Some thoughts on models in thread form:
Climate models are based on physics. Their code describes the fundamental processes that we know drive atmospheric processes: radiative transfer, thermodynamics, the idea gas law, etc. 1/
Climate models are based on physics. Their code describes the fundamental processes that we know drive atmospheric processes: radiative transfer, thermodynamics, the idea gas law, etc. 1/
Despite what you might hear, these models have been thoroughly tested. In fact, I’ve spent much of my career looking at model output and comparing to observations and I am constantly amazed how well climate models do. 2/
Now this doesn’t mean models do everything well. Some process are not simulated from first-principles — e.g., cloud microphysics, which occurs on too small a spatial scale for models to resolve. 4/
Climate models handle this by parameterization: they assume these results of these processes can be described as a simple function of the quantities that the model does simulate, such as grid-average temperature and water vapor. 5/
This is probably the weak point in climate models, and a lot of effort has gone into improving the parameterizations. While not perfect, they are good enough that the model performance is quite impressive. 6/
Not perfect, mind you. And a determined denier can always find something that the model does not simulate well. However, for the big things that we care about, the models do well. 7/
What about the COVID models? Those are a completely different class of model. They are a few differential equations that rely on the key parameters being specified: e.g., latency period, transmission rate, … 8/
These models are not based on simple physics, but are essentially sophisticated curve-fitting exercises. This doesn’t mean they’re not useful for predictions, but it does mean they’re different from climate models. 9/
Despite the differences, though, one thing is key: if you want to estimate what might happen in the future, you need a model. You don’t have to like them, but models are the only way to evaluate different scenarios/policies. 10/
That’s what makes the attacks on models so pernicious. If we can discredit models, then we remove one way of evaluating different policies and hamstring the ability of science to help us make good decisions. 11/
