disba

disba: layered surface-wave dispersion

Use disba for 1D isotropic layered-model dispersion and sensitivity kernels. It is a forward solver; extracting observed dispersion from raw waveforms or performing tomography requires separate processing and inversion design.

Model and period conventions

• Supply four equal-length one-dimensional arrays: thickness in km, Vp/Vs in km/s, and density in g/cm³. A row-per-layer table must be transposed once: PhaseDispersion(*model.T).
• Do not use *zip(thickness, vp, vs, rho): with four layers it silently interchanges properties, and other layer counts give the wrong arguments.
• Treat the last layer as the half-space and write its thickness as 0.0. Other layers have positive thickness. The final thickness is not a finite bottom boundary, even if a nonzero value is accepted by the library.
• Use finite positive periods in increasing order. Density must be positive; for the solid models below require vs > 0 and vp**2 > (4/3) * vs**2. Preserve measured units and model provenance.

Phase and group velocity

This complete example uses a solid layered model. The first call may take longer while Numba compiles the solver.

import numpy as np
from disba import PhaseDispersion, GroupDispersion

thickness = np.array([0.5, 1.0, 2.0, 0.0])  # km
vp = np.array([1.5, 2.5, 4.0, 6.0])          # km/s
vs = np.array([0.8, 1.4, 2.3, 3.5])          # km/s
rho = np.array([1.8, 2.0, 2.3, 2.6])         # g/cm³
periods = np.linspace(0.1, 5.0, 50)          # s, increasing

phase_solver = PhaseDispersion(thickness, vp, vs, rho)
group_solver = GroupDispersion(thickness, vp, vs, rho)
rayleigh = phase_solver(periods, mode=0, wave='rayleigh')
love = phase_solver(periods, mode=0, wave='love')
group = group_solver(periods, mode=0, wave='rayleigh')
period_s, phase_km_s = rayleigh.period, rayleigh.velocity

Each result is a DispersionCurve named tuple, not a velocity array. Use result.period and result.velocity together: unavailable roots may be omitted, especially for higher modes. len(result) counts tuple fields, not computed periods; use result.period.size.