def write_gmm_data_file(model_name, mag, dist, result_type,
periods, file_out,
component_type="AVERAGE_HORIZONTAL",):
"""
Create a file of input and output parameters for the sommerville GMM.
params:
model_name: The ground motion model, as a string.
mag: dictionary, key - the mag column name, values, the mag vectors,
as a list
dist: dictionary, key - the distance column name, value,
the distance vectors, as a list.
result_type: MEAN or TOTAL_STDDEV
periods: A list of periods requiring SA values.
The first value has to be 0.0.
Mag, distance and periods will be iterated over to give a single SA for
each combination.
file_out: The file name and location of the produced data file.
"""
assert periods[0] == 0.0
handle = open(file_out, 'wb')
writer = csv.writer(handle, delimiter=',', quoting=csv.QUOTE_NONE)
# write title
title = [mag[0], dist[0], 'result_type', 'component_type'] + periods[1:] + \
['pga']
writer.writerow(title)
# prepare the coefficients
model = Ground_motion_specification(model_name)
coeff = model.calc_coefficient(periods)
coeff = reshape(coeff, (coeff.shape[0], 1, 1, coeff.shape[1]))
sigma_coeff = model.calc_sigma_coefficient(periods)
sigma_coeff = reshape(sigma_coeff, (sigma_coeff.shape[0], 1, 1,
sigma_coeff.shape[1]))
# Iterate
for magi in mag[1]:
for disti in dist[1]:
dist_args = {'mag': array([[[magi]]]),
dist[0]: array([[[disti]]]),
'coefficient': coeff,
'sigma_coefficient': sigma_coeff}
log_mean, log_sigma = model.distribution(**dist_args)
sa_mod = list(log_mean.reshape(-1))
sa_mod = [math.exp(x) for x in sa_mod]
sigma_mod = list(log_sigma.reshape(-1))
if result_type == 'MEAN':
row = [magi, disti, result_type, component_type] + \
sa_mod[1:] + \
[sa_mod[0]]
else:
row = [magi, disti, result_type, component_type] + \
sigma_mod[1:] + \
[sigma_mod[0]]
writer.writerow(row)
handle.close()
def write_gmm_data_file_depth(
model_name, mag, dist, depth, result_type, periods, file_out, component_type="AVERAGE_HORIZONTAL"
):
"""
Create a file of input and output parameters for the sommerville GMM.
params:
model_name: The ground motion model, as a string.
mag: dictionary, key - the mag column name, values, the mag vectors,
as a list
dist: dictionary, key - the distance column name, value,
the distance vectors, as a list.
depth: depth in km.
result_type: MEAN or TOTAL_STDDEV
periods: A list of periods requiring SA values.
The first value has to be 0.0.
Mag, distance and periods will be iterated over to give a single SA for
each combination.
file_out: The file name and location of the produced data file.
"""
assert periods[0] == 0.0
handle = open(file_out, "wb")
writer = csv.writer(handle, delimiter=",", quoting=csv.QUOTE_NONE)
# write title
title = [depth[0], mag[0], dist[0], "result_type", "component_type"] + periods[1:] + ["pga"]
writer.writerow(title)
# prepare the coefficients
model = Ground_motion_specification(model_name)
coeff = model.calc_coefficient(periods)
coeff = reshape(coeff, (coeff.shape[0], 1, 1, coeff.shape[1]))
sigma_coeff = model.calc_sigma_coefficient(periods)
sigma_coeff = reshape(sigma_coeff, (sigma_coeff.shape[0], 1, 1, sigma_coeff.shape[1]))
# Iterate
for depi in depth[1]:
for magi in mag[1]:
for disti in dist[1]:
dist_args = {
"mag": array([[[magi]]]),
dist[0]: array([[[disti]]]),
"depth": array([[[depi]]]),
"coefficient": coeff,
"sigma_coefficient": sigma_coeff,
}
log_mean, log_sigma = model.distribution(**dist_args)
sa_mod = list(log_mean.reshape(-1))
sa_mod = [math.exp(x) for x in sa_mod]
sigma_mod = list(log_sigma.reshape(-1))
if result_type == "MEAN":
row = [depi, magi, disti, result_type, component_type] + sa_mod[1:] + [sa_mod[0]]
else:
row = [depi, magi, disti, result_type, component_type] + sigma_mod[1:] + [sigma_mod[0]]
writer.writerow(row)
handle.close()
class Ground_motion_calculator(object):
"""Ground_motion_calculator instances are used to calculate the ground
motion given a ground motion specifiction.
