Use of Four-Dimensional Data Assimilation by Newtonian Relaxation and Latent-Heat Forcing to Improve a Mesoscale-Model Precipitation Forecast: A Case Study

View More View Less
  • 1 Department of Meteorology, The Pennsylvania State University, University Park, Pennsylvania
© Get Permissions
Full access

Abstract

The Penn State/NCAR mesoscale model has been used in a study of special static- and dynamic-initialization techniques that improve a very-short-range forecast of the heavy convective rainfall that occurred in Texas, Oklahoma and Kansas during 9–10 May 1979, the SESAME IV study period. In this study, the model is initialized during the precipitation event. Two types of four-dimensional data assimilation (FDDA) procedures are used in the dynamic-initialization experiments in order to incorporate data during a 12-hour preforecast period. With the first type, FDDA by Newtonian relaxation is used to incorporate sounding data during the preforecast period. With the second FDDA procedure, radar-based precipitation-rate estimates and hourly raingage data are used to define a three-dimensional latent-heating rate field that contributes to the diabatic heating term in the model's thermodynamic equation during the preforecast period. This latent-heating specification procedure is also employed in static-initialization experiments, where the observed rainfall rate and radar echo pattern near the initial time of the forecast are used to infer a latent-heating rate that is specified in the mesoscale model's thermodynamic equation during the early part of the actual forecast. The precipitation forecasts from these static- and dynamic-initialization experiments are compared with the forecast produced when only operational radiosonde data are used in a conventional static initialization.

The conventional (control) initialization procedure that used only operational radiosonde data produced a precipitation prediction for the first three to four hours of the forecast period that would have been inadequate in an operational setting. Whereas at the initial time of the forecast there was substantial convective precipitation observed in a band near the edge of an elevated mixed layer, the model did not initiate the heavy rainfall until about the fourth hour of the forecast.

The use of the experimental static initialization with prescribed latent heating during the first forecast hour produced greatly improved rainfall rates during the first three to four hours. The success of the technique was shown to be not especially sensitive to moderate variations in the duration, intensity and vertical distribution of the imposed heating. Applications of the Newtonian-relaxation procedure during the preforecast period, that relaxed the model solution toward the initial large-scale analysis, produced a better precipitation forecast than did the control, with a maximum in approximately the correct position, but the intensities were too small. Combined use of either the preforecast or in-forecast latent-heat forcing with the Newtonian relaxation produced an improved forecast of rainfall intensity compared to use of the Newtonian relaxation alone. Even though both the experimental static- and dynamic-initialization procedures produced considerably improved very-short-range precipitation forecasts, compared to the control, the experimental static-initialization procedure that used latent-heat forcing during the first forecast hour did slightly better for this case.

Abstract

The Penn State/NCAR mesoscale model has been used in a study of special static- and dynamic-initialization techniques that improve a very-short-range forecast of the heavy convective rainfall that occurred in Texas, Oklahoma and Kansas during 9–10 May 1979, the SESAME IV study period. In this study, the model is initialized during the precipitation event. Two types of four-dimensional data assimilation (FDDA) procedures are used in the dynamic-initialization experiments in order to incorporate data during a 12-hour preforecast period. With the first type, FDDA by Newtonian relaxation is used to incorporate sounding data during the preforecast period. With the second FDDA procedure, radar-based precipitation-rate estimates and hourly raingage data are used to define a three-dimensional latent-heating rate field that contributes to the diabatic heating term in the model's thermodynamic equation during the preforecast period. This latent-heating specification procedure is also employed in static-initialization experiments, where the observed rainfall rate and radar echo pattern near the initial time of the forecast are used to infer a latent-heating rate that is specified in the mesoscale model's thermodynamic equation during the early part of the actual forecast. The precipitation forecasts from these static- and dynamic-initialization experiments are compared with the forecast produced when only operational radiosonde data are used in a conventional static initialization.

The conventional (control) initialization procedure that used only operational radiosonde data produced a precipitation prediction for the first three to four hours of the forecast period that would have been inadequate in an operational setting. Whereas at the initial time of the forecast there was substantial convective precipitation observed in a band near the edge of an elevated mixed layer, the model did not initiate the heavy rainfall until about the fourth hour of the forecast.

The use of the experimental static initialization with prescribed latent heating during the first forecast hour produced greatly improved rainfall rates during the first three to four hours. The success of the technique was shown to be not especially sensitive to moderate variations in the duration, intensity and vertical distribution of the imposed heating. Applications of the Newtonian-relaxation procedure during the preforecast period, that relaxed the model solution toward the initial large-scale analysis, produced a better precipitation forecast than did the control, with a maximum in approximately the correct position, but the intensities were too small. Combined use of either the preforecast or in-forecast latent-heat forcing with the Newtonian relaxation produced an improved forecast of rainfall intensity compared to use of the Newtonian relaxation alone. Even though both the experimental static- and dynamic-initialization procedures produced considerably improved very-short-range precipitation forecasts, compared to the control, the experimental static-initialization procedure that used latent-heat forcing during the first forecast hour did slightly better for this case.

Save