Importance of Minor‐Looking Treatments in Global Climate Models

Abstract

Global climate models (GCMs) are key tools used for climate simulations of the earth. GCMs with horizontal resolutions typically of the order of one hundred kilometers (several tens to a few hundreds of kilometers) have been used for the Fifth Phase of the Coupled Model Intercomparison Project (CMIP5; Taylor et al., 2012) simulations and the Sixth Phase (CMIP6; Eyring et al., 2016). GCMs consist of, for instance, atmospheric models, ocean models, aerosol models, and chemical composition models. Atmospheric models contain dynamics and physics schemes including cloud, convection, turbulence, radiation, and gravity wave drag schemes.Physics schemes critically control the performance of GCMs, which can be measured in terms of biases in temperature, precipitation, radiation fluxes, and sea ice coverage and the reproduction of temperature increase in the 20 th century. Therefore, a lot of time and effort has been devoted to improving the physics schemes, and the performances of GCMs are gradually improving. In general, model developers tend to attribute the improvement of the model performance to the introduction of more advanced or sophisticated physics schemes. An example is the introduction of prognostic variables of higher moment; for instance, the introduction of prognostic number concentration of cloud particles in addition to prognostic cloud water content (CWC) (two-moment schemes) in contrast to schemes with only prognostic CWC (one-moment schemes). Another example is using more hydrometeor categories as prognostic variables: for instance, using snow, graupel, and hail in addition to the ice crystals used in simpler schemes. In turbulence schemes, the inclusion of turbulent moment terms of higher order is considered more theoretically sophisticated. Furthermore, the introduction of missing effects is also discussed in many papers; for example, the radiative effect of snow, the effect of subgrid inhomogeneity of cloud water content on radiation, the pressure gradient effect on convective momentum transport, and the influence of wind shear on the cloud overlap assumption. Modifying models in this way based on theory or observations is doubtlessly important for developing more advanced GCMs. However, such improvements do not always have major impacts.

A Perspective on the Future of CMIP

“…CMIP6 had less spark than CMIP5, and for reasons that are now easier to understand, even less can be expected from a possible CMIP7. Not only have the systematics of the class of models CMIP targets become more stable, we now also know that projections of warming and regional changes are sensitive to “minor treatments” (Kawai et al., 2022; Zhu et al., 2022), albeit not systematically so (McWilliams, 2007). By refuting the idea of large‐scale control, these sensitivities undermine a long‐standing ambition of CMIP, which has been to support parameterization development.…”

Addressing Complexity in Global Aerosol Climate Model Cloud Microphysics

“…As the model is built up sequentially, every added process deepens the generative entrenchment (meaning an entanglement in climate model evolution with development steps depending on each other so that modeling options depend on previous choices (Lenhard and Winsberg (2010) adopting a concept introduced by Wimsatt (2007) for climate science); also termed path dependence or legacy effect by Babel (2019)). Equifinality More processes mean more free parameters, which need to be set via tuning and may allow for multiple equally plausible model realizations with similar or indistinguishable results (Beven, 2006; Beven & Freer, 2001; Mülmenstädt et al., 2020; Tapiador et al., 2019). Overinterpretation Including more processes or more sophisticated schemes brings the danger of overinterpreting the processes that are represented while neglecting the impact of “minor‐looking treatments” such as thresholds (Kawai et al., 2022) or of those processes that are not represented (Mülmenstädt & Feingold, 2018). Provocatively put, the research into and representation of more and more processes may even be acting as an “engine of distraction,” meaning that it may obscure elemental relationships or other study objects and that thus the detail produces ignorance (Proctor and Schiebinger (2008) citing Wes Jackson for the term on p. 24). No reduction in uncertainty At the same time, the increase in model and process complexity may not be decreasing uncertainty (Carslaw et al., 2018; Fiedler et al., 2019; Knutti & Sedláček, 2013; Lahsen, 2005; Mauritsen et al., 2012; Puy et al., 2022; Stevens & Bony, 2013), increasing the abilities of the model (see e.g., Zelinka et al.…”

“…3. Overinterpretation Including more processes or more sophisticated schemes brings the danger of overinterpreting the processes that are represented while neglecting the impact of "minor-looking treatments" such as thresholds (Kawai et al, 2022) or of those processes that are not represented (Mülmenstädt & Feingold, 2018). Provocatively put, the research into and representation of more and more processes may even be acting as an "engine of distraction," meaning that it may obscure elemental relationships or other study objects and that thus the detail produces ignorance (Proctor and Schiebinger (2008) citing Wes Jackson for the term on p. 24).…”