In this paper, we proposed a method of adaptive variation amplitude, which can be used to extract medium and short-term anomalies of apparent resistivity. The Fourier sliding method is applied to calculate the annual variation component of the daily or monthly average data, and the observed value is subtracted from the annual variation to obtain the annual-subtraction data. Then the wavelet decomposition method is used to extract the trend change of the annual-subtraction data. When the annual-subtraction data curve crosses the trend line each time, the corresponding data in the annual-subtraction data set is regarded as the starting point. Then the relative variation amplitude of the later data relative to the starting point is calculated. The abnormal threshold line is obtained when the cumulative mean square deviation of the change range in the entire analysis period reaches 2.5 times high. This method effectively alleviates the problem of defining the abnormal start time in traditional original curve analysis. The abnormal threshold line based on the multi-year normal variation range can be used to extract medium and short-term anomalies in which variation range is less than 1%.