This documentation is not for the latest stable Salvus version.
periodhere) and re-running the whole notebook could also be used to run realistic and large-scale simulations.
obspyPython libraries which can be installed either via
%matplotlib inline %config Completer.use_jedi = False import os import obspy.clients.fdsn import pyasdf from salvus.flow import api, simple_config from salvus.mesh import simple_mesh SALVUS_FLOW_SITE_NAME = os.environ.get("SITE_NAME", "local")
# Controls the dominant period of the mesh and the width # of the source time function. It is given in seconds. period = 4000.0 # We'll first build a mesh using the simple_mesh interface. m = simple_mesh.Globe3D() m.basic.min_period_in_seconds = period # At these period we don't require a crust. Adding a 3D model # is the topic of another tutorial. m.basic.model = "prem_ani_no_crust" # Higher order shapes and models better approximate the sphere. # With order 4 we achieve a very good approximation of it # even with only very few elements. m.advanced.tensor_order = 4 # In order to make it a bit more interesting we'll create an # elliptic mesh. This is the WGS84 ellipsoid. m.spherical.ellipticity = 0.0033528106647474805 # This is an important setting. The more elementes per wavelength # the more accurate the solution. 2 is a conservative value and # the default. Many global seismologist only use 1 element per # wavelength which ends up being 16 times cheaper in terms of # simulation cost but is still usable in many scenarios. m.basic.elements_per_wavelength = 2.0 # Tohuko-Oki earthquake. Information is taken from the GCMT catalog # which unfortunately does not offer a proper web service. source = simple_config.source.seismology.SideSetMomentTensorPoint3D( latitude=37.5200, longitude=143.0500, depth_in_m=20000, side_set_name="r1", mrr=1.730000e22, mtt=-2.810000e21, mpp=-1.450000e22, mrt=2.120000e22, mrp=4.550000e22, mtp=-6.570000e21, source_time_function=simple_config.stf.GaussianRate( half_duration_in_seconds=period / 2 ), ) # Download GSN stations via IRIS. _GSN is the virtual GSN network # which groups a number of actual seismic network. inv = obspy.clients.fdsn.Client("IRIS").get_stations( network="_GSN", level="station", format="text" ) # Create the simulation object and combine all the information. w = simple_config.simulation.Waveform(mesh=m.create_mesh()) # Sources and receivers will be placed exactly relative to the # local mesh surface. Please refer to the sources and receivers # documentation for more details. w.add_sources(source) w.add_receivers( simple_config.receiver.seismology.parse( inv, dimensions=3, fields=["displacement"] ) ) # Visualize it. w
<salvus.flow.simple_config.simulation.Waveform object at 0x7ff98222d690>
# We use SalvusFlow to run the simulation. The site determines # where it will run in the end. Might be the local machine, or # a large remote cluster. api.run( input_file=w, site_name=SALVUS_FLOW_SITE_NAME, output_folder="global_simulation", )
Job `job_2106221210205681_56c28dbd50` running on `local` with 4 rank(s). Site information: * Salvus version: 0.11.34-29-g8106d88f * Floating point size: 32
* Downloaded 35.1 MB of results to `global_simulation`. * Total run time: 13.85 seconds. * Pure simulation time: 12.04 seconds.
<salvus.flow.sites.salvus_job.SalvusJob at 0x7ffa0c83ced0>
# Now we'll just randomly select a waveform to plot. ds = pyasdf.ASDFDataSet("./global_simulation/receivers.h5") ds.waveforms.IU_ANMO.displacement.plot()
<Figure size 800x750 with 3 Axes>