Development of the COAMPS adjoint mesoscale modeling system for assimilating microwave radiances within hurricanes
An adjoint mesoscale modeling system based on the Naval Research Laboratory's Coupled Ocean Atmosphere Mesoscale Prediction System (COAMPS) atmospheric model was created for use in sensitivity and data assimilation experiments. In addition to the tangent linear and adjoint models of the dynamical core of the COAMPS model, the system includes the tangent linear and adjoint models of the boundary layer turbulent kinetic energy, cumulus, and explicit moist physics parameterizations. The inclusion of these adjoint model physics schemes allows for assimilation experiments involving rain-affected observations such as microwave radiances.
A radiative transfer model which includes the effects of hydrometeors on atmospheric radiation was linked to the adjoint modeling system to assimilate microwave radiance observations. Probability distribution functions of model-produced and SSM/I observed brightness temperatures show that the mesoscale prediction overestimates the areas of precipitation, but overall matches the microwave observations quite well. Furthermore, estimates of vertical background error covariance matrices for the hydrometeor variables were calculated using differences between model forecasts which utilized different explicit moisture schemes. The statistics of the differences between the forecasts were assumed to be the same as the statistics of the background error for these variables. The inverse of these matrices (which are needed for data assimilation) were computed using Singular Value Decomposition. Only the largest singular value was kept in calculating the inverse. This ensured that all of the elements of the inverse matrix were non-negative.
Finally, microwave radiance observations for Hurricane Bonnie (1998) were assimilated in a 4-dimensional variational data assimilation framework using the COAMPS adjoint model. The model-produced radiances calculated from the analysis fields after the assimilation process match the observations well for the lower frequency channels which are sensitive to liquid precipitation and water vapor. In the highest frequency channel, where the presence of frozen hydrometeors can have a large impact on the radiance value, the match between the analysis and the observations was not as good. The forecasted hurricane was slightly stronger after the assimilation of microwave radiances in terms of both maximum surface windspeed and minimum central sea level pressure, and some improvement was seen in radiance space as well. More observations from within the hurricane, which will improve the analysis of other variables, will most likely be needed to see a greater forecast impact from the assimilation of these observations.