The Calibration Of Wind Speed Ensemble Forecasts And The Analysis Of Forecast Inconsistency About The Wind Speed Forecasts

Based on the 10 m wind speed forecasts during the summer of 2012 from the ECMWF, CMA and NCEP in the TIGGE datasets, a Kalman filter bias-correction which combines with a sliding weight method has been done to calibrate the ensemble perturbed forecasts. The effect of this calibration method will be examined by RMSE and Talagrand Distribution. At the same time, the concept of forecast inconsistency will be introduced. The characteristics of the forecast inconsistency for the original 10 m wind speed perturbed ensemble forecasts(from ECMWF) and the revised 10 m wind speed ensemble perturbed forecasts(from ECMWF) have been conducted by using Jumpiness index and other different forecast jumps-the "flip", "flip-flop", "flip-flop-flip" and so on. On the one hand we can have a base understanding of the characteristics of the forecast inconsistency for the 10 m wind speed, on the other hand we introduce the forecast inconsistency as an evaluation standard of the deterministic forecast to examine the calibration effect of Kalman filter bias-correction method. By the results of the present study, we can conclude the conclusions as following:In general, for the all three centers, the RMSE of the 10 m wind speed ensemble forecasts which have different start time have be decreased by the calibration during all the lead time (12 h-240 h). But for ECMWF center and CMA center, when the start time is 12 UTC, the revised results are better. For the NCEP center, there is no apparent difference between the two start time. In terms of the spatial distribution of RMSE, the revised results are better in the low latitudes area no matter what the lead time. When the start time is 00 UTC, for ECMWF center and CMA center, the revised results at extend time forecasts are better than the results at short time forecasts. For NCEP center, the results are completely opposite to the other two centers. When the start time is 12 UTC, the revised results at short time forecasts are better than the results at extend time forecasts for all the centers. Compared with the 00 UTC, the revised results in high latitudes area are much better when the start time is 12 UTC. The calibration method has a good effect on improving the dispersion of ensemble members. The Talagrand pictures show that the U or L distribution are less appearance after calibration.Secondly, in terms of the period-average inconsistency features of the 10 m wind speed ensemble forecasts from ECMWF center, all average the period-average inconsistency indices increase with the forecast range in agreement with the practical experience that the forecasts are usually more consistent at short forecast ranges. After calibration the period-average inconsistency indices of ensemble mean are lower than before. This result shows that Kalman filter bias-correction method can reduce the forecast inconsistency of the 10 m wind speed ensemble forecasts. Thirdly, in frequency statistics of the inconsistency, the frequencies of the "flip", "flip-flop" and "flip-flop-flip" are in descending order, and the frequencies of the "flip", "flip-flop" and "flip-flop-flip" are lower after calibration. The calibration method has a better effect on reducing the "flip" and "flip-flop" than "flip-flop-flip".