velestuser.ektext

这是一个不错的源码

least three different initial velocity models for any model geometry (layer 

thickness): one with extremly low crustal velocities, one with extremly high and 

one with intermediate crustal velocities.



After many VELEST runs with different initial model parameters (layer 

thickness and velocities), hypocentral parameters, and various control parameters 

you will get a feeling of how parts of the RMS misfit function (Fig. 2b) for your 

problem-data set look like. You will also see if the problem is reasonably well 

determined by the data. You may then decide on the best model layering 

based on the results of the previous VELEST runs and based on the depth 

distribution of the earthquakes. Choose a simple model by combining layers 

where velocities are very similar, unless you want to mimic a gradient. 



Note: Topmost layers are mostly subvertically and bottom layers are mostly 

subhorizontally penetrated. Therefore, the resolution in these layers is generally 

lower than in the central layers that contain the hypocenters.





APPROACHING A MINIMUM



If one finds a local minimum RMS region one would hope that the velocity model 

part of the solution is similar to the situation in Fig. 4, where the final velocity 

models (B1 and B2) within the well-resolved depth range are almost identical for 

different initial models (A1 and A2). If the near-surface velocities and the Moho 

depth are well-known one could increase the damping for these layers in the 

model file (*.mod).



Note: Fix the Moho depth by increased damping of the sub-Moho velocities only, 

do not overdamp both layer velocities below and above the Moho.



Once a local minimum in the RMS-function (Fig. 2b) has been found one may 

want to check for low velocity layers (LVL's). For reason of apriori information 

or else one might want to allow LVL's only at certain depth ranges or one might 

want to fix some layer velocities obtained by earlier VELEST runs. The 

additional damping parameter for each layer provided by the input velocity 

model file allows to have variable layer velocity damping. 



Note: Do not fix (overdamp) near-surface layer velocities and also overdamp 

station delays unless a Minimum 1-D model with corresponding station delays has 

been reached.





     



Figure 4.	Final velocity models (B1 and B2) resulting from two identical 

inversion procedures of same data with two different initial velocity 

models (A1 and A2).





In many cases it is advantageous to calculate two runs with fewer iterations each 

than one VELEST run with double the number of iterations since in the first case 

one may select among the output files (velocity model, station delays, earthquake 

locations) data for use as input in the second run. 







TESTING A MINIMUM 1D MODEL



There are numerous tests that may be performed and some might already have 

been performed during the previous VELEST runs. Here I just list a few 

possiblities.



USING SHOTS and QUARY BLASTS



Because these sources normally lay at or near the Earth's surface raypaths for 

shots and blasts travel twice (at either end) through the near-surface structure. 

Thus shots and blasts are much harder to locate by a well-distributed seismic 

array (Kissling 1988). We take advantage of this and of the known source 

parameters by using shot and blast data to estimate absolute location errors 

(Kissling 1988, Kradolfer 1989). When performing this test, however, shot 

data should NOT be treated differently from earthquake data in routine 

location procedure where the Minimum 1D model with the station corrections 

is fixed (INVERTRATIO=ITTMAX+1) or at least heavily damped 

(VTHET=STATHET=999.)

Note: Normally, shot and quarry blast data should NOT be used to derive a 

Minimum 1D model other than providing information for the model layering 

(from refraction type experiments) for two reasons. First, (as noted above) 

the errors resulting from near-surface heterogeneity are larger for shots than 

for earthquakes. Second, the test described above is only valid if the tested 

model is independent of the test data.

Shots recorded along a seismic refraction line are excellent to densely sample 

the wave field in a specific direction and to derive a model of the underlying 

structure. The station distribution of such an experiment, however, rate 

among the worst possible ways to deploy a seismic station array for 

earthquake location. Obviously, data from such different experiments should 

not be mixed and treated in a similar way by a trial-and-error process that 

involves many inversions. Rather, these data should both be used in different 

ways that exploit their strengths, as suggested above.



COMPARING INITIAL and FINAL HYPOCENTERS



When dealing with a large (>200) data set of well-locatable earthquakes 

(gap<<180; nobs>10) one might want to check for a bias in the velocity model 

and station corrections as a result of systematic (though possibly small) shift of 

the hypocenters.

Use the Minimum 1D model with station corrections as initial model 

parameters and apply damping of VTHET=1., STATHET=0.1, XYTHET= 

ZTHET=OTHET=0.01 with ITTMAX=9 and INVERTRATIO=2. To obtain 

initial hypocenters you write a short routine that shifts all hypocenters 

randomly (!) by 5 km or 10 km (avoid "air quakes") from the locations 

obtained with the Minimum 1D model. If the Minimum 1D model with 

stations corrections denotes a very robust minimum, the results of this test 

should be a small or negligable variation in the Minimum 1D model and 

station corrections while the hypocenters should be more or less relocated 

back to their original (correct) positions. Hypocenters that remain in the 

wrong locations need to be checked individually. If the final model and/or 

station corrections obtained in this test deviate significantly from the 

Minimum 1D model, the solution space has not been sampled completely and 

the search for another (or better) Minimum 1D model must continue.



COMPARING MINIMUM 1-D MODEL and STATION DELAYS 

with CRUSTAL STRUCTURE



In the Minimum 1D model, the layer velocities will approximately equal the 

average velocity of the 3D structure within the same depth range that has been 

sampled by the data. Note that it is not the spatial average of all the area under 

study. Rather, the velocities of the 1D model approach the average of the 3D 

structure in each layer weighted by the total ray length.

