-
Strain-drop, which is an important earthquake source parameter, is the reduction of average elastic strain over the earthquake
rupture during earthquake. Strain-drop together with seismic moment provides fundamental information on the physics of earthquake
rupture. The only way to obtain strain-drop is to study the seismic waveform. The observed seismic waveform is the convolution
among source, propagation path and site terms, and we can only obtain strain-drop from the source term. Traditionally, researchers use the
waveform of a smaller event, which occurred very close to a large event, to represent the path and site term, and then de-convolute
it from waveform of the large event to obtain the separated source term, but such method has several limitations and could not be
generally applied to large data set. Shearer et al (2006) proposed several procedures to obtain stress drops (strain-drop times rigidity)
in a large data set. I generally followed their procedures, and independently wrote the codes for all the calculations to obtain
strain-drops from P- phase waveform window. Working with Prof. Yehuda Ben-Zion and Prof. Zhigang Peng, I applied these procedures to
waveforms in the Izmit-Duzce aftershock data set with ~26000 small earthquakes, and obtained strain-drops to ~7500 well-recorded events
with good signal quality (Yang et al, 2009). Based on the above infrastructure, I continue to
analyze the source parameters associated with S- phase waveform window, so that the seismic radiated energy could be estimated once
we know both the source parameters of P- and S- wave. Due to the fact that the scattered P- wave coda complicates the analysis of S- wave from
a single time window around the S- arrival, I stack the S- coda waves in multiple windows to increase the S- wave signal and obtain satisfied
source parameters of S-wave (Yang et al, 2009, in preparation).
Figure 1a. A distribution of strain-drops of 7498 events along the Karadere-Duzce branch of the NAF.
Figure 1b. The fit for strain-drops of twelve M2.1 events with their 201 neighbouring events respectively. Number at the up-right in each panel corresponds to the same number in 1A with location.
Seismicity and Associated Geophysical Properties
-
Earthquake catalog analysis constitutes the basis for seismic hazard evaluation and earthquake forecasting test.
In 1997, I began to work with Prof. Ma Li in Center for Analysis and Prediction, China Seismological Bureau to
analyze Chinese earthquake catalog and test an earthquake prediction model: Accelerating Moment Release model
(Yang and Ma, 1999). As a continuation of this work, I went to New Zealand to work with Prof. David Vere-Jones
and Dr. Russell Robinson in 1998. During this half a year period, I put forward a method to locate the region of future earthquake based on the
critical earthquake concept, and applied this method to study earthquake catalogs in China and New Zealand
(Yang et al, 2001).
I also assisted Prof. Vere-Jones to use statistical methods to evaluate the AMR model (Vere-Jones et al, 2001).
Figure 2. A critical earthquake "cloudy map" for China using data from 1995 to 1999 with M>3.0. This work was finished in 2001 using the method discussed in Yang et al (2001)
Begining at 2005, I worked with Prof. Yehuda Ben-Zion to analyze aftershock sequences in southern California. The purpose is to test the predicted results from a damage rheology model proposed by Ben-Zion and Lyakhovsky (2006), and this model shows that the statistical features of regional seismicity are controlled by the background physical conditions. Using individual aftershock sequences and stacked aftershock sequences in different geophysical regions, we concluded that the aftershock productivity is inversely correlated with heat flow and ambient sediment layer thickness, and such conclusion matches with the prediction from the damage rheology model. This part of work was documented in a paper (Yang and Ben-Zion, 2009). A by-product of this study is that we can infer the seismic coupling of an unknown region from its value in the known region and the ratio of aftershock productivities in both regions.
Figure 3a. A distribution of earthquakes in Southern California (1984 - 2002, M>2.0) and USGS heat flow data. The Imperial Valley and Coso regions (A and E) have relatively high heat flow levels, the largest and the smallest areas and regular-to-moderate sedimentary covers. The Landers and Hector-Mine area (B) has low heat flow, relatively large area and very thin sedimentary cover (order 100–200 m) that is above the seismicity. The San Bernardino region (C), which includes both the San Bernardino valley and mountains, has low average heat flow, moderate area and moderate average sedimentary cover. The Ventura Basin region (D) has low heat flow, moderate area and thick sedimentary cover (order 10 km) that extends nearly to the bottom of the seismogenic zone.
Figure 3b. Stacked aftershock productivities in each of the five areas (left) and their relationship with heat flow and thickness of sediment cover.
Earthquake Location and Network Maintaining
-
From 2002 to the middle of 2004, I worked with Prof. Richard Aster to locate seismic events occurred around New Mexico, and maintained
the Socorro regional seismic network. An Earthworm real-time system was set to process the data flow automatically.
However, a large part of the automatically triggered events are not seismic events, so I manually identified seismic events, picked phases and
located them. I was trying to build a reference event database, and use cross-correlation technique to locate
incoming unknown events. The Socorro Magma Body is a very interesting "hot point" underneath the Socorro
county inside the Colorado Plateau.
Every 2-3 years, there should have a big swarms occurred around it. Working with Prof. Susan Bilek, Prof. Rick Aster and Darren Harte,
I am trying to study the associated source properties of these SMB earthquakes.
Clipped Waveforms Detection and Correction
-
Clipped waveforms exist widely in near field data set due to the amplitude of seismic signal goes beyond instrumental maximum amplitude range.
They are often regarded as unusable data and discarded and very little study had been done to them. I got to know them when I had to remove them
in the work of strain-drop study. Based on the special characters of clipped waveforms, I wrote a simple algorithm to detect one type of clipped
waveforms in the Izmit-Duzce aftershock data set. Lately, I found no all the clipped waveforms are unusable. About 70% of the detected clipped
waveforms have only one sample clipped, with linear interpretation method (e.g. Kriging method), we can correct such clipped waveforms. I also
propose another method of using similar waveforms to correct the clipped waveforms, and make a comparison between the two methods (Yang and Ben-Zion,
2009, in preparation).