摘要 : An L2 regularized logistic regression model is developed in this study to predict weekly tropical cyclone (TC) genesis over the western North Pacific (WNP) Ocean and subregions of the WNP including the South China Sea (SCS), the w... 展开 An L2 regularized logistic regression model is developed in this study to predict weekly tropical cyclone (TC) genesis over the western North Pacific (WNP) Ocean and subregions of the WNP including the South China Sea (SCS), the western WNP (WWNP), and the eastern WNP (EWNP). The potential predictors for the TC genesis model include a time-varying TC genesis climatology, the Madden-Julian oscillation (MJO), the quasi-biweekly oscillation (QBWO), and ENSO. The relative importance of the predictors in a constructed L2 regression model is justified by a forward stepwise selection procedure for each region from a 0-week to a 7-week lead. Cross-validated hindcasts are then generated for the corresponding prediction schemes out to a 7-week lead. The TC genesis climatology generally improves the regional model skill, and the importance of intraseasonal oscillations and ENSO is regionally dependent. Over the WNP, there is increased model skill over the time-varying climatology in predicting weekly TC genesis out to a 4-week lead by including the MJO and QBWO, whereas ENSO has a limited impact. On a regional scale, ENSO and then either the MJO or QBWO are the two most important predictors over the EWNP andWWNP after the TC genesis climatology. The MJO is found to be the most important predictor over the SCS. The logistic regression model is shown to have comparable reliability and forecast skill scores to the ECMWF dynamical model on intraseasonal time scales. 收起