China is one of the most affected countries by the earthquake disaster in the world. The earthquake not only damaged people’s lives and property safety, but also triggered a series of secondary disasters such as landslides, mud-rock flows and collapses. Earthquake prediction is one of the difficult scientific problems in the world which needs to be explored for a long time. Earthquake precursor research is the key to earthquake prediction.

Earthquake is the result of instability rupture after the strain on fault accumulates to the limit under the action of tectonic stress. The borehole strain gauge can detect the small stress load change before rock fracture, so borehole strain observation can not only record long period strain change, but also provide high frequency strain information, which is the basis of earthquake precursor research. In the borehole strain observation, the strain response caused by solid tide, air pressure, water level and other environmental factors will drown out the pre-earthquake strain anomaly. Therefore, it is of great significance to study the environmental response and anomaly extraction method of borehole strain data, and obtain the earthquake related strain anomaly accurately and reliably from the strain observation data.

At present, many scholars have conducted researches on environmental response removal and pre-earthquake anomaly extraction. This paper based on the environmental observation data, the strain response caused by solid tide, air pressure and water level was solved and removed to obtain the crustal strain. solid tide is the periodic deformation of the earth under the gravitational force of the sun and the moon, which is the main reason for the periodic variation of diurnal wave and semi-diurnal wave in the borehole strain observation data. Air pressure and water level have immediate negative correlation to strain data. Time Series Decomposition is to decompose the time series with complex changes into several sub-components. The long-term background trend and periodic trend of the observed data before and after the Lushan earthquake （January 2011 to January 2014） were removed by Time Series Decomposition.Multi-channel Singular Spectrum Analysis can use the correlation between signals of different channels to decompose the data. In this paper, the trace matrix is constructed of borehole strain residual, air pressure and water level three-channel measurement signal for Multi-channel Singular Spectrum Analysis, and obtain the strain response of air pressure and water level. The results show that Time Series Decomposition can effectively remove the periodic changes dominated by solid tides. The correlation coefficient of the water level and its strain response is −0.97, and in 96.1% of the days the absolute value of the correlation coefficient of the daily air pressure and its strain response is greater than 0.9, which verify removal of the environmental response the proposed method effectively.

Normal crustal strain, which is not affected by strong earthquakes and environment, is a short-period random signal. According to its characteristic of Gauss, this paper extracts the crustal strain data by day and accumulates the number of anomalies. It is found that the accumulation of negentropy anomalies of the Lushan earthquake showed a linear increase -- acceleration increase -- a small amount of anomalies -- acceleration increase. Then the negentropy anomalies accumulation is compared with the Benioff strain accumulation. It is found that the negentropy anomaly accumulation trend is consistent with the Benioff strain accumulation trend. Combined with the fracture evolution process of rock stress loading （initial micro-fracture -- extensional fracture -- stress lock -- earthquake eruption）, it was guessed that the abnormal acceleration increase of the Lushan earthquake 4 to 6 months before the earthquake may be an earthquake precursor and related to the extensional fracture.

In order to further analyze the advantages of environmental response removal and negentropy anomaly extraction methods, this paper compared the method in this paper and difference processing. A large amount of precursor anomaly information is contained in the high-frequency components of the observed data, and the crustal strain is also dominated by high-frequency information. The difference process can remove the low-frequency information in the observed data, and one-dimensional difference processing is carried out on the standardized strain observation data to make a comparison with the crustal strain after removing the influence of environmental factors in this paper. The negentropy anomaly accumulation and non-Gaussian distribution days accumulation showed the environmental response removal method can effectively remove the disturbance caused by air pressure, and the crustal strain data after the removal of environmental response can more effectively extract the earthquake precursor anomaly.

The above conclusions sufficiently indicate that the methods in this paper are effective for removing the environmental response of borehole strain observation data and extracting earthquake precursor anomalies.