dsmpy 1.0a2

Python wrapper for DSM

dsmpy

Python package for computation of synthetic waveforms in spherically homogeneous transversely isotropic (VTI) media using the Direct Solution Method (DSM; Kawai et al. 2006).

Documentation and tutorials can be found here.

INSTALLATION

Requirements

1. As dsmpy relies on mpi4py, you need to have a working MPI implementation installed (see mpi4py documentation)
2. gfotran >= 4.8

Using pip

1. In a shell, type

pip install dsmpy

This will download the dsmpy package and required dependencies from the PyPI repository.

2. Check that dsmpy has been installed succesfully:

python -c "import dsmpy"

Note: Fortran sources for the DSM will be compiled during the installation (using numpy.f2py and the GNU Fortran compiler). If you get compilation errors, check the following:

• gfortran >= 4.8 is required for succesful compilation, because of the optimization flag '-Ofast'
• If you have gfortran <4.8, you should change the compiler flag from '-Ofast' to '-O3' in <path_of_dsmpy_folder>/pydsm/__init__.py

GETTING STARTED

To get started, you should at least run python test_tipsv.py and python test_tish.psv located in in <path_of_pydsm_folder>/pydsm/tests. These scripts check dsmpy against pre-computed synthetics using the DSM (Fortran).

EXAMPLES

1. Running dsmpy using an input file (run on multiple CPUs). A template input file is in <path_of_pydsm_folder>/dsmpy/tests/input_files/template.txt:

sac_files ~/git/dsmpy/tests/sac_files/*T
output_folder ~/git/dsmpy/tests/sac_files
# duration of synthetics (in seconds)
tlen 3276.8
# number of points of frequency-domain synthetics
# minimum period Tmin = tlen / nspc (s)
nspc 256 
# sampling frequency for time-domain synthetics
sampling_hz 20
# prem, ak135
seismic_model prem 
# 0: P-SV+SH, 1: P-SV, 2: SH (default: 0)
mode 0
# 0: quiet, 1: talkative, 2: debug (default: 0)
verbose 0

To run this input file on 2 CPUs:

1. open a Terminal
2. change the current directory to the dsmpy directory
3. paste:

mpiexec -n 2 python pydsm/main.py tests/input_files/template.txt

2. Running dsmpy from a python script. Below is an example of python script using dsmpy to compute synthetics:

from dsmpy import dsm, seismicmodel
from dsmpy.event import Event
from dsmpy.station import Station
from dsmpy.utils.cmtcatalog import read_catalog
# load gcmt catalog
catalog = read_catalog()
# get event from catalog
event = Event.event_from_catalog(
    catalog, '200707211534A')
# define station FCC
stations = [
    Station(
        name='FCC', network='CN',
        latitude=58.7592, longitude=-94.0884), 
    ]
# load (anisotropic) PREM model
seismic_model = seismicmodel.SeismicModel.prem()
tlen = 3276.8 # duration of synthetics (s)
nspc = 256 # number of points in frequency domain
sampling_hz = 20 # sampling frequency for sythetics
# create input parameters for pydsm
input = dsm.PyDSMInput.input_from_arrays(
    event, stations, seismic_model, tlen, nspc, sampling_hz)
# compute synthetics in frequency domain calling DSM Fortran
output = dsm.compute(input)
output.to_time_domain() # perform inverse FFT
output.filter(freq=0.04) # apply a 25 seconds low-pass filter
us = output.us # synthetics. us.shape = (3,nr,tlen)
ts = output.ts # time points [0, tlen]
# brackets can be used to access component and station
u_Z_FCC = output['Z', 'FCC_CN']
# to plot a three-component record section, use
output.plot()
plt.show()
# to write synthetics to SAC files, use
output.write(root_path='.', format='sac')

3. Running dsmpy using a (Fortran) DSM input file. Methods are provided to run dsmpy using an input file for the (Fortran) DSM:

from pydsm import dsm, rootdsm_sh
parameter_file = rootdsm_sh + 'AK135_SH.inf'
inputs = dsm.PyDSMInput.input_from_file(parameter_file, mode=2)
outputs = dsm.compute(inputs, mode=2)
outputs.to_time_domain()
us = outputs.us    # us.shape = (3,nr,tlen)
ts = outputs.ts    # len(ts) = tlen
stations = outputs.stations        # len(stations) = nr
components = outputs.components    # len(components) = 3