OpenHVSR User Manual (Ver. 2.0.

6)

Samuel Bignardi Ph. D.

December 10, 2015

Last update: October 1, 2018.

 1

About OpenHVSR

Credits
OpenHVSR is an algorithm developed in Matlab® (Release 2010a) by Ph.D.
Samuel Bignardi. If you use this code for research purposes, please consider to
support my research by citing:

ˆ Bignardi, S.; Yezzi, A. J.; Fiussello, S.; Comelli, A. (2018). OpenHVSR -

Processing Toolkit: Enhanced HVSR processing of distributed microtremor
measurements and spatial variation of their informative content.
Comput-
ers & Geosciences. vol. 120, p. 10-20. DOI: 10.1016/j.cageo.2018.07.006

ˆ Bignardi, S. (2017). The uncertainty of estimating the thickness of soft

sediments with the HVSR method: A computational point of view on weak
lateral variations.
Journal of Applied Geophysics, Vol. 145, p. 28-38.
DOI:10.1016/j.jappgeo.2017.07.017

ˆ Bignardi, S.; Mantovani, A.; Abu Zeid, N. (2016). OpenHVSR: imaging

the subsurface 2D/3D elastic properties through multiple HVSR modeling
and inversion. Computers & Geosciences.

ˆ Abu Zeid, N.; Corradini, E.; Bignardi, S.; Nizzo, V.; Santarato, G. (2017).
The passive seismic technique HVSR as a reconnaissance tool for map-
ping paleo-soils: the case of the Pilastri archaeological site, Northern Italy.
Archaeological Prospection. DOI: 10.1002/arp.1568

ˆ Bignardi, S.; Fedele, F.; Santarato, G.; Yezzi, A. J.; Rix, G. J. (2013).
Surface waves in laterally heterogeneous media. Journal of engineering me-
chanics. Vol. 139; p. 1158-1165, ISSN: 0733-9399, DOI: 10.1061/(ASCE)
EM.1943-889.0000566.

ˆ Abu Zeid, N., Corradini, E., Bignardi, S., Santarato, G., 2016, Unusual
geophysical techniques in archaeology - HVSR and induced polarization,
a case history, 22nd European Meeting of Environmental and Engineering
Geophysics, NSAG-2016, DOI 10.3997/2214-4609.201602027.

ˆ Bignardi, S., Abu Zeid, N., Corradini, E., Santarato, G., 2017, The HVSR
technique from array data, speeding up mapping of paleo-surfaces and
buried remains: The case of the Bronze-Age site of Pilastri (Italy), SEG
Technical Program Exp. Abstracts 2017, 5119-5124, DOI 10.1190/segam2017-
17746745.1.

ˆ Abu Zeid, N., Bignardi, S., Santarato, G., Peresani, M., 2017b, Exploring
the paleolithic cave of Fumane (Italy): Geophysical methods as planning
tool for archaeology, SEG Technical Program Expanded Abstracts 2017,
5125-5129, DOI 10.1190/segam2017-17729320.1.
 2

ˆ Bignardi, S., Abu Zeid, N., Corradini, E., Santarato, G., 2017, The HVSR
technique from array data, speeding up mapping of paleo-surfaces and
buried remains: The case of the Bronze-Age site of Pilastri (Italy), SEG
Technical Program Exp. Abstracts 2017, 5119-5124, DOI 10.1190/segam2017-
17746745.1.

ˆ Bignardi, S. Fiussello, S., Yezzi, A., 2018, Free and improved computer
codes for HVSR processing and inversions, 31st Symposium on the Appli-
cation of Geophysics to Engineering and Environmental Problems, (SAGEEP
2018), Nashville Tennessee, USA March 25-29.

ˆ Obradovic, M., Abu Zeid, N., Bignardi, S., Bolognesi, M., Peresani, M.,
Russo, P., Santarato, G., 2015; High resolution geophysical and topo-
graphical surveys for the characterization of Fumane Cave Prehistoric Site,
Italy, Near Surface Geoscience 2015, DOI, 10.3997/2214-4609.201413676.

ˆ Herak, M, (2008). ModelHVSRA Matlab tool to model horizontal-to-

vertical spectral ratio of ambient noise. Computers & Geosciences vol. 34,
pp 1514-1526.

ˆ Lunedei, E. and Albarello, D. (2010). Theoretical hvsr curves from full

waveeld modelling of ambient vibrations in a weakly dissipative layered
earth. Geophys. J. Int., 181:10931108.

License
OpenHVSR is free software: you can redistribute it and/or modify it 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.
OpenHVSR is distributed in the hope that it will be useful, but WITHOUT
ANY WARRANTY; without even the implied warranty of MERCHANTABIL-
ITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General
Public License for more details.
You should have received a copy of the GNU General Public License along
with Nome-Programma. If not, see <http://www.gnu.org/licenses/>.

