A Study Of Cloud Physical Initialization In Rapid Refresh Assimilation And Prediction System

The cloud initialization technology is aimed at the physical quantity of water condensate,which is not easily observed in practice.Through the radar reflectivity observation data,the initial field of the water condensate and the related model physical and dynamic variables are adjusted,so that the cloud related information and characteristics at the initial time are as close to the balance of the dynamic and physical patterns as possible.It is an important way to solve the "spin-up" problem of numerical weather prediction and improve the accuracy of prediction.By using the second generation of SMS-WARR Shanghai Meteorological Service-WRF ADAS Rapid Refresh System,the contrast test and ideal test were analyzed with an emphasis of the influence of the cloud initialization and the related physical processes.An improved scheme of cloud initialization is proposed and a nunerical experiment is carried out.The main contents and conclusions show as follows:1)The comparison between results of with/without cloud initialization shows that the forecast using cloud initialization can greatly increase 00-06 hour rainfall T score,and the position of the forecasted precipitation is closer to the observation as well.Further analysis indicates that the cloud initialization can significantly improve the initial fields of hydrometeors,provide more accurate related information of the convection system,obtain the better forecasts of hydrometeors,vertical velocity,etc.over severe convection area,and largely avoid delay of precipitation occurrence during early stage.In the system,cloud analysis constructs the initial state of cloud hydrometeors and adjusts the temperature and humidity field according to corresponding physical constraints,which mitigate the inconsistency between the microphysics and dynamics at initial time,then shorten the"spin-up" time due to the lack of the initial microphysical information.However,results with/without cloud initialization from integrations of more than 6 hours are quite similar to each other.Cloud initialization can substantially improve the initial humidity analysis over the cloud areas,and result in a better position of the mesoscale vortex and low-level jet.A tornado occurred on June 23,2016 in Funing county,Jiangsu Province is simulated.A tornado parent storm is identified near Funing in the model output,which occurs in area with very strong convective activity.In a comparing simulation without rapid data refresh(cloud initialization),there is no supercell storm recognized showing the key role played by hourly data assimilation in the simulation of the tornado.2)Using the radar reflectivity which is retrieved from the model prediction results and two different precipitation particles calculation schemes of Kessler and Ferrier,the distribution of hydrometeors of the cloud initialization is diagnosed and analyzed.The results show that,although the inverse radar reflectivity is more coordinated with the background field,the adjustment of the three-dimensional cloud field in the background field is smaller,but in the new initial field of the cloud analysis,the content of cloud ice and cloud water is significantly greater than that in the background field,especially the increase of cloud ice content is very obvious,cloud particles(cloud ice and cloud water)of Kessler and Ferrier results are close The precipitation particles(snow water,shotgun and rainwater)of Kessler and Ferrier schemes have little difference,and are reduced than the precipitation particles in the background field.3)To improve the cloud initialization,a convective-scale velocity determined by surface sensible heat and latent heat flux as a criterion of convection is introduced into the shceme to estimating cloud ice and cloud water for stratiform and convective clouds separately.The results showed that,some cloud analysis grids are determined to be non-convection grids with the convection criterion?The distribution of the convection grids is similar to that of upward sensible heat flux,and display a diurnal variation at land.With separated stratiform and convective grids as well as different cloud water and ice schemes,the mixing ratios of cloud ice and cloud water are significantly reduced,which leads to weakening of radical adjustment of cloud particles in initial stage of the model integration especially in the land area.The precipitation intensity of 00-01h and 00-06h model forecast using improved cloud initialization scheme is reduced.In the cycle experiments,the area with strong precipitation tends to be smaller and the intensity is significantly weakened.