Simultaneous Nonlinear Estimation of Model Physics Uncertainty and Model State in Simulations of Deep Convection

Errors and/or uncertainty in model physics parameterizations are a primary source of forecast error in weather and climate prediction. In particular, simplifying assumptions about the form of the particle size distribution of ice and liquid condensate have an important effect on the details of cloud and precipitation development and feedback on the radiative fluxes, heating rates, and thermodynamic environment. Consequently, efforts are being undertaken to constrain the uncertainty in model physics parameterizations, largely via estimation of the empirical parameters that represent subgrid cloud variability. Nonlinearity in the parameter-state functional relationship presents a significant challenge to model uncertainty characterization using linear Gaussian methodologies.

Previous work has shown that model uncertainty can be effectively characterized in off-line one-dimensional models of deep convection using a nonlinear Markov chain Monte Carlo algorithm. In this presentation, model parameter uncertainty is evaluated for model configurations in which changes in parameters are allowed to feed back on the convective scale dynamics and environment. When model parameter values are constrained with observations, many aspects of the deep convective structure are also uniquely determined. Where this is not the case, the results of the nonlinear parameter estimation provide guidance as to which observations are required to produce robust estimates of convective dynamics and environment, as well as their uncertainty characteristics.