State-of-the art climate models used for future projections of climate change have become huge tangled webs of code representing thousands of physical processes. In many of these processes, we have to simplify or in some cases even make an educated guess as to how things are related – these simplifications are called parameterisations. Some are based on experimental data in the laboratory or in the atmosphere; others are based on detailed models too expensive to include in our climate model that must be able to run for centuries of model time. Different models do this in slightly different ways, and this is one reason why climate simulations can differ from each other. To understand exactly why the models behave differently, it is sometimes necessary to strip back to the basics the part of the model we are interested in. This is what we have done in a recent paper for one very important simplification – the effect of aerosol particles on the size of cloud drops, and hence on the brightness of our planet.
Aerosols are the seeds of cloud droplets. Without them, it would take forever (literally) for enough water molecules to stick together to form a cloud droplet of a few tens of microns across. Specks of sea-salt, or dissolved chemicals such as ammonium sulphate, can pull water molecules in and so kickstart the growth process. In general, more of these cloud condensation nuclei lead to more cloud droplets, and given a limited amount of water in any given cloud, each one is smaller. Smaller droplets are more reflective and scatter more sunlight back to space, making the clouds brighter. Shiptracks are a good example of this process; shipping fuel being rich in sulphate, the exhausts leave a trail of brighter cloud behind them (an example is shown in Figure 1).
Figure 1. Increased numbers of smoke particles from ships funnels interact with the local low-level cloud to produce smaller more reflective cloud droplets along distinctive ship tracks off the north-west coast of the US in January 2015. Image from NASA Earth Observatory.
These processes happen on very small scales and can’t be represented in a climate model. The last Intergovernmental Panel on Climate Change (IPCC) report said that “aerosols and their interactions with clouds continue to contribute the largest uncertainty to the total climate forcing estimate”. Models usually describe this process in two steps. Firstly, they relate the mass of aerosol (or certain types of aerosol) to the number of cloud condensation nuclei, and then they have an experimentally derived relationship to predict the size of the resulting cloud droplets and the reflectivity of the cloud. We wanted to know how much of the difference seen in aerosol-cloud interactions in climate models is due to these relationships, and how much is due to other things, for example different meteorology leading to different aerosol or indeed cloud patterns. We took four models from the latest international climate model intercomparison project (CMIP5) and from the output created “functional forms” that mimic the step going from number of cloud seeds to size of cloud droplets. These simplified versions then allowed us to give one “model” the aerosol change, pre-industrial load or cloud pattern from a different model in a way not possible with full climate models.
To cut a long story short (and simplify the complexity), we found that the form of the aerosol-cloud interaction was responsible for a large part of the full climate model differences, but that other differences between models, particularly aerosol amount, brought the full climate models closer together than we would have expected from just the parameterisations (Figure 2). We could never have done these experiments with the full climate models. But, because we have stripped this interaction back to basics, we have learnt something new about how complex climate models behave, and given a hint as to where future effort should be concentrated.
Figure 2. Radiative forcing (change in energy budget of the climate system) from pre-industrial to present day changes in aerosols from simplified versions of four different models (left hand panel). The differences between these estimates are smaller than the differences due to the aerosol-cloud interaction simplification only, which we see when we use the same simplified models with the change in aerosol seen in one of the models (right hand panel). This suggests that other differences between the models, e.g. aerosol change, pre-industrial aerosol, compensate for the differences in the parameterisations.