Earthquake hypocentre and origin time

Focus and Epicenter Focus (hypocentre) This is the actual location where fault movement begins.  Almost every earthquake has its focus located below the earth's surface.   Epicentre This is the point on the land surface directly above the focus and is the location normally reported in the news or shown on maps. Note that an epicentre needs not be located on a fault line.

Focus and Epicenter

Earthquake Locations

Definitions Depth Classification:   Shallow - 0-70 km depth Intermediate - 70-300 km depth Deep - 300-700 km depth   Origin Time - Time of initial break of the fault. Travel time - The time required for a wave train to travel from its source to a point of observation. Travel Time Curves - Expected travel times of various waves from the hypocenter to stations at various distances. Seismograph Station - A site at which one or more seismographs are set up and routinely monitored. Hypocentral distance – Distance from an hypocenter to any seismic station. Epicentral distance – Distance from an epicenter to any seismic station.  

Earthquake Locations The earthquake location is defined by: the earthquake hypocenter (x0, y0, z0): longitude (x0) [degrees or km] latitude (y0) [degrees or km] depth below the surface (z0) [km] the origin time t0 [hh:mm:ss.ss] (GMT)

Earthquake Location - methods Graphical methods Computational methods Advanced methods

Earthquake Location - methods Graphical methods Single-stations method Method of circles/chords Wadati diagram

Single-stations method BAZ Total horizontal component Apparent angle of incidence

Single-stations method The arrival time of P- and S-wave Eliminate t0 For VP = 6 km/s, and VP/VS = 1.73 D = 8 (tSg – tPg) D = 10 (tSn – tPn) For local and regional earthquakes < 1000 km

Multiple-stations method (circles/chords) S1, S2, S3 – stations d1, d2, d3 – epicentral distances

Multiple-stations method (circles)

The arrival time of P- and S-wave Wadati diagram The arrival time of P- and S-wave Eliminate D

Wadati graph for estimation of the origin time To Assumptions: vp/vs is a constant and the P and S phases are of the same type like Pg and Sg or Pn and Sn

Computer location Manual location methods provide insight into the location problems, however in practice we use computer methods. The calculated arrival time at station i can be written as: where T is the travel time as a function of the location of the station (xi, yi, zi), the hypocenter (x0, y0, z0), and the origin time t0. This equation has 4 unknowns, so in principle 4 arrival-time observations from at least 3 stations are needed in order to determine the hypocenter and origin time. If we have n observations, there will be n equations of the above type and the system is over determined and has to be solved in such a way that the misfit or residual ri at each station is minimized.

Computer location The residual ri is defined as the difference between the observed and calculated travel times which is the same as the difference between the observed and calculated arrival times - observed time - calculated time

Computer location The travel-time function T is a nonlinear function of the model parameters, it is not possible to solve the Equation above with any analytical methods. The non-linearity is evident even in a simple 2-D epicentre determination where the travel time ti from the point (x, y) to a station (xi, yi) can be calculated as where v is the velocity. It is obvious that ti does not scale linearly with either x or y so it is not possible to use any set of linear equations to solve the problem and standard linear methods cannot be used.

Computer location Grid Search Iterative methods

Grid Search Grid search over all possible locations and origin times (x0j,y0j,z0j,t0j) and computation the arrival time at each station The hypocentral location and origin time would then be the point with the best agreement between the observed and calculated times. Measure of best agreement: The most common approach is to use the least squares solution  find the minimum of the sum of the squared residuals e from the n observations: The root mean squared residual RMS, is defined as

Grid Search stations j Local minimum  ambiguity Particularly nasty problem is the existence of outliers, i.e., individual large residuals. A residual of 4 will contribute 16 times more to the misfit e, than a residual of 1. Using the sum of the absolute residuals as a norm for the misfit can partly solve this problem:

Iterative methods Based on linearization of the problem In order to linearize the problem, it is now assumed that the true hypocenter is close enough to the guessed value so that travel-time residuals at the trial hypocenter are a linear function of the correction we have to make in the hypocentral distance.

Iterative methods The calculated arrival times at station i are: with residuals: We now assume that these residuals are due to the error in the trial solution and the corrections needed to make them zero are x,  y,  z, and  t. If the corrections are small, we can calculate the corresponding corrections in travel times by approximating the travel time function by a Taylor series and using only the first term. The residual can now be written:

Iterative methods In matrix form: Residual Matrix of Unknown vector partial correction derivatives vector in location and time This is a set of linear equations with 4 unknowns (corrections to hypocenter and origin time), and there is one equation for each observed phase time. Normally there would be many more equations than unknowns (e.g., 4 stations with 3 phases each would give 12 equations).

Iterative methods Solution – using standard least squares techniques Iteration 1 – Trial solution  new solution (1) + RMS(1) Iteration 2 – New trial solution  new solution (2) + RMS(2) Iteration 3 – New trial solution  new solution (3) + RMS(3) RMS(1) > RMS(2) > RMS(3) … Geiger (1910) method of earthquake location. Problems: 1) The iterative process converges to a local minimum 2) The solution depends on the start location (sometimes)

Iterative methods (example) Real Estimated

Iterative methods (problems) Starting Point (2) Starting Point (1) Starting Point (3) Local minimum  ambiguity

Advanced methods Double-difference (Waldhauser and Ellsworth, 2000) Residuals between observed and theoretical traveltime differences (or double-differences) are minimised for pairs of earthquakes at each station, while linking together all observed event-station pairs. Global differential evolution algorithm (Ruzek and Kvasnicka, 2001) Advanced grid search procedures for highly homogeneous medium (Lomax et al., 2001)

Advanced methods (example) Double-difference (Waldhauser and Ellsworth, 2000) Examples of improving the ABCE locations for earthquake clusters (red dots) from regional networks of seismic stations (triangles) in China by relocating the events with the double-difference location algorithm (courtesy of Paul G. Richards).

