Applied Research on Adaptive Observation for Identifying Sensitive Regions Based upon TIGGE Data

Accurate prediction of high impact weather is very important. Adaptive observation has an immediate significance to improve the quality of high impact weather forecast. THORPEX (THe Observing system Research and Predictability EXperiment) is a ten year international atmospheric research program with the primary objective to improve the accuracy of 1 day to 2 week high impact weather forecasts for the benefit of society and economy. THORPEX is developed under four sub programs, and two of them are related to adaptive observation, which are observing systems, data assimilation and observing strategies, so adaptive observation is the foundation and prerequisite to achieve the ultimate goal of THORPEX. Adaptive observation is a type of observation interactive between weather forecast (or weather service) and observation. The key to adaptive observation lies in identifying sensitive regions. Sensitive regions are defined as localized regions from where the analysis errors grow significantly and thereby the forecast skills are degraded. As a new method for identification of sensitive regions which is based on ensemble predictions, ETKF (Ensemble Transform Kalman Filter) has many advantages, such as the program simplicity, independency and so on. On the basis of an ensemble prediction solely, ETKF solves Kalman equation in the space of ensemble, and it can estimate the reductions of prediction error covariance caused by adding observation immediately. The research of ETKF has a proactive effect on both adaptive observation and ensemble predictions. On the basis of TIGGE (THORPEX Interactive Grand Global Ensemble) data, through identification and comparison of the sensitive regions of two different types of heavy precipitation events, specific aspects in the actual application of ETKF method to the adaptive observation are analyzed in detail. The results indicate that the appropriate horizontal resolution and coverage of ensemble data can be used for a reasonable result and shorter computational time, and the reasonable result shows that the geographical distribution of the signal variance maxima calculated from different resolutions and coverages are basically coincident and a relatively small coverage can even be used for the heavy precipitation event under regional weather circulation comparatively to that under macroscale weather circulation; identified sensitive regions are more credible by using ensemble predictions available which are initialized more recently; identified sensitive regions calculated from different meteorological centers' ensemble data are more consistent and reliable for the heavy precipitation event under clear macroscale weather circulation than under regional weather circulation; a calculation of signal variance using nine adjacent grids at one time makes the wide distribution of signal variance maxima, and makes it not conducive to locate sensitive regions more precisely. To the result of sensitive regions, the effects could negligible if the analysis errors of routine observation are changed moderately. Sensitive regions under different circulation are dependent on the selected metric to some extent. In conclusion, the sensitive regions identified by ETKF are understandable and reasonable.