Abstract: Global climate change is a complicated scientific problem that involves changes in the climate system or the Earth’s system. It is necessary to consider changes of each system component and their complex interactions, as well as multiple physical and geobiochemical processes. This paper will focus on the fundermental physical problems in the following three aspects. (1) Radiative transfer and global radiation balance. Global climate changes will occur once the global radiation balance is destroyed, either due to natural causes (solar activity and eruption of vocanoes), or due to human activities, such as an increase in emission of greenhouse gases which leads to enhanced greenhouse effects. These are the main drivers responsible for global climate changes. (2) Atmospheric waves and internal climate variability. Besides the above-mentioned external forcings, climate change can also be caused by internal climate variability, including atmospheric internal variability and coupling of the climate variability. These mainly exert their effects through various kinds of atmospheric waves. Among them, the Rossby wave is a most important mode. Their propagation and unstable development is a cause leading to climate change. (3) The chaotic nature and predictability of the climate system. The prediction of future climate change based on a climate model is very sensitive to initial values, i.e., small errors of initial fields can lead to completely different results. At the same time, the fidelity of climate modes can also cause significant prediction errors.This chaotic nature of the atmosphere limits weather prediction to a range of two weeks. However, due to significant weakening of the chaotic nature under external forcings, climate prediction may extend up to a month, a season, a year, or even multi-decades and multi-centuries. In order to narrow down predictive uncertainty caused by initial errors and incompleteness of the physical model, climate prediction now uses ensemble prediction techniques, thus leading to probabilistic predictions.