How to code, validate, parallelize and utilize models that integrate dynamics, surface fluxes, aerosol-microphysics, and radiative processes to generate of clouds to understand aerosol effects on rainfall and climate?

A recent paper by Stevens et al. [1] compared simulations from 10 different cloud-resolving models with the DYCOMS-II aircraft observations found

" ... a compelling need for treating cloud top processes and significant leaps in the understanding of subgrid-scale physics, that is still beyond the reach of most centers."

Fortunately, we at Los Alamos anticipated this modeling challenge and had already begun taking this significant leap by developing a high-resolution cloud-resolving model by focusing on numerical details, formulations of aerosol-cloud microphysics, comparisons with observational effects, capturing subgrid-scale effects, and exploiting our massively parallel supercomputing platforms.

As part of our project, the three key components - numerics, experiments, and observations - needed to accurately model cloud-aerosol interactions at high spatial resolutions are currently being developed and integrated. Once validated, our cloud-aerosol model will be used to understand how aerosols modify mass, momentum, and energy budgets within the boundary-layer, thus providing critical information to develop parameterizations for both direct and indirect aerosol effects in coarser resolution global climate models. The numerical component involves solving the cloud-aerosol equation set via a nonlinear Newton- Krylov (NK) approach. This unique numerical approach is unlike most approaches used for cloud modeling in that it requires cloud parameterizations be smooth on the dynamical time scale of the problem [2]. This ensures that the temporal error is small and bounded during a NK simulation. We stress that temporal errors in more traditional approaches can become so large that model interpretation becomes essentially meaningless. Another critical numerical component of our work is limiting evaporation at cloud boundaries, without which long- lived stratus clouds can spuriously disappear [3]. Such problems are responsible for the rather pessimistic outlook on LES models reached by Stevens et al. [1]. We have solved a key problem in our cloud resolving LES by inventing and introducing a scheme that limits cloud evaporation. We evaluate the evaporation time constant using the dynamical eddy mixing time constant as the clock. When mixing is slower than evaporation, evaporation rates are reduced. Our simulations reproduce the DYCON-2 observations, which the 10 community LES models had, trouble with, as shown in the figure below.

LANL simulations of cloud water distributions (gm/Kg) for conditions measured during DYACOMS-II. As the tuning parameter "phi" that scales the cloud evaporation-limiting scheme in our LES model is increased the simulated clouds persist longer over larger areas, consistent with observations. "Phi" between 0.5-1 yields the best fit with observations.

Two new limiting approaches have also been developed, a continuous probability distribution function-based approach [3] commonly used in combustion modeling and a discrete approach based upon a stochastic particle model. The more traditional bulk and bin aerosol-cloud approaches are also included in our model to facilitate meaningful comparisons. Furthermore, Parameters in our aerosol microphysics model (e.g. what fraction of aerosol type (e. g. salt, sulfate, carbonaceous) are effective as a cloud condensation nuclei, and how effective are these at taking up water) are being tuned using laboratory experiments on water uptake on various aerosol types. Furthermore, we have utilized high- resolution observations of temperature and water distributions in shallow cumulus clouds from DOE's Multispectral Thermal Imager (MTI) satellite over Oklahoma's ARM site to validate our models.

References:

[1] Stevens, B., et. al., Evaluation of large-Eddy simulations via observations of nocturnal marine stratocumulus; Monthly Weather Review, June 2005; 133(6), pp. 1443-1462.

[2] Reisner, J., et al., Microphysical Cloud Resolving LES Model That Integrates Robust Numerical, Experimental, and Observational Components to Predict Aerosol Effects on Climate, AGU Spring Meeting, New Orleans, May 2005, Paper A41-F03.

[3] Jeffery, C. A., and Reisner, J., A study of cloud mixing and evolution using PDF methods: Cloud front propagation and evaporation, LA-UR-05-5421, under review J. Atmos. Sci., July 2005.