Location errors RMS depends on the number of stations and does not give any indication of error (less stations  smaller RMS) Contouring the grid search RMS – indication of uncertainty Error ellipse – contours within which there is a 67% probability (or any other desired probability) of finding the epicentre. Minor axis Wrong! (ISC, PDE) Major axis The only proper way to report error is to give the full specification of the error ellipsoid.

Location errors Depend on: The arrival time reading errors The station configuration The velocity model uncertainties

Location errors Dependence on the arrival time reading errors Real Estimated Inversion of arrival times with a 0.1 s standard reading error. Hypocenter is the correct location. Start is the start location, and the locations are shown after the three following iterations. e is the misfit

Relative location methods Master event technique Joint hypocenter determination

Relative location methods Master event technique (in case of clustering) Relative location to one particularly well-located event (master event) Ideally the same stations and phases should be used for the location of the master event and the other events.

Relative location methods Joint hypocentre location (Douglas, 1967) Joint determination of m hypocentres and origin times and n station corrections by adding the station residuals and writing the equations for all m earthquakes (index j): Since the matrix G of this equation is now much larger than the 4 x 4 matrix for a single event location, efficient inversion schemes must be used. If we use e.g., 20 stations with 2 phases each for 10 events, there will be 20 *10 *2 = 400 equations and 80 unknowns (10 hypocenters and origin times, and 20 station residuals).

Relative location methods Joint hypocentre location (example) Comparison of earthquake locations using the normal procedure at ISC (left) and JHD relocations (right). The events are located in the Kurile subduction zone along the rupture zones of large thrust events in 1963 and 1958. The vertical cross sections shown traverse the thrust zone from left to right. Note that the JHD solutions reduce the scatter and make it possible to define a dipping plane (from Schwartz et al., 1989).

Practical considerations in earthquake location More different phases  better constrain on the solution Examples of significant improvement of hypocenter location for teleseismic events by including secondary phases. Left: hypocenter locations using only P phases; middle: by including S phases; right: by including also depth phases and core reflections with a different sign of ∂T/∂z (modified from Schöffel and Das, J. Geophys. Res., Vol. 104, No. B6, page 13,104, Figure 2; Ó 1999).

Practical considerations in earthquake location Starting location – very important because of the local minimum Hypocentral depth – very difficult to define. We need: A station close by (no more than 2 times the depth) Depth phases (pP, sP, pPmP, etc.) The travel-time derivative with respect to depth changes very slowly as function of depth) unless the station is very close to the epicenter. In other words, the depth can be moved up and down without changing the travel time much. The figure shows a shallow (ray 1) and a deeper event (ray 2). It is clear that the travel-time derivative with respect to depth is nearly zero for ray 1 but not for ray 2. The depth – distance trade off in the determination of focal depth.

Practical considerations in earthquake location Weighting schemes – depending on: Residuals Distance Type of waves (P is better than S) Uncertainty in the phase reading (e,i)

Importance of velocity model 1-D model 2-D model (blocks, lateral inhomogeneity) 3-D model Many studies have shown (e.g., Kissling, 1988) that accuracy of locating hypocenters can be improved by using a well-constrained minimum 1-D velocity model with station corrections and is better than using a regional 1-D model. However, Spallarossa et al. (2001) recently showed that in strongly heterogeneous local areas even a 1-D model with station corrections does not significantly improve the accuracy of the location parameters. High-precision location in such cases can be achieved only by using a 3-D model.

Importance of velocity model 1-D model 3-D model Epicentral distribution of aftershocks of the Cariaco earthquake (Ms=6.8) on July 9, 1997 in NE Venezuela. Top: results from HYPO71 based on a one-dimensional velocity-depth distribution. Bottom: Relocation of the aftershocks on the basis of a 3-D model derived from a tomographic study of the aftershock region (courtesy of M. Baumbach, H. Grosser and A. Rietbrock).

Importance of velocity model 3-D distribution of the P-wave velocity in the focal region of the 1997 Cariaco earthquake as derived from a tomographic study. The horizontal section shows the velocity distribution in the layer between 2 km and 4 km depth. Red and blue dots mark the epicenters of the aftershocks. The red ones were chosen because of their suitability for the tomography. The six vertical cross sections show the depths' distribution of the aftershocks (green dots) together with the deviations of the P-wave velocity from the average reference model. The depth range and the lateral changes of fault dip are obvious (courtesy of M. Baumbach, H. Grosser and A. Rietbrock).

Importance of velocity model Illustration of the systematic mislocation of earthquakes along a fault with strong lateral velocity contrast. vo is the assumed model velocity with v2 > vo > v1.

Simultaneous determination of the velocity model and the locations VELEST (Kissling et al., 1995) - program to invert simultaneously for 1-D velocities (Vp & Vs) and hypocentral parameter. The goal of VELEST is to establish a so called minimum 1-D model with station corrections used for high precision earthquake location SIMULPS12 - 3-D velocity model determination and hypocentral location with local earthquake data. Full inversion, not tomography, this is the program written by Cliff Thurber and modified by others (most recently Donna Eberhart-Phillips).

Bulletin example – NEIC/USGS

Bulletin example – BGR/Germany

Bulletin example – GSC

Example of epicentral map

Example of vertical cross-section

Tonga region hypocenters