| Because the losses caused by meteorological disasters are becoming more and more serious in recent years. It is urgent for meteorological departments to provid extended range forecast with longer period of validity than mid-term prediction from the point of materials and manpower used to prevent disasters. NCEP /NCAR reanalysis data(200 h Pa, 500 h Pa and 700 h Pa) with 5 components( height, humidity, U component, V component and vertical velocity) during the period 1980-2014 was used to reserve low frequency signal with the period of 30-60 days and calculate the climate state of low frequency meteorological element field, realize Butterworth band pass filter. Then extend the low frequency data of element fields day by day from 150 days before to the next 30 days for summer months during the period 1980-2009 with the methods of period analysis, scoring double standard screening and optimal subset regression methods. Third, test the fitting results by analyzing climatic anomaly correlation coefficient and comparing actual figure and fitted figure. Fourth, extended-range(10-30days) precipitatipn forecast and results testing are carried out for 44 stations located in the middle region of northwest China with the method of gradually- filtering analog method. The results are as follows:(1) For low frequency wave forecast of 10-30 days, climate anomaly correlation coefficients at 200 h Pa, 500 h Pa and 700 h Pa levels display a decreasing trend with the extending of the period of validity. It decreases rapidly in early 10 days, and tends to be stable in later 10-30 days. 5 low frequency meteorological element fields at three different levels(200 h Pa, 500 h Pa and 700 h Pa) pass 95% confidence level in 6 key regions(West Siberia, Xinjiang,Lake Baikal, Central Asia, the bay of Bengal, the Tibetan Plateau and the Western Pacific Region),which have important effects on the Extended-range precipitation of the middle region of northwest C hina.(2) The results of the climate anomaly correlation coefficients of 5 low frequency element fields at three different levels showed that the coefficients in the Northern Hemisphere were greater than those on global scale, and the coefficients on global scale were greater than those in the Southern Hemisphere. Furthermore, the coefficients in equatorial region(10°S-10°N) were greater than those in low latitudes(30°S-30°N), meantime the coefficients in low latitudes(30°S-30°N) were greater than those in the middle latitudes(60°S-60°N).(3) We found that the fitted system configuration, center, the location and range of systems with high and low values of low frequency obtained from the methods of period analysis, scoring double standard screening and optimal subset regression methods are all in good line with actual results by comparing the fitted figures and actual figures. However, the fitting on the systems of long and narrow low frequency sometimes suggest phenomena of fracture and the fitted values for centers usually lower than actual values.(4) The results of gradually- filtering analog method showed that: when the number of precipitation simila r cases of each station α≥3, the Cs score was the highest. Based on the Cs scores of precipitation forcast(R≥0.1mm) for 44 stations located in the middle region of northwest China, we knew that 6 stations were located in northwest Q inghai with Cs mean values less than 0.3, 3 stations(Wuwei, Jingtai and Jingyuan) with Cs mean values ranging from 0.3 to 0.5, other 35 stations with Cs scores greater than 0.5 and accounting for 79.50% of the total number of stations.(5) As a whole, the Cs mean value for precipitation forecast(R≥0.1mm) for 44 stations located in the middle region of northwest China was 0.57, the CS mean value was 0.52 when R≥1mm, the Cs mean value was 0.34 when R≥5mm. results of Northwest Midwest precipitation forecasts by gradually- filtering analog method showed that the mean value of Cs score of the average prediction accuracy of 44 stations was 0.57 when R≥0.1mm, the mean value of Cs score was 0.52 the mean value of Cs score was 0.34 when R≥5mm. The mean values of 3 different kinds Cs scores were all higher than climatic probability value, and Cs mean values tend to be stable with the forecast time extending.In conclusion, the methods of period analysis, scoring double standard screening and optimal subset regression methods were used to extend low frequency signa l in this paper and Extended-range forecast experiments were carried out with the method of gradually- filtering analog method, the results are believable and methods involved in this study are feasible. |
PDF Full Text Request