The Gridpoint Statistical Interpolation (GSI) and Weather Research Forecasting (WRF) models will be used to carry out assimilation and forecast runs. We used 12-hour assimilation with a 90 minute assimilation window. Control experiment involves assimilating a set of baseline data (eg. conventional data, AMSU-A and B, MHS, and HIRS). In the experiment run, clear sky AIRS radiances are added. Constraints by the model top, only 83 channels from the AIRS Near Real Time (NRT) channel set are assimilated. Decision that a channel peaks above the model top is based on adjoint calculations using the Tropical climatological profile. Channels that are contaminated by ozone and solar radiation and are flag noisy/bad are also discarded.
Forecasts are run out to 48 hours and verified against radiosondes, satellite observations not assimilated and precipitation will be carried out. Forecast impact with the assimilation of AIRS will also be evaluated.