Wavelet-based Methods For Detecting Earthquake Information In Remote Sensing Data

Satellite remote sensing technology has been applied to various fields of earthquake science, but due to the lack of effective methods of data processing technology for earthquake remote sensing data, most information obtained from earthquake remote sensing data has not been fully utilized. So an effective technology of processing earthquake remote sensing data can be used to detect earthquake anomalies, monitor and predict earthquakes with remote sensing information available.A wavelet method is an analysis method with a fixed size window but its shape can be changed with time and frequency of local transformation. That is, in the low-frequency part the higher frequency resolution has the lower temporal resolution, by contrast in the high-frequency part, the higher time resolution has lower frequency resolution, it is thus called as a mathematical microscope. More recently wavelet methods have been widely applied in signal analysis, image processing, computer classification and identification, fault diagnosis of large-scale machinery, and studying earth science.Based on the basic principle of wavelet methods, we adapted the four mathematical methods of wavelet maxima, wavelet coefficients, wavelet decomposition and wavelet holder exponent. By considering the Wenchuan earthquake (Ms8.0) occurred on May 12th, 2008 and the Pu'er earthquake (Ms6.4) occurred on June 3rd, 2007, we analyzed and processed electromagnetic data observed by the Demeter electromagnetic satellite of France and long-wave radiation (OLR) data observed by the NOAA meteorological satellite of the United States. A series of comparative analysis results provide the validity of the four methods in detecting seismic anomalies within remote sensing data.1,Wavelet maxima for detecting anomalies in OLR data observed by the NOAA satellite OLR is the radiation of flux energy reflected to the sky after the Earth surface absorbs solar radiation. Due to constraints of physical periodic activities in the Solar-Earth system, OLR from the Earth has the various characteristics of periodicities; In addition, Earth interior activities, such as earthquakes and volcanoes, have scatter and outward emission, which will form OLR. Such activities normally are non-periodic, discontinuous and isolated. Based on the physical causes of long-wave radiation mentioned above, this section uses wavelet maxima to detect seismic anomalies within OLR data.Specific studying method includes:1) By taking account of the tectonic background, we define an experimental area, we use an epicenter as the center and divide the experimental area into 9 grids by 300km;2) By employing one-dimensional continuous wavelet methods and the method of singularity detection, we calculated wavelet maxima of OLR in each gird, and draw wavelet maxima curves;3) Maxima curve in each grid are rearranged onto one diagram to form a wavelet maxima sequence curve, and earthquake anomalies are identified in the curves through analyzing wavelet maxima distributing over the wavelet maxima sequence curves.4) Detecting earthquake remote sensing information through a series of comparative analysis experiments over the Wenchuan earthquake (Ms8.0) and the Pu'er earthquake (Ms6.4).5) Taking account of the great energy and large influence sphere of the Wenchuan earthquake with 8.0 magnitude, we comparatively analyze disturbance area and non-disturbance area, based on the history of seismic activities and geological features. In this study we selected three experimental areas, namely region 1, region 2 and the Wenchuan region, the Wenchuan region covers the earthquake epicenter, region 1 is close to the fault zone of the Wenchuan earthquake, region 2 is far away from the seismic zone, which is a non-seismic area. The specific geographic locations of these three regions are as follows: Wenchuan region: from 30°N to 33°N, 103°E to 106°E, Region 1: from 28°N to 31°N, 105°E to 108°E, about 200 kilometers away from the Wenchuan region, Region 2: from 44°N to 47°N, 113°E to 116°E, about 1,400 kilometers away from the Wenchuan region.6) In order to make forcible experimental results, We use the epicenter of the Wenchuan earthquake as the center and divide the experimental area (21°N to 41°N, 93°E to 113°E), which is 2000km by 2000km, the used wavelet analysis method is db1, wavelet scales are 16, and the study OLR mean data, which is from October 2006 to September 2008.7) We use the epicenter of the Pu'er earthquake as the center and divide the experimental area (13°N to 