基于ARMA模型的地磁偏角缺数处理方法

针对地震仪器记录的地磁数据存在单点缺失和连续多点缺失而不利于地震数据处理和地震预报的问题，同时为了快速处理非震异常值，本文提出将时间序列自回归移动平均（ARMA）预测模型用于地磁数据插值处理，并与均值插值、线性插值进行对比分析。结果表明，均值插值、线性插值和 ARMA 模型单点缺失的平均标准误差分别为 0.110 2、0.006 9 和 0.000 1，连续多点缺失的平均标准误差分别为 0.258 23、0.194 2 和 0.004 86，说明 ARMA 模型在单点缺失和连续多点缺失时均具有较低标准误差，且能很好地保持实际观测序列的曲线形态，插值效果较好，有望成为地磁数据序列处理的一种新方法。

Processing Method of Missing Number of Geomagnetic Declination Based on ARMA Model

There are single point missing and continuous multi-point missing in geomagnetic data recorded by seismic instruments, which is not conducive to seismic data processing and earthquake prediction. In order to quickly process the non-seismic abnormal data, we propose a time series autoregressive moving average (ARMA) prediction model for geomagnetic data interpolation processing. We compare the ARMA model with mean interpolation and linear interpolation. The results show that the mean standard errors of missing single point of mean interpolation, linear interpolation and ARMA model are 0.110 2, 0.006 9 and 0.000 1, and the mean standard errors of missing continuous multi-point are 0.258 23, 0.194 2 and 0.004 86, respectively. The results indicate that ARMA model has a low standard error in single point missing and continuous multi-point missing, which can well maintain the curve shape of the actual observation sequence, and the interpolation effect is better. It is expected to become a new method of geomagnetic data sequence processing.