In order to enhance the prediction accuracy of 0-2 h nowcasting of short-time heavy precipitation and thunderstorm gale, this paper proposes a severe convection probability nowcasting method based on the vertical profile characteristics of the dual-polarization radar (referred to as the CSCPVP method). By using the improved Bayesian probability method, the vertical profile characteristics of polarization radar of two types of severe convection disasters are introduced into the extrapolation model to realize the advanced identification of severe convection attributes. The broad constraints of regional model prediction are integrated to ensure that the proximity extrapolation prediction results of severe convection are more consistent with the actual dynamic and microphysical characteristics. The evaluation results of the entire rainy season (June-September 2023) in Zhejiang show that the CSCPVP method has significantly improved the nowcasting ability of two types of severe convection than the existing operational methods. Overall, the new method has improved 0-2 h nowcasting ability of systematic, local and distributed severe convection. The critical success index of short-time heavy precipitation increases from 8%-16% to 22%-26%, and the critical success index of thunderstorm gale increases from 7% to 10%-11%. Thus, the problem of false alarms and missing prediction has been effectively improved.