The Analysis And Test Of The Predictability Of T213L31 Objective

In order to checkout the T213 model's capability, the quality of its products and its predictability, in this paper, we calculate standard error (SE) and correlation coefficient (R) of the given fields using the model's two years and five months' data from January 1, 2002 to May 31, 2004. In addition, a scheme of 5×5 latitude and longitude interval fields is used to calculate SE and R of five key elements in different layers, and the features of the fields in different seasons are also discussed. After finding out the features, rules and correlation factors for the change forecast skill (namely anomaly correlation coefficient), we research its changing mechanism primarily. Considering the complexity of the atmosphere's space-time change, we design models based on Support Vector Machine (SVM) for different seasons, respectively.The above-mentioned results show: (1) SE and R of different elements' fields in different seasons and space-time have obvious different features. SE is relatively better in Autumn than other seasons, and SE in Spring and Winter takes the second place, in Summer, SE behave more unstable. R has the similar features; From forecast effect, SE and R of five elements in different layers gradually become worse with forecast time elapsing; SE in different layers shows different features; From low-levels to high-levels, the SE of temperature fields become less as a whole, however, the relative humidity fields' forecast results are reverse. SE of height fields and wind vector fields of u and v are bigger in the middle layers as a whole, which are relatively smaller in low layers and high layers. (2) Models based on the SVM are proposed which are used to forecast the forecast skill. The results show that the forecast values can preferably reflect the forecast skill's change trend. It proves that using SVM to forecast the forecast??ll is feasible and significant.