Objective Techniques For Forecasting Tropical Cyclones Motion Over South China Sea In Summer

Basing on the sample of Tropical Cyclone(TC) in summer (including three months of July, August and September) from 1960 to 2003 for 44 years that heads west enters the South China Sea,and according to the characteristic of TC and its mutual function of the environment flow field, we setting up the multiple regression which selecting the predictors by condition number both in the predictors of climatology persistance and Numerical forecasting(NWP) products to predict the TC motion for 24 hours.First, we selected two kinds of elmetary factor-group as the foundation of setting up the multiple regression which performed on 24h forecasts to the route of TC. These factors involed in climatology persistence and NWP. The factors of climatology persistence are according to the characteristic of the durative transformation TC. The further are using the NCEP/NCAR Global Reanalysis Data and taking the moves TC as the center, then select some grid of the region about the motion TC as the physical quantity predictors (NWP). Second, we singled out some factors in these elmetary factors again, there are two methods to ascertain: one scheme is selecting several important predictors by the best correlation with TC considering the characteristic that their good instruction to predict the way of TC,the other scheme is choosing anther predictors by condition number in all the rest elmetary factors. Finally we setting up a new prognostic equation what based on those factors which chosen deliberately to predict the TC motion. The result indicated, the new model what based on selecting factors by condition number is good, its independent samples test error for 24 hours of this three months (July, August and September) is 153.9 km. The results show a large improvement over other objective forecasting methods due to the condition number reducing and controlling the Multi-Collinearity among the independent variable efficiency.In addition, we contrast the new model what ground on selecting factors by condition number to the traditionary method names stepwise regression based on the same samples. The predictors which in the experiment used the primary factor-groupwhich rested on with the new method has been completely consistent. In reason, we try to change the F to setting up several different equations to predict the TC motion. The mean error for 24 hours in the same independent samples of this three months by traditionary model is beyond 200 km. The distance are far bigger than the new model's and it suggest that the traditionary model's ability is badly because the forecasting equations are influenced by the Multi-Collinearity among the independent variable efficiency.Furthermore, we determinant the skill level of the new model to the CLIPER model which the standard to checking the other methods,and the method mainly establish the equations according to TC's predictors of climatology persistance. In the experiment, we setting up the CLIPER model based on the same elmetary factors of climatology persistence as that the new model rest on. We also control the F to setting up several different CLIPER equations.The results show the skill of the new scheme what based on selecting factors by condition number is superior to the CLIPER scheme in the same samples,and it suggests that the new model could be utilizes to forecast the route of TC in actual.According to the above results it's successful to get the new model in this paper with selecting several important predictors by the best correlation and choosing the other predictors by condition number in all the elmetary factors of climatology persistance and Numerical forecasting(NWP) products to predict the TC motion for 24 hours. It's an effective way for the operation because of the better experiment is based on the mass modeling samples and independency samples.