def test_interp_no_extrapolate_low_bad_ordering(self):
functx = asarray([2, 1, 0.8])
functy = asarray([20, 10, 8])
x = asarray([1.3,1.5,1.7])
y = interp(x, functy, functx, extrapolate_low=False)
y_ans = x * 10.0
self.assert_(allclose(y, y_ans, rtol=0.05))
x = asarray([0.5,1.5,2.5])
y = interp(x, functy, functx, extrapolate_low=False)
y_ans = x * 10.0
self.assert_(allclose(y, [8.,15.,25.], rtol=0.05))
def test_interp_no_extrapolate_low(self):
functx = asarray([1,2])
functy = asarray([10,20])
x = asarray([1.3,1.5,1.7])
y = interp(x, functy, functx, extrapolate_low=False)
y_ans = x * 10.0
self.assert_(allclose(y, y_ans, rtol=0.05))
x = asarray([0.5,1.5,2.5])
y = interp(x, functy, functx, extrapolate_low=False)
y_ans = x * 10.0
self.assert_(allclose(y, [10.,15.,25.], rtol=0.05))
def test_interp(self):
functx = asarray([1,2])
functy = asarray([10,20])
x = asarray([0.5,1.5,2.5])
y = interp(x, functy, functx)
y_ans = x * 10.0
self.assert_(allclose(y, y_ans, rtol=0.05))
def test_interp_no_extrapolate_high_local_max(self):
functx = asarray([2, 1, 3])
functy = asarray([30, 10, 20])
x = asarray([0.5, 1.5, 2.5, 3.0, 4.0])
y = interp(x, functy, functx, extrapolate_high=False)
#print "y", y
self.assert_(allclose(y, [0., 20., 25., 20., 20.], rtol=0.05))
def distribution(self, ground_motion, site_classes,
Mw, event_periods,
event_activity=None):
"""
Calculate the distribution of surface motion, given log of ground
(bedrock) motion, site classes of events, event magnitudes,
and event periods.
ground_motion: ndarray [site, event, period]
returns: (log_mean, log_sigma). ndarrays with same shape as ground_motion
Implementation:
bin all the pga and magnitudes
loop over the models site classes:
get the amplification this site_class
get the indices of the sites that fall in the current site_class
get the gm, pga, and mag for the sites in this site_class
interpolate model amplification and std, then take the amp factors
... and std from the right bins in the interpolated model
Dimension of final distribution = [ground_motion_samples]*[sites]*...
If you sample it, it will return
[samples * ground_motion_samples]*[sites]*...
parameters:
event_activity: Being phased out of here. Don't use.
Don't clean up though, just incase we need to spawn.
"""
if not Mw.size == ground_motion.shape[1]:
raise ValueError
# bin all the pga and magnitudes
event_pga_bins_indices = self._bin_indices(ground_motion[:, :, 0],
self.pga_bins)
event_mag_bins_indices = self._bin_indices(Mw,
self.moment_magnitude_bins)
# initialise mean and sigma
log_amplification = zeros(ground_motion.shape, dtype=float)
log_sigma = zeros(ground_motion.shape, dtype=float)
for site_class in self.log_amplifications.keys():
# interpolate model amplification and std, then take the amp
# factors and std from the right bins in the interpolated model
log_amp = interp(
event_periods, self.log_amplifications[site_class],
self.periods, axis=2)
log_stds = interp(event_periods, self.log_stds[site_class],
self.periods, axis=2)
num_periods = len(event_periods)
num_events = len(Mw)
num_sites = len(site_classes)
try:
in_zone = [s == site_class for s in site_classes]
in_zone = array(in_zone) * 1 # upcast from boolean
code = """
int m,n;
for (int i=0; i<num_sites; ++i){
if (in_zone(i)==1){
for (int j=0; j<num_events; ++j){
m=event_mag_bins_indices(j);
n=event_pga_bins_indices(i,j);
for (int k=0; k<num_periods; ++k){
log_amplification(i,j,k)=log_amp(m,n,k);
log_sigma(i,j,k)=log_stds(m,n,k);
}
}
}
}
return_val = 0;
"""
try:
weave.inline(code,
['event_pga_bins_indices',
'event_mag_bins_indices',
'log_amp', 'log_stds',
'log_amplification', 'log_sigma',
'num_sites', 'num_events',
'num_periods',
'in_zone'],
type_converters=weave_converters.eqrm,
compiler='gcc')
except IOError:
raise util.WeaveIOError
except:
print "No gcc found. Running Regolith_amplification_model.\
distribution_function in pure python"
for i in range(num_sites):
if site_classes[i] == site_class:
for j in range(num_events):
m = event_mag_bins_indices[j]
n = event_pga_bins_indices[i, j]
for k in range(num_periods):
log_amplification[i, j, k] = log_amp[m, n, k]
log_sigma[i, j, k] = log_stds[m, n, k]
log_mean = log(ground_motion) + log_amplification
return log_mean, log_sigma
def distribution(self,
ground_motion,
site_classes,
Mw,
event_periods,
event_activity=None):
"""
Calculate the distribution of surface motion, given log of ground
(bedrock) motion, site classes of events, event magnitudes,
and event periods.