The station delays are the average values for the azimuthally and radially 

varying time delays at these stations relative to the near-surface velocities of 

the Minimum 1D model. In teleseismic studies, the main data signal to be 

interpreted and inverted is the distribution of travel times delays at seismic 

stations with respect to the azimuth of the incoming wave fronts. With local 

earthquake data the azimuthal and radial dependences are never as uniform for 

all stations as for the teleseismic data if the hypocenters are well distributed 

over the area.







5.  Overview of Input/Output files





5.1  INPUT-FILES



Content:	default name



VELEST-control-file	VELEST.CMN ,

Model-file	INTIAL.MOD ,

Station-file (coords,elev,etc.)	LIST.STA ,

Earthquake data (hypocenters and travel times)	PHASEDATA.CNV

                                              or (in SED-format)	PHASEDATA.SED

                                        or (in archive-format)	DATA.ARCVEL

shot data (optional)	SHOTS.CNV



 (for single_event_mode:)

Region names (Swiss & Flinn-Engdahl)	REGIONSNAMEN.DAT

Region coordinates (Swiss & Flinn-Engdahl)	REGIONSKOORD.DAT

Seismo-file (for Magnitude-calculation; optional)	SEISMO.PARAM





5.2  OUTPUT-FILES



Content:	default name



main print-output 	VELEST.OUT



final hypocenters and travel times	VELOUT.CNV

    (may be used as input for another iteration)

new station list with new sta. corrections	VELOUT.STA

    (may be used as input for another iteration)



 (in single_event_mode:)

'HYPO71'-compatible output of the locations	VELOUT.ARCVEL

    (print-output in single_event_mode)



Optional Output:



file011 - summary cards of final hypocenters (for plotting)	VELOUT.SMP

file013 - raypoints of all the rays	VELOUT.RAY

file021 - partial derivatives of all the rays	VELOUT.DRV

file075 - coordinates and ALE of all located events	VELOUT.ALE

file076 - coordinates and DSPR of all located events	VELOUT.DSPR

file077 - coordinates and residual of reflected ray 	VELOUT.RFL

file078 - coordinates and residual of refracted ray	VELOUT.RFR

file079 - residual and according distance	VELOUT.RES







6.  Overview of content of INPUT files





6.1  Explanation of Control Parameters (file VELEST.CMN)



This input file contains the control-parameters only, model- and station-

parameters are read in from seperate files (their names are specified in 

the control-file). All parameters are read in free format.





OLAT, OLON, ROTATE:

Latitude and Longitude (in degrees) or center of cartesian coordinate system 

and of short distance conversion. ROTATE denotes the angle of clockwise 

rotated y-axis from North.



ICOORDSYSTEM:	Default value: 0

= 0	for short distance conversion to cartesian coordi-

nate system all over the Earth with positive values 

representing Longitude West

= 1	for short distance conversion to cartesian coordi-

nate system all over the Earth with positive values 

representing Longitude East (CAUTION: this option 

has not yet been fully tested!)

=2	for Swiss coordinate system



Latitudes and Longitudes are used throughout the program (and its 

input/output) in decimal-degrees. In all the I/O the decimal degrees are 

characterized with N(orth), S(outh), E(ast) and W(est), respectively. 

Internally, the program handles latitude North positive and South 

negative, longitude West positive and longitude East negative. Cartesian 

coordinates are used in a right-hand system: positive X towards West, 

positive Y towards North and positive Z towards the center of the Earth. 

The exception are Swiss data when Swiss-coordinates are used with the 

center of projection being the city of Berne (x=600.,y=200), positive X-

axis towards East, the positive Y-axis towards North, and Z downwards.



ZSHIFT:	Default value: 0.0

ZSHIFT is used to systematically shift all hypocenters in depth relative 

to the depth given in their summary cards (see Fig. 5). This option has 

been added because some routine location programs (HYPO71 and 

others) do not account for station elevation and, therefore, the 

hypocentral depth is calculated with respect to the average station 

elevation and not relative to sea level. Normal usage of VELEST accounts 

for station elevations with zero elevation in the model refering to mean 

sea level. [Example: If the earthquakes have been located with HYPO71 

and a station network of average station elevation of 900m while you 

now want to run VELEST while using station elevations, you would put 

ZSHIFT=0.9].

    



Figure 5. Purpose of switches zshift and iuselev.



ITRIAL, ZTRIAL:	Default values: 0 ,  0.0

Controling the trial hypocenter in the single_event_mode.

itrial=0	use hypocenter coordinates provided by 

	summary card.

        =1	trial hypocenter equals station coordinates of 

	earliest observation and ztrial for depth.



ISED:	Default value: 0

Controling the input format of the earthquake data.

=0	in converted (*.CNV) archive format

=1	VELEST archive type format

=2	SED format (SED= Swiss Seismological Service)

If program VELEST should read in other than one of the above formats, 

the user may add an additional subroutine, which reads the desired 

format. Any subroutine of this kind should always read the phase data 

of one entire event. The call of the routine may look as follows:

call INPUTxyz(iunit,nobsread,sta,iyr(i),imo(i),iday(i),ihr(i),imin(i),sec,

rmk1,rmk2,cphase,ipwt,amx,prx,xlat,alon,emag(i),depth,e(1,i),i1,

ifixsolution,i3,eventtype)



NEQ:	number of earthquakes



NSHOT:	number of shots or blasts



ISINGLE:	Default value: 0

Switch controling mode of VELEST.

=0	simultaneous mode

=1	single_event_mode



IRESOLCALC:	Default value: 0

Resolution matrix calculation in single_event_mode.



DMAX:	Default value: 200.0

Maximal epicentral distance for use of phase. Observations at stations at 

greater distances are neglected.