HPC-parallelized Bayesian Inversion for Near-surface Imaging (NEOPSY)

Multizonal Transdimensional Inversion (MTI) to infer near-surface 1D layered velocity models.

---

A unique software package NEOPSY for Multizonal Transdimensional Inversion (MTI) of near-surface 1D layered velocity models within a full Bayesian framework (Hallo et al., 2021). The framework is designed as transdimensional and data-driven, meaning the model complexity (number of layers) is inferred directly from the data. This is implemented by a mathematical Occam's razor inherent to the Bayesian formulation. The transdimensional model space is sampled using Markov Chain Monte Carlo (MCMC) with Parallel Tempering, enabling efficient exploration of high-dimensional and non-linear parameter spaces. The code is parallelized in Fortran using MPI (CPU), specifically optimized for deployment on High-Performance Computing (HPC) clusters and large-scale seismic inversions.

Important

This code was used for the inversion of near-surface structure on Mars, published in Nature Communications (Hobiger, Hallo, Schmelzbach et al., 2021).

1 METHODOLOGY

The suite uses the theory by Hallo et al. (2021).

Hallo, M., Imperatori, W., Panzera, F., Fäh, D. (2021). Joint multizonal transdimensional Bayesian inversion of surface wave dispersion and ellipticity curves for local near-surface imaging, Geophysical Journal International, 226 (1), 627-659. https://doi.org/10.1093/gji/ggab116

2 TECHNICAL IMPLEMENTATION

Bayesian Inference, Markov Chain Monte Carlo (MCMC), Uncertainty Quantification, High-Performance Computing (HPC), Code Parallelization by MPI (CPU), Transdimensional Bayesian Inference, Data-driven Inversion, Occam's razor

• Input Data:
  • Surface wave dispersion and ellipticity curves (both Rayleigh and Love waves)
  • Observed data from SPAC, MASW, RayDec, and other methods
  • Supports Rayleigh fundamental and higher modes (1st, 2nd, and 3rd), Love fundamental and higher modes (1st, 2nd, and 3rd), as well as Rayleigh wave ellipticity (H/V) and ellipticity angle
• Key Features:
  • Multimodal Joint Inversion: Simultaneously inverts various wave types and modes
  • Bayesian Framework: Infers the posterior probability density function to provide solutions with a rigorous uncertainty estimate
  • Transdimensional Inversion: Automatically adjusts the number of dimensions (number of layers) based on the Occam's razor principle
  • Multizonal Prior: Allows users to define depth-dependent zones for layer parameters (V_S, V_P, mass density, and Poisson's ratio)
  • Robust LVZ Support: Capable of inversion for complex velocity profiles containing low-velocity zones (inverse velocity gradients)
• Outputs:
  • High-quality figures and standard ASCII text files
  • Posterior probability density functions for: V_S, V_P, mass density, Poisson's ratio, V_S30, depth of layer interfaces, quarter-wavelength parameters (velocity, depth, impedance), SH-wave transfer functions, and amplification relative to a reference rock model
  • Representative 1D models: Maximum Likelihood (ML), Maximum A Posteriori (MAP), Marginal Maximum A Posteriori (MAX), and posterior mean (AM) with its associated uncertainty

3 RELEASE HISTORY (MAJOR VERSIONS)

• 4.0 (Fjord) — Public Release | May 2026

  • Public open-source release
  • Integration with Geopsy 3.X
  • New Outputs: Output ASCII files with layer interfaces, MAP model uncertainty, amplification, and V_S30
  • Advanced Visualization: Improved all plotting scripts, high-resolution resultant figures
  • Localization: Added option for user-specific reference rock velocity model (for amplification); Japanese and Swiss reference models present in default

• 3.0 (Eagle) — Geopsy Update | May 2021

  • Internal-only version
  • Integration with Geopsy 3.X
  • Geopsy Update: Forward problem routines changed to Geopsy 3.X
  • Enhanced Engine: Support for irregular sampling of observed data in the frequency domain, improved speed by using binary data transfer between MCMC and Geopsy cores
  • Key Publications: Eagle version used in publications by Glueer et al. (2024, https://doi.org/10.3390/geosciences14020028), van Ginkel et al. (2025, https://doi.org/10.5194/tc-19-1469-2025)

