| "Spin-up" is one of the most serious problems for short-range precipitation forecast. It isrelated to the cloud information deficiencies in the first guess field. This paper focuses oncloud information initialization and assimilation in GRAPES-MESO NWP model. Themotivation for this research is to provide a reference for GRAPES-MOSO operationalshort-range precipitation forecasting.Firstly, the characteristics of cloud and precipitation occurrence and evolution are analyzedin the GRAPES model. The model parameters used is the same with operational numericalprediction models. The results shows:(1) Because of the initial cloud information deficiencies,cloud water appears after10-15minutes; rain water is produced after30-40minutes; rain fallsto the ground after50-70minutes; and cloud water and rain water are developed toquasi-equilibrium state in1-1.5hours.(2) Compared with the observation, the modelprecipitation organization structure is still insufficient in2hours. There is a significantdifference between the predicted precipitation and the observation rain. It could hardly meetthe demands of the operational now-casting system.The cloud analysis scheme is described in detail in the next section. The scheme includesgrid could fraction retrieving technique, the could particles (cloud water, rain water)quantitative analysis technique, and the technique of classifying the could classification andthe quantitative calculating of participation particles. The radar reflectivity data, Geostationarysatellite data, and surface observation are used to construct three-cloud and precipitation fields.The feasibility of the cloud analysis scheme is studied through the "718" heavy rain case inJinan city, Shandong province. The results show that with this cloud analysis method canprovide significant improvement in precipitation range and intensity. The cloud "warm start",in which the previous time forecast cloud particles is used as the initial field, is also tested bycompared with the cloud cold start. It is found that the forecast cloud information could notimprove the short range precipitation prediction.The rationality of cloud analysis scheme introduced into GRAPES_MESO model wasanalyzed in cloud macroscopic and microscopic characteristics. The cloud macroscopiccharacteristics is diagnosed through (1) cloud base tested by sounding data collected perminute;(2)cloud top pressure tested by MODIS data;(3) cloud total amount tested by FYGeostationary satellite products;(4) cloud liquid water path tested by MODIS data. The cloud microscopic properties are investigated by TRMM data, CLOUDSAT data and radar data. Thediagnostic content includes cloud hydrometeors distribution at different horizontal levels;cloud hydrometeors vertical distribution, etc. The diagnostic results are summarized:(1) Thecloud macroscopic characteristics on distribution of cloud amount, cloud top pressure,hydrometers are reasonable.(2) The cloud base analysis is not quite correct.(3) Thebackground cloud fraction is overestimated.(4) Stratus cloud droplets are also overvalued.The cloud analysis scheme is optimized to make up the defects that are revealed in theabove results. The sounding data collected per minute is introduced into cloud analysis.Surface observation cloud base analysis method is updated. A new relative humidity threshold,which is matching the model high horizontal resolution, is adopted in background relativehumidity cloud retrieve algorithm. A more suitable stratiform rain cloud droplets particlequantitative analysis program is introduced.The cloud analysis scheme optimized is examined in GRAPES meso-scale model by a casestudy. A set of experiments are carried out. The experiments with or without cloud analysiswere designed to assess the sensitivities of cloud information. The cloud analysis usingdifferent kind of observation type was designed to test the sensitivities of those data. And theexperiments with cycle or not to investigate the impact of cloud analysis in cycle. It is foundthat the application of the cloud analysis has a significant positive impact on short rangeprecipitation prediction. In one hour, the precipitation forecast is more close to observation.And the positive effect can last over six hours. Especially initialized cloud information incycle forecast, prediction rain hourly evolution is more accurate. As for the impactsensitivities for the cloud analysis scheme, the radar data has a positive influence onprecipitation prediction within the radar coverage or downstream field. The surfaceobservation and minutely sounding play an important role in precipitation prediction, whilethe effect by using satellite data is not readily noticeable. Nevertheless the effect of short-termprecipitation with cloud analysis depends on the background quality. If a false alarm ofprecipitation prediction exists in the background, the relative superfluous part of the modelprecipitation prediction is difficult to be removed.In order to meet the requirements of now-casting, the radar VAD data assimilation isresearched. The radar VAD wind is mean horizontal wind at different height in cloud and rainfield.. Weather it will play a positive impact or not in data assimilation, the key problem is theVAD data quality control. The multiple iterative fitting quality control scheme is developed.This method could purify the data gradually to satisfy the harmonic characteristics, and thenegative bias compared with the background is also corrected by using it. Furthermore, aseries of comprehensive quality control is carried out for different error source. Case andmonthly continuous experiments reveal that the retrieved VAD data could satisfy the demands of quality control for data assimilation.To make up the disadvantage of radar VAD without pressure information, a method toretrieve the pressure for VAD wind referring to the ground radar observation is proposed. Themethod involves the introducing polytrophic atmosphere into the pressure–altitude formula inwhich the surface information is obtained from the surface radar observation. One year testindicates that the surface radar observation reference method reduces the error obviouslycompared with the traditional climate statistic method. Especially the vertical temperaturelapse rate varying with time is better than other lapse rate.The squall line case study is presented. The result of assimilating the VAD data withquality control in GAPES RAFS (Rapid analysis and forecast system) shows positive impacton wind and rain forecast.
|
PDF Full Text Request