An Idealized Assimilation Experiment of Satellite-Based Observations for the Analysis and Prediction of Tropical Multiscale Weather Systems

The present study explores the potential improvement in practical predictability of tropical multiscale weather systems from the assimilation of satellite-based observations using a perfect-model observing system simulation experiment (OSSE). The investigated observing networks include the retrieval products of atmospheric temperature and humidity profiles from Advanced TIROS Operational Vertical Sounder (ATOVS) and Global Positioning System Radio Occultation (GPSRO), Atmospheric Motion Vectors (AMV), Meteosat-7 infrared brightness temperature (Met7-Tb), and Cyclone Global Navigation Satellite System (CYGNSS) surface wind speed. All observations are assimilated every 3 h for a 17-day test period using Ensemble Kalman Filter (EnKF) and a regional Weather Research and Forecasting (WRF) model to evaluate their impact. The tropical weather consists of large-scale equatorial Rossby (ER), Kelvin, mixed-Rossby-gravity (MRG), and inertio-gravity (IG) waves, as well as IG waves at smaller scales. For the regional model, the boundary condition provides some predictability for the large-scale waves even without data assimilation, although large errors may still occur in wave amplitude. The assimilation of ATOVS and AMV improves analysis accuracy of wind, temperature, humidity and hydrometeors for the large- to intermediate-scale waves and extend the predictability limit by ~4 days. The Met7-Tb further improves the model representation of hydrometeor near cloud top at smaller scales, although the 3-h assimilation interval is not enough to retain this extra prediction skill. CYGNSS is able to further improve large-scale low-level wind and temperature. The sensitivity to the choice of localization distances and to varying observation density is also tested.