Research Institute, Tsukuba, Ibaraki, Japan
Abstract. Precise understanding of Greenland temperature variability is important in two ways. First, Greenland ice sheet melting associated with rising temperature is a major global sea level forcing, potentially affecting large populations in coming centuries. Second, Greenland temperatures are highly affected by North Atlantic Oscillation/Arctic Oscillation (NAO/AO) and Atlantic multidecadal oscillation (AMO). In our earlier study, we found that Greenland temperature deviated negatively (positively) from northern hemispheric (NH) temperature trend during stronger (weaker) solar activity owing to changes in atmospheric/oceanic changes (e.g. NAO/AO) over the past 800 yr (Kobashi et al., 2013). Therefore, a precise Greenland temperature record can provide important constraints on the past atmospheric/oceanic circulation in the region and beyond. Here, we investigated Greenland temperature variability over the past 4000 yr reconstructed from argon and nitrogen isotopes from trapped air in a GISP2 ice core, using a one-dimensional energy balance model with orbital, solar, volcanic, greenhouse gas, and aerosol forcings. The modelled northern Northern Hemisphere (NH) temperature exhibits a cooling trend over the past 4000 yr as observed for the reconstructed Greenland temperature through decreasing annual average insolation. With consideration of the negative influence of solar variability, the modelled and observed Greenland temperatures agree with correlation coefficients of r = 0.34–0.36 (p = 0.1–0.04) in 21 yr running means (RMs) and r = 0.38–0.45 (p = 0.1–0.05) on a centennial timescale (101 yr RMs). Thus, the model can explain 14 to 20% of variance of the observed Greenland temperature in multidecadal to centennial timescales with a 90–96% confidence interval, suggesting that a weak but persistent negative solar influence on Greenland temperature continued over the past 4000 yr. Then, we estimated the distribution of multidecadal NH and northern high-latitude temperatures over the past 4000 yr constrained by the climate model and Greenland temperatures. Estimated northern NH temperature and NH average temperature from the model and the Greenland temperature agree with published multi-proxy temperature records with r = 0.35–0.60 in a 92–99% confidence interval over the past 2000 yr. We found that greenhouse gases played two important roles over the past 4000 yr for the rapid warming during the 20th century and slightly cooler temperature during the early period of the past 4000 yr. Lastly, our analysis indicated that the current average temperature (1990–2010) or higher temperatures occurred at a frequency of 1.3 times per 1000 yr for northern high latitudes and 0.36 times per 4000 yr for NH temperatures, respectively, indicating that the current multidecadal NH temperature (1990–2010) is more likely unprecedented than not (p = 0.36) for the past 4000 yr.