2D Synthetic dataset of numerical simulations of long-period seismicity in a volcanic edifice and related sensitivity kernels☆

This work describes the data used in the EPSL research article “Quantifying strong seismic propagation effects in the upper volcanic edifice using sensitivity kernels”. The dataset is generated in order to investigate to what extent the seismic signals recorded on volcanoes are affected by near surface velocity structure. Data were calculated using the computational spectral elements scheme SPECFEM2D, where the wave propagation beneath Mount Etna volcano, Italy, was simulated in both homogeneous and heterogeneous models. The heterogeneous model comprises a low-velocity superficial structure (top several hundred meters) based on the previously published studies. Several different source mechanisms and locations were used in the simulations. The seismic wavefield was “recorded” by 15 surface receivers distributed along the surface of the volcano. The associated sensitivity kernels were also computed. These kernels highlight the region of the velocity model that affects the recorded seismogram within a desired time window. The text files describing the velocity models used in the simulations are also provided. The data may be of interest to volcano seismologists, as well as earthquake seismologists studying path effects and wave propagation through complex media.


a b s t r a c t
This work describes the data used in the EPSL research article "Quantifying strong seismic propagation effects in the upper volcanic edifice using sensitivity kernels". The dataset is generated in order to investigate to what extent the seismic signals recorded on volcanoes are affected by near surface velocity structure. Data were calculated using the computational spectral elements scheme SPECFEM2D, where the wave propagation beneath Mount Etna volcano, Italy, was simulated in both homogeneous and heterogeneous models. The heterogeneous model comprises a low-velocity superficial structure (top several hundred meters) based on the previously published studies. Several different source mechanisms and locations were used in the simulations. The seismic wavefield was "recorded" by 15 surface receivers distributed along the surface of the volcano. The associated sensitivity kernels were also computed. These kernels highlight the region of the velocity model that affects the recorded seismogram within a desired time window. The text files describing the velocity models used in the simulations are also provided. The data may be of interest to volcano seismologists, as well as earthquake seismologists studying path effects and wave propagation through complex media.
© -Heterogeneous model generated from Table 1 below, (b) Digital topography obtained from the CGIAR consortium for spatial information described in file topography.txt , (c) Different source and receiver locations.

Value of the Data
• This synthetic dataset highlights strong path effects on the long-period waveforms due to the shallow structure beneath volcanoes, supporting previous studies showing that, even if longperiod seismicity are often associated with magma movements in resonant conduits (e.g: [2] ), the corresponding wavefield can be significantly distorted by near-surface structure and surface topography (e.g. [3] , [4] ), • The data are useful for volcano seismologists, and seismologists in general studying path effects and wave propagation through complex media, • The data could be used for the investigation of seismic inversions in the presence of low velocity layering, • This type of data usually requires quite a long time to calculate (including setting up the models). Therefore, it may help provide data to communities who do not have easy access to or experience with numerical wave simulation tools.

Data Description
The presented dataset has been calculated with the aim of investigating the highly complex nature of seismic wave propagation in near surface volcanic structures and demonstrate the extent to which resonating LP events can be caused by the structural model of the volcanic edifice. Files can be divided into five groups. The first one contains the input parameters that have been used to run the SPECFEM2D simulations. The second and third groups contain the velocity models and the source-time functions generated by SPECFEM2D based on the material properties and the source characteristics, respectively, defined in the first group. The last two groups contain the seismograms and the sensitivity kernels resulting from the simulations.
• Input parameters (Folder SIMULATIONS_INPUT ) 1. material_properties.txt : This file contains four columns with information about the seismic properties of each of the layers in the 2D model as in Table 1 . The first column indicates the thickness of the layer, and the second, third and fourth columns give the velocities of the P-and S-waves and the density, respectively. 2. topograpy.txt : Contains two columns with the horizontal and vertical coordinates in meters of the 2D model's surface of Mt Etna obtained from the CGIAR consortium for spatial information. 3. stations.txt : Contains two columns with the horizontal and vertical position of the 15 recording stations on the 2D model as in Table 4 . 4. source_location.txt : Contains three columns with the depth and the horizontal and vertical coordinates in meters on the 2D model of the two seismic sources used in the simulations as in Table 2 . 5. source_type.txt : Contains two columns indicating the frequencies in Hertz and mechanisms of the seismic sources used on the simulations as in table Table 3 .  Table 1 Seismic velocities and density structure for the heterogeneous model. The S-wave velocities for the first 360 m depth are from [5] reporting the S-wave velocity structure derived from probabilistic inversion of Rayleigh-wave dispersion data. Deeper velocities come from [6] . Corresponding P-wave velocities are obtained using a Poisson's ratio of 0.25 and densities using the Gardner's relationship ( [7] ). 745  430  1620  60  1334  770  1873  60  1957  1130  2062  120  2304  1330  2148  60  2771  1600  2249  1640  3250  1876  2341  4255  4800  2771  2580   Table 2 Seismic source locations used for the simulations.     Table 1 or file material_properties.txt . coordinates in meters of each point in the 2D model and the next three columns contain the density, P-wave and S-wave sensitivity kernels, respectively. Files are stored and named indicating the parameters used to perform the simulation and the time window for which the kernels are calculated. For example: OUTPUT_KERNELS/90 0 0_30 0 0 /kernel_homo_30 0 0_1551_SVF7e-1Hz_03-06.dat is the kernel corresponding to the time window (3,6) seconds of the seismogram collected by the station located at (30 0 0,1551) from a simulation using the homogeneous velocity model and the 0.7 Hz seismic source applied as a single vertical force at position (90 0 0,30 0 0),

