The Cloud Problem: Why Climate Models Are Wrong

Modern climate models, which directly affect global climate policy and the pace of energy transition, are not able to predict the rate of global warming with sufficient accuracy - it seems that they have hit a fundamental "wall", writes The Wall Street Journal . This means that decisions on the climate agenda will remain evaluative and political for the foreseeable future.

The problem with clouds

The fact that something was wrong with climatological forecasts became clear in 2018, when new climate models (including the most advanced CESM2) began to produce global warming accelerated by a third compared to the previous ones - up to 5.3 degrees Celsius with an increase in double the concentration of CO2 in the atmosphere.

Scientists estimate that carbon dioxide concentrations will double by 2100 if the climate goals of last year's Glasgow Summit are not met . Under these conditions, every tenth degree of warming is important: most experts agree that +5 degrees to the average temperature is the point of no return with the "greenhouse Earth", the gradual melting of Antarctica and the flooding of the coastal strip.

Scientists immediately suspected that the problem was in the new, more advanced model of cloud formation in the atmosphere, which nevertheless unbalanced some climate models in general. This version was finally confirmed last year, writes WSJ. Tests on paleoclimate data showed that "new" cloud models became too sensitive to CO2 and predicted temperatures lower than actual at low concentrations of this greenhouse gas and higher at high concentrations.

For several years, the developers have rebalanced the models and by the Glasgow summit have narrowed the likely range of warming to 2.5 - 4 degrees. This is not the latest version: the updated report of the UN Intergovernmental Commission on Climate Change, which gives governments around the world the scientific basis for the climate agenda, will be released this year.

The limit of knowledge

CESM2 operates on a relatively coarse grid of 100 by 100 kilometers. But for the Earth's atmosphere, this already gives 64,800 "cubes" of air, 72 layers each. Each step of the forecast requires recalculation of 4.6 million data units, and each change in the model requires calculations on the model for 300 simulated years. In order to find and eliminate one of the errors associated with the mechanism of water vapor condensation in clouds on solid particles, it took 10 experts almost five months. In total, five important changes have been made to the model since 2018, and this is far from the end.

All this is complicated by non-obvious and often not fully understood feedbacks. For example, an increase in average temperature in recent decades has caused an increase in average cloud cover in high latitudes and a decrease in tropical latitudes. And ocean warming has reduced the formation of the whitest, best reflecting solar radiation stratocumulus clouds, which together leads to increased accumulation of heat in the atmosphere.

All known effects can theoretically be taken into account, the problem is that the better and more detailed the model, the longer it is calculated, and exponentially. Climatologists at NCAR (US National Center for Atmospheric Research) are already missing the Cheyenne supercomputer, which in 2016 was among the twenty fastest in the world. The agency's new $40 million Derecho supercomputer, to be launched this year, will be 3.5 times as powerful, but barely enough for the next generation of models. And in order to reliably link local cloud systems, cyclones, forest fires and ocean currents into a single simulation, a power increase of three orders of magnitude is needed - about 1000 times, scientists say.

While climate models are key to humanity's survival in a changing world, they hit several walls at once. This is the limit of the complexity of physical models, the limit of computer power, the fundamental uncertainty of chaotic processes and the problem of changing the behavior of the climate with changing concentrations of carbon dioxide, methane and other greenhouse gases, lists the WSJ: despite improvements, the models are still not accurate enough for direct planning.

“Climate models, despite their imperfections, are our best tool for understanding the future,” Andrew Gettelman, one of the developers of GESM2, told the agency. “The fact that they can be wrong doesn’t worry me too much. It scares me that they are true."