Tree rings and volcanic cooling

To the Editor — In their Letter, Mann and colleagues1 claim to have identified a discrepancy between the degree of volcanic cooling in climate model simulations and the analogous cooling indicated in a tree-ring-based Northern Hemisphere temperature reconstruction2, and attribute it to a putative temporary cessation of tree growth at some sites near the temperature limit for growth. They argue that this growth cessation would lead to missing rings in cool years, thus resulting in underestimation of cooling in the tree-ring record. This suggestion implies that periods of volcanic cooling could result in widespread chronological errors in tree-ring-based temperature reconstructions1,3. Mann and colleagues base their conclusions solely on the evidence of a tree-ring-growth model. Here we point to several factors that challenge this hypothesis of missing tree rings; specifically, we highlight problems in their implementation of the tree-ring model used1, a lack of consideration of uncertainty in the amplitude and spatial pattern of volcanic forcing and associated climate responses, and a lack of any empirical evidence for misdating of treering chronologies. Several aspects of their tree-ringgrowth simulations are erroneous. First, they use an algorithm that has not been tested for its ability to reflect actual observations (Supplementary Fig. 1), even though established growth models, such as the Vaganov–Shashkin model4,5, are available. They rely on a minimum growth temperature threshold of 10 °C that is incompatible with real-world observations. This condition is rarely met in regions near the limit of tree growth, where ring formation demonstrably occurs well below this temperature: there is abundant empirical evidence that the temperature limit for tree-ring formation is around 5 °C (refs 6,7). Mann and colleagues arbitrarily and without justification require 26 days with temperatures above their unrealistic threshold for ring formation. Their resulting growing season becomes unusually short, at 50–60 days rather than the more commonly observed 70–137 days4,7. Furthermore, they use a quadratic function to describe growth that has no basis in observation or theory, and they ignore any daylength and moisture constraints on growth. These assumptions all bias Mann and colleagues’ tree-growth model results1 towards erroneously producing missing tree rings. Reconstructing simulated temperatures in the same manner as Mann and colleagues, but using a well-tested tree-ring growth model5 and realistic parameters provides no support for their hypothesis (Fig. 1). Instead we find good agreement between summertime temperatures reconstructed from pseudoproxies and those simulated with a climate model (CSM1.4)8 (Fig. 1a), for the whole record as well as in specific years following major volcanic eruptions (Fig. 1b–d). Mann and colleagues’ principal result arises from their failure to select a realistic minimum temperature for growth, use actual treering chronology locations and recognize Tree rings and volcanic cooling