On the use of model physics parameters as control variables in data assimilation systems
In this presentation, we quantify the functional relationship between model parameters and observations using a Markov chain Monte Carlo (MCMC) algorithm. We examine cloud microphysics and radiation packages from a cloud resolving model that are similar to schemes used in modern regional and general circulation models, and demonstrate how the joint probability distribution returned from MCMC can be used to
• map the functional relationship between changes in model physics parameters and changes in model output,
• identify which parameters have the most significant effect on various model output fields,
• describe the nature of nonlinearity in the parameter-state relationship, and
• explore how changes in the characteristics of observations affect the model state.
The results of the MCMC-based inversion suggest that nonuniqueness in the relationship between changes in parameter values and changes in model output may be the reason for the loss in “parameter identifiability” noted in previous studies. We examine these results in detail and suggest ways in which nonuniqueness can be avoided in the construction of a data assimilation scheme that includes model physics parameters as control variables. We also compare results from simulations of different types of cloud system to how the influence of cloud microphysical parameters changes for different cloud systems.