• 2.0 (Desert) — Numerical Engine Revision | May 2020

  • Internal-only version
  • Integration with Geopsy 2.X
  • Enhanced Engine: Improved search for layers with inverse velocity, improved birth from prior
  • Advanced Visualization: QWL impedance statistic, SH-wave transfer function statistic, high-resolution V_S30
  • Key Publications: Desert version (Desert2 patch) used in publications by Hallo et al. (2021, https://doi.org/10.1093/gji/ggab116), Hobiger, Hallo, Schmelzbach et al. (2021, https://doi.org/10.1038/s41467-021-26957-7)

• 1.0 (Crimson/Coral) — Initial Release | February 2020

  • Internal-only version
  • Integration with Geopsy 2.X
  • Crimson version for Rayleigh waves (onshore), and Coral version for Scholte waves (offshore/lake)
  • Key Publication: Coral version (Coral2 patch) used in publication by Shynkarenko et al. (2021, https://doi.org/10.1093/gji/ggab314)

4 REQUIREMENTS

1. LINUX/UNIX machine with installed timeout program

2. Fortran90 (gfortran or ifort) and MPI (mpif90 or mpiifort) compilers

3. Compiled Parallel Tempering library pt.f90

   • Parallel Tempering (PT) library (M.Sambridge)
   • http://www.iearth.edu.au/codes/ParallelTempering/

4. Installed Geopsy

   • Open source software for geophysical research and applications (M.Wathelet)
   • Version 3.X is required
   • Neopsy makes use of the gpdc and gpell functions (in command line)
   • https://www.geopsy.org/

5. Python 3.12 or MATLAB R2025b for plotting results

   • Python Libraries: matplotlib, numpy, scipy
   • Install Python dependencies via pip:

pip install -r requirements.txt

5 PACKAGE CONTENT

1. data - Directory containing observed data (dispersion and ellipticity SPAC curves) and reference rock velocity models vs_ref_Swiss.ascii and vs_ref_Japan.ascii
2. inv - Work directory for the ongoing inversion
3. lib - Directory with MATLAB function library
4. src - Directory with Fortran source codes of the NEOPSY (includes Makefile)
5. data.para - Settings of input data
6. input.para - Settings of the inversion
7. run_timpi.sh - Run the MTI inversion (step 1)
8. run_tires.sh - Run post-processing of the MTI inversion (step 2)
9. plot_pop.m - Plot resultant figures by MATLAB (step 3)
10. plot_pop.py - Plot resultant figures by Python (step 3)
11. lib.py - Python function library
12. run_clean.sh - Clean working directory after inversion (optional)
13. requirements.txt - pip requirements file for instalation of Python dependencies

6 COMPILATION

1. Compile Parallel Tempering (PT) library in the ./src/pt folder
2. Set your compilers in Makefile in the ./src folder and type:

make

3. Check npdc, tiser, timpi, and tires binary programs in the project root directory
4. If necessary, make binaries executable by

chmod +x npdc tiser timpi tires

7 USAGE

1. Create a work directory ./inv and log directory ./inv/log. The name of the working folder is optional, but it has to be the same as defined in the input.para file (run_timpi.sh creates these directories automatically)
2. Prepare data files with observed dispersion and ellipticity curves in directory ./data
3. Set input parameters of data and inversion (data.para and input.para) in NEOPSY root directory
4. Step 1 - Run NEOPSY (executable timpi or tiser). You can do it manually, but it is strongly recommended to use the prepared bash script run_timpi.sh

bash ./run_timpi.sh

5. The result of NEOPSY is an ensemble of possible velocity models drawn from the transdimensional posterior probability density function. You can wait for the termination of MCMC Markov chains (it can last several hours) or kill timpi at any time (kill -9 PID). NEOPSY saves models continuously, so there is no harm if you kill it prematurely; there will be fewer models in the ensemble (but be careful to wait until the end of the burn-in phase). These models are saved in ./inv directory in binary files (xmodels*.dat). You can check the state of the inversion at any time by browsing log files in ./inv/log. Especially, note number of preformed steps and the best data variance reduction in files stats_*.log.

  MCstep - Number of performed Markov steps
  AccT1 - New accepted sampling models
  AccAll - New accepted models (20 chains)
  Err - Number of cases when Geopsy failed
  aVR[%] - Highest data Variance Reduction
------------------------------------------
  MCstep   AccT1  AccAll     Err    aVR[%]
max_val:     100    2000    2000      100%
------------------------------------------
     100       0    1191       0  -2177.08
     200       0    1215       0    -80.91
     300       0    1154       0     14.60
     400      38    1093      11     46.63
     500      81    1037       1     65.67
	 ...

