Interactive Receiver Functions Forward Modeller (IRFFM)
The Interactive Receiver Function Forward Modeller (IRFFM) is a Java program for
interactive forward modelling of teleseismic receiver functions (the first version, v1.0, was released in 2009).
IRFFM v1.1 is the current version that has been available since November 2010.
IRFFM v1.1 has been compiled so that it is compatible with earlier Java versions (from v1.4, inclusive) and it works on
32-bit computers. Several functionalities have been added since v1.0, most notably: i) the program now comes with a display
indicating the goodness of fit (variance reduction in percentages); ii) an option to print the resulting plots to either a printer
or a jpg file.
A manual describing the program and the main requirements could be obtained by clicking on the pdf icon below. Please read the
user manual for some important information before download and installation.
IRFFM2 (simultaneous forward modelling of receiver functions and surface wave dispersion) will be available for download soon.
Download IRFFM user manual here
IRFFM V1.1 is available for the following operating systems/platforms: Linux Suse, Mac OS X and Solaris 10 (x86
Screen snapshot showing the IRFFMv1.1 interface
Linux version download
Mac version download
Solaris version download
The IRFFM software is presented in:
Tkalčić, H, Y. Chen, R. Liu, Z. Huang, L. Sun and W. Chan, Multi-step modelling of teleseismic receiver functions
combined with constraints from seismic tomography: Crustal structure beneath southeast China, Geophys. J. Int., submitted,
Multi-Step Modelling of Teleseismic Receiver Functions Combined With Constraints From Seismic Tomography: Crustal
Structure Beneath Southeast China
In this study, in which IRFFM software is featured for the first time, we perform a receiver-based study of the lithosphere
of southeast China using the waveform records of excellent quality from fourteen Chinese National Digital Seismic Network (CNDSN) and four Global
Seismic Network (GSN) stations. Receiver functions (RFs) are predominantly sensitive to the gradients in the lithospheric elastic
parameters, and it is impossible to determine a non-unique distribution of seismic parameters such as absolute shear-wave speeds
as a function of depth unless other geophysical data are combined with RFs. Thus we combine RFs with independent information from
shear- and compressional-wave speeds below the Mohorovičić discontinuity, available from the existing tomographic
studies. The preparation of RFs and consequent analysis consist of multiple steps. First, a statistical approach based on a
calculation of the cross-correlation matrix is described and used to estimate averaged RFs for a large number of waveforms
available in this study (see Figure 1 below). Second, an interactive forward modelling software (IRFFM) is introduced and
applied to observed RFs to define a prior, physically acceptable range of elastic parameters in the lithosphere. This is followed
by a grid-search for a simple crustal structure. An initial model for a linearised, iterative inversion is constructed from
multiple constraints, including results from the grid-search for shear-wave speed, the Moho-depth versus vp/vs ratio domain search
and tomography. We obtain 1-D velocity profiles for all eighteen stations. The thickness of the crust constrained by the three
independent techniques appears to be more variable in comparison with tomographic studies, with the crust thinning significantly
towards the east (see Figure 2 below).
We used IRFFM to get a quick understanding of the features present in RFs, as well as a quantitative measure about the
range of parameters that produce theoretical RFs similar to the observed RFs. For example, one can explore how the crustal
thickness and the impedance contrast affect the P to S conversion, seen as the second peek in observed RFs. The estimated model
parameters using IRFFM are in a good agreement with the results from the H-k search.
Figure 1. Radial RFs calculated for one of the stations in the study from the southeastern azimuths for all earthquakes
without rejecting waveforms based on signal-to-noise ratio are shown in black. Mutually coherent waveforms selected using the
cross-correlation matrix approach are shown in blue. The selected waveforms are correlated with the cross-correlation coefficient
0.9 or higher with a) at least 25% of other waveforms and b) at least 50% of other waveforms. The thick red line is the average
calculated from the selected RFs.
Figure 2. Comparison of interpolated maps of crustal thickness (Moho depth) for southeast China using eighteen data points
corresponding to the locations of the stations from this study. a) P-wave tomography (Sun and Toksöz, 2006) and b) this study,
using RFs inversion modelling results
Other studies using similar approaches
Different variations of the method have so far been applied to waveform data:
Arabian Peninsula: multi-step approach on RFs and SW dispersion including modeling of polarization anisotropy (Tkalčić
et al., JGR, 2006)
China: combination of grid-search, H-k method, and linearized inversion (Chen et al., JGR, 2010)
Croatia: combination of grid-search, H-k method, and Monte Carlo inversion (Stipčević et al., GJI, 2010, in
Australia: combination of grid-search and Neighborhood Algorithm inversions (Fontaine et al., PEPI, 2010, in revision).
Australia: multi-step approach on RFs and ambient noise dispersion data, featuring IRFFM2 (Tkalčić et al.,,
2010, in preparation).
--- red --->
Back to homepage