33°N, 91°E to 111°E), which is 2000km by 2000km, the used wavelet analysis method is gaus3, wavelet scales are 16, and the study OLR mean data, which is from May 2006 to April 2008.Preliminary experiment results are presented belowas follows:1) Prior to the Wenchuan earthquake, significant earthquake anomalies can be detected within 300 kilometers from the earthquake epicenter in OLR data; The anomalies detected from wavelet maxima sequence curves for the Wenchuan region are obviously greater than other two regions, comparing with the anomaly magnitudes of the two comparative experiment regions, region 1 is greater than region 2, and basically no anomalies can be found in region 2, although the influence sphere of the Wenchuan earthquake is large, but it decreases with distance, and is limited.2) The decrease in earthquake anomalies can be detected within 2000 kilometers from the earthquake epicenter in OLR data eight months prior to Wenchuan earthquake (September 2007), which continue until January 2008, anomalies intensity increase dramatically four months before Wenchuan earthquake (February 2008), this reflects great energy change during Wenchuan earthquake; anomalies intensity in August 2008 exceeded the mean value of 141.1%, which may caused by a large number of strong aftershocks after Wenchuan earthquake.3) Three non-periodic, discontinuous and isolated anomalies are detected in OLR data prior to Pu'er earthquake; comparing with two years data, from September 3rd, 2006 to September 3rd, 2007 and from September 3rd, 2007 to September 3rd, 2008, the experimental results show that, yearly periodic variation in wavelet maxima of two year data is similar to some extent, but the magnitude of anomalies in OLR data is larger from Pu'er earthquake.4) The decrease in earthquake anomalies can be detected within 2000 kilometers from the earthquake epicenter in OLR data five months prior to Pu'er earthquake (January 2007), anomalies intensity increase dramatically in April 2007 and reach maximum, this reflects great energy change during Pu'er earthquake.5) In general, the magnitude of anomalies and influence sphere in the Wenchuan earthquake is larger than Pu'er earthquake, which show a positive correlation with the earthquake intensity.6) Experimental results are basically consistent with the previous conclusion using other methods, which validate the effectiveness of wavelet maxima.2,Wavelet methods for detecting anomalies in Demeter electromagnetic satellite dataA conventional method for studying the Demeter electromagnetic satellite data is manually to identify anomalies in Demeter level-2 image data, this method is not only easily to cause a miscarriage of justice, but also it may not be able to help to carry out an analysis of Demeter level-1 data in depth, which resultings in not fully utilizing information obtained.Through preliminarily interpreting the level-2 image data of electron density (Ne), electron temperature (Te), ion density (Ni) in the Langmuir probe (ISL) and ion temperature (Ti) in plasma analyzer (IAP) observed by the Demeter satellite, this section selected 10 satellite half-orbits from four earthquakes occurred in China, using wavelet coefficients, wavelet decomposition and wavelet holder exponent to detect anomalies, and summarize anomalies of characteristics.Specific studying method is: by employing one-dimensional continuous wavelet transforms, we calculate wavelet coefficients with different scales and then highlight earthquake anomalies by folding wavelet coefficients of each scale; by decomposing the Demeter satellite data into high-frequency and low-frequency parts, we identify earthquake anomalies from the parts related to earthquakes; by employing one-dimensional complex Morlet wavelet transform and a linear regression, we calculate wavelet holder exponents for detecting hidden earthquake anomalies in the original sensing data.Preliminary explanation and conclusions are as follows:1) Obvious anomalies could be detected about 2000 kilometers within the epicenter area before the Wenchuan earthquake by using wavelet coefficients to process four physical parameters of half orbit 20595-1 original data of the Demeter electromagnetic satellite.2) By using the above three wavelet methods to analyze all spectrum of half-orbit 15440-1 data, obvious anomalies could be detected in five half-orbits of Demeter electromagnetic satellite before the Pu'er earthquake; meanwhile by using wavelet holder exponent, anomalies could also be detected before the Pu'er earthquake in resulting curves of the parameters of the original data.