Model Sensitivity-Determinate Multiparameter Estimation in Coupled Climate Models

While the information estimation theory based on Bayes’ theorem is developed as various data assimilation algorithms for state estimation, it has also been applied to model parameter estimation. The resulted observation-estimated parameters can mitigate model bias. Parameter estimation is very promising for a coupled climate model to constrain its climate drift in climate simulation and prediction. However, given the existence of numerous model parameters, how to systematically perform parameter estimation is a research topic. Linking model sensitivities with the signal-to-noise ratio of parameter estimation, this study develops a physically-based methodology of simultaneous multiple parameter estimation for coupled climate models with biased physics. While either all the parameters within the biased physical scheme or only the most influential physical parameters being optimized, can greatly mitigate the model biases induced by biased physics, better results for climate estimation and prediction will be obtained when using this physically-based methodology of simultaneous multiple parameter estimation. These results provide a guideline when the real observations are assimilated into a coupled general circulation model that includes imperfect physical schemes for improving the performance of climate estimation and prediction by multiple parameter estimation.