Data Assimilation: Current Status and Outlook for the Future

Most extreme precipitation events, regardless of season, are associated with mesoscale phenomena: floods from hurricane rainbands or mesoscale convective systems, bands of heavy snow from cold air passing over warm water or from slant-wise convection, or prolonged freezing rain from overrunning in sharp frontal zones. Accurate forecasts of these events depend upon models that faithfully replicate atmospheric behavior at many scales of motion but also upon effective assimilation of observations relevant to the event. For many years, observations have been provided well in advance of, and sometimes without regard for, our ability to assimilate them. The major barriers to effective assimilation are:

• ineffective quality control measures

• lack of forward models (algorithms for estimating an observed parameter from the model variables)

• imperfect knowledge of error statistics either for the observation source or the short-term model forecast that is to be corrected by the observations

• lack of appropriate constraints to link the mass, motion, and moisture fields in the finished analysis

• what to do when the correction to the background is large (e.g., thunderstorm in the wrong place).

Steps to address these problems are being considered within NOAA and elsewhere.