Synthesis of Observed and Simulated Rain Microphysics to Inform a New Bayesian Statistical Framework for Microphysical Parameterization in Climate Models

Principal Investigator(s):
Marcus van Lier-Walqui, The Trustees of Columbia University in the City of New York
Future progress in the modeling of cloud microphysical processes depends upon our ability to leverage the information provided by new observational data sets. Such efforts are not trivial, and are complicated by the nature of current state-of-the-art microphysical parameterization schemes. To wit, schemes (both bulk and bin) contain many free parameters whose values are chosen ad-hoc, or through tuning using limited or uncertain empirical data. Furthermore, the very structure of schemes (such as the choice of a given parameterized drop size distribution, or functional form for a process rate) introduces additional uncertainty. Recent progress in microphysical parameterization schemes has often exacerbated this problem by increasing complexity and thereby increasing the number of parameters and assumptions that must be supported by empirical or theoretical evidence. On the other hand, recent advances in observational capabilities, such as available ARM polarimetric and zenith-pointing radars, allow for unprecedented information on rain microphysical processes that can be used to improve schemes. We will develop a new bulk microphysical parameterization scheme that facilitates constraint by observations and quantification of uncertainty. The key elements of this scheme are that:

  1. it does not assume an a priori functional form for its raindrop size distribution,
  2. process rates are represented as power-law functions of prognostic size distribution moments,
  3. complexity can be added systematically as required by observed microphysical behavior.

Development of the scheme, and constraint on its free parameters and structural form, will be driven by statistical comparison to both simulated and real polarimetric radar observations of rain microphysics. Simulated observations will be generated by a bin microphysics scheme that has undergone rigorous uncertainty quantification, using Bayesian inference and ARM radar observations from the AMIE/DYNAMO and MC3E field campaigns. Constraint of parameters in the new bulk scheme will use an independent set of AMIE/DYNAMO and MC3E observations with Bayesian inference tools such as Markov Chain Monte Carlo samplers and transdimensional inference methods. The new scheme will then be implemented into the Weather Research and Forecasting model (WRF) and tested for deep convective cases from MC3E and AMIE/DYNAMO, with a focus on convective processes through rain evaporation, cold pool properties, and storm dynamics. Expected Outcomes:

  • Development and observational constraint of a novel bulk rain microphysics parameterization with reduced structural uncertainties. 
  • Estimation of uncertainty in state-of-the-art bin schemes (whose uncertainties have not been robustly estimated with comparison to observations).
  • Investigation of information content in polarimetric radar observations of rain (an effort that will yield recommendations for observational calibration, uncertainty, and scanning strategy requirements).
  • Greater confidence in the ability to simulate microphysical-dynamical linkages in deep convection using an improved, observationally-constrained approach for bulk rain microphysics, and quantification of the impacts of microphysical uncertainty on deep convective processes.