However, ESMs are deficient in simulating the internal multi-decadal variability, and GST time series analysis methods suffer from the data end effect and not explicitly accounting for the strength of the anthropogenic forcing. Here we are developing an observational data-driven model for GST projection, separating natural variability and human-caused warming. This model makes use of the carbon dioxide concentration in the atmosphere to largely eliminate the data end effect.
For the projection at the end of 21st century under the scenarios RCP8.5 and RCP4.5, the natural interannual and multi-decadal variabilities play a minor role because their magnitudes are much smaller than the warming. For the decadal prediction over the next 20 years, both natural multi-decadal variability and the human-caused warming are important, and the robustness of our method is tested through hindcasting over the past 20 years. Uncertainties are quantified by using different GST datasets as well as using the same dataset with different data periods.
In this presentation, we will discuss these results based on our ongoing research and compare them with other studies based on ESMs and GST time series analyses.