Underestimation of volcanic cooling in tree-ring-based reconstructions of hemispheric temperatures

The largest eruption of a tropical volcano during the past millennium occurred in AD 1258–1259. Its estimated radiative forcing was several times larger than the 1991 Pinatubo eruption1. Radiative forcing of that magnitude is expected to result in a climate cooling of about 2 C (refs 2–5). This effect, however, is largely absent from tree-ring reconstructions of temperature6–8, and is muted in reconstructions that employ a mix of tree-rings and other proxy data9,10. This discrepancy has called into question the climate impact of the eruption2,5,11. Here we use a tree-growth model driven by simulated temperature variations to show that the discrepancy between expected and reconstructed temperatures is probably an artefact caused by a reduced sensitivity to cooling in trees that grow near the treeline. This effect is compounded by the secondary effects of chronological errors due to missing growth rings and volcanically induced alterations of diffuse light. We support this conclusion with an assessment of synthetic proxy records created using the simulated temperature variations. Our findings suggest that the evidence from tree rings is consistent with a substantial climate impact2–5 of volcanic eruptions in past centuries that is greater than that estimated by tree-ring-based temperature reconstructions. As with all climate proxies, tree-ring records have their own particular strengths and limitations. They are datable to the precise year, thus resolving specific events such as El Niño episodes or volcanic eruptions, and they often show a strikingly close relationship with climate variables. These relationships, however, typically hold only for the growing season1,6–9. Trees growing near the latitudinal or elevational treeline are typically selected for use in reconstructing past temperature changes7,8 because their growth is primarily limited by temperature. One unintended consequence is that these same trees may not document the full extent of past cooling events, and in particular the response to the immense ad 1258/1259 eruption. We employed simulations of the Northern Hemisphere mean temperature response to natural (volcanic+solar) and anthropogenic radiative forcing over the past millennium (Methods), using both a forced simulation3 of the US National Center for Atmospheric Research (NCAR)CSM1.4 coupled ocean– atmosphere general circulation model (GCM) and simulations of a simpler energy-balance climate model (EBM; refs 12,13). The EBM, though less comprehensive, is more amenable to extensive sensitivity analyses. Driven with estimated radiative forcings, both models capture the main features of the historical temperature record (Fig. 1a), including the abrupt multiyear cooling events following the major historical eruptions (the muted observed cooling following the Krakatau eruption of 1883 may be an artefact of data-quality issues in earlier decades14). Given their success in reproducing volcanic cooling events of the historical era, we