A THORPEX Based Study Of Weather Predicability

In this study, various aspects of predictability of high-impact weather are investi-gated by the use of TIGGE dataset, operational numerical model data archive and observational data from T-PARC field campaign under the THORPEX project.We first conducted an investigation of the characteristics of operational en-semble forecast model of China using TIGGE dataset. Based on the verification result of geopotential height at 500 hPa, the upper limit of our operational con-trol forecast (without perturbation) is about 6 days. By averaging 15 ensemble members with the same weight, the ensemble forecast can outperform the control forecast by increasing the forecast limit by a day. A simple super-ensemble fore-cast composed of ensemble forecasts from China and NCEP can outperform the simple ensemble average from both centers, although the forecast skill of NCEP ensemble is significantly better than that of China. Moreover, the verification of different model variables generally showed different forecast skill with the humid-ity forecast in the lower troposphere having the least forecast skill. The analysis of spatial error indicates that the region over the northwest pacific has the low-est predictability. Further investigation over this region shows that the forecast error mainly comes from the smaller scale perturbations associated with tropical cyclone. A direct assessment of the forecast skill of tropical cyclone shows that current operational ensemble models have very limited forecast skill.Further verification of high resolution global and regional deterministic fore-cast models showed the same characteristics of forecast error. The lower-level humidity-related model variables still suffered from low forecast skill. The spatial and temporal correlation analysis between humidity and other model variables showed strong relations of model error between model variables. A similar as-sessment of the tropical cyclone forecast skill using the high resolution dataset shows no significant improvement over that of the low resolution model. Due to the limited forecast skill of tropical cyclone, an experimental method based on the numerical model output was developed to improve tropical cyclone related heavy-rain weather. This method combines physical and statistical components to forecast the heavy rain. Heavy rain related weather events are the most com-mon type of high-impact weathers. The forecast of heavy precipitation has been a challenge for both research and operational community for decades. Previous studies of precipitation forecast mainly use pure statistical techniques. Although significant improvement has been made, these statistical methods are not intuitive in understanding the relations between predictands and predictors, and further understanding of precipitation related dynamics and physics can not be easily incorporated into the forecast model. In an attempt to overcome the limitation of traditional methods, the idea of ’ingredents’for heavy precipitation is intro-duced as a physical constraint for the resultant forecast model. It is found that the newly developed method can improve the forecast of heavy precipitation by running a series of reforecast from 2004-2007. The early success of such physical regularized forecast model shows the potential of applying physical constrains to statistical model.Beyond the assessment of the predictability of operational numerical model and the development of new method for the heavy precipitation forecast, an ob-servational study on the structure and evolution of tropical cyclone rainband was conducted based on the high-resolution data collected during T-PARC field cam-paign. Since the predictability of tropical cyclone is closely related to its internal structure, the high resolution dataset collected during the T-PARC field cam-paign was used to investigate the structure and dynamics of Typhoon Hagupit’s principal rainband. Radar observations showed intense and highly-organized con-vective cells in Hagupit’s principal rainband which has not been documented be-fore. The along-line averaged vertical structure of Hagupit’s principal rainband showed many characteristics of those in a middle-latitude squall-line. By com-paring structural characteristics of different rainband types, it was noticed that the different rainband structures can be explained by the cold pool dynamics which has been widely applied to squall-lines. Further investigation of the envi- ronmental structure and surface cold pool showed that the two dynamic factors are quasi-balanced. The similarity of structure and dynamics between Hagupit’s principal rainband and squall-lines indicated the role of cold pool dynamics in de-termining the structure of different rainbands. We further conducted a numerical simulation of Typhoon Hagupit. The simulation successfully reproduced intense and highly-organized rainbands, and the investigation of the relative strength of surface cold pool and local vertical wind shear also showed that the two factors are quasi-balanced.