def main():
fault = geoprobe.swfault('/data/nankai/data/swFaults/jdk_oos_splay_large_area_depth.swf')
faultxyz = data.to_world(data.to_xyz(data.fault))
horxyz = data.to_world(data.to_xyz(data.horizons[0]))[::100]
slip = invert_slip(faultxyz, horxyz, alpha=data.alpha)
azimuth = np.degrees(np.arctan2(*slip[::-1]))
azimuth = 90 - azimuth
mag = np.linalg.norm(slip) / 1000
f = FaultModel(fault, horxyz, origxyz=horxyz, ve=3, azimuth=azimuth,
slip=mag, alpha=data.alpha, calc_fault=faultxyz)
f.configure_traits()
def main():
fault = geoprobe.swfault(
'/data/nankai/data/swFaults/jdk_oos_splay_large_area_depth.swf')
faultxyz = data.to_world(data.to_xyz(data.fault))
horxyz = data.to_world(data.to_xyz(data.horizons[0]))[::100]
slip = invert_slip(faultxyz, horxyz, alpha=data.alpha)
azimuth = np.degrees(np.arctan2(*slip[::-1]))
azimuth = 90 - azimuth
mag = np.linalg.norm(slip) / 1000
f = FaultModel(fault,
horxyz,
origxyz=horxyz,
ve=3,
azimuth=azimuth,
slip=mag,
alpha=data.alpha,
calc_fault=faultxyz)
f.configure_traits()
def _invert_inclined_shear(args):
"""Dummy function to work around multiprocessing.pool only accepting
one iterable argument and needing functions to be visible from __main__."""
args = list(args)
numsamples = args[-1]
args = args[:-1]
# "Nudge" solution down-dip (give it a starting direction of downdip)
dipdir = np.radians(data.fault_strike + 90)
dx, dy = np.cos(dipdir), np.sin(dipdir)
direc = np.array([[dx,0],[0,dy]], dtype=np.float)
# Randomly resample fault with replacement
args[0] = subsample(args[0])
# Randomly resample horizon (potentially not with replacement, depending
# on numsamples) (Horizon is over-sampled anwyay. It's okay to subsample)
args[1] = subsample(args[1], numsamples)
kwargs = dict(direc=direc, return_metric=True, overlap_thresh=1)
(dx,dy), var = homogeneous_simple_shear.invert_slip(*args, **kwargs)
return dx, dy, var
def _invert_inclined_shear(args):
"""Dummy function to work around multiprocessing.pool only accepting
one iterable argument and needing functions to be visible from __main__."""
args = list(args)
numsamples = args[-1]
args = args[:-1]
# "Nudge" solution down-dip (give it a starting direction of downdip)
dipdir = np.radians(data.fault_strike + 90)
dx, dy = np.cos(dipdir), np.sin(dipdir)
direc = np.array([[dx, 0], [0, dy]], dtype=np.float)
# Randomly resample fault with replacement
args[0] = subsample(args[0])
# Randomly resample horizon (potentially not with replacement, depending
# on numsamples) (Horizon is over-sampled anwyay. It's okay to subsample)
args[1] = subsample(args[1], numsamples)
kwargs = dict(direc=direc, return_metric=True, overlap_thresh=1)
(dx, dy), var = homogeneous_simple_shear.invert_slip(*args, **kwargs)
return dx, dy, var
models = [[0, 0]]
i = 0
for hor in data.horizons[::-1]:
print hor.name
# Downsample horizon for faster solution...
xyz = data.to_xyz(hor)[::50,:]
xyz = data.to_world(xyz)
# Move this horizon to the last horizon's best fit offset.
# Following the path of all previous horizons...
for model in models:
dx, dy = model
xyz = inclined_shear(fault, xyz, (dx,dy), data.alpha)
model = invert_slip(fault, xyz, data.alpha, guess=(0,0))
models.append(model)
i += 1
models = np.array(models)
movement = models[:,:2].cumsum(axis=0)
x, y = movement.T
plt.plot(x, y, marker='o')
utilities.plot_plate_motion(time=3e5)
plt.axis('equal')
plt.show()
def forced_direction_inversion(fault, xyz, alpha, azimuth, **kwargs):
azimuth = np.radians(90 - azimuth)
dx, dy = np.cos(azimuth), np.sin(azimuth)
direc = [[dx, dy], [dx, dy]]
return invert_slip(fault, xyz, alpha, direc=direc, **kwargs)
fault = data.to_world(data.to_xyz(data.fault))
slips = [(0,0)]
guess = (0,0)
heaves = [(0,0,0)]
planar_variances = []
variances = []
for i, hor in enumerate(data.horizons[::-1]):
print hor.name
xyz = data.to_xyz(hor)[::50]
xyz = data.to_world(xyz)
slip, metric = invert_slip(fault, xyz, alpha=data.alpha, guess=guess,
overlap_thresh=1, return_metric=True)
heave = utilities.calculate_heave(slip, hor)
variances.append(metric)
planar_var = planar_variance(xyz)
planar_variances.append(planar_var)
print metric / planar_var
slips.append(slip)
heaves.append(heave)
x, y = np.array(slips).T
