Optimizing Radiative Transfer Calculations in the CRTM for Clouds, Precipitation and Aerosols

NCEP will soon be exploiting satellite microwave radiance data in clouds and precipitation for its new hybrid variational-ensemble data assimilation (DA) system. These high impact areas present special challenges for the calculation of forward and adjoint radiances for the Community Radiative Transfer Model (CRTM) because of greater computational expense. This added expense can potentially limit the use of these and other satellite data. However, through judicious choice of the number of streams (more streams mean greater accuracy) within the CRTM, this computational cost can be reduced without sacrificing significant accuracy.

This study describes one such way of selecting beforehand the optimal number of streams needed by the CRTM for NWP model profiles containing clouds, precipitation, and/or aerosols. The method is extremely fast and uses the successive order of scattering approximation as a measure of the degree of multiple scattering expected. A higher degree of scattering usually requires a greater number of streams to achieve the same accuracy. We call this measure of scattering a scattering indicator (SI).

Selecting the optimal number of streams will depend on the desired accuracy, the wavelength of radiation, among other factors. This study will focus mainly on infrared and microwave wavelengths. Selection rules for the SI for switching to a different number of streams were determined from CRTM calculations for a wide range of cloud, precipitation and aerosol profiles from Weather Research & Forecasting (WRF) and WRF-Chem model simulations.

Assuming an accuracy of 0.5 K or less, results for simulated Special Sensor Microwave Imager/Sounder and Advanced Microwave Scanning Radiometer-EOS brightness temperatures for hurricane Katrina show that scattering is negligible for the upper atmosphere temperature channels (54-64 GHz), while only 2 or 4 streams are needed for window channels (<50 GHz). However, the water vapor channels (183 GHz) require at least 4 streams and sometimes 16 streams. Additional results will be presented at the symposium.