Attributes:
GM_spec: instance of Ground_motion_specification
coefficient: ground motion coefficient for the given periods
sigma_coefficient: ground motion sigma coefficient for the given periods
periods: the periods
"""
def __init__(self, ground_motion_model_name, periods):
"""
Args:
ground_motion_model_name: A string, naming the ground motion model
periods: The periods that will be used for this simulation. Used
to calculate coefficient and sigma_coefficient.
RESIZING NOTES
Adding lots of extra dimensions.
'distance' had dimension [site]*[events]
'magnitude' had dimension [events]
'coefficient' had dimension [number of coefficients]*[Period]
'sigma_coefficient' had dimension [number of coefficients]*[Period]
Now 'distance' and 'magnitude' have dimension:
[bonus dimension!]*[site]*[events]
once 'coefficient' and 'sigma_coefficient are unpacked (ie
c1,c2,c4,c6,c7,c10=c), they have dimension:
[bonus dimension!]*[site]*[events]
Note that some of these dimensions are degenerate (newaxises),
such as [site] for 'magnitude'.
newaxis is used to broadcast arrays into higher dimensions:
a=array([1,2,3])
b=array([0,1])
a=a[...,newaxis]
print a
>[[1]
> [2]
> [3]]
# if a is added to a 1D array of length n; a will act as:
# [[1, 1, ... (n times],
# [2, 2, ... (n times],
# [3, 3, ... (n times]]
# (this is just broadcasting rules)
a+b
>array([[1, 2],
> [2, 3],
> [3, 4]])
Note that all [bonus dimensions] are degenerate.
They are there because once the distribution is sampled,
I want it to maintain the same number of dimension. I don't
want it to add an extra dimension for the spawnings. If that
happens, then it is hard to address futher samplings (from
soil) or multi-models in a uniform manner.
so ground_motion_from_toro =[gmd_torro]
sampled ground_motion_from_toro = [gmd1,gmd2,gmd3, ... (n samples)]
Uniform behaviour.
"""
self.GM_spec = Ground_motion_specification(ground_motion_model_name)
periods = asarray(periods)
# calc the coefficient and sigma_coefficient for the input periods
coefficient = self.GM_spec.calc_coefficient(periods)
sigma_coefficient = self.GM_spec.calc_sigma_coefficient(periods)
# Adding extra dimensions.
self.coefficient = coefficient[:, newaxis, newaxis, :]
self.sigma_coefficient = sigma_coefficient[:, newaxis, newaxis, :]
def distribution_function(self,
dist_object,
dist_types,
mag_dict,
periods=None,
depth=None,
depth_to_top=None,
fault_type=None,
Vs30=None,
mag_type=None,
Z25=None,
dip=None,
width=None,
event_activity=None):
"""
dist_object must give distance info if dist_object.distance(dist_type)
is called. The distance info must be an array.
Returns:
log_mean - dimensions are
log_sigma
FIXME: Why should we let depth be None?
"""
# dist_type and mag_type are attributes of self.GM_spec
# we shouldn't pass them around.
distances = {}
for dist_type in dist_types:
distances[dist_type] = dist_object.distance(dist_type)
mag = mag_dict[mag_type]
if depth is not None:
depth = asarray(depth)
(mag, depth, depth_to_top, fault_type, dip,
width) = self.resize_mag_depth(mag, depth, depth_to_top, fault_type,
dip, width)
for dist_type in dist_types:
distances[dist_type] = self.resize_dist(distances[dist_type],
mag.size)
# This is calling the distribution functions described in the
# ground_motion_interface module.
# We add the new 'dist_object' parameter to cater to models that
# require more than one distance. Once all existing models use
# the new parameter we can remove the 'distance' parameter.
distribution_args = {
'mag': mag,
'coefficient': self.coefficient,
'sigma_coefficient': self.sigma_coefficient,
'depth': depth,
'depth_to_top': depth_to_top,
'fault_type': fault_type,
'Vs30': Vs30,
'Z25': Z25,
'dip': dip,
'width': width,
'periods': periods
}
for dist_type in dist_types:
distribution_args[dist_type] = distances[dist_type]
(log_mean, log_sigma) = self.GM_spec.distribution(**distribution_args)