plt.plot(x, y, 'bo-')
x, y, z = np.array(heaves).T
def forced_direction_inversion(fault, xyz, alpha, azimuth, **kwargs):
azimuth = np.radians(90 - azimuth)
dx, dy = np.cos(azimuth), np.sin(azimuth)
direc = [[dx, dy], [dx, dy]]
return invert_slip(fault, xyz, alpha, direc=direc, **kwargs)
def forced_direction_inversion(fault, xyz, alpha, azimuth, **kwargs):
azimuth = np.radians(90 - azimuth)
dx, dy = np.cos(azimuth), np.sin(azimuth)
direc = [[dx, dy], [dx, dy]]
return invert_slip(fault, xyz, alpha, direc=direc, **kwargs)
def visualize(slip, faultxyz, horxyz):
fault = geoprobe.swfault('/data/nankai/data/swFaults/jdk_oos_splay_large_area_depth.swf')
azimuth = np.degrees(np.arctan2(*slip[::-1]))
azimuth = 90 - azimuth
mag = np.linalg.norm(slip) / 1000
f = FaultModel(fault, horxyz, origxyz=horxyz, ve=3, azimuth=azimuth,
slip=mag, alpha=data.alpha, calc_fault=faultxyz)
f.configure_traits()
fault = data.world_xyz(data.fault)
for i, hor in enumerate(data.horizons[::-1]):
print hor.name
xyz = data.to_world(data.to_xyz(hor))[::50]
slip, metric = invert_slip(fault, xyz, alpha=data.alpha, guess=(0,0),
overlap_thresh=1, return_metric=True)
# slip, metric = forced_direction_inversion(fault, xyz, data.alpha, data.fault_strike+90,
# guess=guess, return_metric=True)
visualize(slip, fault, xyz)
def invert(fault, xyz, alpha):
return invert_slip(fault, xyz, alpha, overlap_thresh=1, return_metric=True)
def planar_variance(xyz):
vecs, vals = geoprobe.utilities.principal_axes(*xyz.T, return_eigvals=True)
return vals[-1]
hor = data.horizons[0]
fault = data.to_world(data.to_xyz(data.fault))
xyz = data.to_xyz(hor)[::100]
xyz = data.to_world(xyz)
alpha = data.alpha
func = _Shear(fault, xyz, alpha=alpha, overlap_thresh=0.1)
planar_var = planar_variance(xyz)
slip, thresh = invert_slip(fault, xyz, alpha=alpha, return_metric=True)
print planar_var, thresh
dx, dy = slip
width = 2000
height = 2000
xx, yy = np.mgrid[dx-width:dx+width:100, dy-height:dy+height:100]
roughness = np.zeros(xx.shape, dtype=np.float)
for i, j in np.ndindex(xx.shape):
roughness[i,j] = func((xx[i,j],yy[i,j]))
roughness = np.ma.masked_greater(roughness, 1.1 * thresh)
flat_point = np.unravel_index(roughness.argmin(), roughness.shape)
plt.pcolormesh(xx, yy, roughness)
def invert(fault, xyz, alpha):
return invert_slip(fault, xyz, alpha, overlap_thresh=1, return_metric=True)
models = [[0, 0]]
i = 0
for hor in data.horizons[::-1]:
print hor.name
# Downsample horizon for faster solution...
xyz = data.to_xyz(hor)[::50, :]
xyz = data.to_world(xyz)
# Move this horizon to the last horizon's best fit offset.
# Following the path of all previous horizons...
for model in models:
dx, dy = model
xyz = inclined_shear(fault, xyz, (dx, dy), data.alpha)
model = invert_slip(fault, xyz, data.alpha, guess=(0, 0))
models.append(model)
i += 1
models = np.array(models)
movement = models[:, :2].cumsum(axis=0)
x, y = movement.T
plt.plot(x, y, marker='o')
utilities.plot_plate_motion(time=3e5)
plt.axis('equal')
plt.show()
def planar_variance(xyz):
vecs, vals = geoprobe.utilities.principal_axes(*xyz.T, return_eigvals=True)
return vals[-1]
hor = data.horizons[0]
fault = data.to_world(data.to_xyz(data.fault))
xyz = data.to_xyz(hor)[::100]
xyz = data.to_world(xyz)
alpha = data.alpha
func = _Shear(fault, xyz, alpha=alpha, overlap_thresh=0.1)
planar_var = planar_variance(xyz)
slip, thresh = invert_slip(fault, xyz, alpha=alpha, return_metric=True)
print planar_var, thresh
dx, dy = slip
width = 2000
height = 2000
xx, yy = np.mgrid[dx - width:dx + width:100, dy - height:dy + height:100]
roughness = np.zeros(xx.shape, dtype=np.float)
for i, j in np.ndindex(xx.shape):
roughness[i, j] = func((xx[i, j], yy[i, j]))
roughness = np.ma.masked_greater(roughness, 1.1 * thresh)
flat_point = np.unravel_index(roughness.argmin(), roughness.shape)
plt.pcolormesh(xx, yy, roughness)
plt.plot(dx, dy, 'ro')
def forced_direction_inversion(azimuth, fault, xyz, alpha, **kwargs):
"""Forces the inversion to only consider slip along the given azimuth."""
azimuth = np.radians(90 - azimuth)
dx, dy = np.cos(azimuth), np.sin(azimuth)
direc = [[dx, dy], [dx, dy]]
return invert_slip(fault, xyz, alpha, direc=direc, **kwargs)
