An improved method of the Globally Resolved Energy Balance model by the Bayesian networks

The accurate simulation of climate is always critically important and also a challenge. This study introduces an improved method of the Globally Resolved Energy Balance (GREB) model by the Bayesian networks based on the concept of a coarse-fine model. The improved method constructs a coarse-fine structure that combines a dynamical model with a statistical model based on employing the GREB model as the global framework and utilizing a Bayesian network constructed on the interrelationships between internal climate variables of the GREB model to achieve local optimization. To objectively validate the performance and generalization of the improved method, the method is applied to the simulation of surface temperature and temperature of the atmosphere based on the 3.75 degrees x 3.75 degrees global data sets by the National Centers for Environmental Prediction (NCEP) and the National Center for Atmospheric Research (NCAR) from 1985 to 2014. The results demonstrate that the improved model exhibits higher average accuracy and lower spatial differentiation than the original GREB model and is robust in long-term simulations. This approach addresses issues with the accuracy of the GREB model in local areas, which can be attributed to an overreliance on boundary and initial conditions, as well as a lack of fully usable observed data. Additionally, the model overcomes the challenge of poor robustness in statistical models due to ambiguous climate inclusions. Thus, the improved method provides a promising way to give a reliable and stable simulation of climate.