Etkf Assimilation And Application Of Adaptive Observation In Tropospheres

Ensemble-based data assimilation are used and developed in recent years. Ensemble Transform kalman filter (ETKF) is one of these schemes and it can be used in adaptive observation because of its characteristic. Former studies and field tests indicate the effectiveness of this method used in the ensemble data assimilation and adaptive observation for synoptic scale system. ETKF based data assimilation and adaptive observation are applied to Mei-yu heavy rainfall which is a meso-scale system. Observing System Simulation Experiments (OSSE) and real-data experiments demonstrate the effectiveness of ETKF based data assimilation for meso-scale Mei-yu heavy rainfall. Sensitivity experiments are used to investigate the sensitivity to localization scale. The feasibility of ETKF based adaptive observation used in Mei-yu heavy rainfall is explored in the last chaper. Some connection is found between sensitive area and weather systems.Through Observing System Simulation Experiments (OSSE) and real-data experiments, it’s found that ETKF can decrease the forecast error and improves the forecast in a Mei-yu heavy rainfall. The root mean square errors (RMSE) of most variables after24hours decrease about25%in OSSE and the rainfall area is more accurate after assimilation. The ETKF performance is degraded in real-data experiment because of model error. RMSEs of wind decrease less than20%, while temperature and humidity decrease only about5%. However, the forecast of precipitation, especially the6hr accumulated precipitation is improved significant after assimilation. It’s also found that the struction and evolution of error is nearly the same. The basic statements of the weather are nearly the same before or after assimilation, so the structure and evolution of error is nearly the same.Because the ensemble number is small, sampling error is unavoidable. Covariance localization should be used in ensemble-based assimilation. The optimum localization scale is about1500km for synoptic scale system. There are many factors influence the optimum localization scale, and the most important one is the scale of weather system. According to my research, the horizontal optimum scale is about 400km to600km in this case. However, sensitivity experiments demonstrated that the effect of data assimilation is sensitive to horizontal localization scale, and is not sensitive to vertical localization scale。This may relate to relatively higher vertical observation density.Adaptive observation is an idea to improve the quality of short term weather forecast. The idea is to finding the most sensitive area to a weather system which people concerned with, then adding and assimilating observations in this sensitive area to improve forecast effectively. The key point is how we could find the sensitive area accurately. Though OSSE and Exhaustive Attack method, the sensitive area is in Qin Hai province36h ago in the8-9July2007Mei-yu heavy rainfall’s case, which is consistent with the sensitive area calculated by ETKF. As target time closer to verification time, the sensitive area is closer to verification area. It’s found that the sensitive areas have some relationship with the troughs in500hPa. At the same time, the sensitive areas are tilted as the weather systems are tilted. These results indicate that the sensitive area determined by ETKF is trustable.