6. Step 2 - Run the after-process (tires) to create PDFs, population histograms, QWL representation, SH-wave amplification, etc. You can run it manually, or by using the bash script run_tires.sh (recommended). The results are saved in ./inv directory as (out* and in* files)

bash ./run_tires.sh

7. Step 3 - Run the Python script python3 plot_pop.py or MATLAB script plot_pop.m for plotting results. If you run the Python script on a server without a graphical user interface, set export MPLBACKEND=Agg (Linux) or $env:MPLBACKEND="Agg" (Windows) before execution. If you run the MATLAB script on a server without a graphical user interface, use matlab -nodisplay -nosplash -nodesktop -r "run('plot_pop.m'); exit;".
8. (Optional) You can clean the working directory after the inversion using the Bash script run_clean.sh. It deletes temporary files, and if you set the switch deleteEnsemble=1, it will also delete the ensemble of solutions (xmodels*.dat), which takes up the most disk space and is not necessary after post-processing.

bash ./run_clean.sh

Note: See connected example files for data.para, input.para, and observed dispersion curves in ./data. Also, see an example of the resulting posterior marginal probability density function of S-wave velocity, depth of layer interfaces, and V_S30 below.

8 COPYRIGHT

Copyright (C) 2019-2021,2026 ETH Zurich (Architect: Miroslav Hallo)

This program is published under the GNU General Public License (GNU GPL).

This program is free software: you can modify it and/or redistribute it or any derivative version under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

This code is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY. We would like to kindly ask you to acknowledge the authors and don't remove their names from the code.

You should have received a copy of the GNU General Public License along with this program. If not, see http://www.gnu.org/licenses/.

9 CITE AS

If you use NEOPSY, please cite both the original methodology paper (preferred) and the software version as follows:

For the methodology and implementation:

Hallo, M., Imperatori, W., Panzera, F., Fäh, D. (2021). Joint multizonal transdimensional Bayesian inversion of surface wave dispersion and ellipticity curves for local near-surface imaging. Geophysical Journal International, 226 (1), 627-659. https://doi.org/10.1093/gji/ggab116

For the specific software version:

Hallo, M. (2026). NEOPSY: HPC-parallelized Bayesian Inversion for Near-surface Imaging (v4.1) [Software]. Zenodo. https://doi.org/10.5281/zenodo.20174687

---

10 USER MANUAL

OBSERVED DATA

All files with observed data should be in the ./data directory. However, you can use different names and paths for data files, but you must specify them in the data.para file. All these files must be in ASCII (UNIX standard - LF) and with line-fixed structure (i.e., do not add or remove any line from these files and do not change the structure). The code reads these files line by line, and it expects a particular structure. In the data.para file, lines starting with # are comments, and the following text may be changed by the user without any effect on functionality. But it is strictly prohibited to add or remove any additional commentary lines.

The data.para input file consists of 32 lines. There are YES/NO switches on lines 5, 8, and 11, which let the user decide which modes of Rayleigh and Love dispersion curves, and ellipticity curves will be used in the inversion. There are filenames (with path) below, where the code expects the particular data (lines 14-26). If you select 0 (i.e., NO) for a mode, the respective filename doesn't matter. You can use both the absolute or relative paths.

The data files (e.g., Data_R0.dat) consist of three columns. The first is frequency. The frequency range and the sampling rate may be irregular, and they may differ among the modes used. But it is recommended to use continuous regular sampling on a log-scale. At least, the frequency samples should be in increasing order. The second column are measured data (slowness [s/m]; ellipticity [0, Inf]; or ellipticity angle [deg]). The third column contains the expected uncertainty (error) as 1sigma (or multiplication factor in the case of ellipticity). NEOPSY is Bayesian inversion, where the measured data are weighted by the inverse of their uncertainties, so the uncertainty matters! If you wish to decrease the weight of any data, increase the uncertainty. If you wish to ignore some of your data samples in the inversion, give them an arbitrary negative uncertainty (e.g., -1). Note that the first line of the data files is always a comment, and there is only one data file allowed per wave mode. I would like to emphasize that the uncertainty, as the inversion input, should contain a sum of uncertainty of observed data (results from field measurements) and theory (we fit data with a 1D layered model, but nature is more complex).

