Impact of Assimilating SSM/I Rainfall Rates on Numerical Prediction of Winter Cyclones

View More View Less
  • 1 Naval Research Laboratory, Washington, D.C.
  • | 2 Naval Postgraduate School, Monterey, California
© Get Permissions
Full access

Abstract

A series of observing system simulation experiments (OSSE) and real data assimilation experiments were conducted to assess the impact of assimilating Special Sensor Microwave/Imager (SSM/I)-estimated rainfall rates on limited-area model predictions of intense winter cyclones.

For the OSSE, the slow-moving, fronto- and cyclogenesis along the cast coast of United States during the second intensive observation period (IOP 2) of the Genesis of Atlantic Lows Experiment (GALE) (26-28 January 1986) was selected as the test case. The perfect “observed” rainfall rates were obtained by an integration of a version of the Naval Research Laboratory (NRL) limited-area model, whereas the “forecast” was generated by a degraded version of the NRL model. A number of OSSEs were conducted in which the “observed” rainfall rates were assimilated into the “forecast” model. Rainfall rates of various data frequencies, different vertical beating profiles, various assimilation windows, and prescribed systematic errors were assimilated to test the sensitivity of the impact. It was found that assimilation of rainfall rates, in general, improves the forecast in terms of sea level pressure S1 scores when either the “observed” or model-determined vertical beating profiles were used. The improvement was insensitive to the error in rainfall magnitude estimates but was sensitive to errors in geographic locations of the precipitation. More frequent observations (additional sensors in orbits) had positive but gradually diminishing benefits.

Real SSM/I-measured rainfall rates were assimilated for the rapid-moving, intense marine cyclone of IOP 4 of the Experiment on Rapidly Intensifying Cyclones over the Atlantic (ERICA) (4–5 January 1989), which started from an initial offshore disturbance with a minimum pressure of 998 mb at 0000 UTC 4 January and developed into a very intense storm of 937 mb 24 h later. The NRL model simulated a well-behaved but less intense cyclogenesis episode based on the RAFS (Regional Analysis and Forecast System) initial analysis, reaching a minimum sea level pressure of 952 mb at 24 h. The first SSM/I aboard a DMSP (Defense Meteorological Satellite Program) satellite flew over the marine cyclone at 0000, 0930, and 2200 UTC 4 January and measured rainfall rates over portions of the warm and cold fronts associated with the cyclone. The SSM/I rainfall rates at 0000 and 0930 UTC were assimilated into the model as latent heating functions in ±3-h windows with model-determined vertical profiles. Two different methods were used to define the latent heating rates for the model in the assimilation experiments: 1) the model heating rates were defined by the maximum of the model computed and the SSM/I measured, and 2) the model beating rates were replaced by the SSM/I-measured rainfall rates within the SSM/I swath. Results of the assimilation experiments indicated that the assimilation in general leads to better intensity forecasts. The best forecast with assimilation predicted a 24-h minimum surface pressure of 943 mb, cutting the forecast error of the “no sat” forecast by 50%. This most efficient assimilation was carried out with assimilations of two-time SSM/I observations using the swath method. Further analysis indicated that the assimilation also resulted in better track and structure forecasts.

Abstract

A series of observing system simulation experiments (OSSE) and real data assimilation experiments were conducted to assess the impact of assimilating Special Sensor Microwave/Imager (SSM/I)-estimated rainfall rates on limited-area model predictions of intense winter cyclones.

For the OSSE, the slow-moving, fronto- and cyclogenesis along the cast coast of United States during the second intensive observation period (IOP 2) of the Genesis of Atlantic Lows Experiment (GALE) (26-28 January 1986) was selected as the test case. The perfect “observed” rainfall rates were obtained by an integration of a version of the Naval Research Laboratory (NRL) limited-area model, whereas the “forecast” was generated by a degraded version of the NRL model. A number of OSSEs were conducted in which the “observed” rainfall rates were assimilated into the “forecast” model. Rainfall rates of various data frequencies, different vertical beating profiles, various assimilation windows, and prescribed systematic errors were assimilated to test the sensitivity of the impact. It was found that assimilation of rainfall rates, in general, improves the forecast in terms of sea level pressure S1 scores when either the “observed” or model-determined vertical beating profiles were used. The improvement was insensitive to the error in rainfall magnitude estimates but was sensitive to errors in geographic locations of the precipitation. More frequent observations (additional sensors in orbits) had positive but gradually diminishing benefits.

Real SSM/I-measured rainfall rates were assimilated for the rapid-moving, intense marine cyclone of IOP 4 of the Experiment on Rapidly Intensifying Cyclones over the Atlantic (ERICA) (4–5 January 1989), which started from an initial offshore disturbance with a minimum pressure of 998 mb at 0000 UTC 4 January and developed into a very intense storm of 937 mb 24 h later. The NRL model simulated a well-behaved but less intense cyclogenesis episode based on the RAFS (Regional Analysis and Forecast System) initial analysis, reaching a minimum sea level pressure of 952 mb at 24 h. The first SSM/I aboard a DMSP (Defense Meteorological Satellite Program) satellite flew over the marine cyclone at 0000, 0930, and 2200 UTC 4 January and measured rainfall rates over portions of the warm and cold fronts associated with the cyclone. The SSM/I rainfall rates at 0000 and 0930 UTC were assimilated into the model as latent heating functions in ±3-h windows with model-determined vertical profiles. Two different methods were used to define the latent heating rates for the model in the assimilation experiments: 1) the model heating rates were defined by the maximum of the model computed and the SSM/I measured, and 2) the model beating rates were replaced by the SSM/I-measured rainfall rates within the SSM/I swath. Results of the assimilation experiments indicated that the assimilation in general leads to better intensity forecasts. The best forecast with assimilation predicted a 24-h minimum surface pressure of 943 mb, cutting the forecast error of the “no sat” forecast by 50%. This most efficient assimilation was carried out with assimilations of two-time SSM/I observations using the swath method. Further analysis indicated that the assimilation also resulted in better track and structure forecasts.

Save