PalVol v1: a proxy-based semi-stochastic ensemble reconstruction of volcanic stratospheric sulfur injection for the last glacial cycle (140 000–50 BP)

Abstract

Abstract. Perturbations in stratospheric aerosol due to explosive volcanic eruptions are a primary contributor to natural climate variability. Observations of stratospheric aerosol are available for the past decades, and information from ice cores has been used to derive estimates of stratospheric sulfur injections and aerosol optical depth over the Holocene (approximately 10 000 BP to present) and into the last glacial period, extending back to 60 000 BP. Tephra records of past volcanism, compared to ice cores, are less complete but extend much further into the past. To support model studies of the potential impacts of explosive volcanism on climate variability across timescales, we present here an ensemble reconstruction of volcanic stratospheric sulfur injection (VSSI) over the last 140 000 years that is based primarily on terrestrial and marine tephra records. VSSI values are computed as a simple function of eruption magnitude based on VSSI estimates from ice cores and satellite observations for identified eruptions. To correct for the incompleteness of the tephra record, we include stochastically generated synthetic eruptions assuming a constant background eruption frequency from the ice core Holocene record. While the reconstruction often differs from ice core estimates for specific eruptions due to uncertainties in the data used and reconstruction method, it shows good agreement with an ice-core-based VSSI reconstruction in terms of millennial-scale cumulative VSSI variations over the Holocene. The PalVol reconstruction provides a new basis to test the contributions of forced vs. unforced natural variability to the spectrum of climate and the mechanisms leading to abrupt transitions in the palaeoclimate record with low- to high-complexity climate models. The PalVol volcanic forcing reconstruction is available at https://doi.org/10.26050/WDCC/PalVolv1 (Toohey and Schindlbeck-Belo, 2023).