Reply to 'Questions of bias in climate models'

Abstract

I appreciate that Smith et al. support the primary conclusion of my paper that accounting for the geographic distribution of radiative forcing is important in determining the climate response. They describe how the reduced-form model MAGICC represents the response to spatially inhomogeneous forcing via four boxes, one for land and one for ocean in each hemisphere, and I appreciate that clarification. My statement that there will be biases in simple models that do not account for the forcing distribution should have said that there will be biases in simple models that do not adequately account for the forcing distribution. Smith et al. report that the current versions of MAGICC underestimate the hemispheric asymmetry of the temperature response to well-mixed greenhouse gases by ~40% relative to the CMIP5 models. Although MAGICC can be recalibrated, this supports the conclusion that the highly simplified representation of forcing and response distributions in MAGICC contributes to the differences with respect to a CMIP5-generation model seen in a previous analysis of the response to predominantly Northern Hemisphere aerosol forcing, as discussed in my paper. Smith et al. also report that MAGICC captures the hemispheric asymmetry of the temperature response to inhomogeneous (predominantly aerosol) forcing seen in the CMIP5 models reasonably well. This agreement appears coincidental, however, as Smith et al. state that their highly asymmetric response is driven by aerosol forcing being greater over land than ocean, whereas in the CMIP5 models I analyzed, the historical aerosol + ozone forcing is actually greater over the oceans than land. The land responds more strongly in the CMIP5 models despite this, not only because of its inherently faster response time, but also because localized forcing influences climate well beyond the location of the forcing itself, especially in the zonal direction. This process is absent in MAGICC. More generally, I presented the hemispheric temperature responses simply as an example to support my claim that the different response to inhomogeneous forcing relative to homogeneous forcing was largely due to the spatial pattern rather than differences in the effectiveness of those forcing agents, and not to imply that the Northern and Southern hemisphere responses told the whole story. In fact, my study showed an even stronger response in the Northern Hemisphere extratropics than in the Northern Hemisphere as a whole, leading to an even larger asymmetry between the Northern and Southern hemisphere extratropics. This is consistent with the Northern Hemisphere extratropics having the greatest land fraction and fractional area in which powerful snow and ice albedo feedbacks operate, and is in agreement with previous results from multiple models4-6. Hence, climate sensitivity is not simply a function of the average Northern and Southern hemisphere forcing. The CMIP5 models use thousands of boxes in the horizontal and do well at capturing the heterogeneity of the climate system. Although a four-box model might be calibrated to match the CMIP5 models' global mean response to aerosol forcing for certain particular cases, assessments of its skill in capturing the climate response to complex temporally and spatially evolving inhomogeneous forcings (for example, recent shifts in aerosols from more northerly developed to more equatorial developing nations) are surely required. Thus, the conclusions of my paper hold firm, namely that the geographic distribution of radiative forcing plays an important role in determining the transient climate response, and that calculations with simple models and those inferring transient climate response from historical surface temperature observations need to adequately account for this.