Enhancing Cloud Radiative Processes and Radiation Efficiency in the Advanced Research Weather Research and Forecasting (WRF) Model
Project Team
Principal Investigator
Collaborative Institutional Lead
The primary objectives of this project are to evaluate and investigate improvements to the treatment of cloud radiative processes and to implement enhancements in radiation performance in the Advanced Research version of the Weather Research and Forecasting (WRF) model. The availability of the AER longwave and shortwave radiation code, RRTMG, as a radiation option in the standard WRF provides a highly accurate and effective foundation from which to further enhance cloud radiative processes that are critical for achieving accurate simulations of weather and climate. This will be accomplished through direct comparison of modeled cloud properties over diverse geographic areas and time periods with surface and satellite measurements to isolate and to prioritize specific deficiencies. Computational efficiency is equally essential for effective numerical weather prediction, and this research will implement and test a version of RRTMG, to be developed under other support, that will utilize graphics processing unit (GPU) technology to greatly improve its computational performance.
A comprehensive assessment of cloud radiative processes in WRF will be achieved by establishing the accuracy of the individual modeled components of these processes such as cloud amount, cloud physical and optical properties and cloud radiative heating rates through comparison to surface and atmospheric measurements including the DOE Atmospheric System Research (ASR) Climate Modeling Best Estimate (CMBE) data product. Co-located cloud measurements from space will provide additional closure to identify model cloud property discrepancies. Previous research has shown that considerable deficiencies in cloud amounts are present in WRF relative to observations. Accurate representation of the heating profile in clouds is of particular interest, since it is through their radiative heating that clouds feed back into the local dynamics and microphysics, especially as these effects are currently represented in dynamical models. This work will also investigate and seek an improvement to a deficiency (on the order of 10 percent) discovered in previous research in the radiative heating near the top of highly scattering clouds related to the two-stream multiple scattering approach that is typically used in dynamical models. The MIT convection and cloud scheme, which has been integrated into WRF for other research, will be utilized in this project as an alternate means of parameterizing and evaluating convective and large-scale cloud cover.
As more accurate and sophisticated algorithms are developed to simulate physical processes in global models, it is critical that their computational efficiency also be considered. This objective will be addressed in this work by examining a means of accelerating radiative transfer, which is one of the most expensive processes in dynamical models. If used effectively, GPUs can provide a substantial improvement in speed by supporting the parallel computation of large numbers of independent radiative calculations. As a k-distribution model, RRTMG is especially well suited to this procedure due to its relatively large internal pseudo-spectral dimension (of order 100) that, when combined with the horizontal grid vector in the dynamical model, will take great advantage of the GPU capability. A GPU-compatible version of RRTMG will be implemented in WRF, and thorough testing will be performed to ensure that it retains the original level of accuracy.