SETTINGS OF INVERSION

The settings of the NEOPSY are managed by the input.para ASCII file (UNIX standard - LF). This file has a line-fixed structure similarly as data.para file. However, the number of lines in this file may vary based on the number of zones or the number of fixed Voronoi nuclei. Add and remove lines very carefully. The minimum number of lines of input.para file is 54 (see figure below), while there are more lines in the multizonal case. An extension by one zone (i.e., increasing the number on the fixed-line 8) results in additional lines for S-wave, P-wave, density, and Poisson’s ratio limits, and Vs/Vp/rho MCMC steps (i.e., 5 additional lines for each additional zone). The number of fixed Voronoi nuclei then simply say to the code how many lines it should read at the end of this file. The meaning of all parameters is emphasized in the comments of this input file.

RUN INVERSION

Step 1 - NEOPSY can be executed on multiple nodes (program timpi) or a single node (program tiser). Both programs must have two input command-line arguments: (1) path to input.para file; (2) path to data.para file. The path may be either absolute or relative. The single-node version is meant for testing purposes or low-end PCs. Each deployed node makes use of one CPU and same amount of RAM (e.g., 16 nodes = 16 deployed CPUs and 16x the required RAM amount). To run the timpi, you have to use mpirun command. For your convenience, it is strongly recommended to use the pre-prepared bash script run_timpi.sh to run the inversion. This bash script gets the name of the working folder; copy input.para and data.para files into the working folder; execute the inversion with the proper arguments in the background; and save PID of the executed inversion to yymmdd_HHMMSS.pid file. The result of the NEOPSY is an ensemble of possible models drawn from the transdimensional posterior probability density function on model parameters. So, you can wait for the termination of MCMC Markov chains (it can last several hours) or kill the process at any time (kill -9 PID). NEOPSY saves models continuously, so there is no harm if you kill it prematurely; there will be fewer models in the ensemble (but be careful to wait until the end of the burn-in phase). These models are saved in ./inv directory in binary files (xmodels*.dat). You can check the state of the inversion at any time by browsing log files in ./inv/log. Especially, note number of preformed steps and the best data variance reduction in files stats_*.log.

Step 2 - When the inversion is done (finished or killed), you must run the after-process (program tires) to create ASCII files with results. The main purpose of this after-processing is to create histograms, probability density functions, compute quarter-wavelength representation, V_S30, SH-wave transfer function, etc. As with timpi, you can run it manually or by using the bash file run_tires.sh. It requires only one node (one CPU). So, it is not as demanding as timpi. Be sure you run it with input.para and data.para files which are identical to those in Step 1. Results (ASCII files) are saved in ./inv directory as out* and in* files.

Step 3 - To plot the results, run the python3 plot_pop.py (Python) or plot_pop.m (MATLAB) script. The plotting scripts can be executed on local machines with a graphical user interface (GUI) or directly via a command-line server. Output figures are saved automatically. If you use a different working directory or need to customize the figure output behavior, you can easily modify the configuration parameters located at the beginning of the plotting scripts.

ERROR LIST

The NEOPSY software can produce various error/warning messages to inform the user about the source of some problems/irregularities. The purpose of the internal error handling is to allow users to find issues in the input and/or to trace the source of potential problems. Most of the errors are input-related and can be easily solved by the user. Moreover, the program may produce some warnings of information characters (excluded from this list as they are informational only).

