| Analog forecast based on stepwise filter method, which calculate the analog degree between the forecast and historical fields of different elements (physical) fields stepwise, a number of analog cases are obtained, the practical situation of these cases are taken as the basis to make weather forecast. This approach not only has the advantages of traditional analog forecasting methods, but also gets a better combination of the physical diagnosis and the weather situation. So it is now widely used in the medium- range and long- range weather forecast.To meet the actual business demand, this paper improves the current method of analog forecasting with stepwise filter.Using daily NCEP reanalysis data from 1958 to 2002,the T213 numerical forecast products and daily surface observations of 72 stations from 2003 to 2006 around the Bohai region, the medium-range elements predictions of 72 stations were made by using the improved analog filtering programs, and take the programs with highest accuray as the final’optimal’program. Then a visual system of medium-range weather forecasting based on the ’optimal’program has been established, the system can be used as a forecasting and assistant tool for the daily weather forecast in all levels of the meteorological stations, the key functions include the view of 4-10 numerical products forecast fields and the history similar cases, results of the medium-term forecast, and the visual output of forecast conclusions, historical fact, etc.Experiment forecast has been producted by using numerical product from 2007 to 2008, the specific results are as follows:(1) precipitation forecast:according to the TS of experiment forecast, the optimal program has certain ability to forecast precipitation in 4-7d in the winter,4-8 d in autumn,4-5 d in spring and summer.(2) Cloud cover forecast:the accuray of winter half year is higher than that of summer half year; the accuracy of daily mean low cloud cover forecast is higher than that of daily mean total cloud cover, cloud cover forecast accuray in high-latitude area is higher than in lower latitude area; the accuracy of the total cloud and the low cloud cover are higher than 60% in the medium-term forecast, both average RMSE of them are less than 3. (3) 24h temperature change forecast:the descending order of 24h average, maximum and minimum temperature change experiment forecast accuracy is summer, autumn, winter and spring; accuracy of average temperature is the highest throughout the year, the accuracy of maximum temperature is higher than the minimum temperature in winter and autumn, the accuracy of minimum temperature is higher than the maximum temperature in spring and summer. The mean temperature forecast accuracy is more than 60% from 2007 to 2008.(4) Wind Speed Forecast:For the four times daily wind speed forecast, the prediction effect of 0-2 wind scale is the best, its mean TS is 67.5% in 4-10d; the mean TS of wind scale^5 is the lowest, which is 5.5% in 4-10d; For the daily maximum wind speed forecast, the TS of 3-4 wind scale is higher than the wind scale≥5, the former is 55.2% in 4-10d, and the latter is 17.4% in 4-10d; There are significant seasonal variations for wind speed forecast, the TS of 0-2 wind scale is the lower in spring than other seasons, but the TS of 3-4 and the wind scale≥5 is higher in spring than other seasons, a probable cause is that there are high frequency strong winds in spring. |
PDF Full Text Request