Experimental Design, Materials and Methods
The spectral elements based software SPECFEM2D ( [1] ) was employed to perform 2D numerical simulations of seismic wave propagation in 2D models of Mount Etna, Italy, resulting in synthetic displacement and velocity seismograms. In addition to these seismograms, associated travel time sensitivity kernels were calculated using the adjoint methodology presented in [8] and implemented as well in SPECFEM2D. The seismograms were calculated for a selected source-receiver pair and the associated kernels were calculated for every three seconds time windows of the seismograms. These kernels show the areas of the volcano that mostly affect the wavefield arriving at the seismic stations within the seismogram time window under investigation. The calculation procedure can be described in four steps: 1. Creation of the two 2D models of Mount Etna, Italy. The digital topography was obtained from the CGIAR consortium for spatial information ( topograpy.txt ). Two velocity models were used in the calculation: homogeneous and heterogeneous. For the homogeneous model seismic velocities Vp = 3500 m/s and Vs = 2384 m/s and density = 20 0 0 kg/m3 were used. For the heterogeneous one S-wave velocities were extracted from [5] for the first 360 m depth and deeper velocities from [6] ( material_properties.txt ). Since the Poissons ratio is not well constrained (e.g. [9][10][11][12][13] ), the typical value for the Earth's crust of 0.25 was used to calculate the corresponding P wave velocities. Although it may affect details of the calculated seismograms, such a choice did not affect the core findings of this work. The densities were calculated using the Gardner relationship ( [7] ) where v p is measured in m s −1 and density in kg m −3 .
2. Definition of the seismic sources. Since the objective of this work is to study low frequency volcanic seismicity and structure tuning effects, the data were calculated for two different source depths, two different source mechanisms and two different source-time functions ( source_location.txt, source_type.txt ). The shallow source was located at 113 m deep (within the low velocity superficial layer) and the deep one at 3613 m deep. The source was a Ricker wavelets applied either as a single vertical force or as an explosion. We also considered two different source central frequencies, 0.7 Hz and 1 Hz. Such closely spaced frequencies were chosen to test sensitivity to small central frequency differences in the most common range of long period (LP) frequencies. 3. Definition of the seismic network. Seismic network contained 15 stations along the volcano surface with the interstation spacing of 10 0 0 m ( stations.txt ). 4. Running simulations. Utilizing the 2D models, the sources and the stations described previously, systematic forward and adjoint simulations of wave propagation were performed by SPECFEM2D for different combinations of source locations, mechanisms, frequencies and velocity models. Sensitivity kernels were calculated using the same software package. These simulations resulted in the seismograms collected by each of the seismic stations (folder OUTPUT_SEISMOGRAMS ) and the sensitivity kernels for each 3 seconds time windows of the investigated time window. These kernels highlight the regions in the subsurface velocity models which have the strongest effect on each time segments in the seismogram (folder OUTPUT_KERNELS ). Meshed velocity models detailed for each point of the 2D model structure ( homo_velocity_model.txt and hetero_velocity_model.txt ) and the source time function corresponding to each of the two frequency sources ( source_7e-1Hz.txt and source_1e0Hz.txt ) were also provided.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships which have, or could be perceived to have, influenced the work reported in this article.