In my research, I use super computers to simulate climate to predict future global warming. Some people mistakenly think that this means using a computer to crunch a large volume of observational data. That's incorrect. The only data that are input into the computer are basic parameters like the size of the earth, the rotation, the land-sea distribution, and atmospheric components.
After these parameters are applied to calculations to solve physics equations, a computer simulates rain in the rainforest, aridity in the desert, coldness in the polar areas, and weather patterns like low pressure or high. In other words, the simulation automatically generates conditions as they actually occur on earth.

However, even though we have physics equations to describe the earth, that doesn't mean we can accurately predict actual changes in climate.
Global climate is extremely complex and is not yet fully understood by current science. For example, to simulate the occurrence of precipitation in clouds, we need to take account of empirical factors.
Computational physics enables us to reproduce actual atmospheric and oceanic motions with a high degree of accuracy. However, as we look closer, errors tend to crop up.
It is therefore essential to validate simulation against real-world observations to create a computational model that is closer to actual conditions on earth. Future predictions are then based on this model.

Because there is no way to objectively know what kind of society people will build in the future, we have to make climate prediction based on various scenarios*, such as "priority to the economy" or "priority to the environment."
Incidentally, there is a special branch of research concerned with the development of socio-economic models dealing with population, economic activity, and other factors.
Thus global warming prediction is an aggregate effort that involves various fields of research.