初值中云变量对T213预报性能的影响

采用云变量自由变化的方案，在国家气象中心全球业务模式T213L31的初始场中增加有关云变量的信息，通过2005年6—8月和2015年12月—2006年2月各3个月的连续滚动对比试验的统计分析和个例预报分析，研究探讨了全球模式初值中增加云变量对模式预报性能的影响。初步研究结果表明:采用自由变化的方案在初值场中增加云的信息，使模式能够较为合理地描述出模式预报初期与云相关变量分布和变化特征，降低了spin-up现象对模式前期降水预报能力的影响，同时对500 hPa形势场预报也有一定程度提高。

Impacts of the Cloud Initialization on T213 Forecasting Performances

Cloud variables such as cloud water, cloud ice and cloud cover are not treated in most operational data objective analyses and initialization schemes. While cloud variations are used as prognostic variables in a forecast model, they are necessary to be defined at initial time. The common method is to set cloud variations to zero at the initial time in the initial field, and some time of several hours are needed by the forecast model in doing spin up at the beginning of the run, which is sure to affect the forecasting capabilities such as precipitation and patterns and so on. The cloud free variation scheme is adopted to realize the adding of cloud variations information to the model initial fields of T213L31, the operational global model, then both a statistic analysis of a 3 month continuous running of test and operation schemes and a case study are carried out. In order to testify that the cloud free variation scheme does not foil the balance between dynamic and thermal variances at the start of model forecast, the stabilities of the model are checked and confirmed firstly by means of analyses on kinetic energy, temperature, and so on by long time integration. A synoptic case study, which happens on June 18—19, 2005, shows that with cloud information including cloud water, cloud ice and cloud cover being added to the initial model field, the spin up phenomenon disappears which always occurs at the starting period of forecasting, and often exists for 12 to 18 hours or longer, before the cloud free scheme being added in the modeling system. And the characteristics of distribution and variation of cloud related variables can be more reasonably depicted by the model than the operational one so that the improvement of the model forecasting performance is led to especially in the short time precipitation prediction. The 3 month continuous parallel experiment in winter and summer respectively are done together with corresponding operational ones. The results based on summer 3 month statistic analyses show that positive contributions to both precipitation and geopotential height pattern are made by adding cloud variations to initial field, that is obvious in the improvement of 24 hours accumulating precipitation's TS from 36 to 108 hours lead time, and the bias of precipitation increases a little at the same time. The 3 month averaged anomaly correlation coefficients and root mean square error at 500　hPa is better than operational one at 1—4 lead days. The results based on winter experiment show that the advantages of threat score over operational one within 60 hours lead time, and the bias of precipitation are identically better than operational one at each lead time, and the 3 month averaged anomaly correlation coefficients and root mean square error at 500 hPa are better than operational one. Positive impacts on operational model forecasting can be made by reasonably initializing cloud variations in the model field. This research is based on the operational global medium range weather forecast model, which confirms that the cloud free variation scheme is reliable and feasible to the operational application in the near future. In the long term, how much contribution will be made to the operation forecasting need to be further evaluated. Cloud variables such as cloud water, cloud ice and cloud cover are not treated in most operational data objective analyses and initialization schemes. While cloud variations are used as prognostic variables in a forecast model, they are necessary to be defined at initial time. The common method is to set cloud variations to zero at the initial time in the initial field, and some time of several hours are needed by the forecast model in doing spin up at the beginning of the run, which is sure to affect the forecasting capabilities such as precipitation and patterns and so on. The cloud free variation scheme is adopted to realize the adding of cloud variations information to the model initial fields of T213L31, the operational global model, then both a statistic analysis of a 3 month continuous running of test and operation schemes and a case study are arried out. In order to testify that the cloud free variation scheme does not foil the balance between dynamic and thermal variances at the start of model forecast, the stabilities of the model are checked and confirmed firstly by means of analyses on kinetic energy, temperature, and so on by long time integration. A synoptic case study, which happens on June 18—19, 2005, shows that with cloud information including cloud water, cloud ice and cloud cover being added to the initial model field, the spin up phenomenon disappears which always occurs at the starting period of forecasting, and often exists for 12 to 18 hours or longer, before the cloud free scheme being added in the modeling system. And the characteristics of distribution and variation of cloud related variables can be more reasonably depicted by the model than the operational one so that the improvement of the model forecasting performance is led to especially in the short time precipitation prediction. The 3 month continuous parallel experiment in winter and summer respectively are done together with corresponding operational ones. The results based on summer 3 month statistic analyses show that positive contributions to both precipitation and geopotential height pattern are made by adding cloud variations to initial field, that is obvious in the improvement of 24 hours accumulating precipitation's TS from 36 to 108 hours lead time, and the bias of precipitation increases a little at the same time. The 3 month averaged anomaly correlation coefficients and root mean square error at 500　hPa is better than operational one at 1—4 lead days. The results based on winter experiment show that the advantages of threat score over operational one within 60 hours lead time, and the bias of precipitation are identically better than operational one at each lead time, and the 3 month averaged anomaly correlation coefficients and root mean square error at 500 hPa are better than operational one. Positive impacts on operational model forecasting can be made by reasonably initializing cloud variations in the model field. This research is based on the operational global medium range weather forecast model, which confirms that the cloud free variation scheme is reliable and feasible to the operational application in the near future. In the long term, how much contribution will be made to the operation forecasting need to be further evaluated.