Using reduced-complexity volcanic aerosol and climate models to produce large ensemble simulations of Holocene temperature

Abstract. Volcanic eruptions are one of the most important drivers of climate variability, but climate model simulations typically show stronger surface cooling than proxy-based reconstructions. Uncertainties associated with eruption source parameters, aerosol–climate modelling, and internal climate variability might explain those discrepancies, but their quantification using complex global climate models is computationally expensive. In this study, we combine a reduced-complexity volcanic aerosol model (EVA_H) and a climate model (FaIR) to simulate global-mean surface temperature from 6755 BCE to 1900 CE (8705 to 50 BP) accounting for volcanic forcing, solar irradiance, orbital, ice sheet, greenhouse gases, land-use forcing, and anthropogenic aerosols and ozone forcing for the historical period (1750–1900 CE). The negligible computational cost of the models enables us to use a Monte Carlo approach to propagate uncertainties associated with eruption source parameters, aerosol and climate modelling, and internal climate variability. Averaging over the last 9000 years, we obtain a global-mean volcanic forcing of −0.15 W m−2 and an associated surface cooling of 0.12 K. Averaged over the 14 largest eruptions (injecting more than 20 Tg of SO2) of 1250–1900 CE, the mean temperature response in tree-ring-based reconstructions is in good agreement with the our simulations, scaled to Northern Hemisphere summer temperature. For individual eruptions, discrepancies between the simulated and reconstructed surface temperature response are almost always within uncertainties. At multimillennial timescales, our simulations reproduce the Holocene global warming trend typically derived from simulations and data assimilation products but exhibit some discrepancies on centennial to millennial timescales. In particular, the Medieval Climate Anomaly to Little Ice Age transition is weaker in our simulations, and we also do not capture a relatively cool period between 3000 and 1000 BCE (5000 and 3000 BP), visible in climate reanalyses. We discuss how uncertainties in land-use forcing and model limitations might explain these differences. Our study demonstrates the value of reduced-complexity volcanic aerosol–climate models to simulate climate at annual to multimillennial timescales.