| dc.description.abstract | In this study an algorithm is developed for improving the model short-term quantitative precipitation forecasts (QPF). In methodology, the complete algorithm is composed of three main part : a newly designed multiple-Doppler radar synthesis technique, a modified thermodynamic retrieval method originally proposed by Gal-Chen, and a moisture adjustment schene. The performance of the proposed method is first investigated under the framework of OSSE (Observing System Simulation Experiment) including different situations.
Experiments results show that retrieved perturbation pressure and temperature have a good agreement with natural run (truth), but the main feature of perturbation water vapor mixing ratio also can be captured. Also, by assimilating the retrieved three-dimensional winds, thermodynamic, and microphysical parameters into a numerical model, the model QPF can significantly improve the accuracy of forecast for almost 1 hour long.
In sensitivity test, it is feasible to use the model outputs to replace the role played by a sounding for estimating the unknown constant at each altitude, which is a required quantity in the Gal-Chen method for deducing the thermodynamic field, and if one conducts a second assimilation even without sounding, additional improvements can be achieved. Moisture adjustment is found to provide a better rainfall precipitation forecast at the early stage of the model integration after the data assimilation. Different saturated condition in water vapor adjustment, the result seems to have less improvements compared to control run. The last part is time interval test for operational radar scanning, retrieval and forecast results shows a slightly different from natural run.
In real case study, we chose case 2008 SoWMEX IOP8 applied on this data assimilation method to retrieve atmospheric state variables and insert in numerical model for following precipitation forecast. The model after assimilating radar data significantly improves the accuracy of forecast for at least 3 hours compared with one without data assimilation in spite of over-estimated precipitation . In addition, the one without sounding data also can improve QPF as much as the one with sounding. Therefore, this data assimilation method is practicable to be implemented on real case study. | en_US |