No abstract

Solar variability has been hypothesized to be a major driver of North Atlantic millennial-scale climate variations through the Holocene along with orbitally induced insolation change. However, another important climate driver, volcanic forcing has generally been underestimated prior to the past 2,500 years partly owing to the lack of proper proxy temperature records. Here, we reconstruct seasonally unbiased and physically constrained Greenland Summit temperatures over the Holocene using argon and nitrogen isotopes within trapped air in a Greenland ice core (GISP2). We show that a series of volcanic eruptions through the Holocene played an important role in driving centennial to millennial-scale temperature changes in Greenland. The reconstructed Greenland temperature exhibits significant millennial correlations with K+ and Na+ ions in the GISP2 ice core (proxies for atmospheric circulation patterns), and δ18O of Oman and Chinese Dongge cave stalagmites (proxies for monsoon activity), indicating that the reconstructed temperature contains hemispheric signals. Climate model simulations forced with the volcanic forcing further suggest that a series of large volcanic eruptions induced hemispheric-wide centennial to millennial-scale variability through ocean/sea-ice feedbacks. Therefore, we conclude that volcanic activity played a critical role in driving centennial to millennial-scale Holocene temperature variability in Greenland and likely beyond.

Growing evidence from general circulation models (GCMs) indicates that the equilibrium climate sensitivity (ECS) depends on the magnitude of forcing, which is commonly referred to as state‐dependence. We present a comprehensive assessment of ECS state‐dependence in Earth system models of intermediate complexity (EMICs) by analyzing millennial simulations with sustained 2×CO2 and 4×CO2 forcings. We compare different extrapolation methods and show that ECS is smaller in the higher‐forcing scenario in 12 out of 15 EMICs, in contrast to the opposite behavior reported from GCMs. In one such EMIC, the Bern3D‐LPX model, this state‐dependence is mainly due to the weakening sea ice‐albedo feedback in the Southern Ocean, which depends on model configuration. Due to ocean‐mixing adjustments, state‐dependence is only detected hundreds of years after the abrupt forcing, highlighting the need for long model integrations. Adjustments to feedback parametrizations of EMICs may be necessary if GCM intercomparisons confirm an opposite state‐dependence.