Abstract: This paper calculates the vertical intensity polarization by using the second data at 88 geomagnetic stations during 2015 to 2021, and systematically summarizes the calculation method of polarization value and the criterion of polarization anomaly discrimination by combing time-space relationship between high-value anomalies and earthquakes. And on this basis, we extract the quantifiable prediction index, which is defined as: ① more than 20% of stations have polarization values higher than two times of the mean square error and continue three days. ② The anomaly area is a space with normalized zero polarization value greater than 0.2. ③ The dominant time of earthquake occurrence is within six months. ④ The earthquake magnitude is positively related to the size of the anomaly area. According to the above criteria, 16 groups of anomalies (32 abnormal areas) are extracted. The prediction efficiency evaluation results show that R-value (R=0.53) is higher than R₀-value (R₀=0.234) with 97.5% confidence level, based on the analysis of seismic examples of accurate report, false report and missing report. Here, we analyze the magnetic disturbance characteristics before the 2021 Maduo, Qinghai M7.4 earthquake using polarization method. The result shows that the synchronous high-value anomaly occurred 16 days before the earthquake, more than 20% of stations has polarization anomalies that last three days and the anomaly area is 14.1×10⁴ km². This anomaly which is up to the standard of index system can be determined as a pre-earthquake polarization anomaly. The study suggests that the above prediction index system has obvious earthquake prediction significance, can be used to extract the polarization anomaly before strong earthquake, and can provide reference for earthquake situation determination.

Key words: Seismo-magnetic, Vertical intensity polarization, Prediction efficiency, Index system, the 2021 Maduo M7.4 earthquake