# FIXME when will this fail? Maybe let it fail then?
# If it does not fail here it fails in analysis.py"
#, line 427, in main
# assert isfinite(bedrock_SA).all()
# example of log_mean when this failed log_mean [[[ NaN NaN NaN]]]
# An Mw of 0.0 caused it.
# this is a good place to catch this error
assert isfinite(log_mean).all()
assert isfinite(log_sigma).all()
return (log_mean, log_sigma)
def resize_mag_depth(self, mag, depth, depth_to_top, fault_type, dip,
width):
"""
Warning, Toro_1997_midcontinent_distribution assumes
that this occurs. So if resizing is changed,
Toro_1997_midcontinent_distribution needs to be changed as well
"""
if mag.size == 1:
mag = mag.reshape([1])
if depth is not None:
depth = depth.reshape([1]) # Don't know if we have to do this
if depth_to_top is not None:
depth_to_top = depth_to_top.reshape([1])
if fault_type is not None:
fault_type = fault_type.reshape([1])
# resize depth, depth_to_top, etc
if depth is not None:
depth = depth[newaxis, :, newaxis]
# collapsed arrays are a bad idea...
if depth_to_top is not None:
depth_to_top = array(depth_to_top)[newaxis, :, newaxis]
if fault_type is not None:
fault_type = array(fault_type)[newaxis, :, newaxis]
if dip is not None:
dip = array(dip)[newaxis, :, newaxis]
if width is not None:
width = array(width)[newaxis, :, newaxis]
assert len(mag.shape) == 1
mag = mag[newaxis, :, newaxis]
return (mag, depth, depth_to_top, fault_type, dip, width)
def resize_dist(self, dist, mag_size):
"""
Warning, Toro_1997_midcontinent_distribution assumes
that this occurs. So if resizing is changed,
Toro_1997_midcontinent_distribution needs to be changed as well
"""
# [4.5,5.5,6.0] => site * mag * T
assert len(dist.shape) < 3
if not len(dist.shape) == 2:
# if distances is collapsed
if dist.size == 1:
# if distances is size 1
dist = dist.reshape((1, 1))
else:
assert len(dist.shape) == 1
if mag_size > 1:
assert dist.size == mag_size
# therefore distance is 1 site * n magnitudes
dist = dist.reshape((1, mag_size))
else:
# therefore distance is n site * 1 magnitudes
dist = dist.reshape((dist.size, 1))
# collapsed arrays are a bad idea...
dist = dist[:, :, newaxis]
# [[30.0,35.0],[45.0,20.0]]=> [site*mag] * T
return dist
class Ground_motion_calculator(object):
"""Ground_motion_calculator instances are used to calculate the ground
motion given a ground motion specifiction.
Attributes:
GM_spec: instance of Ground_motion_specification
coefficient: ground motion coefficient for the given periods
sigma_coefficient: ground motion sigma coefficient for the given periods
periods: the periods
"""
def __init__(self, ground_motion_model_name, periods):
"""
Args:
ground_motion_model_name: A string, naming the ground motion model
periods: The periods that will be used for this simulation. Used
to calculate coefficient and sigma_coefficient.
RESIZING NOTES
Adding lots of extra dimensions.
'distance' had dimension [site]*[events]
'magnitude' had dimension [events]
'coefficient' had dimension [number of coefficients]*[Period]
'sigma_coefficient' had dimension [number of coefficients]*[Period]
Now 'distance' and 'magnitude' have dimension:
[bonus dimension!]*[site]*[events]
once 'coefficient' and 'sigma_coefficient are unpacked (ie
c1,c2,c4,c6,c7,c10=c), they have dimension:
[bonus dimension!]*[site]*[events]
Note that some of these dimensions are degenerate (newaxises),
such as [site] for 'magnitude'.
newaxis is used to broadcast arrays into higher dimensions:
a=array([1,2,3])
b=array([0,1])
a=a[...,newaxis]
print a
>[[1]
> [2]
> [3]]
# if a is added to a 1D array of length n; a will act as:
# [[1, 1, ... (n times],
# [2, 2, ... (n times],
# [3, 3, ... (n times]]
# (this is just broadcasting rules)
a+b
>array([[1, 2],
> [2, 3],
> [3, 4]])
Note that all [bonus dimensions] are degenerate.
They are there because once the distribution is sampled,
I want it to maintain the same number of dimension. I don't
want it to add an extra dimension for the spawnings. If that
happens, then it is hard to address futher samplings (from
soil) or multi-models in a uniform manner.
so ground_motion_from_toro =[gmd_torro]
sampled ground_motion_from_toro = [gmd1,gmd2,gmd3, ... (n samples)]
Uniform behaviour.
