Impact of a new radiative transfer parameterization in forecast and climate mode.

A major overhaul of the radiative transfer (RT) parameterization, used in the operational NWP models (called GEM), is under way at the Meteorological Service of Canada (MSC). The objective is to replace the operational broad band RT scheme with a correlated k-distribution (CKD) method recently proposed by Li and Barker (2005). In conjunction with the replacement of the RT scheme, several other aspects of the RT parameterization are being revisited; ozone climatology, aerosol climatology, cloud optical properties and cloud subgrid-scale inhomogeneity. The CKD method is capable of achieving an accuracy comparable to line by line models with an extreme reduction in the number of RT operations performed. This is achieved by sorting the highly variable absorption coefficients in ascending order, for a given homogeneous layer and spectral band. Given the now smoothly varying function, representative values of the absorption coefficients can be selected. Li and Barker (2005) have proposed a new CKD model for gaseous transmission that uses much less intervals in the cumulative probability space than other CKD models. Up to 1 mb only 20 (solar) + 22 (infrared) main intervals are needed. This scheme uses 9 frequency intervals for LW and 4 for SW. Note that only the near-IR portion of the SW spectrum is treated using the CKD method, the rest of the SW spectrum is dealt with in frequency space. This scheme is being used at the Canadian Centre for Climate Modelling and Analysis (CCCma) in the GCM version 4 model. At MSC, this new scheme has been tested both in forecast mode and in climate mode. It has been tested within the current operational regional and global GEM model and also within two versions in development (GEM-Meso and GEM-Strato). The GEM-Meso has a significantly increased vertical and horizontal resolution while the model top is increased from 10 to 0.1 hPa in the GEM-Strato version. Results from these tests have shown a much improved temperature structure of the stratosphere as well as the position of the tropopause and the jet streams. A comparison with the global radiosonde network has also shown some systematic improvements in temperature and winds for short and medium range forecasts. Comparison with surface radiative fluxes measured at ARM and SURFRAD sites have shown reduced biases in the model fluxes. The increased downward LW flux reduces the cold surface temperature bias present in the regional model (North America at 14km resolution). At the 12th AMS Conference on Atmospheric Radiation, the impact of this new RT parameterization on short, medium and climate time scale simulations will be presented. Results of a multi year comparison of surface radiative fluxes will also be shown.