A student learned physics from a textbook and is taking an exam. The first two problems are solved brilliantly, but on the third, they err on something not covered in the book. That is roughly what the hybrid atmospheric model NeuralGCM faced when run through three levels of testing: cyclones, El Niño, and anomalous warming.
The model combines a physical core with machine learning that replaces expensive computations of small-scale processes. Instead of calculating every little cloud, the neural network learns from ERA5 reanalysis data and IMERG satellite precipitation from 2001–2018. Then it is asked about things not in that data, such as a uniform 4-degree ocean warming. This tests whether the model has grasped the physics rather than just memorized statistics.
The model also recognizes El Niño: the PNA teleconnection is reproduced with a correlation of 0.86 for 200 hPa geopotential height, within the spread of physical models. However, it overestimates the amplitude of the nonlinear asymmetry in the response. And under uniform warming, global surface temperature rises by 0.86 kelvin per kelvin of ocean heating, and precipitation by 4.14% per kelvin, close to CMIP6 estimates. It would seem the exam is passed.
Yet in the upper troposphere and lower stratosphere, the model produces excessive cooling in the tropics and excessive warming in polar regions. As a result, the jet stream does not shift poleward as it does in physical models. The reason is simple: the model did not receive carbon dioxide and ozone concentrations as inputs, so its stratospheric response is unphysical.
The student who learned physics from a textbook passed the exam, but not with flying colors. NeuralGCM showed that hybrid models can generalize beyond their training data, something purely neural network models could not do. But the stratospheric errors remind us that before trusting such models for climate projections, we need to understand the minimum physics they require. The exam continues.
The NeuralGCM model was trained on ERA5 reanalysis and IMERG satellite precipitation only for 2001–2018, yet it managed to reproduce the response to warming that was not present in that data. It is as if a student who learned physics from a textbook passed an exam on problems not covered in the book.
The idea of hybrid intelligence, where machine learning complements physical laws, echoes Stanisław Lem's novel 'Solaris', in which an ocean-mind generates models of visitors from memories. Here, the model generates climate responses based on learned patterns, but sometimes creates phantoms in the stratosphere.