• ERROR 101 – This occurs if the user calls an executable binary file (timpi, tiser, tires) without two command-line arguments. The first argument (path/)filename is the (path and) filename of the input.para file. The second argument (path/)filename is the (path and) filename of the data.para file. Solution: Include both in your command- line expression, calling these binary files.
• ERROR 102 – This occurs if the input.para file given in the first command-line argument does not exist. Solution: Include a correct filename and path to this file (the path is not mandatory if it is in the same directory as the executable file).
• ERROR 103 – This occurs if the data.para file given in the second command-line argument does not exist. Solution: Include a correct filename and path to this file (the path is not mandatory if it is in the same directory as the executable file).
• ERROR 104 – This occurs if the number of dept-zones in input.para is not a positive integer. Solution: Include a correct number of dept-zones in the input.para file (e.g., 1).
• AUTOC 105 – This occurs if the program performs an autocorrection of the user input to fulfill requirements: The total maximal number of the Voronoi nuclei from input.para must be greater than the number of depth-zones. Solution: Just bear it in mind or increase the maximum number of the Voronoi nuclei in the input.para.
• AUTOC 106 – This occurs if the program performs an autocorrection of the user input to fulfill requirements: The minimum layer-interface depth must be positive and non-zero (because a logarithm of depth is used). Solution: Just bear it in mind or include a valid minimum layer interface depth in the input.para.
• ERROR 107 – This occurs if the profile’s maximum depth is smaller than the minimum depth in the input.para. Solution: Include correct depths in the input.para file (e.g., 0.5 250.0).
• AUTOC 108 – This occurs if the program performs an autocorrection of the user input to fulfill requirements: The transdimensional proposal probability. Solution: Just bear it in mind or include a number < 0.45 in the input.para.
• AUTOC 109 – This occurs if the program performs an autocorrection of the user input to fulfill requirements: The number of Voronoi nuclei in the non-transdimensional inversion. Solution: Just bear it in mind or increase the total maximal number of the Voronoi nuclei in the input.para.
• ERROR 110 – This occurs if the Gaussian random walker step size in the input.para is too large for this profile. Solution: Decrease the log-depth step size in the input.para file.
• ERROR 111 – This occurs if a zone is defined below the max depth of the profile. Solution: Increase the maximal depth of the profile in the input.para file or delete the deepest depthzone.
• SWAP 112 – This occurs if the program performs an autocorrection of the user input to fulfill requirements: The first argument of S-wave velocity thresholds must be smaller than the second one. The program performed an automatic swap of these two values. Solution: Just bear it in mind or swap them manually in the input.para.
• SWAP 113 – This occurs if the program performs an autocorrection of the user input to fulfill requirements: The first argument of P-wave velocity thresholds must be smaller than the second one. The program performed an automatic swap of these two values. Solution: Just bear it in mind or swap them manually in the input.para.
• SWAP 114 – This occurs if the program performs an autocorrection of the user input to fulfill requirements: The first argument of density thresholds must be smaller than the second one. The program performed an automatic swap of these two values. Solution: Just bear it in mind or swap them manually in the input.para.
• SWAP 115 – This occurs if the program performs an autocorrection of the user input to fulfill requirements: The first argument of Poisson ratio thresholds must be smaller than the second one. The program performed an automatic swap of these two values. Solution: Just bear it in mind or swap them manually in the input.para.
• ERROR 116 – This occurs if the Gaussian random walker S-wave velocity step size in the input.para is zero or less. Solution: Set the S-wave velocity step size to a positive non-zero value in the input.para file.
• ERROR 117 – This occurs if the Gaussian random walker P-wave velocity step size in the input.para is zero or less. Solution: Set the P-wave velocity step size to a positive non-zero value in the input.para file.
• ERROR 118 – This occurs if the Gaussian random walker density step size in input.para is zero or less. Solution: Set the density step size to a positive non-zero value in the input.para file.
• AUTOC 119 – This occurs if the program performs an autocorrection of the user input to fulfill requirements: The Gaussian random walker S-wave velocity step size in the input.para was too large and has been decreased. Solution: Just bear it in mind or decrease it manually in the input.para.
• AUTOC 120 – This occurs if the program performs an autocorrection of the user input to fulfill requirements: The Gaussian random walker P-wave velocity step size in the input.para was too large and has been decreased. Solution: Just bear it in mind or decrease it manually in the input.para.
• AUTOC 121 – This occurs if the program performs an autocorrection of the user input to fulfill requirements: The Gaussian random walker density step size in the input.para was too large and has been decreased. Solution: Just bear it in mind or decrease it manually in the input.para.
• ERROR 122 – This occurs if the depth difference between two arbitrary fixed Voronoi nuclei is smaller than a limit (0.1[m] in default). Solution: Increase the depth span between the fixed Voronoi nuclei in the input.para file.
