AMSU-A Bias Correction for COAMPS/NAVDAS

It is generally acknowledged that showing positive impact from direct radiance assimilation is more difficult for mesoscale models than for global models. Aside from the very real difficulties in defining positive impact quantitatively and verifiably, there are statistical issues as well. For our global NWP model NOGAPS (Navy Operational Global Atmospheric Prediction System), two weeks of global radiance data were needed to produce stable bias correction coefficients for the Advanced Microwave Sounding Unit A (AMSU-A) in our 3DVar system, NAVDAS (NRL Atmospheric Variational Data Assimilation System). These coefficients depend on the underlying global model, and so may not be directly useful for our mesoscale model, COAMPS®. Because COAMPS is a limited area model, it takes longer to collect adequate statistics for bias correction, which is problematic if e.g. it takes longer than a season. Increasing the data density is problematic as well, because correlated error may be introduced in the horizontal. Ideally we must carefully combine information on radiance bias from the global model with information from the mesoscale model to produce the best bias correction for a given region. Results from experiments utilizing data and predictors from both NOGAPS and COAMPS will be shown.