Nowadays, topic detection and tracking (TDT) has been widely used. As one research tasks for TDT, new event detection can provide prior knowledge to TDT, so it has great theoretical research significance in the field of TDT. Because LDA model cannot automatically identify new events, and the number of LDA topic had been determined by the artificial, or by repeated experiments, it has low efficiency . Here presented a new method, which considering the correlation of subject terms and report time, it can dynamically adaptive updated topics, then detect the new event. This method can adaptive dynamic upload the number of topics, then determinate reasonable theme number. Experiment results demonstrate that this method has some ad-vantages, increased the sensitivity of new events detection.