Comment on ‘‘Climate forcing by the volcanic eruption of Mount

[1] Douglass and Knox [2005, hereinafter referred to as DK], present an analysis of the observed tropospheric cooling following the 1991 Mt. Pinatubo eruption, and claim that these data imply a very low value for the climate sensitivity (equivalent to 0.6 C equilibrium warming for a CO2 doubling). We show here that their analysis is flawed and their results are incorrect. [2] We begin with a very simple analysis of the response to volcanic forcing. If ‘S’ is the climate sensitivity ( C/Wm ) and the maximum forcing is DQ, then the maximum equilibrium cooling is DTeq = SDQ. Because of oceanic thermal inertia, the actual maximum temperature reduction, DT, will be substantially less than DTeq. The ratio of actual cooling to equilibrium cooling,a=DT/DTeq,may be referred to as the fractional realized cooling. For short term volcanic forcing, conventional values for a are around 0.3. [3] What do DK’s results imply for a? DK have S = 0.15 C/Wm , DQ 3 Wm , and DT 0.5 C (see DK, Figure 3). Here, DQ is obtained by multiplying the peak visible optical depth change of 0.16 (see DK, Figure 2) by a scaling factor of 18.5, their central estimate, and DT is their smoothed value for the maximum cooling. These values imply that a = DT/(SDQ) 1.1. (A larger cooling estimate, such as the unsmoothed value of 0.7 C, would give an even larger value for a.) DK’s results therefore imply that the actual cooling from the Pinatubo eruption was more than the equilibrium cooling! This is an improbable result, and it is difficult to think of a physical mechanism through which it might occur. [4] We can test the validity of the DK approach directly using a model case where we know the climate sensitivity, and see whether their approach can recover the known value. The case we consider is a coupled atmosphere-ocean general circulation model simulation of the effects of volcanic eruptions on climate [Ammann et al., 2003]. The model used is the NCAR/DOE Parallel Climate Model (PCM). This is the same model that was used by DK to obtain the post-Pinatubo optical depth time series (DK, Figure 2). In contrast to the real-world case, the Pinatubo response signal in PCM is very wellcharacterized, because multiple model realizations allow the noise of internally-generated variability to be reduced significantly [see Wigley et al., 2005]. We also know from earlier work [Raper et al., 2001] that the climate sensitivity for this model is 0.46 C/Wm , smaller than in most other models but still substantially greater than the DK result of 0.15 C/Wm . [5] It is a simple matter to fit DK’s analytical solution for the Pinatubo response (DK, equation (6)) to the PCM results for Pinatubo. Their analytical solution contains two free parameters, the climate sensitivity and a response time (t). By minimizing the root-mean-square difference between the PCM ‘observed’ and DK ‘model’ values we obtain S = 0.166 C/Wm 2 and t = 8.3 months for PCM. This best-fit result is shown in Figure 1. (Note that PCM’s peak cooling is slightly less than the best estimates of the observed peak cooling.) It is clear that, while the DK method can provide a reasonable fit to the data, it is unable to recover the known value of S for PCM, underestimating the true sensitivity by a factor of almost three. Given this gross failure, DK’s method is unlikely to be able to estimate a reliable sensitivity value from real-world observational data. [6] The reason for this failure lies in the over-simplified one-box climate model that is used by DK. To see this, we write this simple type of climate model equation in its conventional form [see, e.g., Raper et al., 2001]: