def point_in_domain(self, longitude, latitude):
"""
Checks if a geographic point is placed inside a rotated spherical
section. It simple rotates the point and checks if it is inside the
unrotated domain. It therefore works for all possible projections
and what not.
:param longitude: The longitude of the point.
:param latitude: The latitude of the point.
:return: bool
"""
if self.rotation_angle_in_degree:
# Rotate the point.
r_lat, r_lng = rotations.rotate_lat_lon(
latitude, longitude, self.rotation_axis,
-1.0 * self.rotation_angle_in_degree)
else:
r_lng = longitude
r_lat = latitude
bw = self.boundary_width_in_degree
# Check if in bounds.
if not ((self.min_latitude + bw) <= r_lat <=
(self.max_latitude - bw)) or \
not ((self.min_longitude + bw) <= r_lng <=
(self.max_longitude - bw)):
return False
return True
def point_in_domain(self, longitude, latitude):
"""
Checks if a geographic point is placed inside a rotated spherical
section. It simple rotates the point and checks if it is inside the
unrotated domain. It therefore works for all possible projections
and what not.
:param longitude: The longitude of the point.
:param latitude: The latitude of the point.
:return: bool
"""
if self.rotation_angle_in_degree:
# Rotate the point.
r_lat, r_lng = rotations.rotate_lat_lon(
latitude, longitude, self.rotation_axis,
-1.0 * self.rotation_angle_in_degree)
else:
r_lng = longitude
r_lat = latitude
bw = self.boundary_width_in_degree
# Check if in bounds.
if not ((self.min_latitude + bw) <= r_lat <=
(self.max_latitude - bw)) or \
not ((self.min_longitude + bw) <= r_lng <=
(self.max_longitude - bw)):
return False
return True
def point_in_domain(latitude, longitude, domain,
rotation_axis=[0.0, 0.0, 1.0],
rotation_angle_in_degree=0.0):
"""
Simple function checking if a geographic point is placed inside a
rotated spherical section. It simple rotates the point and checks if it
is inside the non-rotated domain.
Domain is a dictionary containing at least the following keys:
* "minimum_latitude"
* "maximum_latitude"
* "minimum_longitude"
* "maximum_longitude"
* "boundary_width_in_degree"
Returns True or False.
"""
from lasif import rotations
min_latitude = domain["minimum_latitude"] + \
domain["boundary_width_in_degree"]
max_latitude = domain["maximum_latitude"] - \
domain["boundary_width_in_degree"]
min_longitude = domain["minimum_longitude"] + \
domain["boundary_width_in_degree"]
max_longitude = domain["maximum_longitude"] - \
domain["boundary_width_in_degree"]
# Rotate the station and check if it is still in bounds.
r_lat, r_lng = rotations.rotate_lat_lon(
latitude, longitude, rotation_axis, -1.0 * rotation_angle_in_degree)
# Check if in bounds. If not continue.
if not (min_latitude <= r_lat <= max_latitude) or \
not (min_longitude <= r_lng <= max_longitude):
return False
return True
def plot_snapshots(self, component, vmin, vmax, outdir, fprx='wavefield',iter0=200, iterf=17000, \
diter=200, stations=False, res="i", projection='lambert', dpi=300, zoomin=2, geopolygons=None, evlo=None, evla=None ):
"""
Plot snapshots of field component at given depth ranging between "valmin" and "valmax"
================================================================================================
Input parameters:
component - component for plotting
The currently available "components" are:
Material parameters: A, B, C, mu, lambda, rhoinv, vp, vsh, vsv, rho
Velocity field snapshots: vx, vy, vz
Sensitivity kernels: Q_mu, Q_kappa, alpha_mu, alpha_kappa
vmin, vmax - minimum/maximum value for plotting
outdir - output directory
fprx - output file name prefix
iter0, iterf - inital/final iterations for plotting
diter - iteration interval
stations - plot stations or not
res - resolution of the coastline (c, l, i, h, f)
projection - projection type (global, regional_ortho, regional_merc)
dpi - dots per inch (figure resolution parameter)
zoomin - zoom in factor for proj = regional_ortho
geopolygons - geological polygons( basins etc. ) for plot
evlo, evla - event location for plotting
=================================================================================================
"""
if not os.path.isdir(outdir):
os.makedirs(outdir)
iterArr=np.arange(iter0 ,iterf+diter, diter, dtype=int)
use_default_iter=False
for iteration in iterArr:
if not str(iteration) in self[component].keys():
warnings.warn('Velocity Snapshot:'+str(iteration)+' does not exist!', UserWarning, stacklevel=1)
# raise KeyError('Velocity Snapshot:'+str(iteration)+' does not exist!')
use_in_iter=True
if use_default_iter: iterArr = self[component].keys()
self.minlat = self.attrs['lat_min']; self.maxlat = self.attrs['lat_max']
self.minlon = self.attrs['lon_min']; self.maxlon = self.attrs['lon_max']
self.n = self.attrs['rotation_axis']; self.rotangle = self.attrs['rotation_angle']
lat_centre = (self.maxlat+self.minlat)/2.0; lon_centre = (self.maxlon+self.minlon)/2.0
self.lat_centre, self.lon_centre = rotations.rotate_lat_lon(lat_centre, lon_centre, self.n, -self.rotangle)
fig=plt.figure(num=None, figsize=(10, 10), dpi=dpi, facecolor='w', edgecolor='k')
# - Set up the map. ------------------------------------------------------------------------
m=self._get_basemap(projection=projection)
try: geopolygons.PlotPolygon(inbasemap=m)
except: pass
try:
evx, evy=m(evlo, evla)
m.plot(evx, evy, 'yo', markersize=2)
except: pass
for iteration in iterArr:
self._plot_snapshot(inmap=m, component=component, vmin=vmin, vmax=vmax, iteration=iteration, stations=stations)
outfname=outdir+'/'+fprx+'_%06d.png' %(int(iteration))
print outfname, outdir
fig.savefig(outfname, format='png', dpi=dpi)
return
def get_waveforms_synthetic(self, event_name, station_id, long_iteration_name):
"""
Gets the synthetic waveforms for the given event and station as a
:class:`~obspy.core.stream.Stream` object.
:param event_name: The name of the event.
:param station_id: The id of the station in the form NET.STA.
:param long_iteration_name: The long form of an iteration name.
"""
from lasif import rotations
st = self._get_waveforms(event_name, station_id, data_type="synthetic", tag_or_iteration=long_iteration_name)
network, station = station_id.split(".")
iteration = self.comm.iterations.get(long_iteration_name)
# This maps the synthetic channels to ZNE.
synthetic_coordinates_mapping = {"X": "N", "Y": "E", "Z": "Z", "E": "E", "N": "N"}
for tr in st:
tr.stats.network = network
tr.stats.station = station
if tr.stats.channel in ["X"]:
tr.data *= -1.0
tr.stats.starttime = self.comm.events.get(event_name)["origin_time"]
tr.stats.channel = synthetic_coordinates_mapping[tr.stats.channel]
if not "specfem" in iteration.solver_settings["solver"].lower():
# Also need to be rotated.
domain = self.comm.project.domain
# Coordinates are required for the rotation.
coordinates = self.comm.query.get_coordinates_for_station(event_name, station_id)
# First rotate the station back to see, where it was
# recorded.
lat, lng = rotations.rotate_lat_lon(
coordinates["latitude"], coordinates["longitude"], domain["rotation_axis"], -domain["rotation_angle"]
)
# Rotate the synthetics.
n, e, z = rotations.rotate_data(
st.select(channel="N")[0].data,
st.select(channel="E")[0].data,
st.select(channel="Z")[0].data,
lat,
lng,
domain["rotation_axis"],
domain["rotation_angle"],
)
st.select(channel="N")[0].data = n
st.select(channel="E")[0].data = e
st.select(channel="Z")[0].data = z
return st
def center(self):
"""
Get the center of the domain.
"""
c = self.unrotated_center
Point = collections.namedtuple("CenterPoint", ["longitude",
"latitude"])
r_lat, r_lng = rotations.rotate_lat_lon(
c.latitude, c.longitude, self.rotation_axis,
self.rotation_angle_in_degree)
return Point(longitude=r_lng, latitude=r_lat)
def center(self):
"""
Get the center of the domain.
"""
c = self.unrotated_center
Point = collections.namedtuple("CenterPoint",
["longitude", "latitude"])
r_lat, r_lng = rotations.rotate_lat_lon(c.latitude, c.longitude,
self.rotation_axis,
self.rotation_angle_in_degree)
return Point(longitude=r_lng, latitude=r_lat)
def get_depth_profile(self, component, latitude, longitude):
"""
Returns a depth profile of the model at the requested at the
GLL points closest do latitude and longitude.
:param component: The component of the model.
:param latitude: The latitude.
:param longitude: The longitude.
"""
# Need to rotate latitude and longitude.
if hasattr(self.domain, "rotation_axis") and \
self.domain.rotation_axis and \
self.domain.rotation_angle_in_degree:
latitude, longitude = rotations.rotate_lat_lon(
latitude, longitude, self.domain.rotation_axis,
-1.0 * self.domain.rotation_angle_in_degree)
x_index = self.get_closest_gll_index("latitude", latitude)
y_index = self.get_closest_gll_index("longitude", longitude)
data = self.parsed_components[component]
depths = self.collocation_points_depth
values = data[x_index, y_index, :]
lat = self.collocation_points_lats[::-1][x_index]
lng = self.collocation_points_lngs[y_index]
# Rotate back.
if hasattr(self.domain, "rotation_axis") and \
self.domain.rotation_axis and \
self.domain.rotation_angle_in_degree:
lat, lng = rotations.rotate_lat_lon(
lat, lng, self.domain.rotation_axis,
self.domain.rotation_angle_in_degree)
return {
"depths": depths,
"values": values,
"latitude": lat,
"longitude": lng
}
def get_depth_profile(self, component, latitude, longitude):
"""
Returns a depth profile of the model at the requested at the
GLL points closest do latitude and longitude.
:param component: The component of the model.
:param latitude: The latitude.
:param longitude: The longitude.
"""
# Need to rotate latitude and longitude.
if hasattr(self.domain, "rotation_axis") and \
self.domain.rotation_axis and \
self.domain.rotation_angle_in_degree:
latitude, longitude = rotations.rotate_lat_lon(
latitude, longitude, self.domain.rotation_axis,
-1.0 * self.domain.rotation_angle_in_degree)
x_index = self.get_closest_gll_index("latitude", latitude)
y_index = self.get_closest_gll_index("longitude", longitude)
data = self.parsed_components[component]
depths = self.collocation_points_depth
values = data[x_index, y_index, :]
lat = self.collocation_points_lats[::-1][x_index]
lng = self.collocation_points_lngs[y_index]
# Rotate back.
if hasattr(self.domain, "rotation_axis") and \
self.domain.rotation_axis and \
self.domain.rotation_angle_in_degree:
lat, lng = rotations.rotate_lat_lon(
lat, lng, self.domain.rotation_axis,
self.domain.rotation_angle_in_degree)
return {
"depths": depths,
"values": values,
"latitude": lat,
"longitude": lng}
def test_RotateLatLon():
"""
Test the lat/lon rotation on a sphere.
"""
# Rotate north pole to equator.
lat_new, lon_new = rotations.rotate_lat_lon(90.0, 0.0, [0, 1, 0], 90)
np.testing.assert_almost_equal(lat_new, 0.0)
np.testing.assert_almost_equal(lon_new, 0.0)
# Rotate north pole to the south pole.
lat_new, lon_new = rotations.rotate_lat_lon(90.0, 0.0, [0, 1, 0], 180)
np.testing.assert_almost_equal(lat_new, -90.0)
np.testing.assert_almost_equal(lon_new, 0.0)
# Rotate north pole to equator, the other way round.
lat_new, lon_new = rotations.rotate_lat_lon(90.0, 0.0, [0, 1, 0], -90)
np.testing.assert_almost_equal(lat_new, 0.0)
np.testing.assert_almost_equal(lon_new, 180.0)
# Rotate (0/0) to the east
lat_new, lon_new = rotations.rotate_lat_lon(0.0, 0.0, [0, 0, 1], 90)
np.testing.assert_almost_equal(lat_new, 0.0)
np.testing.assert_almost_equal(lon_new, 90.0)
# Rotate (0/0) to the west
lat_new, lon_new = rotations.rotate_lat_lon(0.0, 0.0, [0, 0, 1], -90)
np.testing.assert_almost_equal(lat_new, 0.0)
np.testing.assert_almost_equal(lon_new, -90.0)
# Rotate the west to the South Pole. The longitude can not be tested
# reliably because is varies infinitly fast directly at a pole.
lat_new, lon_new = rotations.rotate_lat_lon(0.0, -90.0, [1, 0, 0], 90)
np.testing.assert_almost_equal(lat_new, -90.0)
def test_RotateLatLon(self):
"""
Test the lat/lon rotation on a sphere.
"""
# Rotate north pole to equator.
lat_new, lon_new = rotations.rotate_lat_lon(90.0, 0.0, [0, 1, 0], 90)
self.assertAlmostEqual(lat_new, 0.0)
self.assertAlmostEqual(lon_new, 0.0)
# Rotate north pole to the south pole.
lat_new, lon_new = rotations.rotate_lat_lon(90.0, 0.0, [0, 1, 0], 180)
self.assertAlmostEqual(lat_new, -90.0)
self.assertAlmostEqual(lon_new, 0.0)
# Rotate north pole to equator, the other way round.
lat_new, lon_new = rotations.rotate_lat_lon(90.0, 0.0, [0, 1, 0], -90)
self.assertAlmostEqual(lat_new, 0.0)
self.assertAlmostEqual(lon_new, 180.0)
# Rotate (0/0) to the east
lat_new, lon_new = rotations.rotate_lat_lon(0.0, 0.0, [0, 0, 1], 90)
self.assertAlmostEqual(lat_new, 0.0)
self.assertAlmostEqual(lon_new, 90.0)
# Rotate (0/0) to the west
lat_new, lon_new = rotations.rotate_lat_lon(0.0, 0.0, [0, 0, 1], -90)
self.assertAlmostEqual(lat_new, 0.0)
self.assertAlmostEqual(lon_new, -90.0)
# Rotate the west to the South Pole. The longitude can not be tested
# reliably because is varies infinitly fast directly at a pole.
lat_new, lon_new = rotations.rotate_lat_lon(0.0, -90.0, [1, 0, 0], 90)
self.assertAlmostEqual(lat_new, -90.0)
def center(self):
"""
Get the center of the domain.
"""
domain = self.comm.project.domain
c = domain.unrotated_center
Point = collections.namedtuple("CenterPoint",
["longitude", "latitude"])
from lasif import rotations
r_lat, r_lng = rotations.rotate_lat_lon(
c.latitude, c.longitude, domain.rotation_axis,
domain.rotation_angle_in_degree)
return Point(longitude=r_lng, latitude=r_lat)
def _get_maximum_bounds(self, min_lat, max_lat, min_lng, max_lng,
rotation_axis, rotation_angle_in_degree):
"""
Small helper function to get the domain bounds of a rotated spherical
section.
:param min_lat: Minimum Latitude of the unrotated section.
:param max_lat: Maximum Latitude of the unrotated section.
:param min_lng: Minimum Longitude of the unrotated section.
:param max_lng: Maximum Longitude of the unrotated section.
:param rotation_axis: Rotation axis as a list in the form of [x, y, z]
:param rotation_angle_in_degree: Rotation angle in degree.
"""
number_of_points_per_side = 50
north_border = np.empty((number_of_points_per_side, 2))
south_border = np.empty((number_of_points_per_side, 2))
east_border = np.empty((number_of_points_per_side, 2))
west_border = np.empty((number_of_points_per_side, 2))
north_border[:, 0] = np.linspace(min_lng, max_lng,
number_of_points_per_side)
north_border[:, 1] = min_lat
south_border[:, 0] = np.linspace(max_lng, min_lng,
number_of_points_per_side)
south_border[:, 1] = max_lat
east_border[:, 0] = max_lng
east_border[:, 1] = np.linspace(min_lat, max_lat,
number_of_points_per_side)
west_border[:, 0] = min_lng
west_border[:, 1] = np.linspace(max_lat, min_lat,
number_of_points_per_side)
# Rotate everything.
for border in [north_border, south_border, east_border, west_border]:
for _i in xrange(number_of_points_per_side):
border[_i, 1], border[_i, 0] = rotations.rotate_lat_lon(
border[_i, 1], border[_i, 0], rotation_axis,
rotation_angle_in_degree)
border = np.concatenate([north_border, south_border, east_border,
west_border])
min_lng, max_lng = border[:, 0].min(), border[:, 0].max()
min_lat, max_lat = border[:, 1].min(), border[:, 1].max()
return min_lat, max_lat, min_lng, max_lng
def _get_maximum_bounds(self, min_lat, max_lat, min_lng, max_lng,
rotation_axis, rotation_angle_in_degree):
"""
Small helper function to get the domain bounds of a rotated spherical
section.
:param min_lat: Minimum Latitude of the unrotated section.
:param max_lat: Maximum Latitude of the unrotated section.
:param min_lng: Minimum Longitude of the unrotated section.
:param max_lng: Maximum Longitude of the unrotated section.
:param rotation_axis: Rotation axis as a list in the form of [x, y, z]
:param rotation_angle_in_degree: Rotation angle in degree.
"""
number_of_points_per_side = 50
north_border = np.empty((number_of_points_per_side, 2))
south_border = np.empty((number_of_points_per_side, 2))
east_border = np.empty((number_of_points_per_side, 2))
west_border = np.empty((number_of_points_per_side, 2))
north_border[:, 0] = np.linspace(min_lng, max_lng,
number_of_points_per_side)
north_border[:, 1] = min_lat
south_border[:, 0] = np.linspace(max_lng, min_lng,
number_of_points_per_side)
south_border[:, 1] = max_lat
east_border[:, 0] = max_lng
east_border[:, 1] = np.linspace(min_lat, max_lat,
number_of_points_per_side)
west_border[:, 0] = min_lng
west_border[:, 1] = np.linspace(max_lat, min_lat,
number_of_points_per_side)
# Rotate everything.
for border in [north_border, south_border, east_border, west_border]:
for _i in range(number_of_points_per_side):
border[_i, 1], border[_i, 0] = rotations.rotate_lat_lon(
border[_i, 1], border[_i, 0], rotation_axis,
rotation_angle_in_degree)
border = np.concatenate(
[north_border, south_border, east_border, west_border])
min_lng, max_lng = border[:, 0].min(), border[:, 0].max()
min_lat, max_lat = border[:, 1].min(), border[:, 1].max()
return min_lat, max_lat, min_lng, max_lng
def filter_location_fct(latitude, longitude):
"""
Simple function checking if a geographic point is placed inside a
rotated spherical section. It simple rotates the point and checks if it
is inside the unrotated domain. The domain specification are passed in
as a closure.
Returns True or False.
"""
# Rotate the station and check if it is still in bounds.
r_lat, r_lng = rotations.rotate_lat_lon(latitude, longitude,
rotation_axis, -1.0 * rotation_angle_in_degree)
# Check if in bounds. If not continue.
if not (min_latitude <= r_lat <= max_latitude) or \
not (min_longitude <= r_lng <= max_longitude):
return False
return True
def _plot_snapshot(self, inmap, component, vmin, vmax, iteration, stations):
"""Plot snapshot, private function used by make_animation
"""
print 'Plotting Snapshot for:',iteration,' steps!'
subgroup=self[component+'/'+str(iteration)]
for key in subgroup.keys():
subdset = subgroup[key]
field = subdset[...]
theta = subdset.attrs['theta']
phi = subdset.attrs['phi']
lats = 90.0 - theta * 180.0 / np.pi
lons = phi * 180.0 / np.pi
lon, lat = np.meshgrid(lons, lats)
if self.rotangle != 0.0:
lat_rot = np.zeros(np.shape(lon),dtype=float)
lon_rot = np.zeros(np.shape(lat),dtype=float)
for idlon in np.arange(len(lons)):
for idlat in np.arange(len(lats)):
lat_rot[idlat,idlon], lon_rot[idlat,idlon] = rotations.rotate_lat_lon(lat[idlat,idlon], lon[idlat,idlon], self.n, -self.rotangle)
lat_rot[idlat,idlon] = 90.0-lat_rot[idlat,idlon]
lon = lon_rot
lat = lat_rot
# - colourmap. ---------------------------------------------------------
cmap = colors.get_colormap('tomo_80_perc_linear_lightness')
x, y = inmap(lon, lat)
im = inmap.pcolormesh(x, y, field, shading='gouraud', cmap=cmap, vmin=vmin, vmax=vmax)
# - Add colobar and title. ------------------------------------------------------------------
cb = inmap.colorbar(im, "right", size="3%", pad='2%')
if component in UNIT_DICT:
cb.set_label(UNIT_DICT[component], fontsize="x-large", rotation=0)
# - Plot stations if available. ------------------------------------------------------------
# if (self.stations == True) & (stations==True):
# x,y = mymap(self.stlons,self.stlats)
# for n in range(self.n_stations):
# plt.text(x[n],y[n],self.stnames[n][:4])
# plt.plot(x[n],y[n],'ro')
return
def point_in_domain(latitude,
longitude,
domain,
rotation_axis=[0.0, 0.0, 1.0],
rotation_angle_in_degree=0.0):
"""
Simple function checking if a geographic point is placed inside a
rotated spherical section. It simple rotates the point and checks if it
is inside the non-rotated domain.
Domain is a dictionary containing at least the following keys:
* "minimum_latitude"
* "maximum_latitude"
* "minimum_longitude"
* "maximum_longitude"
* "boundary_width_in_degree"
Returns True or False.
"""
from lasif import rotations
min_latitude = domain["minimum_latitude"] + \
domain["boundary_width_in_degree"]
max_latitude = domain["maximum_latitude"] - \
domain["boundary_width_in_degree"]
min_longitude = domain["minimum_longitude"] + \
domain["boundary_width_in_degree"]
max_longitude = domain["maximum_longitude"] - \
domain["boundary_width_in_degree"]
# Rotate the station and check if it is still in bounds.
r_lat, r_lng = rotations.rotate_lat_lon(latitude, longitude, rotation_axis,
-1.0 * rotation_angle_in_degree)
# Check if in bounds. If not continue.
if not (min_latitude <= r_lat <= max_latitude) or \
not (min_longitude <= r_lng <= max_longitude):
return False
return True
def lasif_generate_dummy_data(args):
"""
Usage: lasif generate_dummy_data
Generates some random example event and waveforms. Useful for debugging,
testing, and following the tutorial.
"""
import inspect
from lasif import rotations
from lasif.adjoint_sources.utils import get_dispersed_wavetrain
import numpy as np
import obspy
if len(args):
msg = "No arguments allowed."
raise LASIFCommandLineException(msg)
proj = _find_project_root(".")
# Use a seed to make it somewhat predictable.
random.seed(34235234)
# Create 5 events.
d = proj.domain["bounds"]
b = d["boundary_width_in_degree"] * 1.5
event_count = 8
for _i in xrange(8):
lat = random.uniform(d["minimum_latitude"] + b,
d["maximum_latitude"] - b)
lon = random.uniform(d["minimum_longitude"] + b,
d["maximum_longitude"] - b)
depth_in_m = random.uniform(d["minimum_depth_in_km"],
d["maximum_depth_in_km"]) * 1000.0
# Rotate the coordinates.
lat, lon = rotations.rotate_lat_lon(lat, lon,
proj.domain["rotation_axis"], proj.domain["rotation_angle"])
time = obspy.UTCDateTime(random.uniform(
obspy.UTCDateTime(2008, 1, 1).timestamp,
obspy.UTCDateTime(2013, 1, 1).timestamp))
# The moment tensor. XXX: Make sensible values!
values = [-3.3e+18, 1.43e+18, 1.87e+18, -1.43e+18, -2.69e+17,
-1.77e+18]
random.shuffle(values)
mrr = values[0]
mtt = values[1]
mpp = values[2]
mrt = values[3]
mrp = values[4]
mtp = values[5]
mag = random.uniform(5, 7)
scalar_moment = 3.661e+25
event_name = os.path.join(proj.paths["events"],
"dummy_event_%i.xml" % (_i + 1))
cat = obspy.core.event.Catalog(events=[
obspy.core.event.Event(
event_type="earthquake",
origins=[obspy.core.event.Origin(
latitude=lat, longitude=lon, depth=depth_in_m, time=time)],
magnitudes=[obspy.core.event.Magnitude(
mag=mag, magnitude_type="Mw")],
focal_mechanisms=[obspy.core.event.FocalMechanism(
moment_tensor=obspy.core.event.MomentTensor(
scalar_moment=scalar_moment,
tensor=obspy.core.event.Tensor(m_rr=mrr, m_tt=mtt,
m_pp=mpp, m_rt=mrt, m_rp=mrp, m_tp=mtp)))])])
cat.write(event_name, format="quakeml", validate=False)
print "Generated %i random events." % event_count
# Update the folder structure.
proj.update_folder_structure()
names_taken = []
def _get_random_name(length):
while True:
ret = ""
for i in xrange(length):
ret += chr(int(random.uniform(ord("A"), ord("Z"))))
if ret in names_taken:
continue
names_taken.append(ret)
break
return ret
# Now generate 30 station coordinates. Use a land-sea mask included in
# basemap and loop until thirty stations on land are found.
from mpl_toolkits.basemap import _readlsmask
from obspy.core.util.geodetics import gps2DistAzimuth
ls_lon, ls_lat, ls_mask = _readlsmask()
stations = []
# Do not use an infinite loop. One could choose a region with no land.
for i in xrange(10000):
if len(stations) >= 30:
break
lat = random.uniform(d["minimum_latitude"] + b,
d["maximum_latitude"] - b)
lon = random.uniform(d["minimum_longitude"] + b,
d["maximum_longitude"] - b)
# Rotate the coordinates.
lat, lon = rotations.rotate_lat_lon(lat, lon,
proj.domain["rotation_axis"], proj.domain["rotation_angle"])
if not ls_mask[np.abs(ls_lat - lat).argmin()][
np.abs(ls_lon - lon).argmin()]:
continue
stations.append({"latitude": lat, "longitude": lon,
"network": "XX", "station": _get_random_name(3)})
if not len(stations):
msg = "Could not create stations. Pure ocean region?"
raise ValueError(msg)
# Create a RESP file for every channel.
resp_file_temp = os.path.join(os.path.dirname(os.path.abspath(
inspect.getfile(inspect.currentframe()))), os.path.pardir, "tools",
"RESP.template_file")
with open(resp_file_temp, "rt") as open_file:
resp_file_template = open_file.read()
for station in stations:
for component in ["E", "N", "Z"]:
filename = os.path.join(proj.paths["resp"], "RESP.%s.%s.%s.BE%s" %
(station["network"], station["station"], "", component))
with open(filename, "wt") as open_file:
open_file.write(resp_file_template.format(
station=station["station"], network=station["network"],
channel="BH%s" % component))
print "Generated %i RESP files." % (30 * 3)
def _empty_sac_trace():
"""
Helper function to create and empty SAC header.
"""
sac_dict = {}
# floats = -12345.8
floats = ["a", "mag", "az", "baz", "cmpaz", "cmpinc", "b", "depmax",
"depmen", "depmin", "dist", "e", "evdp", "evla", "evlo", "f",
"gcarc", "o", "odelta", "stdp", "stel", "stla", "stlo", "t0", "t1",
"t2", "t3", "t4", "t5", "t6", "t7", "t8", "t9", "unused10",
"unused11", "unused12", "unused6", "unused7", "unused8", "unused9",
"user0", "user1", "user2", "user3", "user4", "user5", "user6",
"user7", "user8", "user9", "xmaximum", "xminimum", "ymaximum",
"yminimum"]
sac_dict.update({key: -12345.0 for key in floats})
# Integers: -12345
integers = ["idep", "ievreg", "ievtype", "iftype", "iinst", "imagsrc",
"imagtyp", "iqual", "istreg", "isynth", "iztype", "lcalda",
"lovrok", "nevid", "norid", "nwfid"]
sac_dict.update({key: -12345 for key in integers})
# Strings: "-12345 "
strings = ["ka", "kdatrd", "kevnm", "kf", "kinst", "ko", "kt0", "kt1",
"kt2", "kt3", "kt4", "kt5", "kt6", "kt7", "kt8", "kt9",
"kuser0", "kuser1", "kuser2"]
sac_dict.update({key: "-12345 " for key in strings})
# Header version
sac_dict["nvhdr"] = 6
# Data is evenly spaced
sac_dict["leven"] = 1
# And a positive polarity.
sac_dict["lpspol"] = 1
tr = obspy.Trace()
tr.stats.sac = obspy.core.AttribDict(sac_dict)
return tr
events = proj.get_all_events()
# Now loop over all events and create SAC file for them.
for _i, event in enumerate(events):
lat, lng = event.origins[0].latitude, event.origins[0].longitude
# Get the distance to each events.
for station in stations:
# Add some perturbations.
distance_in_km = gps2DistAzimuth(lat, lng, station["latitude"],
station["longitude"])[0] / 1000.0
a = random.uniform(3.9, 4.1)
b = random.uniform(0.9, 1.1)
c = random.uniform(0.9, 1.1)
body_wave_factor = random.uniform(0.095, 0.015)
body_wave_freq_scale = random.uniform(0.45, 0.55)
distance_in_km = random.uniform(0.99 * distance_in_km, 1.01 *
distance_in_km)
_, u = get_dispersed_wavetrain(dw=0.001,
distance=distance_in_km, t_min=0, t_max=900, a=a, b=b, c=c,
body_wave_factor=body_wave_factor,
body_wave_freq_scale=body_wave_freq_scale)
for component in ["E", "N", "Z"]:
tr = _empty_sac_trace()
tr.data = u
tr.stats.network = station["network"]
tr.stats.station = station["station"]
tr.stats.location = ""
tr.stats.channel = "BH%s" % component
tr.stats.sac.stla = station["latitude"]
tr.stats.sac.stlo = station["longitude"]
tr.stats.sac.stdp = 0.0
tr.stats.sac.stel = 0.0
path = os.path.join(proj.paths["data"],
"dummy_event_%i" % (_i + 1), "raw")
if not os.path.exists(path):
os.makedirs(path)
tr.write(os.path.join(path, "%s.%s..BH%s.sac" %
(station["network"], station["station"], component)),
format="sac")
print "Generated %i waveform files." % (30 * 3 * len(events))
def plot_depth_slice(self, component, depth_in_km):
"""
Plots a depth slice.
"""
lat_bounds = [rotations.colat2lat(_i)
for _i in self.setup["physical_boundaries_x"][::-1]]
lng_bounds = self.setup["physical_boundaries_y"]
depth_bounds = [6371 - _i / 1000
for _i in self.setup["physical_boundaries_z"]]
data = self.parsed_components[component]
available_depths = np.linspace(*depth_bounds, num=data.shape[2])[::-1]
depth_index = np.argmin(np.abs(available_depths - depth_in_km))
lon, lat = np.meshgrid(
np.linspace(*lng_bounds, num=data.shape[1]),
np.linspace(*lat_bounds, num=data.shape[0]))
if self.rotation_axis and self.rotation_angle_in_degree:
lon_shape = lon.shape
lat_shape = lat.shape
lon.shape = lon.size
lat.shape = lat.size
lat, lon = rotations.rotate_lat_lon(lat, lon, self.rotation_axis,
self.rotation_angle_in_degree)
lon.shape = lon_shape
lat.shape = lat_shape
# Get the center of the map.
lon_0 = lon.min() + lon.ptp() / 2.0
lat_0 = lat.min() + lat.ptp() / 2.0
plt.figure(0)
# Attempt to zoom into the region of interest.
max_extend = max(lon.ptp(), lat.ptp())
extend_used = max_extend / 180.0
if extend_used < 0.5:
x_buffer = 0.2 * lon.ptp()
y_buffer = 0.2 * lat.ptp()
m = Basemap(projection='merc', resolution="l",
#lat_0=lat_0, lon_0=lon_0,
llcrnrlon=lon.min() - x_buffer,
urcrnrlon=lon.max() + x_buffer,
llcrnrlat=lat.min() - y_buffer,
urcrnrlat=lat.max() + y_buffer)
else:
m = Basemap(projection='ortho', lon_0=lon_0, lat_0=lat_0,
resolution="c")
m.drawcoastlines()
m.fillcontinents("0.9", zorder=0)
m.drawmapboundary(fill_color="white")
m.drawparallels(np.arange(-80.0, 80.0, 10.0), labels=[1, 0, 0, 0])
m.drawmeridians(np.arange(-170.0, 170.0, 10.0), labels=[0, 0, 0, 1])
m.drawcountries()
x, y = m(lon, lat)
im = m.pcolormesh(x, y, data[::-1, :, depth_index],
cmap=tomo_colormap)
# Add colorbar and potentially unit.
cm = m.colorbar(im, "right", size="3%", pad='2%')
if component in UNIT_DICT:
cm.set_label(UNIT_DICT[component], fontsize="x-large", rotation=0)
plt.suptitle("Depth slice of %s at %i km" % (component,
int(depth_in_km)), size="large")
def _on_button_press(event):
if event.button != 1 or not event.inaxes:
return
lon, lat = m(event.xdata, event.ydata, inverse=True)
# Convert to colat to ease indexing.
colat = rotations.lat2colat(lat)
x_range = (self.setup["physical_boundaries_x"][1] -
self.setup["physical_boundaries_x"][0])
x_frac = (colat - self.setup["physical_boundaries_x"][0]) / x_range
x_index = int(((self.setup["boundaries_x"][1] -
self.setup["boundaries_x"][0]) * x_frac) +
self.setup["boundaries_x"][0])
y_range = (self.setup["physical_boundaries_y"][1] -
self.setup["physical_boundaries_y"][0])
y_frac = (lon - self.setup["physical_boundaries_y"][0]) / y_range
y_index = int(((self.setup["boundaries_y"][1] -
self.setup["boundaries_y"][0]) * y_frac) +
self.setup["boundaries_y"][0])
plt.figure(1, figsize=(3, 8))
depths = available_depths
values = data[x_index, y_index, :]
plt.plot(values, depths)
plt.grid()
plt.ylim(depths[-1], depths[0])
plt.show()
plt.close()
plt.figure(0)
plt.gcf().canvas.mpl_connect('button_press_event', _on_button_press)
plt.show()
def plot_slice(self,
depth,
min_val_plot=None,
max_val_plot=None,
colormap='tomo',
res='i',
save_under=None,
verbose=False,
lasif_folder=None,
vmin=None,
vmax=None):
"""
plot horizontal slices through an ses3d model
plot_slice(self,depth,colormap='tomo',res='i',save_under=None,
verbose=False)
depth=depth in km of the slice
colormap='tomo','mono'
res=resolution of the map, admissible values are: c, l, i, h f
save_under=save figure as *.png with the filename "save_under".
Prevents plotting of the slice.
"""
import matplotlib.cm
from matplotlib.colors import LogNorm
import matplotlib.pylab as plt
plt.style.use('seaborn-pastel')
if not lasif_folder:
raise NotImplementedError
from lasif.scripts.lasif_cli import _find_project_comm
comm = _find_project_comm(lasif_folder, read_only_caches=False)
plt.figure(figsize=(32, 18))
depth_position_map = {
50: (0, 0),
100: (0, 1),
150: (1, 0),
250: (1, 1),
400: (2, 0),
600: (2, 1)
}
for depth, location in depth_position_map.items():
ax = plt.subplot2grid((3, 5), location)
radius = 6371.0 - depth
# set up a map and colourmap
m = comm.project.domain.plot(ax=ax)
import lasif.colors
my_colormap = lasif.colors.get_colormap(
"tomo_full_scale_linear_lightness")
from lasif import rotations
x, y = np.meshgrid(self.data.longitude, self.data.latitude)
x_shape = x.shape
y_shape = y.shape
lat_r, lon_r = rotations.rotate_lat_lon(
y.ravel(), x.ravel(), comm.project.domain.rotation_axis,
comm.project.domain.rotation_angle_in_degree)
x, y = m(lon_r, lat_r)
x.shape = x_shape
y.shape = y_shape
plot_data = self.data.sel(radius=radius, method="nearest")
plot_data = np.ma.masked_invalid(plot_data.data)
# Overwrite colormap things if given.
if vmin is not None and vmax is not None:
min_val_plot = vmin
max_val_plot = vmax
else:
mean = plot_data.mean()
max_diff = max(abs(mean - plot_data.min()),
abs(plot_data.max() - mean))
min_val_plot = mean - max_diff
max_val_plot = mean + max_diff
# Plotting essentially constant models.
min_delta = 0.01 * abs(max_val_plot)
if (max_val_plot - min_val_plot) < min_delta:
max_val_plot = max_val_plot + min_delta
min_val_plot = min_val_plot - min_delta
# Plot.
im = m.pcolormesh(x,
y,
plot_data,
cmap=my_colormap,
vmin=min_val_plot,
vmax=max_val_plot,
shading="gouraud")
# make a colorbar and title
m.colorbar(im, "right", size="3%", pad='2%')
plt.title(str(depth) + ' km')
# Depth based statistics.
plt.subplot2grid((3, 5), (0, 4), rowspan=3)
plt.title("Depth statistics")
mean = self.data.mean(axis=(0, 1))
std = self.data.std(axis=(0, 1))
_min = self.data.min(axis=(0, 1))
_max = self.data.max(axis=(0, 1))
plt.fill_betweenx(self.data.radius,
mean - std,
mean + std,
label="std",
color="#FF3C83")
plt.plot(mean, self.data.radius, label="mean", color="k", lw=2)
plt.plot(_min, self.data.radius, color="grey", label="min")
plt.plot(_max, self.data.radius, color="grey", label="max")
plt.legend(loc="best")
plt.xlabel("Value")
plt.ylabel("Radius")
# Roughness plots.
plt.subplot2grid((3, 5), (0, 2))
data = np.abs(self.data.diff("latitude", n=1)).sum("latitude").data
plt.title("Roughness in latitude direction, Total: %g" % data.sum())
plt.pcolormesh(self.data.longitude.data,
self.data.radius.data,
data.T,
cmap=matplotlib.cm.viridis,
norm=LogNorm(data.max() * 1E-2, data.max()))
plt.colorbar()
plt.xlabel("Longitude")
plt.ylabel("Radius")
plt.subplot2grid((3, 5), (1, 2))
data = np.abs(self.data.diff("longitude", n=1)).sum("longitude").data
plt.title("Roughness in longitude direction. Total: %g" % data.sum())
plt.pcolormesh(self.data.latitude.data,
self.data.radius.data,
data.T,
cmap=matplotlib.cm.viridis,
norm=LogNorm(data.max() * 1E-2, data.max()))
plt.colorbar()
plt.xlabel("Latitude")
plt.ylabel("Radius")
plt.subplot2grid((3, 5), (2, 2))
data = np.abs(self.data.diff("radius", n=1)).sum("radius").data
plt.title("Roughness in radius direction. Total: %g" % data.sum())
plt.pcolormesh(self.data.longitude.data,
self.data.latitude.data,
data,
cmap=matplotlib.cm.viridis)
plt.colorbar()
plt.xlabel("Longitude")
plt.ylabel("Latitude")
# L2
plt.subplot2grid((3, 5), (0, 3))
data = (self.data**2).sum("latitude").data
plt.title("L2 Norm in latitude direction, Total: %g" % data.sum())
plt.pcolormesh(self.data.longitude.data,
self.data.radius.data,
data.T,
cmap=matplotlib.cm.viridis)
plt.colorbar()
plt.xlabel("Longitude")
plt.ylabel("Radius")
plt.subplot2grid((3, 5), (1, 3))
data = (self.data**2).sum("longitude").data
plt.title("L2 Norm in longitude direction, Total: %g" % data.sum())
plt.pcolormesh(self.data.latitude.data,
self.data.radius.data,
data.T,
cmap=matplotlib.cm.viridis)
plt.colorbar()
plt.xlabel("Latitude")
plt.ylabel("Radius")
plt.subplot2grid((3, 5), (2, 3))
data = (self.data**2).sum("radius").data
plt.title("L2 Norm in radius direction, Total: %g" % data.sum())
plt.pcolormesh(self.data.longitude.data,
self.data.latitude.data,
data,
cmap=matplotlib.cm.viridis)
plt.colorbar()
plt.xlabel("Longitude")
plt.ylabel("Latitude")
plt.suptitle("File %s" % self._filename, fontsize=20)
plt.tight_layout(rect=(0, 0, 1, 0.95))
# save image if wanted
if save_under is None:
plt.show()
else:
plt.savefig(save_under, dpi=150)
plt.close()
def get_value(self):
station_id, coordinates = self.items[self.current_index]
data = Stream()
# Now get the actual waveform files. Also find the
# corresponding station file and check the coordinates.
this_waveforms = {
_i["channel_id"]: _i
for _i in waveforms
if _i["channel_id"].startswith(station_id + ".")
}
marked_for_deletion = []
for key, value in this_waveforms.iteritems():
value["trace"] = read(value["filename"])[0]
data += value["trace"]
value["station_file"] = \
station_cache.get_station_filename(
value["channel_id"],
UTCDateTime(value["starttime_timestamp"]))
if value["station_file"] is None:
marked_for_deletion.append(key)
msg = ("Warning: Data and station information for '%s'"
" is available, but the station information "
"only for the wrong timestamp. You should try "
"and retrieve the correct station file.")
warnings.warn(msg % value["channel_id"])
continue
data[-1].stats.station_file = value["station_file"]
for key in marked_for_deletion:
del this_waveforms[key]
if not this_waveforms:
msg = "Could not retrieve data for station '%s'." % \
station_id
warnings.warn(msg)
return None
# Now attempt to get the synthetics.
synthetics_filenames = []
for name, path in synthetic_files.iteritems():
if (station_id + ".") in name:
synthetics_filenames.append(path)
if len(synthetics_filenames) != 3:
msg = "Found %i not 3 synthetics for station '%s'." % (
len(synthetics_filenames), station_id)
warnings.warn(msg)
return None
synthetics = Stream()
# Read all synthetics.
for filename in synthetics_filenames:
synthetics += read(filename)
for synth in synthetics:
if synth.stats.channel in ["X", "Z"]:
synth.data *= -1.0
synth.stats.channel = SYNTH_MAPPING[synth.stats.channel]
synth.stats.starttime = event_info["origin_time"]
# Process the data.
len_synth = synthetics[0].stats.endtime - \
synthetics[0].stats.starttime
data.trim(synthetics[0].stats.starttime - len_synth * 0.05,
synthetics[0].stats.endtime + len_synth * 0.05)
if data:
max_length = max([tr.stats.npts for tr in data])
else:
max_length = 0
if max_length == 0:
msg = (
"Warning: After trimming the waveform data to "
"the time window of the synthetics, no more data is "
"left. The reference time is the one given in the "
"QuakeML file. Make sure it is correct and that "
"the waveform data actually contains data in that "
"time span.")
warnings.warn(msg)
data.detrend("linear")
data.taper()
new_time_array = np.linspace(
synthetics[0].stats.starttime.timestamp,
synthetics[0].stats.endtime.timestamp,
synthetics[0].stats.npts)
# Simulate the traces.
for trace in data:
# Decimate in case there is a large difference between
# synthetic sampling rate and sampling_rate of the data.
# XXX: Ugly filter, change!
if trace.stats.sampling_rate > (6 *
synth.stats.sampling_rate):
new_nyquist = trace.stats.sampling_rate / 2.0 / 5.0
trace.filter("lowpass",
freq=new_nyquist,
corners=4,
zerophase=True)
trace.decimate(factor=5, no_filter=None)
station_file = trace.stats.station_file
if "/SEED/" in station_file:
paz = Parser(station_file).getPAZ(
trace.id, trace.stats.starttime)
trace.simulate(paz_remove=paz)
elif "/RESP/" in station_file:
trace.simulate(
seedresp={
"filename": station_file,
"units": "VEL",
"date": trace.stats.starttime
})
else:
raise NotImplementedError
# Make sure that the data array is at least as long as the
# synthetics array. Also add some buffer sample for the
# spline interpolation to work in any case.
buf = synth.stats.delta * 5
if synth.stats.starttime < (trace.stats.starttime + buf):
trace.trim(starttime=synth.stats.starttime - buf,
pad=True,
fill_value=0.0)
if synth.stats.endtime > (trace.stats.endtime - buf):
trace.trim(endtime=synth.stats.endtime + buf,
pad=True,
fill_value=0.0)
old_time_array = np.linspace(
trace.stats.starttime.timestamp,
trace.stats.endtime.timestamp, trace.stats.npts)
# Interpolation.
trace.data = interp1d(old_time_array, trace.data,
kind=1)(new_time_array)
trace.stats.starttime = synthetics[0].stats.starttime
trace.stats.sampling_rate = \
synthetics[0].stats.sampling_rate
data.filter("bandpass", freqmin=lowpass, freqmax=highpass)
synthetics.filter("bandpass",
freqmin=lowpass,
freqmax=highpass)
# Rotate the synthetics if nessesary.
if self.rot_angle:
# First rotate the station back to see, where it was
# recorded.
lat, lng = rotations.rotate_lat_lon(
coordinates["latitude"], coordinates["longitude"],
self.rot_axis, -self.rot_angle)
# Rotate the data.
n_trace = synthetics.select(component="N")[0]
e_trace = synthetics.select(component="E")[0]
z_trace = synthetics.select(component="Z")[0]
n, e, z = rotations.rotate_data(n_trace.data, e_trace.data,
z_trace.data, lat, lng,
self.rot_axis,
self.rot_angle)
n_trace.data = n
e_trace.data = e
z_trace.data = z
return {
"data": data,
"synthetics": synthetics,
"coordinates": coordinates
}
def plot_slice(self, depth, min_val_plot=None, max_val_plot=None,
colormap='tomo', res='i', save_under=None, verbose=False,
lasif_folder=None, vmin=None, vmax=None):
"""
plot horizontal slices through an ses3d model
plot_slice(self,depth,colormap='tomo',res='i',save_under=None,
verbose=False)
depth=depth in km of the slice
colormap='tomo','mono'
res=resolution of the map, admissible values are: c, l, i, h f
save_under=save figure as *.png with the filename "save_under".
Prevents plotting of the slice.
"""
import matplotlib.cm
from matplotlib.colors import LogNorm
import matplotlib.pylab as plt
plt.style.use('seaborn-pastel')
if not lasif_folder:
raise NotImplementedError
from lasif.scripts.lasif_cli import _find_project_comm
comm = _find_project_comm(lasif_folder, read_only_caches=False)
plt.figure(figsize=(32, 18))
depth_position_map = {
50: (0, 0),
100: (0, 1),
150: (1, 0),
250: (1, 1),
400: (2, 0),
600: (2, 1)
}
for depth, location in depth_position_map.items():
ax = plt.subplot2grid((3, 5), location)
radius = 6371.0 - depth
# set up a map and colourmap
m = comm.project.domain.plot(ax=ax)
import lasif.colors
my_colormap = lasif.colors.get_colormap(
"tomo_full_scale_linear_lightness")
from lasif import rotations
x, y = np.meshgrid(self.data.longitude, self.data.latitude)
x_shape = x.shape
y_shape = y.shape
lat_r, lon_r = rotations.rotate_lat_lon(
y.ravel(), x.ravel(),
comm.project.domain.rotation_axis,
comm.project.domain.rotation_angle_in_degree)
x, y = m(lon_r, lat_r)
x.shape = x_shape
y.shape = y_shape
plot_data = self.data.sel(radius=radius, method="nearest")
plot_data = np.ma.masked_invalid(plot_data.data)
# Overwrite colormap things if given.
if vmin is not None and vmax is not None:
min_val_plot = vmin
max_val_plot = vmax
else:
mean = plot_data.mean()
max_diff = max(abs(mean - plot_data.min()),
abs(plot_data.max() - mean))
min_val_plot = mean - max_diff
max_val_plot = mean + max_diff
# Plotting essentially constant models.
min_delta = 0.01 * abs(max_val_plot)
if (max_val_plot - min_val_plot) < min_delta:
max_val_plot = max_val_plot + min_delta
min_val_plot = min_val_plot - min_delta
# Plot.
im = m.pcolormesh(
x, y, plot_data,
cmap=my_colormap, vmin=min_val_plot, vmax=max_val_plot,
shading="gouraud")
# make a colorbar and title
m.colorbar(im, "right", size="3%", pad='2%')
plt.title(str(depth) + ' km')
# Depth based statistics.
plt.subplot2grid((3, 5), (0, 4), rowspan=3)
plt.title("Depth statistics")
mean = self.data.mean(axis=(0, 1))
std = self.data.std(axis=(0, 1))
_min = self.data.min(axis=(0, 1))
_max = self.data.max(axis=(0, 1))
plt.fill_betweenx(self.data.radius, mean - std, mean + std,
label="std", color="#FF3C83")
plt.plot(mean, self.data.radius, label="mean", color="k", lw=2)
plt.plot(_min, self.data.radius, color="grey", label="min")
plt.plot(_max, self.data.radius, color="grey", label="max")
plt.legend(loc="best")
plt.xlabel("Value")
plt.ylabel("Radius")
# Roughness plots.
plt.subplot2grid((3, 5), (0, 2))
data = np.abs(self.data.diff("latitude", n=1)).sum("latitude").data
plt.title("Roughness in latitude direction, Total: %g" % data.sum())
plt.pcolormesh(self.data.longitude.data, self.data.radius.data,
data.T, cmap=matplotlib.cm.viridis,
norm=LogNorm(data.max() * 1E-2, data.max()))
plt.colorbar()
plt.xlabel("Longitude")
plt.ylabel("Radius")
plt.subplot2grid((3, 5), (1, 2))
data = np.abs(self.data.diff("longitude", n=1)).sum("longitude").data
plt.title("Roughness in longitude direction. Total: %g" % data.sum())
plt.pcolormesh(self.data.latitude.data, self.data.radius.data, data.T,
cmap=matplotlib.cm.viridis,
norm=LogNorm(data.max() * 1E-2, data.max()))
plt.colorbar()
plt.xlabel("Latitude")
plt.ylabel("Radius")
plt.subplot2grid((3, 5), (2, 2))
data = np.abs(self.data.diff("radius", n=1)).sum("radius").data
plt.title("Roughness in radius direction. Total: %g" % data.sum())
plt.pcolormesh(self.data.longitude.data, self.data.latitude.data,
data, cmap=matplotlib.cm.viridis)
plt.colorbar()
plt.xlabel("Longitude")
plt.ylabel("Latitude")
# L2
plt.subplot2grid((3, 5), (0, 3))
data = (self.data ** 2).sum("latitude").data
plt.title("L2 Norm in latitude direction, Total: %g" % data.sum())
plt.pcolormesh(self.data.longitude.data, self.data.radius.data,
data.T, cmap=matplotlib.cm.viridis)
plt.colorbar()
plt.xlabel("Longitude")
plt.ylabel("Radius")
plt.subplot2grid((3, 5), (1, 3))
data = (self.data ** 2).sum("longitude").data
plt.title("L2 Norm in longitude direction, Total: %g" % data.sum())
plt.pcolormesh(self.data.latitude.data, self.data.radius.data, data.T,
cmap=matplotlib.cm.viridis)
plt.colorbar()
plt.xlabel("Latitude")
plt.ylabel("Radius")
plt.subplot2grid((3, 5), (2, 3))
data = (self.data ** 2).sum("radius").data
plt.title("L2 Norm in radius direction, Total: %g" % data.sum())
plt.pcolormesh(self.data.longitude.data, self.data.latitude.data,
data, cmap=matplotlib.cm.viridis)
plt.colorbar()
plt.xlabel("Longitude")
plt.ylabel("Latitude")
plt.suptitle("File %s" % self._filename, fontsize=20)
plt.tight_layout(rect=(0, 0, 1, 0.95))
# save image if wanted
if save_under is None:
plt.show()
else:
plt.savefig(save_under, dpi=150)
plt.close()
def plot_depth_slice(self, component, depth_in_km):
"""
Plots a depth slice.
"""
lat_bounds = [
rotations.colat2lat(_i)
for _i in self.setup["physical_boundaries_x"][::-1]
]
lng_bounds = self.setup["physical_boundaries_y"]
depth_bounds = [
6371 - _i / 1000 for _i in self.setup["physical_boundaries_z"]
]
data = self.parsed_components[component]
available_depths = np.linspace(*depth_bounds, num=data.shape[2])[::-1]
depth_index = np.argmin(np.abs(available_depths - depth_in_km))
lon, lat = np.meshgrid(np.linspace(*lng_bounds, num=data.shape[1]),
np.linspace(*lat_bounds, num=data.shape[0]))
if self.rotation_axis and self.rotation_angle_in_degree:
lon_shape = lon.shape
lat_shape = lat.shape
lon.shape = lon.size
lat.shape = lat.size
lat, lon = rotations.rotate_lat_lon(lat, lon, self.rotation_axis,
self.rotation_angle_in_degree)
lon.shape = lon_shape
lat.shape = lat_shape
# Get the center of the map.
lon_0 = lon.min() + lon.ptp() / 2.0
lat_0 = lat.min() + lat.ptp() / 2.0
plt.figure(0)
# Attempt to zoom into the region of interest.
max_extend = max(lon.ptp(), lat.ptp())
extend_used = max_extend / 180.0
if extend_used < 0.5:
x_buffer = 0.2 * lon.ptp()
y_buffer = 0.2 * lat.ptp()
m = Basemap(
projection='merc',
resolution="l",
#lat_0=lat_0, lon_0=lon_0,
llcrnrlon=lon.min() - x_buffer,
urcrnrlon=lon.max() + x_buffer,
llcrnrlat=lat.min() - y_buffer,
urcrnrlat=lat.max() + y_buffer)
else:
m = Basemap(projection='ortho',
lon_0=lon_0,
lat_0=lat_0,
resolution="c")
m.drawcoastlines()
m.fillcontinents("0.9", zorder=0)
m.drawmapboundary(fill_color="white")
m.drawparallels(np.arange(-80.0, 80.0, 10.0), labels=[1, 0, 0, 0])
m.drawmeridians(np.arange(-170.0, 170.0, 10.0), labels=[0, 0, 0, 1])
m.drawcountries()
x, y = m(lon, lat)
im = m.pcolormesh(x, y, data[::-1, :, depth_index], cmap=tomo_colormap)
# Add colorbar and potentially unit.
cm = m.colorbar(im, "right", size="3%", pad='2%')
if component in UNIT_DICT:
cm.set_label(UNIT_DICT[component], fontsize="x-large", rotation=0)
plt.suptitle("Depth slice of %s at %i km" %
(component, int(depth_in_km)),
size="large")
def _on_button_press(event):
if event.button != 1 or not event.inaxes:
return
lon, lat = m(event.xdata, event.ydata, inverse=True)
# Convert to colat to ease indexing.
colat = rotations.lat2colat(lat)
x_range = (self.setup["physical_boundaries_x"][1] -
self.setup["physical_boundaries_x"][0])
x_frac = (colat - self.setup["physical_boundaries_x"][0]) / x_range
x_index = int(((self.setup["boundaries_x"][1] -
self.setup["boundaries_x"][0]) * x_frac) +
self.setup["boundaries_x"][0])
y_range = (self.setup["physical_boundaries_y"][1] -
self.setup["physical_boundaries_y"][0])
y_frac = (lon - self.setup["physical_boundaries_y"][0]) / y_range
y_index = int(((self.setup["boundaries_y"][1] -
self.setup["boundaries_y"][0]) * y_frac) +
self.setup["boundaries_y"][0])
plt.figure(1, figsize=(3, 8))
depths = available_depths
values = data[x_index, y_index, :]
plt.plot(values, depths)
plt.grid()
plt.ylim(depths[-1], depths[0])
plt.show()
plt.close()
plt.figure(0)
plt.gcf().canvas.mpl_connect('button_press_event', _on_button_press)
plt.show()
def par2quakeml(Par_filename, QuakeML_filename, rotation_axis=[0.0, 1.0, 0.0],
rotation_angle=-57.5, origin_time="2000-01-01 00:00:00.0",
event_type="other event"):
# initialise event
ev = Event()
# open and read Par file
fid = open(Par_filename, 'r')
fid.readline()
fid.readline()
fid.readline()
fid.readline()
lat_old = 90.0 - float(fid.readline().strip().split()[0])
lon_old = float(fid.readline().strip().split()[0])
depth = float(fid.readline().strip().split()[0])
fid.readline()
Mtt_old = float(fid.readline().strip().split()[0])
Mpp_old = float(fid.readline().strip().split()[0])
Mrr_old = float(fid.readline().strip().split()[0])
Mtp_old = float(fid.readline().strip().split()[0])
Mtr_old = float(fid.readline().strip().split()[0])
Mpr_old = float(fid.readline().strip().split()[0])
# rotate event into physical domain
lat, lon = rot.rotate_lat_lon(lat_old, lon_old, rotation_axis,
rotation_angle)
Mrr, Mtt, Mpp, Mtr, Mpr, Mtp = rot.rotate_moment_tensor(
Mrr_old, Mtt_old, Mpp_old, Mtr_old, Mpr_old, Mtp_old, lat_old, lon_old,
rotation_axis, rotation_angle)
# populate event origin data
ev.event_type = event_type
ev_origin = Origin()
ev_origin.time = UTCDateTime(origin_time)
ev_origin.latitude = lat
ev_origin.longitude = lon
ev_origin.depth = depth
ev.origins.append(ev_origin)
# populte event moment tensor
ev_tensor = Tensor()
ev_tensor.m_rr = Mrr
ev_tensor.m_tt = Mtt
ev_tensor.m_pp = Mpp
ev_tensor.m_rt = Mtr
ev_tensor.m_rp = Mpr
ev_tensor.m_tp = Mtp
ev_momenttensor = MomentTensor()
ev_momenttensor.tensor = ev_tensor
ev_momenttensor.scalar_moment = np.sqrt(Mrr ** 2 + Mtt ** 2 + Mpp ** 2 +
Mtr ** 2 + Mpr ** 2 + Mtp ** 2)
ev_focalmechanism = FocalMechanism()
ev_focalmechanism.moment_tensor = ev_momenttensor
ev_focalmechanism.nodal_planes = NodalPlanes().setdefault(0, 0)
ev.focal_mechanisms.append(ev_focalmechanism)
# populate event magnitude
ev_magnitude = Magnitude()
ev_magnitude.mag = 0.667 * (np.log10(ev_momenttensor.scalar_moment) - 9.1)
ev_magnitude.magnitude_type = 'Mw'
ev.magnitudes.append(ev_magnitude)
# write QuakeML file
cat = Catalog()
cat.append(ev)
cat.write(QuakeML_filename, format="quakeml")
# clean up
fid.close()
def plot_raydensity(map_object, station_events, domain):
"""
Create a ray-density plot for all events and all stations.
This function is potentially expensive and will use all CPUs available.
Does require geographiclib to be installed.
"""
import ctypes as C
from lasif import rotations
from lasif.domain import RectangularSphericalSection
from lasif.tools.great_circle_binner import GreatCircleBinner
from lasif.utils import Point
import multiprocessing
import progressbar
from scipy.stats import scoreatpercentile
if not isinstance(domain, RectangularSphericalSection):
raise NotImplementedError(
"Raydensity currently only implemented for rectangular domains. "
"Should be easy to implement for other domains. Let me know.")
# Merge everything so that a list with coordinate pairs is created. This
# list is then distributed among all processors.
station_event_list = []
for event, stations in station_events:
if domain.rotation_angle_in_degree:
# Rotate point to the non-rotated domain.
e_point = Point(*rotations.rotate_lat_lon(
event["latitude"], event["longitude"], domain.rotation_axis,
-1.0 * domain.rotation_angle_in_degree))
else:
e_point = Point(event["latitude"], event["longitude"])
for station in stations.itervalues():
# Rotate point to the non-rotated domain if necessary.
if domain.rotation_angle_in_degree:
p = Point(*rotations.rotate_lat_lon(
station["latitude"], station["longitude"],
domain.rotation_axis,
-1.0 * domain.rotation_angle_in_degree))
else:
p = Point(station["latitude"], station["longitude"])
station_event_list.append((e_point, p))
circle_count = len(station_event_list)
# The granularity of the latitude/longitude discretization for the
# raypaths. Attempt to get a somewhat meaningful result in any case.
lat_lng_count = 1000
if circle_count < 1000:
lat_lng_count = 1000
if circle_count < 10000:
lat_lng_count = 2000
else:
lat_lng_count = 3000
cpu_count = multiprocessing.cpu_count()
def to_numpy(raw_array, dtype, shape):
data = np.frombuffer(raw_array.get_obj())
data.dtype = dtype
return data.reshape(shape)
print "\nLaunching %i greatcircle calculations on %i CPUs..." % \
(circle_count, cpu_count)
widgets = ["Progress: ", progressbar.Percentage(),
progressbar.Bar(), "", progressbar.ETA()]
pbar = progressbar.ProgressBar(widgets=widgets,
maxval=circle_count).start()
def great_circle_binning(sta_evs, bin_data_buffer, bin_data_shape,
lock, counter):
new_bins = GreatCircleBinner(
domain.min_latitude, domain.max_latitude,
lat_lng_count, domain.min_longitude,
domain.max_longitude, lat_lng_count)
for event, station in sta_evs:
with lock:
counter.value += 1
if not counter.value % 25:
pbar.update(counter.value)
new_bins.add_greatcircle(event, station)
bin_data = to_numpy(bin_data_buffer, np.uint32, bin_data_shape)
with bin_data_buffer.get_lock():
bin_data += new_bins.bins
# Split the data in cpu_count parts.
def chunk(seq, num):
avg = len(seq) / float(num)
out = []
last = 0.0
while last < len(seq):
out.append(seq[int(last):int(last + avg)])
last += avg
return out
chunks = chunk(station_event_list, cpu_count)
# One instance that collects everything.
collected_bins = GreatCircleBinner(
domain.min_latitude, domain.max_latitude,
lat_lng_count, domain.min_longitude,
domain.max_longitude, lat_lng_count)
# Use a multiprocessing shared memory array and map it to a numpy view.
collected_bins_data = multiprocessing.Array(C.c_uint32,
collected_bins.bins.size)
collected_bins.bins = to_numpy(collected_bins_data, np.uint32,
collected_bins.bins.shape)
# Create, launch and join one process per CPU. Use a shared value as a
# counter and a lock to avoid race conditions.
processes = []
lock = multiprocessing.Lock()
counter = multiprocessing.Value("i", 0)
for _i in xrange(cpu_count):
processes.append(multiprocessing.Process(
target=great_circle_binning, args=(chunks[_i], collected_bins_data,
collected_bins.bins.shape, lock,
counter)))
for process in processes:
process.start()
for process in processes:
process.join()
pbar.finish()
stations = chain.from_iterable((
_i[1].values() for _i in station_events if _i[1]))
# Remove duplicates
stations = [(_i["latitude"], _i["longitude"]) for _i in stations]
stations = set(stations)
title = "%i Events, %i unique raypaths, "\
"%i unique stations" % (len(station_events), circle_count,
len(stations))
plt.title(title, size="xx-large")
data = collected_bins.bins.transpose()
if data.max() >= 10:
data = np.log10(np.clip(data, a_min=0.5, a_max=data.max()))
data[data >= 0.0] += 0.1
data[data < 0.0] = 0.0
max_val = scoreatpercentile(data.ravel(), 99)
else:
max_val = data.max()
cmap = cm.get_cmap("gist_heat")
cmap._init()
cmap._lut[:120, -1] = np.linspace(0, 1.0, 120) ** 2
# Slightly change the appearance of the map so it suits the rays.
map_object.fillcontinents(color='#dddddd', lake_color='#dddddd', zorder=2)
lngs, lats = collected_bins.coordinates
# Rotate back if necessary!
if domain.rotation_angle_in_degree:
for lat, lng in zip(lats, lngs):
lat[:], lng[:] = rotations.rotate_lat_lon(
lat, lng, domain.rotation_axis,
domain.rotation_angle_in_degree)
ln, la = map_object(lngs, lats)
map_object.pcolormesh(ln, la, data, cmap=cmap, vmin=0, vmax=max_val,
zorder=10)
# Draw the coastlines so they appear over the rays. Otherwise things are
# sometimes hard to see.
map_object.drawcoastlines(zorder=3)
map_object.drawcountries(linewidth=0.2, zorder=3)
def Iter2snapshot(snapD, evlo, evla, component, vmin, vmax, stations, fprx, projection, outdir, dpi, \
lat_min, lat_max, lon_min, lon_max, lat_centre, lon_centre, res, zoomin, geopolygons, dt):
"""Plot snapshot, used by plot_snapshots_mp
"""
print 'Plotting Snapshot for:',snapD.iteration,' step!'
# - Set up the map. ------------------------------------------------------------------------
if projection=='global':
m=Basemap(projection='ortho', lon_0=lon_centre, lat_0=lat_centre, resolution=res)
m.drawparallels(np.arange(-80.0,80.0,10.0),labels=[1,0,0,1])
m.drawmeridians(np.arange(-170.0,170.0,10.0),labels=[1,0,0,1])
elif projection=='regional_ortho':
m1 = Basemap(projection='ortho', lon_0=lon_min, lat_0=lat_min, resolution='l')
m = Basemap(projection='ortho',lon_0=lon_min,lat_0=lat_min, resolution=res,\
llcrnrx=0., llcrnry=0., urcrnrx=m1.urcrnrx/zoomin, urcrnry=m1.urcrnry/3.5)
elif projection=='regional_merc':
m=Basemap(projection='merc',llcrnrlat=lat_min,urcrnrlat=lat_max,llcrnrlon=lon_min,urcrnrlon=lon_max,lat_ts=20,resolution=res)
m.drawparallels(np.arange(np.round(lat_min),np.round(lat_max),d_lat),labels=[1,0,0,1])
m.drawmeridians(np.arange(np.round(lon_min),np.round(lon_max),d_lon),labels=[1,0,0,1])
elif projection=='lambert':
distEW, az, baz=obspy.geodetics.gps2dist_azimuth(lat_min, lon_min,
lat_min, lon_max) # distance is in m
distNS, az, baz=obspy.geodetics.gps2dist_azimuth(lat_min, lon_min,
lat_max+2., lon_min) # distance is in m
m = Basemap(width=distEW, height=distNS, rsphere=(6378137.00,6356752.3142), resolution='l', projection='lcc',\
lat_1=lat_min, lat_2=lat_max, lon_0=lon_centre, lat_0=lat_centre+1.)
m.drawparallels(np.arange(-80.0,80.0,10.0), linewidth=1, dashes=[2,2], labels=[1,0,0,0], fontsize=15)
m.drawmeridians(np.arange(-170.0,170.0,10.0), linewidth=1, dashes=[2,2], labels=[0,0,1,1], fontsize=15)
m.drawcoastlines()
m.fillcontinents(lake_color='#99ffff',zorder=0.2)
m.drawmapboundary(fill_color="white")
m.drawcountries()
try:
evx, evy=m(evlo, evla)
m.plot(evx, evy, 'yo', markersize=2)
except: pass
for procD in snapD:
field = procD.field
lats = procD.lat
lons = procD.lon
lon, lat = np.meshgrid(lons, lats)
if snapD.rotangle != 0.0:
lat_rot = np.zeros(np.shape(lon),dtype=float)
lon_rot = np.zeros(np.shape(lat),dtype=float)
for idlon in np.arange(len(lons)):
for idlat in np.arange(len(lats)):
lat_rot[idlat,idlon], lon_rot[idlat,idlon] = rotations.rotate_lat_lon(lat[idlat,idlon], lon[idlat,idlon], snapD.n, -snapD.rotangle)
lat_rot[idlat,idlon] = 90.0-lat_rot[idlat,idlon]
lon = lon_rot
lat = lat_rot
# - colourmap. ---------------------------------------------------------
cmap = colors.get_colormap('tomo_80_perc_linear_lightness')
x, y = m(lon, lat)
im = m.pcolormesh(x, y, field, shading='gouraud', cmap=cmap, vmin=vmin, vmax=vmax)
# - Add colobar and title. ------------------------------------------------------------------
cb = m.colorbar(im, "right", size="3%", pad='2%')
if component in UNIT_DICT:
cb.set_label(UNIT_DICT[component], fontsize="x-large", rotation=0)
try:
geopolygons.PlotPolygon(inbasemap=m)
except:
pass
outfname=outdir+'/'+fprx+'_%06d.png' %(int(snapD.iteration))
savefig(outfname, format='png', dpi=dpi)
del m, lon, lat
return
def plot_snapshots_mp(self, component, vmin, vmax, outdir, fprx='wavefield',iter0=100, iterf=17100, diter=200,
stations=False, res="i", projection='lambert', dpi=300, zoomin=2, geopolygons=None, evlo=None, evla=None, dt=None ):
"""Multiprocessing version of plot_snapshots
================================================================================================
Input parameters:
component - component for plotting
The currently available "components" are:
Material parameters: A, B, C, mu, lambda, rhoinv, vp, vsh, vsv, rho
Velocity field snapshots: vx, vy, vz
Sensitivity kernels: Q_mu, Q_kappa, alpha_mu, alpha_kappa
depth - depth for plot (km)
vmin, vmax - minimum/maximum value for plotting
outdir - output directory
fprx - output file name prefix
iter0, iterf - inital/final iterations for plotting
diter - iteration interval
stations - plot stations or not
res - resolution of the coastline (c, l, i, h, f)
projection - projection type (global, regional_ortho, regional_merc)
dpi - dots per inch (figure resolution parameter)
zoomin - zoom in factor for proj = regional_ortho
geopolygons - geological polygons( basins etc. ) for plot
evlo, evla - event location for plotting
=================================================================================================
"""
if not os.path.isdir(outdir):
os.makedirs(outdir)
iterArr=np.arange(iter0 ,iterf+diter, diter, dtype=int)
use_default_iter=False
for iteration in iterArr:
if not str(iteration) in self[component].keys():
warnings.warn('Velocity Snapshot:'+str(iteration)+' does not exist!', UserWarning, stacklevel=1)
use_default_iter=True
if use_default_iter: iterArr = self[component].keys()
lat_min = self.attrs['lat_min']; lat_max = self.attrs['lat_max']
lon_min = self.attrs['lon_min']; lon_max = self.attrs['lon_max']
self.n = self.attrs['rotation_axis']; self.rotangle = self.attrs['rotation_angle']
lat_centre = (lat_max+lat_min)/2.0; lon_centre = (lon_max+lon_min)/2.0
lat_centre, lon_centre = rotations.rotate_lat_lon(lat_centre, lon_centre, self.n, -self.rotangle)
print '================================= Start preparing generating snapshots =================================='
dataLst=[]
for iteration in iterArr:
subgroup=self[component+'/'+str(iteration)]
snapD=snap_data()
for key in subgroup.keys():
subdset = subgroup[key]
field = subdset[...]
theta = subdset.attrs['theta']
phi = subdset.attrs['phi']
lats = 90.0 - theta * 180.0 / np.pi
lons = phi * 180.0 / np.pi
snapD.append(proc_data(field=field, lon=lons, lat=lats))
snapD.n=self.n; snapD.rotangle=self.rotangle; snapD.iteration=iteration
dataLst.append(snapD)
self.close()
print '============================= Start multiprocessing generating snapshots ==============================='
PLOTSNAP = partial(Iter2snapshot, evlo=evlo, evla=evla, component=component, \
vmin=vmin, vmax=vmax, stations=stations, fprx=fprx, projection=projection, \
outdir=outdir, dpi=dpi, lat_min=lat_min, lat_max=lat_max, lon_min=lon_min, lon_max=lon_max,\
lat_centre=lat_centre, lon_centre=lon_centre, res=res, \
zoomin=zoomin, geopolygons=geopolygons, dt=dt)
pool=multiprocessing.Pool()
pool.map(PLOTSNAP, dataLst) #make our results with a map call
pool.close() #we are not adding any more processes
pool.join() #tell it to wait until all threads are done before going on
print '============================== End multiprocessing generating snapshots ================================='
return
def plot_depth_slice(self, component, vmin, vmax, iteration=0,
res="l", projection='lambert', zoomin=2, geopolygons=None, evlo=None, evla=None):
"""
Plot depth slices of field component at given depth ranging between "valmin" and "valmax"
================================================================================================
Input parameters:
component - component for plotting
The currently available "components" are:
Material parameters: A, B, C, mu, lambda, rhoinv, vp, vsh, vsv, rho
Velocity field snapshots: vx, vy, vz
Sensitivity kernels: Q_mu, Q_kappa, alpha_mu, alpha_kappa
vmin, vmax - minimum/maximum value for plotting
iteration - iteration step for snapshot
res - resolution of the coastline (c, l, i, h, f)
proj - projection type (global, regional_ortho, regional_merc)
zoomin - zoom in factor for proj = regional_ortho
geopolygons - geological polygons( basins etc. ) for plot
=================================================================================================
"""
# - Some initialisations. ------------------------------------------------------------------
fig=plt.figure()
self.minlat = self.attrs['lat_min']; self.maxlat = self.attrs['lat_max']
self.minlon = self.attrs['lon_min']; self.maxlon = self.attrs['lon_max']
self.n = self.attrs['rotation_axis']; self.rotangle = self.attrs['rotation_angle']
lat_centre = (self.maxlat+self.minlat)/2.0; lon_centre = (self.maxlon+self.minlon)/2.0
self.lat_centre, self.lon_centre = rotations.rotate_lat_lon(lat_centre, lon_centre, self.n, -self.rotangle)
# - Set up the map. ------------------------------------------------------------------------
m=self._get_basemap(projection=projection)
try: geopolygons.PlotPolygon(inbasemap=m)
except: pass
try:
evx, evy=m(evlo, evla)
m.plot(evx, evy, 'yo', markersize=2)
except: pass
group=self[component]
subgroup=group[str(iteration)]
for key in subgroup.keys():
subdset = subgroup[key]
field = subdset[...]
theta = subdset.attrs['theta']
phi = subdset.attrs['phi']
lats = 90.0 - theta * 180.0 / np.pi
lons = phi * 180.0 / np.pi
lon, lat = np.meshgrid(lons, lats)
if self.rotangle != 0.0:
lat_rot = np.zeros(np.shape(lon),dtype=float)
lon_rot = np.zeros(np.shape(lat),dtype=float)
for idlon in np.arange(len(lons)):
for idlat in np.arange(len(lats)):
lat_rot[idlat,idlon],lon_rot[idlat,idlon] = rotations.rotate_lat_lon(lat[idlat,idlon], lon[idlat,idlon], self.n, -self.rotangle)
lat_rot[idlat,idlon] = 90.0-lat_rot[idlat,idlon]
lon = lon_rot
lat = lat_rot
# - colourmap. ---------------------------------------------------------
cmap = colors.get_colormap('tomo_80_perc_linear_lightness')
x, y = m(lon, lat)
im = m.pcolormesh(x, y, field, shading='gouraud', cmap=cmap, vmin=vmin, vmax=vmax)
# - Add colobar and title. ------------------------------------------------------------------
cb = m.colorbar(im, "right", size="3%", pad='2%')
if component in UNIT_DICT:
cb.set_label(UNIT_DICT[component], fontsize="x-large", rotation=0)
# plt.suptitle("Depth slice of %s at %i km" % (component, r_effective), fontsize=20)
# - Plot stations if available. ------------------------------------------------------------
# if self.stations and stations:
# x,y = m(self.stlons,self.stlats)
# for n in range(self.n_stations):
# plt.text(x[n],y[n],self.stnames[n][:4])
# plt.plot(x[n],y[n],'ro')
plt.show()
print "minimum value: "+str(vmin)+", maximum value: "+str(vmax)
return
def get_waveforms_synthetic(self, event_name, station_id,
long_iteration_name):
"""
Gets the synthetic waveforms for the given event and station as a
:class:`~obspy.core.stream.Stream` object.
:param event_name: The name of the event.
:param station_id: The id of the station in the form ``NET.STA``.
:param long_iteration_name: The long form of an iteration name.
"""
from lasif import rotations
import lasif.domain
iteration = self.comm.iterations.get(long_iteration_name)
st = self._get_waveforms(event_name,
station_id,
data_type="synthetic",
tag_or_iteration=iteration.long_name)
network, station = station_id.split(".")
formats = list(set([tr.stats._format for tr in st]))
if len(formats) != 1:
raise ValueError(
"The synthetics for one Earthquake must all have the same "
"data format under the assumption that they all originate "
"from the same solver. Found formats: %s" % (str(formats)))
format = formats[0].lower()
# In the case of data coming from SES3D the components must be
# mapped to ZNE as it works in XYZ.
if format == "ses3d":
# This maps the synthetic channels to ZNE.
synthetic_coordinates_mapping = {"X": "N", "Y": "E", "Z": "Z"}
for tr in st:
tr.stats.network = network
tr.stats.station = station
# SES3D X points south. Reverse it to arrive at ZNE.
if tr.stats.channel in ["X"]:
tr.data *= -1.0
# SES3D files have no starttime. Set to the event time.
tr.stats.starttime = \
self.comm.events.get(event_name)["origin_time"]
tr.stats.channel = \
synthetic_coordinates_mapping[tr.stats.channel]
# Rotate if needed. Again only SES3D synthetics need to be rotated.
domain = self.comm.project.domain
if isinstance(domain, lasif.domain.RectangularSphericalSection) \
and domain.rotation_angle_in_degree and \
"ses3d" in iteration.solver_settings["solver"].lower():
# Coordinates are required for the rotation.
coordinates = self.comm.query.get_coordinates_for_station(
event_name, station_id)
# First rotate the station back to see, where it was
# recorded.
lat, lng = rotations.rotate_lat_lon(
lat=coordinates["latitude"],
lon=coordinates["longitude"],
rotation_axis=domain.rotation_axis,
angle=-domain.rotation_angle_in_degree)
# Rotate the synthetics.
n, e, z = rotations.rotate_data(
st.select(channel="N")[0].data,
st.select(channel="E")[0].data,
st.select(channel="Z")[0].data, lat, lng,
domain.rotation_axis, domain.rotation_angle_in_degree)
st.select(channel="N")[0].data = n
st.select(channel="E")[0].data = e
st.select(channel="Z")[0].data = z
st.sort()
# Apply the project function that modifies synthetics on the fly.
fct = self.comm.project.get_project_function("process_synthetics")
return fct(st,
iteration=iteration,
event=self.comm.events.get(event_name))
def plot_depth_slice(self,
component,
depth_in_km,
m,
absolute_values=True):
"""
Plots a depth slice.
:param component: The component to plot.
:type component: basestring
:param depth_in_km: The depth in km to plot. If the exact depth does
not exists, the nearest neighbour will be plotted.
:type depth_in_km: integer or float
:param m: Basemap instance.
"""
depth_index = self.get_closest_gll_index("depth", depth_in_km)
# No need to do anything if the currently plotted slice is already
# plotted. This is useful for interactive use when the desired depth
# is changed but the closest GLL collocation point is still the same.
if hasattr(m, "_plotted_depth_slice"):
# Use a tuple of relevant parameters.
if m._plotted_depth_slice == (self.directory, depth_index,
component, absolute_values):
return None
data = self.parsed_components[component]
depth = self.collocation_points_depth[depth_index]
lngs = self.collocation_points_lngs
lats = self.collocation_points_lats
# Rotate data if needed.
lon, lat = np.meshgrid(lngs, lats)
if hasattr(self.domain, "rotation_axis") and \
self.domain.rotation_axis and \
self.domain.rotation_angle_in_degree:
lon_shape = lon.shape
lat_shape = lat.shape
lon.shape = lon.size
lat.shape = lat.size
lat, lon = rotations.rotate_lat_lon(
lat, lon, self.domain.rotation_axis,
self.domain.rotation_angle_in_degree)
lon.shape = lon_shape
lat.shape = lat_shape
x, y = m(lon, lat)
depth_data = data[::-1, :, depth_index]
# Plot values relative to AK135.
if not absolute_values:
cmp_map = {"rho": "density", "vp": "vp", "vsh": "vs", "vsv": "vs"}
factor = {
"rho": 1000.0,
"vp": 1.0,
"vsh": 1.0,
"vsv": 1.0,
}
if component not in cmp_map:
vmin, vmax = depth_data.min(), depth_data.max()
vmedian = np.median(depth_data)
offset = max(abs(vmax - vmedian), abs(vmedian - vmin))
if vmax - vmin == 0:
offset = 0.01
vmin = vmedian - offset
vmax = vmedian + offset
else:
reference_value = self.one_d_model.get_value(
cmp_map[component], depth) * factor[component]
depth_data = (depth_data - reference_value) / reference_value
depth_data *= 100.0
offset = np.abs(depth_data)
try:
offset = offset[offset < 50].max()
except BaseException:
offset = offset.max()
vmin = -offset
vmax = offset
else:
vmin, vmax = depth_data.min(), depth_data.max()
vmedian = np.median(depth_data)
offset = max(abs(vmax - vmedian), abs(vmedian - vmin))
min_delta = abs(vmax * 0.005)
if (vmax - vmin) < min_delta:
offset = min_delta
vmin = vmedian - offset
vmax = vmedian + offset
# Remove an existing pcolormesh if it exists. This does not hurt in
# any case but is useful for interactive use.
if hasattr(m, "_depth_slice"):
m._depth_slice.remove()
del m._depth_slice
im = m.pcolormesh(x,
y,
depth_data,
cmap=tomo_colormap,
vmin=vmin,
vmax=vmax)
m._depth_slice = im
# Store what is currently plotted.
m._plotted_depth_slice = (self.directory, depth_index, component,
absolute_values)
return {"depth": depth, "mesh": im, "data": depth_data}
def _plot_hdf5_model_horizontal(f,
component,
output_filename,
vmin=None,
vmax=None):
import matplotlib.cm
import matplotlib.pylab as plt
data = xarray.DataArray(f["data"][component][:],
[("latitude", 90.0 - f["coordinate_0"][:]),
("longitude", f["coordinate_1"][:]),
("radius", f["coordinate_2"][:] / 1000.0)])
plt.style.use('seaborn-pastel')
from lasif.domain import RectangularSphericalSection
domain = RectangularSphericalSection(**dict(f["_meta"]["domain"].attrs))
plt.figure(figsize=(32, 18))
depth_position_map = {
50: (0, 0),
100: (0, 1),
150: (1, 0),
250: (1, 1),
400: (2, 0),
600: (2, 1)
}
for depth, location in depth_position_map.items():
ax = plt.subplot2grid((3, 5), location)
radius = 6371.0 - depth
# set up a map and colourmap
m = domain.plot(ax=ax, resolution="c", skip_map_features=True)
import lasif.colors
my_colormap = lasif.colors.get_colormap(
"tomo_full_scale_linear_lightness")
from lasif import rotations
x, y = np.meshgrid(data.longitude, data.latitude)
x_shape = x.shape
y_shape = y.shape
lat_r, lon_r = rotations.rotate_lat_lon(
y.ravel(), x.ravel(), domain.rotation_axis,
domain.rotation_angle_in_degree)
x, y = m(lon_r, lat_r)
x.shape = x_shape
y.shape = y_shape
plot_data = data.sel(radius=radius, method="nearest")
plot_data = np.ma.masked_invalid(plot_data.data)
# Overwrite colormap things if given.
if vmin is not None and vmax is not None:
min_val_plot = vmin
max_val_plot = vmax
else:
mean = plot_data.mean()
max_diff = max(abs(mean - plot_data.min()),
abs(plot_data.max() - mean))
min_val_plot = mean - max_diff
max_val_plot = mean + max_diff
# Plotting essentially constant models.
min_delta = 0.001 * abs(max_val_plot)
if (max_val_plot - min_val_plot) < min_delta:
max_val_plot = max_val_plot + min_delta
min_val_plot = min_val_plot - min_delta
# Plot.
im = m.pcolormesh(x,
y,
plot_data,
cmap=my_colormap,
vmin=min_val_plot,
vmax=max_val_plot,
shading="gouraud")
# make a colorbar and title
m.colorbar(im, "right", size="3%", pad='2%')
plt.title(str(depth) + ' km')
# Depth based statistics.
plt.subplot2grid((3, 5), (0, 4), rowspan=3)
plt.title("Depth statistics")
mean = data.mean(axis=(0, 1))
std = data.std(axis=(0, 1))
_min = data.min(axis=(0, 1))
_max = data.max(axis=(0, 1))
plt.fill_betweenx(data.radius,
mean - std,
mean + std,
label="std",
color="#FF3C83")
plt.plot(mean, data.radius, label="mean", color="k", lw=2)
plt.plot(_min, data.radius, color="grey", label="min")
plt.plot(_max, data.radius, color="grey", label="max")
plt.legend(loc="best")
plt.xlabel("Value")
plt.ylabel("Radius")
plt.hlines(data.radius,
plt.xlim()[0],
plt.xlim()[1],
color="0.8",
zorder=-10,
linewidth=0.5)
# Roughness plots.
plt.subplot2grid((3, 5), (0, 2))
_d = np.abs(data.diff("latitude", n=1)).sum("latitude").data
plt.title("Roughness in latitude direction, Total: %g" % _d.sum())
plt.pcolormesh(data.longitude.data,
data.radius.data,
_d.T,
cmap=matplotlib.cm.viridis)
try:
plt.colorbar()
except:
pass
plt.xlabel("Longitude")
plt.ylabel("Radius")
plt.subplot2grid((3, 5), (1, 2))
_d = np.abs(data.diff("longitude", n=1)).sum("longitude").data
plt.title("Roughness in longitude direction. Total: %g" % data.sum())
plt.pcolormesh(data.latitude.data,
data.radius.data,
_d.T,
cmap=matplotlib.cm.viridis)
try:
plt.colorbar()
except:
pass
plt.xlabel("Latitude")
plt.ylabel("Radius")
plt.subplot2grid((3, 5), (2, 2))
_d = np.abs(data.diff("radius", n=1)).sum("radius").data
plt.title("Roughness in radius direction. Total: %g" % _d.sum())
plt.pcolormesh(data.longitude.data,
data.latitude.data,
_d,
cmap=matplotlib.cm.viridis)
try:
plt.colorbar()
except:
pass
plt.xlabel("Longitude")
plt.ylabel("Latitude")
# L2
plt.subplot2grid((3, 5), (0, 3))
_d = (data**2).sum("latitude").data
plt.title("L2 Norm in latitude direction, Total: %g" % _d.sum())
plt.pcolormesh(data.longitude.data,
data.radius.data,
_d.T,
cmap=matplotlib.cm.viridis)
try:
plt.colorbar()
except:
pass
plt.xlabel("Longitude")
plt.ylabel("Radius")
plt.subplot2grid((3, 5), (1, 3))
_d = (data**2).sum("longitude").data
plt.title("L2 Norm in longitude direction, Total: %g" % _d.sum())
plt.pcolormesh(data.latitude.data,
data.radius.data,
_d.T,
cmap=matplotlib.cm.viridis)
try:
plt.colorbar()
except:
pass
plt.xlabel("Latitude")
plt.ylabel("Radius")
plt.subplot2grid((3, 5), (2, 3))
_d = (data**2).sum("radius").data
plt.title("L2 Norm in radius direction, Total: %g" % _d.sum())
plt.pcolormesh(data.longitude.data,
data.latitude.data,
_d,
cmap=matplotlib.cm.viridis)
try:
plt.colorbar()
except:
pass
plt.xlabel("Longitude")
plt.ylabel("Latitude")
plt.suptitle("Component %s - File %s" % (component, output_filename),
fontsize=20)
plt.tight_layout(rect=(0, 0, 1, 0.95))
plt.savefig(output_filename, dpi=150)
plt.close()
def finalize_adjoint_sources(self, iteration_name, event_name):
"""
Finalizes the adjoint sources.
"""
import numpy as np
from lasif import rotations
window_manager = self.comm.windows.get(event_name, iteration_name)
event = self.comm.events.get(event_name)
iteration = self.comm.iterations.get(iteration_name)
iteration_event_def = iteration.events[event["event_name"]]
iteration_stations = iteration_event_def["stations"]
# For now assume that the adjoint sources have the same
# sampling rate as the synthetics which in LASIF's workflow
# actually has to be true.
dt = iteration.get_process_params()["dt"]
# Current domain and solver.
domain = self.comm.project.domain
solver = iteration.solver_settings["solver"].lower()
adjoint_source_stations = set()
if "ses3d" in solver:
ses3d_all_coordinates = []
event_weight = iteration_event_def["event_weight"]
output_folder = self.comm.project.get_output_folder(
type="adjoint_sources",
tag="ITERATION_%s__%s" % (iteration_name, event_name))
l = sorted(window_manager.list())
for station, windows in itertools.groupby(
l, key=lambda x: ".".join(x.split(".")[:2])):
if station not in iteration_stations:
continue
print(".", end=' ')
station_weight = iteration_stations[station]["station_weight"]
channels = {}
try:
for w in windows:
w = window_manager.get(w)
channel_weight = 0
srcs = []
for window in w:
ad_src = window.adjoint_source
if not ad_src["adjoint_source"].ptp():
continue
srcs.append(ad_src["adjoint_source"] * window.weight)
channel_weight += window.weight
if not srcs:
continue
# Final adjoint source for that channel and apply all
# weights.
adjoint_source = np.sum(srcs, axis=0) / channel_weight * \
event_weight * station_weight
channels[w.channel_id[-1]] = adjoint_source
except LASIFError as e:
print(("Could not calculate adjoint source for iteration %s "
"and station %s. Repick windows? Reason: %s" %
(iteration.name, station, str(e))))
continue
if not channels:
continue
# Now all adjoint sources of a window should have the same length.
length = set(len(v) for v in list(channels.values()))
assert len(length) == 1
length = length.pop()
# All missing channels will be replaced with a zero array.
for c in ["Z", "N", "E"]:
if c in channels:
continue
channels[c] = np.zeros(length)
# Get the station coordinates
coords = self.comm.query.get_coordinates_for_station(
event_name, station)
# Rotate. if needed
rec_lat = coords["latitude"]
rec_lng = coords["longitude"]
# The adjoint sources depend on the solver.
if "ses3d" in solver:
# Rotate if needed.
if domain.rotation_angle_in_degree:
# Rotate the adjoint source location.
r_rec_lat, r_rec_lng = rotations.rotate_lat_lon(
rec_lat, rec_lng, domain.rotation_axis,
-domain.rotation_angle_in_degree)
# Rotate the adjoint sources.
channels["N"], channels["E"], channels["Z"] = \
rotations.rotate_data(
channels["N"], channels["E"],
channels["Z"], rec_lat, rec_lng,
domain.rotation_axis,
-domain.rotation_angle_in_degree)
else:
r_rec_lat = rec_lat
r_rec_lng = rec_lng
r_rec_depth = 0.0
r_rec_colat = rotations.lat2colat(r_rec_lat)
# Now once again map from ZNE to the XYZ of SES3D.
CHANNEL_MAPPING = {"X": "N", "Y": "E", "Z": "Z"}
adjoint_source_stations.add(station)
adjoint_src_filename = os.path.join(
output_folder, "ad_src_%i" % len(adjoint_source_stations))
ses3d_all_coordinates.append(
(r_rec_colat, r_rec_lng, r_rec_depth))
# Actually write the adjoint source file in SES3D specific
# format.
with open(adjoint_src_filename, "wt") as open_file:
open_file.write("-- adjoint source ------------------\n")
open_file.write(
"-- source coordinates (colat,lon,depth)\n")
open_file.write("%f %f %f\n" %
(r_rec_colat, r_rec_lng, r_rec_depth))
open_file.write("-- source time function (x, y, z) --\n")
# Revert the X component as it has to point south in SES3D.
for x, y, z in zip(-1.0 * channels[CHANNEL_MAPPING["X"]],
channels[CHANNEL_MAPPING["Y"]],
channels[CHANNEL_MAPPING["Z"]]):
open_file.write("%e %e %e\n" % (x, y, z))
open_file.write("\n")
elif "specfem" in solver:
s_set = iteration.solver_settings["solver_settings"]
if "adjoint_source_time_shift" not in s_set:
warnings.warn(
"No <adjoint_source_time_shift> tag in the "
"iteration XML file. No time shift for the "
"adjoint sources will be applied.", LASIFWarning)
src_time_shift = 0
else:
src_time_shift = float(s_set["adjoint_source_time_shift"])
adjoint_source_stations.add(station)
# Write all components. The adjoint sources right now are
# not time shifted.
for component in ["Z", "N", "E"]:
# XXX: M band code could be different.
adjoint_src_filename = os.path.join(
output_folder, "%s.MX%s.adj" % (station, component))
adj_src = channels[component]
l = len(adj_src)
to_write = np.empty((l, 2))
to_write[:, 0] = \
np.linspace(0, (l - 1) * dt, l) + src_time_shift
# SPECFEM expects non-time reversed adjoint sources and
# the sign is different for some reason.
to_write[:, 1] = -1.0 * adj_src[::-1]
np.savetxt(adjoint_src_filename, to_write)
else:
raise NotImplementedError(
"Adjoint source writing for solver '%s' not yet "
"implemented." % iteration.solver_settings["solver"])
if not adjoint_source_stations:
print("Could not create a single adjoint source.")
return
if "ses3d" in solver:
with open(os.path.join(output_folder, "ad_srcfile"), "wt") as fh:
fh.write("%i\n" % len(adjoint_source_stations))
for line in ses3d_all_coordinates:
fh.write("%.6f %.6f %.6f\n" % (line[0], line[1], line[2]))
fh.write("\n")
elif "specfem" in solver:
adjoint_source_stations = sorted(list(adjoint_source_stations))
with open(os.path.join(output_folder, "STATIONS_ADJOINT"),
"wt") as fh:
for station in adjoint_source_stations:
coords = self.comm.query.get_coordinates_for_station(
event_name, station)
fh.write("{sta} {net} {lat} {lng} {ele} {dep}\n".format(
sta=station.split(".")[1],
net=station.split(".")[0],
lat=coords["latitude"],
lng=coords["longitude"],
ele=coords["elevation_in_m"],
dep=coords["local_depth_in_m"]))
print("Wrote adjoint sources for %i station(s) to %s." %
(len(adjoint_source_stations), os.path.relpath(output_folder)))
def get_value(self):
station_id, coordinates = self.items[self.current_index]
data = Stream()
# Now get the actual waveform files. Also find the
# corresponding station file and check the coordinates.
this_waveforms = {_i["channel_id"]: _i for _i in waveforms
if _i["channel_id"].startswith(station_id + ".")}
marked_for_deletion = []
for key, value in this_waveforms.iteritems():
value["trace"] = read(value["filename"])[0]
data += value["trace"]
value["station_file"] = \
station_cache.get_station_filename(
value["channel_id"],
UTCDateTime(value["starttime_timestamp"]))
if value["station_file"] is None:
marked_for_deletion.append(key)
msg = ("Warning: Data and station information for '%s'"
" is available, but the station information "
"only for the wrong timestamp. You should try "
"and retrieve the correct station file.")
warnings.warn(msg % value["channel_id"])
continue
data[-1].stats.station_file = value["station_file"]
for key in marked_for_deletion:
del this_waveforms[key]
if not this_waveforms:
msg = "Could not retrieve data for station '%s'." % \
station_id
warnings.warn(msg)
return None
# Now attempt to get the synthetics.
synthetics_filenames = []
for name, path in synthetic_files.iteritems():
if (station_id + ".") in name:
synthetics_filenames.append(path)
if len(synthetics_filenames) != 3:
msg = "Found %i not 3 synthetics for station '%s'." % (
len(synthetics_filenames), station_id)
warnings.warn(msg)
return None
synthetics = Stream()
# Read all synthetics.
for filename in synthetics_filenames:
synthetics += read(filename)
for synth in synthetics:
if synth.stats.channel in ["X", "Z"]:
synth.data *= -1.0
synth.stats.channel = SYNTH_MAPPING[synth.stats.channel]
synth.stats.starttime = event_info["origin_time"]
# Process the data.
len_synth = synthetics[0].stats.endtime - \
synthetics[0].stats.starttime
data.trim(synthetics[0].stats.starttime - len_synth * 0.05,
synthetics[0].stats.endtime + len_synth * 0.05)
if data:
max_length = max([tr.stats.npts for tr in data])
else:
max_length = 0
if max_length == 0:
msg = ("Warning: After trimming the waveform data to "
"the time window of the synthetics, no more data is "
"left. The reference time is the one given in the "
"QuakeML file. Make sure it is correct and that "
"the waveform data actually contains data in that "
"time span.")
warnings.warn(msg)
data.detrend("linear")
data.taper()
new_time_array = np.linspace(
synthetics[0].stats.starttime.timestamp,
synthetics[0].stats.endtime.timestamp,
synthetics[0].stats.npts)
# Simulate the traces.
for trace in data:
# Decimate in case there is a large difference between
# synthetic sampling rate and sampling_rate of the data.
# XXX: Ugly filter, change!
if trace.stats.sampling_rate > (6 *
synth.stats.sampling_rate):
new_nyquist = trace.stats.sampling_rate / 2.0 / 5.0
trace.filter("lowpass", freq=new_nyquist, corners=4,
zerophase=True)
trace.decimate(factor=5, no_filter=None)
station_file = trace.stats.station_file
if "/SEED/" in station_file:
paz = Parser(station_file).getPAZ(trace.id,
trace.stats.starttime)
trace.simulate(paz_remove=paz)
elif "/RESP/" in station_file:
trace.simulate(seedresp={"filename": station_file,
"units": "VEL", "date": trace.stats.starttime})
else:
raise NotImplementedError
# Make sure that the data array is at least as long as the
# synthetics array. Also add some buffer sample for the
# spline interpolation to work in any case.
buf = synth.stats.delta * 5
if synth.stats.starttime < (trace.stats.starttime + buf):
trace.trim(starttime=synth.stats.starttime - buf,
pad=True, fill_value=0.0)
if synth.stats.endtime > (trace.stats.endtime - buf):
trace.trim(endtime=synth.stats.endtime + buf, pad=True,
fill_value=0.0)
old_time_array = np.linspace(
trace.stats.starttime.timestamp,
trace.stats.endtime.timestamp,
trace.stats.npts)
# Interpolation.
trace.data = interp1d(old_time_array, trace.data,
kind=1)(new_time_array)
trace.stats.starttime = synthetics[0].stats.starttime
trace.stats.sampling_rate = \
synthetics[0].stats.sampling_rate
data.filter("bandpass", freqmin=lowpass, freqmax=highpass)
synthetics.filter("bandpass", freqmin=lowpass,
freqmax=highpass)
# Rotate the synthetics if nessesary.
if self.rot_angle:
# First rotate the station back to see, where it was
# recorded.
lat, lng = rotations.rotate_lat_lon(
coordinates["latitude"], coordinates["longitude"],
self.rot_axis, -self.rot_angle)
# Rotate the data.
n_trace = synthetics.select(component="N")[0]
e_trace = synthetics.select(component="E")[0]
z_trace = synthetics.select(component="Z")[0]
n, e, z = rotations.rotate_data(n_trace.data, e_trace.data,
z_trace.data, lat, lng, self.rot_axis, self.rot_angle)
n_trace.data = n
e_trace.data = e
z_trace.data = z
return {"data": data, "synthetics": synthetics,
"coordinates": coordinates}
def plot_raydensity(map_object, station_events, domain):
"""
Create a ray-density plot for all events and all stations.
This function is potentially expensive and will use all CPUs available.
Does require geographiclib to be installed.
"""
import ctypes as C
from lasif import rotations
from lasif.domain import RectangularSphericalSection
from lasif.tools.great_circle_binner import GreatCircleBinner
from lasif.utils import Point
import multiprocessing
import progressbar
from scipy.stats import scoreatpercentile
if not isinstance(domain, RectangularSphericalSection):
raise NotImplementedError(
"Raydensity currently only implemented for rectangular domains. "
"Should be easy to implement for other domains. Let me know.")
# Merge everything so that a list with coordinate pairs is created. This
# list is then distributed among all processors.
station_event_list = []
for event, stations in station_events:
if domain.rotation_angle_in_degree:
# Rotate point to the non-rotated domain.
e_point = Point(*rotations.rotate_lat_lon(
event["latitude"], event["longitude"], domain.rotation_axis,
-1.0 * domain.rotation_angle_in_degree))
else:
e_point = Point(event["latitude"], event["longitude"])
for station in stations.itervalues():
# Rotate point to the non-rotated domain if necessary.
if domain.rotation_angle_in_degree:
p = Point(*rotations.rotate_lat_lon(
station["latitude"], station["longitude"],
domain.rotation_axis, -1.0 *
domain.rotation_angle_in_degree))
else:
p = Point(station["latitude"], station["longitude"])
station_event_list.append((e_point, p))
circle_count = len(station_event_list)
# The granularity of the latitude/longitude discretization for the
# raypaths. Attempt to get a somewhat meaningful result in any case.
lat_lng_count = 1000
if circle_count < 1000:
lat_lng_count = 1000
if circle_count < 10000:
lat_lng_count = 2000
else:
lat_lng_count = 3000
cpu_count = multiprocessing.cpu_count()
def to_numpy(raw_array, dtype, shape):
data = np.frombuffer(raw_array.get_obj())
data.dtype = dtype
return data.reshape(shape)
print "\nLaunching %i greatcircle calculations on %i CPUs..." % \
(circle_count, cpu_count)
widgets = [
"Progress: ",
progressbar.Percentage(),
progressbar.Bar(), "",
progressbar.ETA()
]
pbar = progressbar.ProgressBar(widgets=widgets,
maxval=circle_count).start()
def great_circle_binning(sta_evs, bin_data_buffer, bin_data_shape, lock,
counter):
new_bins = GreatCircleBinner(domain.min_latitude, domain.max_latitude,
lat_lng_count, domain.min_longitude,
domain.max_longitude, lat_lng_count)
for event, station in sta_evs:
with lock:
counter.value += 1
if not counter.value % 25:
pbar.update(counter.value)
new_bins.add_greatcircle(event, station)
bin_data = to_numpy(bin_data_buffer, np.uint32, bin_data_shape)
with bin_data_buffer.get_lock():
bin_data += new_bins.bins
# Split the data in cpu_count parts.
def chunk(seq, num):
avg = len(seq) / float(num)
out = []
last = 0.0
while last < len(seq):
out.append(seq[int(last):int(last + avg)])
last += avg
return out
chunks = chunk(station_event_list, cpu_count)
# One instance that collects everything.
collected_bins = GreatCircleBinner(domain.min_latitude,
domain.max_latitude, lat_lng_count,
domain.min_longitude,
domain.max_longitude, lat_lng_count)
# Use a multiprocessing shared memory array and map it to a numpy view.
collected_bins_data = multiprocessing.Array(C.c_uint32,
collected_bins.bins.size)
collected_bins.bins = to_numpy(collected_bins_data, np.uint32,
collected_bins.bins.shape)
# Create, launch and join one process per CPU. Use a shared value as a
# counter and a lock to avoid race conditions.
processes = []
lock = multiprocessing.Lock()
counter = multiprocessing.Value("i", 0)
for _i in xrange(cpu_count):
processes.append(
multiprocessing.Process(target=great_circle_binning,
args=(chunks[_i], collected_bins_data,
collected_bins.bins.shape, lock,
counter)))
for process in processes:
process.start()
for process in processes:
process.join()
pbar.finish()
stations = chain.from_iterable(
(_i[1].values() for _i in station_events if _i[1]))
# Remove duplicates
stations = [(_i["latitude"], _i["longitude"]) for _i in stations]
stations = set(stations)
title = "%i Events, %i unique raypaths, "\
"%i unique stations" % (len(station_events), circle_count,
len(stations))
plt.title(title, size="xx-large")
data = collected_bins.bins.transpose()
if data.max() >= 10:
data = np.log10(np.clip(data, a_min=0.5, a_max=data.max()))
data[data >= 0.0] += 0.1
data[data < 0.0] = 0.0
max_val = scoreatpercentile(data.ravel(), 99)
else:
max_val = data.max()
cmap = cm.get_cmap("gist_heat")
cmap._init()
cmap._lut[:120, -1] = np.linspace(0, 1.0, 120)**2
# Slightly change the appearance of the map so it suits the rays.
map_object.fillcontinents(color='#dddddd', lake_color='#dddddd', zorder=0)
lngs, lats = collected_bins.coordinates
# Rotate back if necessary!
if domain.rotation_angle_in_degree:
for lat, lng in zip(lats, lngs):
lat[:], lng[:] = rotations.rotate_lat_lon(
lat, lng, domain.rotation_axis,
domain.rotation_angle_in_degree)
ln, la = map_object(lngs, lats)
map_object.pcolormesh(ln, la, data, cmap=cmap, vmin=0, vmax=max_val)
# Draw the coastlines so they appear over the rays. Otherwise things are
# sometimes hard to see.
map_object.drawcoastlines()
map_object.drawcountries(linewidth=0.2)
def _plot_hdf5_model_horizontal(f, component, output_filename,
vmin=None, vmax=None):
import matplotlib.cm
import matplotlib.pylab as plt
data = xarray.DataArray(
f["data"][component][:], [
("latitude", 90.0 - f["coordinate_0"][:]),
("longitude", f["coordinate_1"][:]),
("radius", f["coordinate_2"][:] / 1000.0)])
plt.style.use('seaborn-pastel')
from lasif.domain import RectangularSphericalSection
domain = RectangularSphericalSection(**dict(f["_meta"]["domain"].attrs))
plt.figure(figsize=(32, 18))
depth_position_map = {
50: (0, 0),
100: (0, 1),
150: (1, 0),
250: (1, 1),
400: (2, 0),
600: (2, 1)
}
for depth, location in depth_position_map.items():
ax = plt.subplot2grid((3, 5), location)
radius = 6371.0 - depth
# set up a map and colourmap
m = domain.plot(ax=ax, resolution="c", skip_map_features=True)
import lasif.colors
my_colormap = lasif.colors.get_colormap(
"tomo_full_scale_linear_lightness")
from lasif import rotations
x, y = np.meshgrid(data.longitude, data.latitude)
x_shape = x.shape
y_shape = y.shape
lat_r, lon_r = rotations.rotate_lat_lon(
y.ravel(), x.ravel(),
domain.rotation_axis,
domain.rotation_angle_in_degree)
x, y = m(lon_r, lat_r)
x.shape = x_shape
y.shape = y_shape
plot_data = data.sel(radius=radius, method="nearest")
plot_data = np.ma.masked_invalid(plot_data.data)
# Overwrite colormap things if given.
if vmin is not None and vmax is not None:
min_val_plot = vmin
max_val_plot = vmax
else:
mean = plot_data.mean()
max_diff = max(abs(mean - plot_data.min()),
abs(plot_data.max() - mean))
min_val_plot = mean - max_diff
max_val_plot = mean + max_diff
# Plotting essentially constant models.
min_delta = 0.001 * abs(max_val_plot)
if (max_val_plot - min_val_plot) < min_delta:
max_val_plot = max_val_plot + min_delta
min_val_plot = min_val_plot - min_delta
# Plot.
im = m.pcolormesh(
x, y, plot_data,
cmap=my_colormap, vmin=min_val_plot, vmax=max_val_plot,
shading="gouraud")
# make a colorbar and title
m.colorbar(im, "right", size="3%", pad='2%')
plt.title(str(depth) + ' km')
# Depth based statistics.
plt.subplot2grid((3, 5), (0, 4), rowspan=3)
plt.title("Depth statistics")
mean = data.mean(axis=(0, 1))
std = data.std(axis=(0, 1))
_min = data.min(axis=(0, 1))
_max = data.max(axis=(0, 1))
plt.fill_betweenx(data.radius, mean - std, mean + std,
label="std", color="#FF3C83")
plt.plot(mean, data.radius, label="mean", color="k", lw=2)
plt.plot(_min, data.radius, color="grey", label="min")
plt.plot(_max, data.radius, color="grey", label="max")
plt.legend(loc="best")
plt.xlabel("Value")
plt.ylabel("Radius")
plt.hlines(data.radius, plt.xlim()[0], plt.xlim()[1], color="0.8",
zorder=-10, linewidth=0.5)
# Roughness plots.
plt.subplot2grid((3, 5), (0, 2))
_d = np.abs(data.diff("latitude", n=1)).sum("latitude").data
plt.title("Roughness in latitude direction, Total: %g" % _d.sum())
plt.pcolormesh(data.longitude.data, data.radius.data,
_d.T, cmap=matplotlib.cm.viridis)
try:
plt.colorbar()
except:
pass
plt.xlabel("Longitude")
plt.ylabel("Radius")
plt.subplot2grid((3, 5), (1, 2))
_d = np.abs(data.diff("longitude", n=1)).sum("longitude").data
plt.title("Roughness in longitude direction. Total: %g" % data.sum())
plt.pcolormesh(data.latitude.data, data.radius.data, _d.T,
cmap=matplotlib.cm.viridis)
try:
plt.colorbar()
except:
pass
plt.xlabel("Latitude")
plt.ylabel("Radius")
plt.subplot2grid((3, 5), (2, 2))
_d = np.abs(data.diff("radius", n=1)).sum("radius").data
plt.title("Roughness in radius direction. Total: %g" % _d.sum())
plt.pcolormesh(data.longitude.data, data.latitude.data,
_d, cmap=matplotlib.cm.viridis)
try:
plt.colorbar()
except:
pass
plt.xlabel("Longitude")
plt.ylabel("Latitude")
# L2
plt.subplot2grid((3, 5), (0, 3))
_d = (data ** 2).sum("latitude").data
plt.title("L2 Norm in latitude direction, Total: %g" % _d.sum())
plt.pcolormesh(data.longitude.data, data.radius.data,
_d.T, cmap=matplotlib.cm.viridis)
try:
plt.colorbar()
except:
pass
plt.xlabel("Longitude")
plt.ylabel("Radius")
plt.subplot2grid((3, 5), (1, 3))
_d = (data ** 2).sum("longitude").data
plt.title("L2 Norm in longitude direction, Total: %g" % _d.sum())
plt.pcolormesh(data.latitude.data, data.radius.data, _d.T,
cmap=matplotlib.cm.viridis)
try:
plt.colorbar()
except:
pass
plt.xlabel("Latitude")
plt.ylabel("Radius")
plt.subplot2grid((3, 5), (2, 3))
_d = (data ** 2).sum("radius").data
plt.title("L2 Norm in radius direction, Total: %g" % _d.sum())
plt.pcolormesh(data.longitude.data, data.latitude.data,
_d, cmap=matplotlib.cm.viridis)
try:
plt.colorbar()
except:
pass
plt.xlabel("Longitude")
plt.ylabel("Latitude")
plt.suptitle("Component %s - File %s" % (component, output_filename),
fontsize=20)
plt.tight_layout(rect=(0, 0, 1, 0.95))
plt.savefig(output_filename, dpi=150)
plt.close()
def get_waveforms_synthetic(self, event_name, station_id,
long_iteration_name):
"""
Gets the synthetic waveforms for the given event and station as a
:class:`~obspy.core.stream.Stream` object.
:param event_name: The name of the event.
:param station_id: The id of the station in the form ``NET.STA``.
:param long_iteration_name: The long form of an iteration name.
"""
from lasif import rotations
import lasif.domain
iteration = self.comm.iterations.get(long_iteration_name)
st = self._get_waveforms(event_name, station_id,
data_type="synthetic",
tag_or_iteration=iteration.long_name)
network, station = station_id.split(".")
formats = list(set([tr.stats._format for tr in st]))
if len(formats) != 1:
raise ValueError(
"The synthetics for one Earthquake must all have the same "
"data format under the assumption that they all originate "
"from the same solver. Found formats: %s" % (str(formats)))
format = formats[0].lower()
# In the case of data coming from SES3D the components must be
# mapped to ZNE as it works in XYZ.
if format == "ses3d":
# This maps the synthetic channels to ZNE.
synthetic_coordinates_mapping = {"X": "N", "Y": "E", "Z": "Z"}
for tr in st:
tr.stats.network = network
tr.stats.station = station
# SES3D X points south. Reverse it to arrive at ZNE.
if tr.stats.channel in ["X"]:
tr.data *= -1.0
# SES3D files have no starttime. Set to the event time.
tr.stats.starttime = \
self.comm.events.get(event_name)["origin_time"]
tr.stats.channel = \
synthetic_coordinates_mapping[tr.stats.channel]
# Rotate if needed. Again only SES3D synthetics need to be rotated.
domain = self.comm.project.domain
if isinstance(domain, lasif.domain.RectangularSphericalSection) \
and domain.rotation_angle_in_degree and \
"ses3d" in iteration.solver_settings["solver"].lower():
# Coordinates are required for the rotation.
coordinates = self.comm.query.get_coordinates_for_station(
event_name, station_id)
# First rotate the station back to see, where it was
# recorded.
lat, lng = rotations.rotate_lat_lon(
lat=coordinates["latitude"], lon=coordinates["longitude"],
rotation_axis=domain.rotation_axis,
angle=-domain.rotation_angle_in_degree)
# Rotate the synthetics.
n, e, z = rotations.rotate_data(
st.select(channel="N")[0].data,
st.select(channel="E")[0].data,
st.select(channel="Z")[0].data,
lat, lng,
domain.rotation_axis,
domain.rotation_angle_in_degree)
st.select(channel="N")[0].data = n
st.select(channel="E")[0].data = e
st.select(channel="Z")[0].data = z
st.sort()
# Apply the project function that modifies synthetics on the fly.
fct = self.comm.project.get_project_function("process_synthetics")
return fct(st, iteration=iteration,
event=self.comm.events.get(event_name))
def par2quakeml(Par_filename,
QuakeML_filename,
rotation_axis=[0.0, 1.0, 0.0],
rotation_angle=-57.5,
origin_time="2000-01-01 00:00:00.0",
event_type="other event"):
# initialise event
ev = Event()
# open and read Par file
fid = open(Par_filename, 'r')
fid.readline()
fid.readline()
fid.readline()
fid.readline()
lat_old = 90.0 - float(fid.readline().strip().split()[0])
lon_old = float(fid.readline().strip().split()[0])
depth = float(fid.readline().strip().split()[0])
fid.readline()
Mtt_old = float(fid.readline().strip().split()[0])
Mpp_old = float(fid.readline().strip().split()[0])
Mrr_old = float(fid.readline().strip().split()[0])
Mtp_old = float(fid.readline().strip().split()[0])
Mtr_old = float(fid.readline().strip().split()[0])
Mpr_old = float(fid.readline().strip().split()[0])
# rotate event into physical domain
lat, lon = rot.rotate_lat_lon(lat_old, lon_old, rotation_axis,
rotation_angle)
Mrr, Mtt, Mpp, Mtr, Mpr, Mtp = rot.rotate_moment_tensor(
Mrr_old, Mtt_old, Mpp_old, Mtr_old, Mpr_old, Mtp_old, lat_old, lon_old,
rotation_axis, rotation_angle)
# populate event origin data
ev.event_type = event_type
ev_origin = Origin()
ev_origin.time = UTCDateTime(origin_time)
ev_origin.latitude = lat
ev_origin.longitude = lon
ev_origin.depth = depth
ev.origins.append(ev_origin)
# populte event moment tensor
ev_tensor = Tensor()
ev_tensor.m_rr = Mrr
ev_tensor.m_tt = Mtt
ev_tensor.m_pp = Mpp
ev_tensor.m_rt = Mtr
ev_tensor.m_rp = Mpr
ev_tensor.m_tp = Mtp
ev_momenttensor = MomentTensor()
ev_momenttensor.tensor = ev_tensor
ev_momenttensor.scalar_moment = np.sqrt(Mrr**2 + Mtt**2 + Mpp**2 + Mtr**2 +
Mpr**2 + Mtp**2)
ev_focalmechanism = FocalMechanism()
ev_focalmechanism.moment_tensor = ev_momenttensor
ev_focalmechanism.nodal_planes = NodalPlanes().setdefault(0, 0)
ev.focal_mechanisms.append(ev_focalmechanism)
# populate event magnitude
ev_magnitude = Magnitude()
ev_magnitude.mag = 0.667 * (np.log10(ev_momenttensor.scalar_moment) - 9.1)
ev_magnitude.magnitude_type = 'Mw'
ev.magnitudes.append(ev_magnitude)
# write QuakeML file
cat = Catalog()
cat.append(ev)
cat.write(QuakeML_filename, format="quakeml")
# clean up
fid.close()
def finalize_adjoint_sources(self, iteration_name, event_name):
"""
Finalizes the adjoint sources.
"""
from itertools import izip
import numpy as np
from lasif import rotations
window_manager = self.comm.windows.get(event_name, iteration_name)
event = self.comm.events.get(event_name)
iteration = self.comm.iterations.get(iteration_name)
iteration_event_def = iteration.events[event["event_name"]]
iteration_stations = iteration_event_def["stations"]
# For now assume that the adjoint sources have the same
# sampling rate as the synthetics which in LASIF's workflow
# actually has to be true.
dt = iteration.get_process_params()["dt"]
# Current domain and solver.
domain = self.comm.project.domain
solver = iteration.solver_settings["solver"].lower()
adjoint_source_stations = set()
if "ses3d" in solver:
ses3d_all_coordinates = []
event_weight = iteration_event_def["event_weight"]
output_folder = self.comm.project.get_output_folder(
type="adjoint_sources",
tag="ITERATION_%s__%s" % (iteration_name, event_name))
l = sorted(window_manager.list())
for station, windows in itertools.groupby(
l, key=lambda x: ".".join(x.split(".")[:2])):
if station not in iteration_stations:
continue
print ".",
station_weight = iteration_stations[station]["station_weight"]
channels = {}
try:
for w in windows:
w = window_manager.get(w)
channel_weight = 0
srcs = []
for window in w:
ad_src = window.adjoint_source
if not ad_src["adjoint_source"].ptp():
continue
srcs.append(ad_src["adjoint_source"] * window.weight)
channel_weight += window.weight
if not srcs:
continue
# Final adjoint source for that channel and apply all
# weights.
adjoint_source = np.sum(srcs, axis=0) / channel_weight * \
event_weight * station_weight
channels[w.channel_id[-1]] = adjoint_source
except LASIFError as e:
print("Could not calculate adjoint source for iteration %s "
"and station %s. Repick windows? Reason: %s" % (
iteration.name, station, str(e)))
continue
if not channels:
continue
# Now all adjoint sources of a window should have the same length.
length = set(len(v) for v in channels.values())
assert len(length) == 1
length = length.pop()
# All missing channels will be replaced with a zero array.
for c in ["Z", "N", "E"]:
if c in channels:
continue
channels[c] = np.zeros(length)
# Get the station coordinates
coords = self.comm.query.get_coordinates_for_station(event_name,
station)
# Rotate. if needed
rec_lat = coords["latitude"]
rec_lng = coords["longitude"]
# The adjoint sources depend on the solver.
if "ses3d" in solver:
# Rotate if needed.
if domain.rotation_angle_in_degree:
# Rotate the adjoint source location.
r_rec_lat, r_rec_lng = rotations.rotate_lat_lon(
rec_lat, rec_lng, domain.rotation_axis,
-domain.rotation_angle_in_degree)
# Rotate the adjoint sources.
channels["N"], channels["E"], channels["Z"] = \
rotations.rotate_data(
channels["N"], channels["E"],
channels["Z"], rec_lat, rec_lng,
domain.rotation_axis,
-domain.rotation_angle_in_degree)
else:
r_rec_lat = rec_lat
r_rec_lng = rec_lng
r_rec_depth = 0.0
r_rec_colat = rotations.lat2colat(r_rec_lat)
# Now once again map from ZNE to the XYZ of SES3D.
CHANNEL_MAPPING = {"X": "N", "Y": "E", "Z": "Z"}
adjoint_source_stations.add(station)
adjoint_src_filename = os.path.join(
output_folder, "ad_src_%i" % len(adjoint_source_stations))
ses3d_all_coordinates.append(
(r_rec_colat, r_rec_lng, r_rec_depth))
# Actually write the adjoint source file in SES3D specific
# format.
with open(adjoint_src_filename, "wt") as open_file:
open_file.write("-- adjoint source ------------------\n")
open_file.write(
"-- source coordinates (colat,lon,depth)\n")
open_file.write("%f %f %f\n" % (r_rec_colat, r_rec_lng,
r_rec_depth))
open_file.write("-- source time function (x, y, z) --\n")
# Revert the X component as it has to point south in SES3D.
for x, y, z in izip(-1.0 * channels[CHANNEL_MAPPING["X"]],
channels[CHANNEL_MAPPING["Y"]],
channels[CHANNEL_MAPPING["Z"]]):
open_file.write("%e %e %e\n" % (x, y, z))
open_file.write("\n")
elif "specfem" in solver:
s_set = iteration.solver_settings["solver_settings"]
if "adjoint_source_time_shift" not in s_set:
warnings.warn("No <adjoint_source_time_shift> tag in the "
"iteration XML file. No time shift for the "
"adjoint sources will be applied.",
LASIFWarning)
src_time_shift = 0
else:
src_time_shift = float(s_set["adjoint_source_time_shift"])
adjoint_source_stations.add(station)
# Write all components. The adjoint sources right now are
# not time shifted.
for component in ["Z", "N", "E"]:
# XXX: M band code could be different.
adjoint_src_filename = os.path.join(
output_folder, "%s.MX%s.adj" % (station, component))
adj_src = channels[component]
l = len(adj_src)
to_write = np.empty((l, 2))
to_write[:, 0] = \
np.linspace(0, (l - 1) * dt, l) + src_time_shift
# SPECFEM expects non-time reversed adjoint sources and
# the sign is different for some reason.
to_write[:, 1] = -1.0 * adj_src[::-1]
np.savetxt(adjoint_src_filename, to_write)
else:
raise NotImplementedError(
"Adjoint source writing for solver '%s' not yet "
"implemented." % iteration.solver_settings["solver"])
if not adjoint_source_stations:
print("Could not create a single adjoint source.")
return
if "ses3d" in solver:
with open(os.path.join(output_folder, "ad_srcfile"), "wt") as fh:
fh.write("%i\n" % len(adjoint_source_stations))
for line in ses3d_all_coordinates:
fh.write("%.6f %.6f %.6f\n" % (line[0], line[1], line[2]))
fh.write("\n")
elif "specfem" in solver:
adjoint_source_stations = sorted(list(adjoint_source_stations))
with open(os.path.join(output_folder, "STATIONS_ADJOINT"),
"wt") as fh:
for station in adjoint_source_stations:
coords = self.comm.query.get_coordinates_for_station(
event_name, station)
fh.write("{sta} {net} {lat} {lng} {ele} {dep}\n".format(
sta=station.split(".")[1],
net=station.split(".")[0],
lat=coords["latitude"],
lng=coords["longitude"],
ele=coords["elevation_in_m"],
dep=coords["local_depth_in_m"]))
print "Wrote adjoint sources for %i station(s) to %s." % (
len(adjoint_source_stations), os.path.relpath(output_folder))
def plot_depth_slice(self, component, depth_in_km):
"""
Plots a depth slice.
:param component: The component to plot.
:type component: basestring
:param depth_in_km: The depth in km to plot. If the exact depth does
not exists, the nearest neighbour will be plotted.
:type depth_in_km: integer or float
"""
lat_bounds = [rotations.colat2lat(_i)
for _i in self.setup["physical_boundaries_x"][::-1]]
lng_bounds = self.setup["physical_boundaries_y"]
depth_bounds = [6371 - _i / 1000.0
for _i in self.setup["physical_boundaries_z"]]
data = self.parsed_components[component]
available_depths = np.linspace(*depth_bounds, num=data.shape[2])
depth_index = np.argmin(np.abs(available_depths - depth_in_km))
actual_depth = available_depths[depth_index]
lngs = self._get_collocation_points_along_axis(
lng_bounds[0], lng_bounds[1], data.shape[1])
lats = self._get_collocation_points_along_axis(
lat_bounds[0], lat_bounds[1], data.shape[0])
lon, lat = np.meshgrid(lngs, lats)
if self.rotation_axis and self.rotation_angle_in_degree:
lon_shape = lon.shape
lat_shape = lat.shape
lon.shape = lon.size
lat.shape = lat.size
lat, lon = rotations.rotate_lat_lon(lat, lon, self.rotation_axis,
self.rotation_angle_in_degree)
lon.shape = lon_shape
lat.shape = lat_shape
# Get the center of the map.
lon_0 = lon.min() + lon.ptp() / 2.0
lat_0 = lat.min() + lat.ptp() / 2.0
plt.figure(0)
# Attempt to zoom into the region of interest.
max_extend = max(lon.ptp(), lat.ptp())
extend_used = max_extend / 180.0
if extend_used < 0.5:
# Calculate approximate width and height in meters.
width = lon.ptp()
height = lat.ptp()
width *= 110000 * 1.1
height *= 110000 * 1.1
# Lambert azimuthal equal area projection. Equal area projections
# are useful for interpreting features and this particular one also
# does not distort features a lot on regional scales.
m = Basemap(projection='laea', resolution="l",
width=width, height=height,
lat_0=lat_0, lon_0=lon_0)
else:
m = Basemap(projection='ortho', lon_0=lon_0, lat_0=lat_0,
resolution="c")
m.drawcoastlines()
m.fillcontinents("0.9", zorder=0)
m.drawmapboundary(fill_color="white")
m.drawparallels(np.arange(-80.0, 80.0, 10.0), labels=[1, 0, 0, 0])
m.drawmeridians(np.arange(-170.0, 170.0, 10.0), labels=[0, 0, 0, 1])
m.drawcountries()
x, y = m(lon, lat)
depth_data = data[::-1, :, depth_index]
vmin, vmax = depth_data.min(), depth_data.max()
vmedian = np.median(depth_data)
offset = max(abs(vmax - vmedian), abs(vmedian - vmin))
if vmax - vmin == 0:
offset = 0.01
vmin = vmedian - offset
vmax = vmedian + offset
im = m.pcolormesh(x, y, depth_data, cmap=tomo_colormap, vmin=vmin,
vmax=vmax)
# Add colorbar and potentially unit.
cm = m.colorbar(im, "right", size="3%", pad='2%')
if component in UNIT_DICT:
cm.set_label(UNIT_DICT[component], fontsize="x-large", rotation=0)
plt.suptitle("Depth slice of %s at %i km" % (
component, int(round(actual_depth))), size="large")
border = rotations.get_border_latlng_list(
rotations.colat2lat(self.setup["physical_boundaries_x"][0]),
rotations.colat2lat(self.setup["physical_boundaries_x"][1]),
self.setup["physical_boundaries_y"][0],
self.setup["physical_boundaries_y"][1],
rotation_axis=self.rotation_axis,
rotation_angle_in_degree=self.rotation_angle_in_degree)
border = np.array(border)
lats = border[:, 0]
lngs = border[:, 1]
lngs, lats = m(lngs, lats)
m.plot(lngs, lats, color="black", lw=2, path_effects=[
PathEffects.withStroke(linewidth=4, foreground="white")])
def _on_button_press(event):
if event.button != 1 or not event.inaxes:
return
lon, lat = m(event.xdata, event.ydata, inverse=True)
# Convert to colat to ease indexing.
colat = rotations.lat2colat(lat)
x_range = (self.setup["physical_boundaries_x"][1] -
self.setup["physical_boundaries_x"][0])
x_frac = (colat - self.setup["physical_boundaries_x"][0]) / x_range
x_index = int(((self.setup["boundaries_x"][1] -
self.setup["boundaries_x"][0]) * x_frac) +
self.setup["boundaries_x"][0])
y_range = (self.setup["physical_boundaries_y"][1] -
self.setup["physical_boundaries_y"][0])
y_frac = (lon - self.setup["physical_boundaries_y"][0]) / y_range
y_index = int(((self.setup["boundaries_y"][1] -
self.setup["boundaries_y"][0]) * y_frac) +
self.setup["boundaries_y"][0])
plt.figure(1, figsize=(3, 8))
depths = available_depths
values = data[x_index, y_index, :]
plt.plot(values, depths)
plt.grid()
plt.ylim(depths[-1], depths[0])
plt.show()
plt.close()
plt.figure(0)
plt.gcf().canvas.mpl_connect('button_press_event', _on_button_press)
plt.show()
def plot_depth_slice(self, component, depth_in_km, m,
absolute_values=True):
"""
Plots a depth slice.
:param component: The component to plot.
:type component: basestring
:param depth_in_km: The depth in km to plot. If the exact depth does
not exists, the nearest neighbour will be plotted.
:type depth_in_km: integer or float
:param m: Basemap instance.
"""
depth_index = self.get_closest_gll_index("depth", depth_in_km)
# No need to do anything if the currently plotted slice is already
# plotted. This is useful for interactive use when the desired depth
# is changed but the closest GLL collocation point is still the same.
if hasattr(m, "_plotted_depth_slice"):
# Use a tuple of relevant parameters.
if m._plotted_depth_slice == (self.directory, depth_index,
component, absolute_values):
return None
data = self.parsed_components[component]
depth = self.collocation_points_depth[depth_index]
lngs = self.collocation_points_lngs
lats = self.collocation_points_lats
# Rotate data if needed.
lon, lat = np.meshgrid(lngs, lats)
if hasattr(self.domain, "rotation_axis") and \
self.domain.rotation_axis and \
self.domain.rotation_angle_in_degree:
lon_shape = lon.shape
lat_shape = lat.shape
lon.shape = lon.size
lat.shape = lat.size
lat, lon = rotations.rotate_lat_lon(
lat, lon, self.domain.rotation_axis,
self.domain.rotation_angle_in_degree)
lon.shape = lon_shape
lat.shape = lat_shape
x, y = m(lon, lat)
depth_data = data[::-1, :, depth_index]
# Plot values relative to AK135.
if not absolute_values:
cmp_map = {
"rho": "density",
"vp": "vp",
"vsh": "vs",
"vsv": "vs"
}
factor = {
"rho": 1000.0,
"vp": 1.0,
"vsh": 1.0,
"vsv": 1.0,
}
if component not in cmp_map:
vmin, vmax = depth_data.min(), depth_data.max()
vmedian = np.median(depth_data)
offset = max(abs(vmax - vmedian), abs(vmedian - vmin))
if vmax - vmin == 0:
offset = 0.01
vmin = vmedian - offset
vmax = vmedian + offset
else:
reference_value = self.one_d_model.get_value(
cmp_map[component], depth) * factor[component]
depth_data = (depth_data - reference_value) / reference_value
depth_data *= 100.0
offset = np.abs(depth_data)
try:
offset = offset[offset < 50].max()
except:
offset = offset.max()
vmin = -offset
vmax = offset
else:
vmin, vmax = depth_data.min(), depth_data.max()
vmedian = np.median(depth_data)
offset = max(abs(vmax - vmedian), abs(vmedian - vmin))
min_delta = abs(vmax * 0.005)
if (vmax - vmin) < min_delta:
offset = min_delta
vmin = vmedian - offset
vmax = vmedian + offset
# Remove an existing pcolormesh if it exists. This does not hurt in
# any case but is useful for interactive use.
if hasattr(m, "_depth_slice"):
m._depth_slice.remove()
del m._depth_slice
im = m.pcolormesh(x, y, depth_data, cmap=tomo_colormap, vmin=vmin,
vmax=vmax)
m._depth_slice = im
# Store what is currently plotted.
m._plotted_depth_slice = (self.directory, depth_index, component,
absolute_values)
return {
"depth": depth,
"mesh": im,
"data": depth_data
}
def generate_input_files(self, iteration_name, event_name,
simulation_type):
"""
Generate the input files for one event.
:param iteration_name: The name of the iteration.
:param event_name: The name of the event for which to generate the
input files.
:param simulate_type: The type of simulation to perform. Possible
values are: 'normal simulate', 'adjoint forward', 'adjoint
reverse'
"""
from wfs_input_generator import InputFileGenerator
# =====================================================================
# read iteration xml file, get event and list of stations
# =====================================================================
iteration = self.comm.iterations.get(iteration_name)
# Check that the event is part of the iterations.
if event_name not in iteration.events:
msg = ("Event '%s' not part of iteration '%s'.\nEvents available "
"in iteration:\n\t%s" %
(event_name, iteration_name, "\n\t".join(
sorted(iteration.events.keys()))))
raise ValueError(msg)
event = self.comm.events.get(event_name)
stations_for_event = iteration.events[event_name]["stations"].keys()
# Get all stations and create a dictionary for the input file
# generator.
stations = self.comm.query.get_all_stations_for_event(event_name)
stations = [{"id": key, "latitude": value["latitude"],
"longitude": value["longitude"],
"elevation_in_m": value["elevation_in_m"],
"local_depth_in_m": value["local_depth_in_m"]}
for key, value in stations.iteritems()
if key in stations_for_event]
# =====================================================================
# set solver options
# =====================================================================
solver = iteration.solver_settings
# Currently only SES3D 4.1 is supported
solver_format = solver["solver"].lower()
if solver_format not in ["ses3d 4.1", "ses3d 2.0",
"specfem3d cartesian", "specfem3d globe cem"]:
msg = ("Currently only SES3D 4.1, SES3D 2.0, SPECFEM3D "
"CARTESIAN, and SPECFEM3D GLOBE CEM are supported.")
raise ValueError(msg)
solver_format = solver_format.replace(' ', '_')
solver_format = solver_format.replace('.', '_')
solver = solver["solver_settings"]
# =====================================================================
# create the input file generator, add event and stations,
# populate the configuration items
# =====================================================================
# Add the event and the stations to the input file generator.
gen = InputFileGenerator()
gen.add_events(event["filename"])
gen.add_stations(stations)
if solver_format in ["ses3d_4_1", "ses3d_2_0"]:
# event tag
gen.config.event_tag = event_name
# Time configuration.
npts = solver["simulation_parameters"]["number_of_time_steps"]
delta = solver["simulation_parameters"]["time_increment"]
gen.config.number_of_time_steps = npts
gen.config.time_increment_in_s = delta
# SES3D specific configuration
gen.config.output_folder = solver["output_directory"].replace(
"{{EVENT_NAME}}", event_name.replace(" ", "_"))
gen.config.simulate_type = simulation_type
gen.config.adjoint_forward_wavefield_output_folder = \
solver["adjoint_output_parameters"][
"forward_field_output_directory"].replace(
"{{EVENT_NAME}}", event_name.replace(" ", "_"))
gen.config.adjoint_forward_sampling_rate = \
solver["adjoint_output_parameters"][
"sampling_rate_of_forward_field"]
# Visco-elastic dissipation
diss = solver["simulation_parameters"]["is_dissipative"]
gen.config.is_dissipative = diss
# Only SES3D 4.1 has the relaxation parameters.
if solver_format == "ses3d_4_1":
gen.config.Q_model_relaxation_times = \
solver["relaxation_parameter_list"]["tau"]
gen.config.Q_model_weights_of_relaxation_mechanisms = \
solver["relaxation_parameter_list"]["w"]
# Discretization
disc = solver["computational_setup"]
gen.config.nx_global = disc["nx_global"]
gen.config.ny_global = disc["ny_global"]
gen.config.nz_global = disc["nz_global"]
gen.config.px = disc["px_processors_in_theta_direction"]
gen.config.py = disc["py_processors_in_phi_direction"]
gen.config.pz = disc["pz_processors_in_r_direction"]
gen.config.lagrange_polynomial_degree = \
disc["lagrange_polynomial_degree"]
# Configure the mesh.
domain = self.comm.project.domain
gen.config.mesh_min_latitude = \
domain["bounds"]["minimum_latitude"]
gen.config.mesh_max_latitude = \
domain["bounds"]["maximum_latitude"]
gen.config.mesh_min_longitude = \
domain["bounds"]["minimum_longitude"]
gen.config.mesh_max_longitude = \
domain["bounds"]["maximum_longitude"]
gen.config.mesh_min_depth_in_km = \
domain["bounds"]["minimum_depth_in_km"]
gen.config.mesh_max_depth_in_km = \
domain["bounds"]["maximum_depth_in_km"]
# Set the rotation parameters.
gen.config.rotation_angle_in_degree = domain["rotation_angle"]
gen.config.rotation_axis = domain["rotation_axis"]
# Make source time function
gen.config.source_time_function = \
iteration.get_source_time_function()["data"]
elif solver_format == "specfem3d_cartesian":
gen.config.NSTEP = \
solver["simulation_parameters"]["number_of_time_steps"]
gen.config.DT = \
solver["simulation_parameters"]["time_increment"]
gen.config.NPROC = \
solver["computational_setup"]["number_of_processors"]
if simulation_type == "normal simulation":
msg = ("'normal_simulate' not supported for SPECFEM3D "
"Cartesian. Please choose either 'adjoint_forward' or "
"'adjoint_reverse'.")
raise NotImplementedError(msg)
elif simulation_type == "adjoint forward":
gen.config.SIMULATION_TYPE = 1
elif simulation_type == "adjoint reverse":
gen.config.SIMULATION_TYPE = 2
else:
raise NotImplementedError
solver_format = solver_format.upper()
elif solver_format == "specfem3d_globe_cem":
cs = solver["computational_setup"]
gen.config.NPROC_XI = cs["number_of_processors_xi"]
gen.config.NPROC_ETA = cs["number_of_processors_eta"]
gen.config.NCHUNKS = cs["number_of_chunks"]
gen.config.NEX_XI = cs["elements_per_chunk_xi"]
gen.config.NEX_ETA = cs["elements_per_chunk_eta"]
gen.config.OCEANS = cs["simulate_oceans"]
gen.config.ELLIPTICITY = cs["simulate_ellipticity"]
gen.config.TOPOGRAPHY = cs["simulate_topography"]
gen.config.GRAVITY = cs["simulate_gravity"]
gen.config.ROTATION = cs["simulate_rotation"]
gen.config.ATTENUATION = cs["simulate_attenuation"]
gen.config.ABSORBING_CONDITIONS = True
if cs["fast_undo_attenuation"]:
gen.config.PARTIAL_PHYS_DISPERSION_ONLY = True
gen.config.UNDO_ATTENUATION = False
else:
gen.config.PARTIAL_PHYS_DISPERSION_ONLY = False
gen.config.UNDO_ATTENUATION = True
gen.config.GPU_MODE = cs["use_gpu"]
gen.config.SOURCE_TIME_FUNCTION = \
iteration.get_source_time_function()["data"]
if simulation_type == "normal simulation":
gen.config.SIMULATION_TYPE = 1
gen.config.SAVE_FORWARD = False
elif simulation_type == "adjoint forward":
gen.config.SIMULATION_TYPE = 1
gen.config.SAVE_FORWARD = True
elif simulation_type == "adjoint reverse":
gen.config.SIMULATION_TYPE = 2
gen.config.SAVE_FORWARD = True
else:
raise NotImplementedError
# Use the current domain setting to derive the bounds in the way
# SPECFEM specifies them.
domain = self.comm.project.domain
lat_range = domain["bounds"]["maximum_latitude"] - \
domain["bounds"]["minimum_latitude"]
lng_range = domain["bounds"]["maximum_longitude"] - \
domain["bounds"]["minimum_longitude"]
c_lat = \
domain["bounds"]["minimum_latitude"] + lat_range / 2.0
c_lng = \
domain["bounds"]["minimum_longitude"] + lng_range / 2.0
# Rotate the point.
c_lat_1, c_lng_1 = rotations.rotate_lat_lon(
c_lat, c_lng, domain["rotation_axis"],
domain["rotation_angle"])
# SES3D rotation.
A = rotations._get_rotation_matrix(
domain["rotation_axis"], domain["rotation_angle"])
latitude_rotation = -(c_lat_1 - c_lat)
longitude_rotation = c_lng_1 - c_lng
# Rotate the latitude. The rotation axis is latitude 0 and
# the center longitude + 90 degree
B = rotations._get_rotation_matrix(
rotations.lat_lon_radius_to_xyz(0.0, c_lng + 90, 1.0),
latitude_rotation)
# Rotate around the North pole.
C = rotations._get_rotation_matrix(
[0.0, 0.0, 1.0], longitude_rotation)
D = A * np.linalg.inv(C * B)
axis, angle = rotations._get_axis_and_angle_from_rotation_matrix(D)
rotated_axis = rotations.xyz_to_lat_lon_radius(*axis)
# Consistency check
if abs(rotated_axis[0] - c_lat_1) >= 0.01 or \
abs(rotated_axis[1] - c_lng_1) >= 0.01:
axis *= -1.0
angle *= -1.0
rotated_axis = rotations.xyz_to_lat_lon_radius(*axis)
if abs(rotated_axis[0] - c_lat_1) >= 0.01 or \
abs(rotated_axis[1] - c_lng_1) >= 0.01:
msg = "Failed to describe the domain in terms that SPECFEM " \
"understands"
raise LASIFError(msg)
gen.config.ANGULAR_WIDTH_XI_IN_DEGREES = lng_range
gen.config.ANGULAR_WIDTH_ETA_IN_DEGREES = lat_range
gen.config.CENTER_LATITUDE_IN_DEGREES = c_lat_1
gen.config.CENTER_LONGITUDE_IN_DEGREES = c_lng_1
gen.config.GAMMA_ROTATION_AZIMUTH = angle
gen.config.MODEL = cs["model"]
pp = iteration.get_process_params()
gen.config.RECORD_LENGTH_IN_MINUTES = \
(pp["npts"] * pp["dt"]) / 60.0
solver_format = solver_format.upper()
else:
msg = "Unknown solver '%s'." % solver_format
raise NotImplementedError(msg)
# =================================================================
# output
# =================================================================
output_dir = self.comm.project.get_output_folder(
"input_files___ITERATION_%s__%s__EVENT_%s" % (
iteration_name, simulation_type.replace(" ", "_"),
event_name))
gen.write(format=solver_format, output_dir=output_dir)
print "Written files to '%s'." % output_dir
def lasif_generate_dummy_data(args):
"""
Usage: lasif generate_dummy_data
Generates some random example event and waveforms. Useful for debugging,
testing, and following the tutorial.
"""
import inspect
from lasif import rotations
from lasif.adjoint_sources.utils import get_dispersed_wavetrain
import numpy as np
import obspy
if len(args):
msg = "No arguments allowed."
raise LASIFCommandLineException(msg)
proj = _find_project_root(".")
# Use a seed to make it somewhat predictable.
random.seed(34235234)
# Create 5 events.
d = proj.domain["bounds"]
b = d["boundary_width_in_degree"] * 1.5
event_count = 8
for _i in xrange(8):
lat = random.uniform(d["minimum_latitude"] + b,
d["maximum_latitude"] - b)
lon = random.uniform(d["minimum_longitude"] + b,
d["maximum_longitude"] - b)
depth_in_m = random.uniform(d["minimum_depth_in_km"],
d["maximum_depth_in_km"]) * 1000.0
# Rotate the coordinates.
lat, lon = rotations.rotate_lat_lon(lat, lon,
proj.domain["rotation_axis"],
proj.domain["rotation_angle"])
time = obspy.UTCDateTime(
random.uniform(
obspy.UTCDateTime(2008, 1, 1).timestamp,
obspy.UTCDateTime(2013, 1, 1).timestamp))
# The moment tensor. XXX: Make sensible values!
values = [
-3.3e+18, 1.43e+18, 1.87e+18, -1.43e+18, -2.69e+17, -1.77e+18
]
random.shuffle(values)
mrr = values[0]
mtt = values[1]
mpp = values[2]
mrt = values[3]
mrp = values[4]
mtp = values[5]
mag = random.uniform(5, 7)
scalar_moment = 3.661e+25
event_name = os.path.join(proj.paths["events"],
"dummy_event_%i.xml" % (_i + 1))
cat = obspy.core.event.Catalog(events=[
obspy.core.event.Event(
event_type="earthquake",
origins=[
obspy.core.event.Origin(latitude=lat,
longitude=lon,
depth=depth_in_m,
time=time)
],
magnitudes=[
obspy.core.event.Magnitude(mag=mag, magnitude_type="Mw")
],
focal_mechanisms=[
obspy.core.event.FocalMechanism(
moment_tensor=obspy.core.event.MomentTensor(
scalar_moment=scalar_moment,
tensor=obspy.core.event.Tensor(m_rr=mrr,
m_tt=mtt,
m_pp=mpp,
m_rt=mrt,
m_rp=mrp,
m_tp=mtp)))
])
])
cat.write(event_name, format="quakeml", validate=False)
print "Generated %i random events." % event_count
# Update the folder structure.
proj.update_folder_structure()
names_taken = []
def _get_random_name(length):
while True:
ret = ""
for i in xrange(length):
ret += chr(int(random.uniform(ord("A"), ord("Z"))))
if ret in names_taken:
continue
names_taken.append(ret)
break
return ret
# Now generate 30 station coordinates. Use a land-sea mask included in
# basemap and loop until thirty stations on land are found.
from mpl_toolkits.basemap import _readlsmask
from obspy.core.util.geodetics import gps2DistAzimuth
ls_lon, ls_lat, ls_mask = _readlsmask()
stations = []
# Do not use an infinite loop. One could choose a region with no land.
for i in xrange(10000):
if len(stations) >= 30:
break
lat = random.uniform(d["minimum_latitude"] + b,
d["maximum_latitude"] - b)
lon = random.uniform(d["minimum_longitude"] + b,
d["maximum_longitude"] - b)
# Rotate the coordinates.
lat, lon = rotations.rotate_lat_lon(lat, lon,
proj.domain["rotation_axis"],
proj.domain["rotation_angle"])
if not ls_mask[np.abs(ls_lat - lat).argmin()][np.abs(ls_lon -
lon).argmin()]:
continue
stations.append({
"latitude": lat,
"longitude": lon,
"network": "XX",
"station": _get_random_name(3)
})
if not len(stations):
msg = "Could not create stations. Pure ocean region?"
raise ValueError(msg)
# Create a RESP file for every channel.
resp_file_temp = os.path.join(
os.path.dirname(
os.path.abspath(inspect.getfile(inspect.currentframe()))),
os.path.pardir, "tools", "RESP.template_file")
with open(resp_file_temp, "rt") as open_file:
resp_file_template = open_file.read()
for station in stations:
for component in ["E", "N", "Z"]:
filename = os.path.join(
proj.paths["resp"], "RESP.%s.%s.%s.BE%s" %
(station["network"], station["station"], "", component))
with open(filename, "wt") as open_file:
open_file.write(
resp_file_template.format(station=station["station"],
network=station["network"],
channel="BH%s" % component))
print "Generated %i RESP files." % (30 * 3)
def _empty_sac_trace():
"""
Helper function to create and empty SAC header.
"""
sac_dict = {}
# floats = -12345.8
floats = [
"a", "mag", "az", "baz", "cmpaz", "cmpinc", "b", "depmax",
"depmen", "depmin", "dist", "e", "evdp", "evla", "evlo", "f",
"gcarc", "o", "odelta", "stdp", "stel", "stla", "stlo", "t0", "t1",
"t2", "t3", "t4", "t5", "t6", "t7", "t8", "t9", "unused10",
"unused11", "unused12", "unused6", "unused7", "unused8", "unused9",
"user0", "user1", "user2", "user3", "user4", "user5", "user6",
"user7", "user8", "user9", "xmaximum", "xminimum", "ymaximum",
"yminimum"
]
sac_dict.update({key: -12345.0 for key in floats})
# Integers: -12345
integers = [
"idep", "ievreg", "ievtype", "iftype", "iinst", "imagsrc",
"imagtyp", "iqual", "istreg", "isynth", "iztype", "lcalda",
"lovrok", "nevid", "norid", "nwfid"
]
sac_dict.update({key: -12345 for key in integers})
# Strings: "-12345 "
strings = [
"ka", "kdatrd", "kevnm", "kf", "kinst", "ko", "kt0", "kt1", "kt2",
"kt3", "kt4", "kt5", "kt6", "kt7", "kt8", "kt9", "kuser0",
"kuser1", "kuser2"
]
sac_dict.update({key: "-12345 " for key in strings})
# Header version
sac_dict["nvhdr"] = 6
# Data is evenly spaced
sac_dict["leven"] = 1
# And a positive polarity.
sac_dict["lpspol"] = 1
tr = obspy.Trace()
tr.stats.sac = obspy.core.AttribDict(sac_dict)
return tr
events = proj.get_all_events()
# Now loop over all events and create SAC file for them.
for _i, event in enumerate(events):
lat, lng = event.origins[0].latitude, event.origins[0].longitude
# Get the distance to each events.
for station in stations:
# Add some perturbations.
distance_in_km = gps2DistAzimuth(lat, lng, station["latitude"],
station["longitude"])[0] / 1000.0
a = random.uniform(3.9, 4.1)
b = random.uniform(0.9, 1.1)
c = random.uniform(0.9, 1.1)
body_wave_factor = random.uniform(0.095, 0.015)
body_wave_freq_scale = random.uniform(0.45, 0.55)
distance_in_km = random.uniform(0.99 * distance_in_km,
1.01 * distance_in_km)
_, u = get_dispersed_wavetrain(
dw=0.001,
distance=distance_in_km,
t_min=0,
t_max=900,
a=a,
b=b,
c=c,
body_wave_factor=body_wave_factor,
body_wave_freq_scale=body_wave_freq_scale)
for component in ["E", "N", "Z"]:
tr = _empty_sac_trace()
tr.data = u
tr.stats.network = station["network"]
tr.stats.station = station["station"]
tr.stats.location = ""
tr.stats.channel = "BH%s" % component
tr.stats.sac.stla = station["latitude"]
tr.stats.sac.stlo = station["longitude"]
tr.stats.sac.stdp = 0.0
tr.stats.sac.stel = 0.0
path = os.path.join(proj.paths["data"],
"dummy_event_%i" % (_i + 1), "raw")
if not os.path.exists(path):
os.makedirs(path)
tr.write(os.path.join(
path, "%s.%s..BH%s.sac" %
(station["network"], station["station"], component)),
format="sac")
print "Generated %i waveform files." % (30 * 3 * len(events))
def finalize_adjoint_sources(self, iteration_name, event_name):
"""
Finalizes the adjoint sources.
"""
from itertools import izip
import numpy as np
from lasif import rotations
all_coordinates = []
_i = 0
window_manager = self.comm.windows.get(event_name, iteration_name)
event = self.comm.events.get(event_name)
iteration = self.comm.iterations.get(iteration_name)
iteration_event_def = iteration.events[event["event_name"]]
iteration_stations = iteration_event_def["stations"]
event_weight = iteration_event_def["event_weight"]
output_folder = self.comm.project.get_output_folder(
"adjoint_sources__ITERATION_%s__%s" % (iteration_name, event_name))
l = sorted(window_manager.list())
for station, windows in itertools.groupby(
l, key=lambda x: ".".join(x.split(".")[:2])):
if station not in iteration_stations:
continue
station_weight = iteration_stations[station]["station_weight"]
channels = {}
for w in windows:
w = window_manager.get(w)
channel_weight = 0
srcs = []
for window in w:
ad_src = window.adjoint_source
if not ad_src["adjoint_source"].ptp():
continue
srcs.append(ad_src["adjoint_source"] * window.weight)
channel_weight += window.weight
if not srcs:
continue
# Final adjoint source for that channel and apply all weights.
adjoint_source = np.sum(srcs, axis=0) / channel_weight * \
event_weight * station_weight
channels[w.channel_id[-1]] = adjoint_source
if not channels:
continue
# Now all adjoint sources of a window should have the same length.
length = set(len(v) for v in channels.values())
assert len(length) == 1
length = length.pop()
# All missing channels will be replaced with a zero array.
for c in ["Z", "N", "E"]:
if c in channels:
continue
channels[c] = np.zeros(length)
# Get the station coordinates
coords = self.comm.query.get_coordinates_for_station(event_name,
station)
# Rotate. if needed
rec_lat = coords["latitude"]
rec_lng = coords["longitude"]
domain = self.comm.project.domain
if domain["rotation_angle"]:
# Rotate the adjoint source location.
r_rec_lat, r_rec_lng = rotations.rotate_lat_lon(
rec_lat, rec_lng, domain["rotation_axis"],
-domain["rotation_angle"])
# Rotate the adjoint sources.
channels["N"], channels["E"], channels["Z"] = \
rotations.rotate_data(
channels["N"], channels["E"],
channels["Z"], rec_lat, rec_lng,
domain["rotation_axis"],
-domain["rotation_angle"])
else:
r_rec_lat = rec_lat
r_rec_lng = rec_lng
r_rec_depth = 0.0
r_rec_colat = rotations.lat2colat(r_rec_lat)
CHANNEL_MAPPING = {"X": "N", "Y": "E", "Z": "Z"}
_i += 1
adjoint_src_filename = os.path.join(output_folder,
"ad_src_%i" % _i)
all_coordinates.append((r_rec_colat, r_rec_lng, r_rec_depth))
# Actually write the adjoint source file in SES3D specific format.
with open(adjoint_src_filename, "wt") as open_file:
open_file.write("-- adjoint source ------------------\n")
open_file.write("-- source coordinates (colat,lon,depth)\n")
open_file.write("%f %f %f\n" % (r_rec_colat, r_rec_lng,
r_rec_depth))
open_file.write("-- source time function (x, y, z) --\n")
for x, y, z in izip(-1.0 * channels[CHANNEL_MAPPING["X"]],
channels[CHANNEL_MAPPING["Y"]],
channels[CHANNEL_MAPPING["Z"]]):
open_file.write("%e %e %e\n" % (x, y, z))
open_file.write("\n")
# Write the final file.
with open(os.path.join(output_folder, "ad_srcfile"), "wt") as fh:
fh.write("%i\n" % _i)
for line in all_coordinates:
fh.write("%.6f %.6f %.6f\n" % (line[0], line[1], line[2]))
fh.write("\n")
print "Wrote %i adjoint sources to %s." % (
_i, os.path.relpath(output_folder))
def generate_input_files(self, iteration_name, event_name,
simulation_type):
"""
Generate the input files for one event.
:param iteration_name: The name of the iteration.
:param event_name: The name of the event for which to generate the
input files.
:param simulate_type: The type of simulation to perform. Possible
values are: 'normal simulate', 'adjoint forward', 'adjoint
reverse'
"""
from wfs_input_generator import InputFileGenerator
# =====================================================================
# read iteration xml file, get event and list of stations
# =====================================================================
iteration = self.comm.iterations.get(iteration_name)
# Check that the event is part of the iterations.
if event_name not in iteration.events:
msg = ("Event '%s' not part of iteration '%s'.\nEvents available "
"in iteration:\n\t%s" %
(event_name, iteration_name, "\n\t".join(
sorted(iteration.events.keys()))))
raise ValueError(msg)
event = self.comm.events.get(event_name)
stations_for_event = list(
iteration.events[event_name]["stations"].keys())
# Get all stations and create a dictionary for the input file
# generator.
stations = self.comm.query.get_all_stations_for_event(event_name)
stations = [{
"id": key,
"latitude": value["latitude"],
"longitude": value["longitude"],
"elevation_in_m": value["elevation_in_m"],
"local_depth_in_m": value["local_depth_in_m"]
} for key, value in stations.items() if key in stations_for_event]
# =====================================================================
# set solver options
# =====================================================================
solver = iteration.solver_settings
# Currently only SES3D 4.1 is supported
solver_format = solver["solver"].lower()
if solver_format not in [
"ses3d 4.1", "ses3d 2.0", "specfem3d cartesian",
"specfem3d globe cem"
]:
msg = ("Currently only SES3D 4.1, SES3D 2.0, SPECFEM3D "
"CARTESIAN, and SPECFEM3D GLOBE CEM are supported.")
raise ValueError(msg)
solver_format = solver_format.replace(' ', '_')
solver_format = solver_format.replace('.', '_')
solver = solver["solver_settings"]
# =====================================================================
# create the input file generator, add event and stations,
# populate the configuration items
# =====================================================================
# Add the event and the stations to the input file generator.
gen = InputFileGenerator()
gen.add_events(event["filename"])
gen.add_stations(stations)
if solver_format in ["ses3d_4_1", "ses3d_2_0"]:
# event tag
gen.config.event_tag = event_name
# Time configuration.
npts = solver["simulation_parameters"]["number_of_time_steps"]
delta = solver["simulation_parameters"]["time_increment"]
gen.config.number_of_time_steps = npts
gen.config.time_increment_in_s = delta
# SES3D specific configuration
gen.config.output_folder = solver["output_directory"].replace(
"{{EVENT_NAME}}", event_name.replace(" ", "_"))
gen.config.simulation_type = simulation_type
gen.config.adjoint_forward_wavefield_output_folder = \
solver["adjoint_output_parameters"][
"forward_field_output_directory"].replace(
"{{EVENT_NAME}}", event_name.replace(" ", "_"))
gen.config.adjoint_forward_sampling_rate = \
solver["adjoint_output_parameters"][
"sampling_rate_of_forward_field"]
# Visco-elastic dissipation
diss = solver["simulation_parameters"]["is_dissipative"]
gen.config.is_dissipative = diss
# Only SES3D 4.1 has the relaxation parameters.
if solver_format == "ses3d_4_1":
gen.config.Q_model_relaxation_times = \
solver["relaxation_parameter_list"]["tau"]
gen.config.Q_model_weights_of_relaxation_mechanisms = \
solver["relaxation_parameter_list"]["w"]
# Discretization
disc = solver["computational_setup"]
gen.config.nx_global = disc["nx_global"]
gen.config.ny_global = disc["ny_global"]
gen.config.nz_global = disc["nz_global"]
gen.config.px = disc["px_processors_in_theta_direction"]
gen.config.py = disc["py_processors_in_phi_direction"]
gen.config.pz = disc["pz_processors_in_r_direction"]
gen.config.lagrange_polynomial_degree = \
disc["lagrange_polynomial_degree"]
# Configure the mesh.
domain = self.comm.project.domain
gen.config.mesh_min_latitude = domain.min_latitude
gen.config.mesh_max_latitude = domain.max_latitude
gen.config.mesh_min_longitude = domain.min_longitude
gen.config.mesh_max_longitude = domain.max_longitude
gen.config.mesh_min_depth_in_km = domain.min_depth_in_km
gen.config.mesh_max_depth_in_km = domain.max_depth_in_km
# Set the rotation parameters.
gen.config.rotation_angle_in_degree = \
domain.rotation_angle_in_degree
gen.config.rotation_axis = domain.rotation_axis
# Make source time function
gen.config.source_time_function = \
iteration.get_source_time_function()["data"]
elif solver_format == "specfem3d_cartesian":
gen.config.NSTEP = \
solver["simulation_parameters"]["number_of_time_steps"]
gen.config.DT = \
solver["simulation_parameters"]["time_increment"]
gen.config.NPROC = \
solver["computational_setup"]["number_of_processors"]
if simulation_type == "normal simulation":
msg = ("'normal_simulate' not supported for SPECFEM3D "
"Cartesian. Please choose either 'adjoint_forward' or "
"'adjoint_reverse'.")
raise NotImplementedError(msg)
elif simulation_type == "adjoint forward":
gen.config.SIMULATION_TYPE = 1
elif simulation_type == "adjoint reverse":
gen.config.SIMULATION_TYPE = 2
else:
raise NotImplementedError
solver_format = solver_format.upper()
elif solver_format == "specfem3d_globe_cem":
cs = solver["computational_setup"]
gen.config.NPROC_XI = cs["number_of_processors_xi"]
gen.config.NPROC_ETA = cs["number_of_processors_eta"]
gen.config.NCHUNKS = cs["number_of_chunks"]
gen.config.NEX_XI = cs["elements_per_chunk_xi"]
gen.config.NEX_ETA = cs["elements_per_chunk_eta"]
gen.config.OCEANS = cs["simulate_oceans"]
gen.config.ELLIPTICITY = cs["simulate_ellipticity"]
gen.config.TOPOGRAPHY = cs["simulate_topography"]
gen.config.GRAVITY = cs["simulate_gravity"]
gen.config.ROTATION = cs["simulate_rotation"]
gen.config.ATTENUATION = cs["simulate_attenuation"]
gen.config.ABSORBING_CONDITIONS = True
if cs["fast_undo_attenuation"]:
gen.config.PARTIAL_PHYS_DISPERSION_ONLY = True
gen.config.UNDO_ATTENUATION = False
else:
gen.config.PARTIAL_PHYS_DISPERSION_ONLY = False
gen.config.UNDO_ATTENUATION = True
gen.config.GPU_MODE = cs["use_gpu"]
gen.config.SOURCE_TIME_FUNCTION = \
iteration.get_source_time_function()["data"]
if simulation_type == "normal simulation":
gen.config.SIMULATION_TYPE = 1
gen.config.SAVE_FORWARD = False
elif simulation_type == "adjoint forward":
gen.config.SIMULATION_TYPE = 1
gen.config.SAVE_FORWARD = True
elif simulation_type == "adjoint reverse":
gen.config.SIMULATION_TYPE = 2
gen.config.SAVE_FORWARD = True
else:
raise NotImplementedError
# Use the current domain setting to derive the bounds in the way
# SPECFEM specifies them.
domain = self.comm.project.domain
lat_range = domain.max_latitude - \
domain.min_latitude
lng_range = domain.max_longitude - \
domain.min_longitude
c_lat = \
domain.min_latitude + lat_range / 2.0
c_lng = \
domain.min_longitude + lng_range / 2.0
# Rotate the point.
c_lat_1, c_lng_1 = rotations.rotate_lat_lon(
c_lat, c_lng, domain.rotation_axis,
domain.rotation_angle_in_degree)
# SES3D rotation.
A = rotations._get_rotation_matrix(domain.rotation_axis,
domain.rotation_angle_in_degree)
latitude_rotation = -(c_lat_1 - c_lat)
longitude_rotation = c_lng_1 - c_lng
# Rotate the latitude. The rotation axis is latitude 0 and
# the center longitude + 90 degree
B = rotations._get_rotation_matrix(
rotations.lat_lon_radius_to_xyz(0.0, c_lng + 90, 1.0),
latitude_rotation)
# Rotate around the North pole.
C = rotations._get_rotation_matrix([0.0, 0.0, 1.0],
longitude_rotation)
D = A * np.linalg.inv(C * B)
axis, angle = rotations._get_axis_and_angle_from_rotation_matrix(D)
rotated_axis = rotations.xyz_to_lat_lon_radius(*axis)
# Consistency check
if abs(rotated_axis[0] - c_lat_1) >= 0.01 or \
abs(rotated_axis[1] - c_lng_1) >= 0.01:
axis *= -1.0
angle *= -1.0
rotated_axis = rotations.xyz_to_lat_lon_radius(*axis)
if abs(rotated_axis[0] - c_lat_1) >= 0.01 or \
abs(rotated_axis[1] - c_lng_1) >= 0.01:
msg = "Failed to describe the domain in terms that SPECFEM " \
"understands. The domain definition in the output " \
"files will NOT BE CORRECT!"
warnings.warn(msg, LASIFWarning)
gen.config.ANGULAR_WIDTH_XI_IN_DEGREES = lng_range
gen.config.ANGULAR_WIDTH_ETA_IN_DEGREES = lat_range
gen.config.CENTER_LATITUDE_IN_DEGREES = c_lat_1
gen.config.CENTER_LONGITUDE_IN_DEGREES = c_lng_1
gen.config.GAMMA_ROTATION_AZIMUTH = angle
gen.config.MODEL = cs["model"]
pp = iteration.get_process_params()
gen.config.RECORD_LENGTH_IN_MINUTES = \
(pp["npts"] * pp["dt"]) / 60.0
solver_format = solver_format.upper()
else:
msg = "Unknown solver '%s'." % solver_format
raise NotImplementedError(msg)
# =================================================================
# output
# =================================================================
output_dir = self.comm.project.get_output_folder(
type="input_files",
tag="ITERATION_%s__%s__EVENT_%s" %
(iteration_name, simulation_type.replace(" ", "_"), event_name))
gen.write(format=solver_format, output_dir=output_dir)
print("Written files to '%s'." % output_dir)
