A Short-Period Surface Wave Dispersion Dataset for Model Assessment of Africa’s Crust: ADAMA.D1 Tolulope Olugboji, and Siyu Xue
Please cite as: Xue, S and Olugboji, T. (2021). URseismology/ADAMA: ADAMA.2 (v0.2). Zenodo. https://doi.org/10.5281/zenodo.6047730
Includes our metadata and measurements:
- ADAMA_stalist: Station metadata
- ADAMA_gvib: Displacement seismograms at 1 Hz
- ADAMA_staconns: List of station pairs that are connected - 150K
- ADAMA_ncfs: NCFs for connected station pairs
- ADAMA_raw: Phase & group velocity extracted from NCFs (Raw)
- ADAMA_clean: Same as (5) but downsampled (11 periods) and cleaned
Codes to build HDF5 files from all data and read from HDF5 files:
- buildADAMAncf.m – compiles the dataset from ncf results
- read_ADAMA_ncfs.m – read in the ncf of a station pair as HDF5 files
- buildADAMAraw.py – compiles the dataset from raw AkiEstimate files
- read_ADAMA_raw.m – read in the Aki phase velocity as HDF5 files
- buildADAMAclean.m – convert dispersion results to ASCII files
All scripts for making plots for manuscript, are provided as:
- fig1_heatmaps.m
- fig2_AfrAllStaQlty.m
- fig3_AkiEst.m
- fig4_orientationTest.m
- fig5_CCFintime.m
- fig6_pathavgModel.m
- fig7_Aki03PVresults.m
- fig7_getAkiData.m
- fig8_comparePVResults.m
- fig9_AFRTomoMap.m
- fig9_GDM52Map.m
- fig9_LithoMap.m
- fig10_outlierRayPath.m
- figS3_ConnDuraSNR.m
- plotplates.m
- saveFig.m
- Outlier_analysis.m
Scripts to download raw data as SEED files from IRIS using the rover system, downsample, remove instrument response, rotate, and postprocess the SEED to SAC files and automate using SLURM bash scripts on the High-performance computing (HPC) platform bluehive. Finally MATLAB scripts to compute NCFS using the Russell codes generalized for HPC on bluehive.
Scripts that use NCFs from 4 above, converted to ASCII files, and produce phase velocities Co. Again, SLURM bash scripts that run the whole process using the HPC system:
- run01code.sh & run01code.slurm: run Aki01 code in parallel.
- run02code.sh & run02code.slurm: run Aki02 code in parallel.
- run03code.sh & run03code.slurm: run Aki03 code in parallel.