ground_motion: ndarray [site, event, period]
returns: (log_mean, log_sigma). ndarrays with same shape as ground_motion
Implementation:
bin all the pga and magnitudes
loop over the models site classes:
get the amplification this site_class
get the indices of the sites that fall in the current site_class
get the gm, pga, and mag for the sites in this site_class
interpolate model amplification and std, then take the amp factors
... and std from the right bins in the interpolated model
Dimension of final distribution = [ground_motion_samples]*[sites]*...
If you sample it, it will return
[samples * ground_motion_samples]*[sites]*...
parameters:
event_activity: Being phased out of here. Don't use.
Don't clean up though, just incase we need to spawn.
"""
if not Mw.size == ground_motion.shape[1]:
raise ValueError
# bin all the pga and magnitudes
event_pga_bins_indices = self._bin_indices(ground_motion[:, :, 0],
self.pga_bins)
event_mag_bins_indices = self._bin_indices(Mw,
self.moment_magnitude_bins)
# initialise mean and sigma
log_amplification = zeros(ground_motion.shape, dtype=float)
log_sigma = zeros(ground_motion.shape, dtype=float)
for site_class in self.log_amplifications.keys():
# interpolate model amplification and std, then take the amp
# factors and std from the right bins in the interpolated model
log_amp = interp(event_periods,
self.log_amplifications[site_class],
self.periods,
axis=2)
log_stds = interp(event_periods,
self.log_stds[site_class],
self.periods,
axis=2)
num_periods = len(event_periods)
num_events = len(Mw)
num_sites = len(site_classes)
try:
in_zone = [s == site_class for s in site_classes]
in_zone = array(in_zone) * 1 # upcast from boolean
code = """
int m,n;
for (int i=0; i<num_sites; ++i){
if (in_zone(i)==1){
for (int j=0; j<num_events; ++j){
m=event_mag_bins_indices(j);
n=event_pga_bins_indices(i,j);
for (int k=0; k<num_periods; ++k){
log_amplification(i,j,k)=log_amp(m,n,k);
log_sigma(i,j,k)=log_stds(m,n,k);
}
}
}
}
return_val = 0;
"""
try:
weave.inline(code, [
'event_pga_bins_indices', 'event_mag_bins_indices',
'log_amp', 'log_stds', 'log_amplification',
'log_sigma', 'num_sites', 'num_events', 'num_periods',
'in_zone'
],
type_converters=weave_converters.eqrm,
compiler='gcc')
except IOError:
raise util.WeaveIOError
except:
print "No gcc found. Running Regolith_amplification_model.\
distribution_function in pure python"
for i in range(num_sites):
if site_classes[i] == site_class:
for j in range(num_events):
m = event_mag_bins_indices[j]
n = event_pga_bins_indices[i, j]
for k in range(num_periods):
log_amplification[i, j, k] = log_amp[m, n, k]
log_sigma[i, j, k] = log_stds[m, n, k]
log_mean = log(ground_motion) + log_amplification
return log_mean, log_sigma
