| Ensemble forecasting is an important direction for the development of numerical weather prediction. Opposed to a single deterministic forecast model it considers the uncertainty of the initial state and model. It results can reflect a variety of future weather conditions and provides more valuable information for the users. The more mature the ensemble prediction systems developed the more superior showed in forecasting capabilities. So the evaluation of ensemble forecast is becoming more and more important, the comparison and analysis will reflect the advantages of the ensemble forecasting visually and quantitatively. Numerical prediction system of China has developed from a single short-term forecasting model to numerical prediction systems that includes short-time and short-range, medium-range prediction model, and then the ensemble forecasting. Researches on the current performance and forecasting capabilities of T213L31 ensemble prediction system(EPS) is necessary. The other researches indicate that compared to advanced countries, there still has a long way to develop our EPS. Deepen the understanding of its performance and forecasting capabilities and the analytical investigation of the advantages and disadvantages will improve our knowledge of the system then helpful for the future improvements and it will to be a guide for operational forecasting.Using the data from THORPEX(The observing System Research and Predictability Experiment) Interactive Grand Global Ensemble (TIGGE),this paper studies the performance of the ensemble prediction system (T213L31 EPS) of China Meteorological Administration (CMA) over the East Asia area. The data includes the analysis and 1-to-10 day forecast data of the 500hPa geopotential height from 2007 to 2011 and the NCEP FNL data. RMSE, correlation coefficient and the rank histogram methods have been used. The ensemble-mean forecast of the T213L31 EPS is more accurate and skillful than its corresponding forecast, as the lead time increases, the superiority of ensemble-mean forecast is more significant especially from t+4 day. The T213L31 EPS still tends to underestimate the ensemble spread for longer lead times for the whole period. Seasonal analysis showed that due to the limited mode which T213L31 EPS based on, the lowest level of the forecast is still in summer, but the EPS is better than the corresponding forecast and has improved the forecast skill in summer. And the analysis shows that the systematic errors are different in four seasons.Prediction of the large-scale systems is more accurate than the meso-and micro-scale weather systems for numerical models. As a global medium-range forecasting system T213L31 EPS is usually for the 3-15 days medium range weather forecasts, but studies for the prediction of different synoptic-scales should be learned more for the ensemble forecasting and the deterministic forecasting. This paper uses nine point smoothing operator to separating the EPS and the deterministic prediction data into synoptic scale and subsynoptic scale to compare the prediction results. And the verification seems that for synoptic-scale system both have the better forecast than subsnoptic scale system. The results also shows that maybe the comparative advantage that ensemble mean has is in the improvement for forecasting the subsnoptic scale system. But for more than 7days forecast the ensemble forecasting itself is still insufficient to predict the subsynoptic scale system. The analytical results of the chosen case are confirmed the earlier conclusions and indicating that the ensemble mean prediction performance is better than the deterministic forecast, as well as for the subsynoptic scale system. With the increasing lead-time the capacity of the forecast is declining, and the deviation of the system strength lies mainly in the forecast bias. And it is possible that Ensemble mean is a better choice for forecast for each single ensemble member is not good from the analysis of RMSE and correlation coefficient. |
PDF Full Text Request