• ERROR 123 – This occurs if the working directory does not exist (./inv/ in default). Solution: create the working directory (with name and path asset in the input.para file).
• ERROR 124 – This occurs if the logging directory does not exist (./inv/log in default). Solution: create the logging directory (a subdirectory named “log” in the working directory).
• AUTOC 125 – This occurs if the program performs an autocorrection of the user input to fulfill requirements: The transdimensional proposal probability must be equal to 0.0 in the case of non-transdimensional inversion. Solution: Just bear it in mind or set the transdimensional proposal probability to 0.0 in the input.para.
• AUTOC 126 – This occurs if the program performs an autocorrection of the user input to fulfill requirements: The transdimensional proposal probability must be equal to 0.0 in the case of non-transdimensional inversion. Solution: Just bear it in mind or set the transdimensional proposal probability to 0.0 in the input.para.
• ERROR 127 – This occurs if the settings in the input.para suggests a non-transdimensional inversion (transdimensional proposal probability is equal to 0.0), but the number of fixed Voronoi nuclei (or the fixed number of layers) is set to zero. Solution: set the transdimensional proposal probability OR the number of required fixed nuclei to a positive number.
• ERROR 128 – This occurs if the settings in the input.para suggests a non-transdimensional inversion (transdimensional proposal probability is equal to 0.0), but there is more than one depth zone defined. The depth zones are a feature of the transdimensional inversion, and it is impossible to mix them with a fixed number of Voronoi nuclei (or layers). Solution: delete all the superfluous depth zones.
• ERROR 129 – This occurs if the settings in the input.para suggests using user-defined initial model (expert mode settings) together with a multizonal inversion. Solution: turn off the respective expert switch OR delete all depth zones.
• ERROR 130 – This occurs if the settings in the input.para suggests using user-defined initial model (expert mode settings) together with a non-transdimensional inversion. Solution: turn off the respective expert switch OR select to use single-zonal transdimensional inversion.
• WARNING 201 – This occurs if no mode of Rayleigh or Love wave dispersion curves in the data.para. Solution: turn on at least one switch for the surface wave dispersion modes (set at least one to 1).
• ERROR 202 – This occurs if there is a switch on for both ellipticity and ellipticity angle. Solution: turn one of these two switches to zero in the data.para.
• ERROR 203-206 – This occurs if a Rayleigh wave data file is given in the data.para does not exist (R0, R1, R2, R3). Solution: Include a correct filename and path to this file (the path is not mandatory if it is in the same directory as the executable file).
• ERROR 207-210 – This occurs if a Love wave data file is given in the data.para does not exist (L0, L1, L2, L3). Solution: Include a correct filename and path to this file (the path is not mandatory if it is in the same directory as the executable file).
• ERROR 211-212 – Occurs if the ellipticity (or ellipticity angle) data file is given in the data.para does not exist. Solution: Include a correct filename and path to this file (the path is not mandatory if it is in the same directory as the executable file).
• ERROR 213 – Occurs if the file with the initial model (expert mode settings), which is given in the data.para, does not exist. Solution: Include a correct filename and path to this file OR turn off the respective expert switch.
• ERROR 214 – This occurs if the initial model (expert mode settings) given in the data.para has a corrupted or unknown structure. Solution: Include the initial model file with the correct structure OR turn off the respective expert switch.
• ERROR 215 – Occurs if the file with the reference velocity model does not exist. Solution: Include a correct filename and path to this file.
• ERROR 301 – This occurs if the generation of initial random models takes too many interactions. It means that the model space is too large and over-conditioned. Solution: Reduce the number of higher modes of observed data; increase the uncertainty of the observed data; decrease the span of search for S, P-wave velocity, density, and Poisson ratio; and/or check if observed data are OK.
• ERROR 302 – This occurs when performing a non-transdimensional inversion (transdimensional proposal probability is equal to 0.0), but the number of fixed Voronoi nuclei (or the fixed number of layers) is set to zero. Solution: set the transdimensional proposal probability OR the number of required fixed nuclei to a positive number.
• ERROR 303 – This occurs if the number of required layers in a non-transdimensional inversion exceeds the maximum number of allowed Voronoi nuclei. Solution: Increase the maximum number of allowed Voronoi nuclei in the input.para file and run again.
• ERROR 304-309 – Occurs if a problem occurred during reading and processing of the user-defined initial model (expert mode settings) given in the data.para. For details, see the message coming with the error code. Solution: Include the initial model file with the correct structure, OR turn off the respective expert switch.
• ERROR 401 – This occurs if there is an error in the forward problem during the after-processing of the saved models. There should be none as the saved models have been processed in the main run already. Solution: Check, if you are after-processing results from the correct main run (check the changed date of xmodels*.dat files and check if they are non-empty); check if you are using the same settings in input.para and data.para in the after-process as in the main run; if nothing helps, clean working directory and run the inversion carefully again.