Can we predict global warming using only statistics?
Not if it is unprecedented and nonlinear. You can't do statistics with a sample size of less than one. Science is our only hope.
R. Saravanan
If you want to make a prediction, you better make sure that you’re either in sample, or that you know the differential equation. 1
In discussions of global warming, you sometimes hear arguments that you can’t trust the complex climate models, that we should rely solely on data to predict the future, and that we should observationally constrain model predictions. This leads us to ask: Can we predict global warming using only statistics? We also ask a related question: Can we identify the causes of global warming using pure statistics?
Let us start by defining temperature as that which is measured by a thermometer, and global warming as a rise in the global average temperature of the order of 2°C over 200 years. It is the typical magnitude of warming that is expected to occur, say, between 1900 to 2100.
Is global warming unprecedented, i.e., has the planet spontaneously warmed in such a manner in the past? Reliable thermometers were only invented about 300 years ago, and accurate global measurements are available only since the late 19th century. So strictly speaking, we only have records for one event of global warming, the current one that is still ongoing. Therefore, the statistical sample count for the global warming events in the available data is a fraction that is less than one! This is true even though we have over a hundred years of temperature measurements. The measured temperature is the sum of the slow global warming event (the signal) superimposed with many fast events, such as El Niño (the noise) (Figure 1). We need multiple independent samples of global warming to separate the signal properties from the noise.
https://metamodel.blog/posts/using-only-statistics/
