This documentation is not for the latest stable Salvus version.

This tutorial is presented as Python code running inside a Jupyter Notebook, the recommended way to use Salvus. To run it yourself you can copy/type each individual cell or directly download the full notebook, including all required files.

Model Interpolation

This notebook shows how to interpolate a given model defined on a regular grid to a Salvus mesh using the Marmousi model as an example.
# initialize notebook
%matplotlib inline
import matplotlib.pyplot as plt
import numpy as np
import obspy
from salvus.mesh import mesh_block

plt.rcParams["figure.figsize"] = (10, 8)
The Marmousi model is provided in a segy file, which we open with obspy and map to numpy arrays.
def get_marmousi():
    nx, ny = 2301, 751
    dx, dy = 4.0, 4.0

    rho = np.empty((ny, nx))
    st ="data/marmousi_density.segy.gz")
    for _i, tr in enumerate(st):
        rho[_i, :] =

    vp = np.empty((ny, nx))
    st ="data/marmousi_velocity.segy.gz")
    for _i, tr in enumerate(st):
        vp[_i, :] =

    x = np.arange(nx) * dx
    y = np.arange(ny) * dy

    return x, y, vp, rho
Plot the model to verify we got it in correctly:
x, y, vp, rho = get_marmousi()
fig, axes = plt.subplots(2, 1)
Build a rectilinear mesh for a subregion of the model. While this might be simplistic, note that the model interpolation afterwards would be the same for any unstructured mesh.
fmax = 10.0  # maximum frequency in Hz
elements_per_wavelength = 2.0  # resolution criterion

# region in the marmousi model to mesh
mesh_min_x = x.max() * 0.2
mesh_max_x = x.max() * 0.7
mesh_min_y = y.min()
mesh_max_y = y.max()

hmax = vp.min() / fmax / elements_per_wavelength
nelem_x = int((mesh_max_x - mesh_min_x) / hmax) + 1
nelem_y = int((mesh_max_y - mesh_min_y) / hmax) + 1

sg = mesh_block.generators.cartesian.rectangle_2d(
m = sg.get_unstructured_mesh()
Interpolate the model to the mesh's element nodes using a scipy function and attach it to the mesh for plotting:
from scipy.interpolate import RectBivariateSpline

f_vp = RectBivariateSpline(x, y, vp.T[:, ::-1])
f_rho = RectBivariateSpline(x, y, rho.T[:, ::-1])

nodes = m.get_element_nodes()

nodes_x = nodes[:, :, 0]
nodes_y = nodes[:, :, 1]

vp_nodes = f_vp(nodes_x, nodes_y, grid=False)
rho_nodes = f_rho(nodes_x, nodes_y, grid=False)

m.attach_field("VP", vp_nodes)
m.attach_field("RHO", rho_nodes)
m.attach_field("fluid", np.ones(m.nelem))