Decomposing the response of the stratospheric Brewer–Dobson circulation to an abrupt quadrupling in CO<sub>2</sub>

Abstract

Abstract. We perform time-slice experiments using HadGEM3-A to decompose the long-term (years 101–150) response of the Brewer–Dobson circulation (BDC) to an abrupt quadrupling in CO2 (4×CO2) into (1) a rapid atmospheric adjustment, (2) a contribution from the global-average sea surface temperature (SST) change (+3.4 K), and (3) an SST pattern effect. The SST fields are derived from the CMIP5 multi-model ensemble. Two further experiments explore the impact on the BDC of the spread in global-average SST response to 4×CO2 across the CMIP5 models (range 2.1–4.9 K). At 70 hPa (10 hPa) the annual-mean tropical upward mass flux increases by 45 % (35 %) due to the 4×CO2 perturbation. At 70 hPa, around 70 % of the increase is from the global-uniform SST warming, with the remainder coming in similar contributions from the rapid adjustment and SST pattern effect. In contrast, at 10 hPa the increase in mass flux comes mainly from the rapid adjustment (∼40 %) and the uniform SST warming (∼45 %), with a small contribution from the SST pattern. At 10 hPa, the effect of the multi-model spread in global-mean SST is comparable in magnitude to the rapid adjustment. Conversely, at 70 hPa the effect of spread in global-mean SST is substantially larger than both the rapid adjustment and the SST pattern effect. We derive an approximately linear sensitivity of the tropical upward mass flux to global surface air temperature change of 0.62×109 kg s−1 K−1 (9 % K−1) at 70 hPa and 0.10×109 kg s−1 K−1 (6 % K−1) at 10 hPa. The results confirm the most important factor for the acceleration of the BDC in the lower stratosphere under increased CO2 is global SST changes. We also quantify for the first time that the rapid adjustment to CO2 is of similar importance to SSTs for the increased BDC in the upper stratosphere. This demonstrates a potential for a fast and slow timescale of the response of the BDC to greenhouse gas forcing, with the relative prominence of those timescales being height dependent.