However, future projections of R_s based on climate models contain large uncertainties due to internal climate variability, model uncertainty, and scenario uncertainty, which have not been eliminated by previous studies. Moreover, the model biases in R_s and the underlying drivers have yet to be quantified.

By integrating the high-quality observations and the latest Coupled Model Intercomparison Project Phase 6 (CMIP6) model data of R_s, a research team led by Prof. Kun Yang (Department of Earth System Science, Tsinghua University) has explored the model bias of CMIP6 models in R_s and quantified the physical causes of the model bias in China. The systematic bias in R_s in CMIP6 models is revealed to be caused by clouds and aerosols, resulting in largely uncertain projections for future changes in R_s.

To correct this effect, the team used historical biases of models to constrain the future projections of R_s under three possible future scenarios based on emergent constraints, an approach with a solid physical basis for narrowing the uncertainties of future climate projections through the combination of an ensemble of climate simulations with contemporary measurements.

The research is published in the journal National Science Review.

The constrained results substantially reduce the projection uncertainties by about 56% in the mid-21st century. Moreover, the team found that the constraints using the combined effect of the TCC and R_s-clear biases can account for about 81% of the projection uncertainties using R.