UCVM svm1d

GTL

Adding a new SVM gtl for UCVMC

• svmgtl

depth profiles

Target point: -118.4,34

echo "-118.4 34.0 " | basin_query -m cvms5 -f ../conf/ucvm.conf 

returns the Z1.0 at 580.0

Profile plots,

cvms5 background

cvms5 baseline

cvms5 with svmgtl -z 0,200

cvms5 with svmgtl -z 0,350

cvms5 with svmgtl -Z 1000

commands used :

./plot_depth_profile.py -s 34,-118.4 -b 0  -e 40000 -d vs,vp,density -v 100 -c cvms5 -o cvms5_depth_nogtl_bkg.png
./plot_depth_profile.py -s 34,-118.4 -b 0  -e 800 -d vs -v 10 -c cvms5 -o cvms5_depth_nogtl_base.png
./plot_depth_profile.py -s 34,-118.4 -b 0  -e 800 -d vs -v 10 -c cvms5,svmgtl:svm -z 0,200 -o cvms5_depth_svmgtl_200.png
./plot_depth_profile.py -s 34,-118.4 -b 0  -e 800 -d vs -v 10 -c cvms5,svmgtl:svm -z 0,350 -o cvms5_depth_svmgtl_350.png
./plot_depth_profile.py -s 34,-118.4 -b 0  -e 800 -d vs -v 10 -c cvms5,svmgtl:svm -Z 1000 -o cvms5_depth_svmgtl_Z1.png

Raw Datafiles

Near Garner Valley

Horizontal plots near Garner Valley,

Loc => -b 33.40,-116.859 -u 33.80,-116.30

This plot shows the Z1.0 crossing values ranges from 0 to 80,

cvms5 Z10 map

Horizontal plots at 50m depth with just cvms5,

cvms5 50m depth

cvms5 50m depth

Horizontal plots ad 30m and 50m with cvms5 and svm,

cvms5&svm 30m depth

cvms5&svm 50m depth

Comparing at the same depth of 80m (there are very slight differences),

cvms5 80m depth

cvms5&svm 80m depth

Comparing at the same depth of 100m (wiki said they are identical image),

cvms5 100m

cvms5&svm 100m

Raw Datafiles

Somewhere by the water

Near westside with cvms5, internal GTL disabled,

cvms5 Z10 map

cvms5 Z25 map

Depth cross plot,

sample commands used :

./plot_cross_section.py -b 34,-118.5 -u 34,-117.5 -h 700 -v 10 -d vs -c cvms5 -a d -s 0 -e 800 -o cvms5_cross_base.png

and,

./plot_cross_section.py -b 34,-118.5 -u 34,-117.5 -h 500 -v 10 -d vs -c cvms5,svmgtl:svm -a d -Z 1000 -s 0 -e 800 -o cvms5_cross_svmgtl_Z1.png

cvms5 background

cvms5 baseline

cvms5 with elygtl -z 0,200

cvms5 with elygtl -z 0,350

cvms5 with elygtl -Z 1000

cvms5 with svmgtl -z 0,200

cvms5 with svmgtl -z 0,350

cvms5 with svmgtl -Z 1000

Raw Datafiles

Additional Info

SVM interpolation function extracted by Elnaz from Jian's class_svm.py from https://github.com/jsh9/PySeismoSoil

// zmax would be the Z1.0 if there is one and vs30
// in data->gtl.vs

double calc_z1_from_Vs30(double vs30) {
  double z1 = 140.511*exp(-0.00303*vs30); // [m]
  return z1;
}

double calc_rho (double vs, double z) {
  if (z == 0.0) {
    z = 0.0001;
  }
  double lb = 1.65; //lower bound  [g/cm^3]
  double rho  = 1000.0 * max(lb, 1.0 + 1.0/ (0.614 + 58.7 * (log(z) + 1.095) / vs)); //[kg/m^3]
  return rho;
}


/* SVM interpolation method */
int ucvm_interp_svm(double zmin, double zmax, ucvm_ctype_t cmode, ucvm_point_t *pnt, ucvm_data_t *data) {

  // curve fitting parameters for SVM model
  double p1 = -2.1688E-04;
  double p2 =  0.5182    ;
  double p3 = 69.452     ;

  double r1 = -59.67     ;
  double r2 = -0.2722    ;
  double r3 = 11.132     ;

  double s1 =  4.110     ;
  double s2 = -1.0521E-04;
  double s3 = -10.827    ;
  double s4 = -7.6187E-03;

  double zstar = 2.5 ; // [m]
  double z1    = zmax; // z at which vs = 1000
  double vz1;

  double z     = data->depth; // interpolation depth

  double vs, k, n, vs0; // vs profiling parameters

  double vscap = 1000.0;
  double eta   = 0.9   ;
  double zeta, veta    ;
  
  double vs30   = data->gtl.vs;

  double nu = 0.3; // poisson ratio
  double vp_vs  = sqrt(2.0*(1.0-nu)/(1.0-2.0*nu));

  //~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  switch (cmode) {
    case UCVM_COORD_GEO_DEPTH:
    case UCVM_COORD_GEO_ELEV:
    break;
    default:
    fprintf(stderr, "Unsupported coord type\n");
    return(UCVM_CODE_ERROR);
    break;
  }

  if (z < 0.0) {
    return(UCVM_CODE_NODATA);
  }

  // if no z1 data, compute empirically
  if (z1 == 0.0 || z1 == -1.0) {
    z1 = calc_z1_from_Vs30(vs30);
  }

  // query in crustal properties
  if (z >= z1) {
    data->cmb.vp     = data->crust.vp;
    data->cmb.vs     = data->crust.vs;
    data->cmb.rho    = data->crust.rho;
    data->cmb.source = UCVM_SOURCE_CRUST;
    return(UCVM_CODE_SUCCESS);
  }

  // z is between 0 and z1: query in SVM model
  vs0 = p1*pow(vs30,2.0) + p2*vs30 + p3;

  if (z1<=zstar) {
    vs  = min(vs0,vscap);
  }
  else { // z1 > zstar
    k   = exp(r1*pow(vs30,2.0) + r3);                    //  <<=== r2 ???
    n   = max(1.0, s1*exp(s2*vs30) + s3*exp(s4*vs30));
    vz1 = vs0*pow(1.0+k*(z1-zstar),1.0/n); // vs @ z1

    if (vz1 <= vscap) { // no need to cap the model
      if (z<=zstar) {
        vs = vs0;
      }
      else {
        vs = vs0*pow(1.0+k*(z-zstar),1.0/n);
      }  
    }
    else { // vz1 > vscap -> need to cap the model with linear interpolation from zeta to z1
      veta = eta*vscap;
      zeta = (1.0/k)*(pow(veta/vs0,n)-1.0)+zstar; // depth at which vs = eta*vscap
      if (z <= zeta) {
        if (z<=zstar) {
          vs = vs0;
        }
        else {
          vs = vs0*pow(1.0+k*(z-zstar),1.0/n);
        }
      }
      else { // z>zeta -> linear interpolation
        vs   = veta + ((vcap-veta)/(z1-zeta))*(z-zeta);
      }
    }
  }

  data->cmb.vs     = vs;
  data->cmb.vp     = vp_vs * vs;
  data->cmb.rho    =  calc_rho(vs,z);
  data->cmb.source = data->gtl.source;
  return(UCVM_CODE_SUCCESS);
}

Related Links

• UCVM