"""
self.GM_spec = Ground_motion_specification(ground_motion_model_name)
periods = asarray(periods)
# calc the coefficient and sigma_coefficient for the input periods
coefficient = self.GM_spec.calc_coefficient(periods)
sigma_coefficient = self.GM_spec.calc_sigma_coefficient(periods)
# Adding extra dimensions.
self.coefficient = coefficient[:, newaxis, newaxis, :]
self.sigma_coefficient = sigma_coefficient[:, newaxis, newaxis, :]
def distribution_function(self, dist_object, dist_types, mag_dict,
periods=None, depth=None, depth_to_top=None,
fault_type=None, Vs30=None, mag_type=None,
Z25=None, dip=None, width=None,
event_activity=None):
"""
dist_object must give distance info if dist_object.distance(dist_type)
is called. The distance info must be an array.
Returns:
log_mean - dimensions are
log_sigma
FIXME: Why should we let depth be None?
"""
# dist_type and mag_type are attributes of self.GM_spec
# we shouldn't pass them around.
distances = {}
for dist_type in dist_types:
distances[dist_type] = dist_object.distance(dist_type)
mag = mag_dict[mag_type]
if depth is not None:
depth = asarray(depth)
(mag, depth, depth_to_top, fault_type,
dip, width) = self.resize_mag_depth(mag, depth, depth_to_top,
fault_type, dip, width)
for dist_type in dist_types:
distances[dist_type] = self.resize_dist(distances[dist_type],
mag.size)
# This is calling the distribution functions described in the
# ground_motion_interface module.
# We add the new 'dist_object' parameter to cater to models that
# require more than one distance. Once all existing models use
# the new parameter we can remove the 'distance' parameter.
distribution_args = {'mag': mag,
'coefficient': self.coefficient,
'sigma_coefficient': self.sigma_coefficient,
'depth': depth,
'depth_to_top': depth_to_top,
'fault_type': fault_type,
'Vs30': Vs30,
'Z25': Z25,
'dip': dip,
'width': width,
'periods': periods}
for dist_type in dist_types:
distribution_args[dist_type] = distances[dist_type]
(log_mean, log_sigma) = self.GM_spec.distribution(**distribution_args)
# FIXME when will this fail? Maybe let it fail then?
# If it does not fail here it fails in analysis.py"
#, line 427, in main
# assert isfinite(bedrock_SA).all()
# example of log_mean when this failed log_mean [[[ NaN NaN NaN]]]
# An Mw of 0.0 caused it.
# this is a good place to catch this error
assert isfinite(log_mean).all()
assert isfinite(log_sigma).all()
return (log_mean, log_sigma)
def resize_mag_depth(self, mag, depth, depth_to_top,
fault_type, dip, width):
"""
Warning, Toro_1997_midcontinent_distribution assumes
that this occurs. So if resizing is changed,
Toro_1997_midcontinent_distribution needs to be changed as well
"""
if mag.size == 1:
mag = mag.reshape([1])
if depth is not None:
depth = depth.reshape([1]) # Don't know if we have to do this
if depth_to_top is not None:
depth_to_top = depth_to_top.reshape([1])
if fault_type is not None:
fault_type = fault_type.reshape([1])
# resize depth, depth_to_top, etc
if depth is not None:
depth = depth[newaxis, :, newaxis]
# collapsed arrays are a bad idea...
if depth_to_top is not None:
depth_to_top = array(depth_to_top)[newaxis, :, newaxis]
if fault_type is not None:
fault_type = array(fault_type)[newaxis, :, newaxis]
if dip is not None:
dip = array(dip)[newaxis, :, newaxis]
if width is not None:
width = array(width)[newaxis, :, newaxis]
assert len(mag.shape) == 1
mag = mag[newaxis, :, newaxis]
return (mag, depth, depth_to_top, fault_type, dip, width)
def resize_dist(self, dist, mag_size):
"""
Warning, Toro_1997_midcontinent_distribution assumes
that this occurs. So if resizing is changed,
Toro_1997_midcontinent_distribution needs to be changed as well
"""
# [4.5,5.5,6.0] => site * mag * T
assert len(dist.shape) < 3
if not len(dist.shape) == 2:
# if distances is collapsed
if dist.size == 1:
# if distances is size 1
dist = dist.reshape((1, 1))
else:
assert len(dist.shape) == 1
if mag_size > 1:
assert dist.size == mag_size
# therefore distance is 1 site * n magnitudes
dist = dist.reshape((1, mag_size))
else:
# therefore distance is n site * 1 magnitudes
dist = dist.reshape((dist.size, 1))
# collapsed arrays are a bad idea...
dist = dist[:, :, newaxis]
# [[30.0,35.0],[45.0,20.0]]=> [site*mag] * T
return dist
