This paper analyzed statistical characteristics of 22 convective indexes in the 1645 samples from April to September, 2001-2009 at Yangkou Harbor and their correlation with lightning activities. The results have shown the monthly variations and different coefficients of the indexes, and then the elimination and forecasting equations were set up by means of bi level logistic regression using the multiple thresholds of indexes pw， SI， TT， JI for elimination and those of SSI， tcl_t， dt_58 for prediction with consideration of monthly diversity of each index which is in close correlation with lightning activities. The fitting test has confirmed good forecasting effects of the equations above.