Aknowledgements
OpenHVSR includes some publicly available routines:

ˆ To evaluate the condence on the results, we included the CONF_LIM

routine, available in the program ModelHVSR [4]. ModelHVSR is avail-
able for download from multiple sites (for e.g. https://sites.google.com/
site/prinstessa/software-freeware ) or by contactinh the author.

ˆ he modeling routine implements the same modeling strategy present in

ModelHVSR [4] , which, in turn, is based based on [6, 7]. While in Mod-
elHVSR a library compiled for 32 bit computers was provided along with
 3

the corresponding mex-fortran source code; here that code was ported to
Matlab so that users on 64 bit machines can avoid dealing with mex les
compilation and to ensure compatibility with any future release of Matlab.

ˆ Routines for smoothing 2D data were written by Carlos Adrián Vargas

Aguilera, Universidad de Guadalajara, México. They were downloaded
from the Matlab Central, http://www.mathworks.com/matlabcentral/leexchange/.

What is OpenHVSR
OpenHVSR is a computer program developed in the Matlab environment, de-
signed for the simultaneous modeling and/or inversion of large, Horizontal-to-
Vertical Spectral Ratio (HVSR or H/V) datasets and with the purpose of con-
structing 2D/3D subsurface models (topography included). The program is
designed to provide a high level of interactive experience to the user and still to
be of intuitive use
Contents
0.0.1 To start the program . . . . . . . . . . . . . . . . . . . . . 4
0.0.2 Knowh issues . . . . . . . . . . . . . . . . . . . . . . . . . 5
0.0.3 Load an example . . . . . . . . . . . . . . . . . . . . . . . 6
0.0.4 Create your own project . . . . . . . . . . . . . . . . . . . 7
0.0.4.1 The project-le . . . . . . . . . . . . . . . . . . 7
0.0.4.2 Data-les . . . . . . . . . . . . . . . . . . . . . . 9
0.0.4.3 Subsurface-les . . . . . . . . . . . . . . . . . . . 10
0.1 Interface description and use . . . . . . . . . . . . . . . . . . . . 10
0.1.1 Tab 1. Main parameter settings and 2D prole optimization 10
0.1.2 Tab 2. Local 1D model optimization . . . . . . . . . . . . 13
0.1.3 Tab 3. Prole view . . . . . . . . . . . . . . . . . . . . . . 16
0.1.4 Tab 4. Condence . . . . . . . . . . . . . . . . . . . . . . 16
0.1.5 Tab 5. Single parameter sensitivity . . . . . . . . . . . . . 17
0.2 Menus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
0.2.1 Menu Files . . . . . . . . . . . . . . . . . . . . . . . . . . 18
0.2.2 Settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
0.2.3 view . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
0.3 Suggestions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

Bibliography 20

0.0.1 To start the program

Place the program at any location in your computer, say for example it is placed
in /my/folder/OpenHVSR/, then

1. Start Matlab (R2010a or higher).

2. Navigate to the /my/folder/OpenHVSR/ folder

3. open the le START_OpenHVSR.m

4. The program is set to 2D mode by default. If you want enable the 3D

mode, just uncomment the line as follows:

4
CONTENTS 5

mode = ' 2D ' ; mode 2D e n a b l e d

%mode = ' 3D ' ;

or

%mode = ' 2D ' ;

mode = ' 3D ' ; mode 3D e n a b l e d

5. run START_OpenHVSR.m

0.0.2 Knowh issues

The program is continuously improved and, despite I strive to carefully test
every new feature, some minor error may occur. Known issues are typically
related to the interface and does not impact the computational routines, which
are carefully tested.
Some issues that the user may experience:
ˆ Text used throughout this program, especially as input les are supposed
to be encoded in plain ISCII or 'UTF-8'. Other encodings may cause
misbehavior.
ˆ Due to changes between dierent Matlab versions, issues may arise when
the program is used on Matlab versions other than 2012b. Typically con-
nected with the Matlab functions uitab, griddata, and boundary.
ˆ From time to time, Matworks declares some of their functions as depre-
cated. If this occurs, the command window will display some warnings
to inform the user than the related functions will be removed in future
Matlab releases. The user should not be concerned with these warnings.
ˆ I implementd right-click context menus on most the images on the program
interface. To take advantage of Matlab's graphical tools, I retained the
Matlab graphical toolbar on the top part of the program interface. This
toolbar contains functions such as: Zoom In, Zoom out, and others
(Figure [g:Matlab-toolbar]). User must be advised that, if any of these
tools is selected, right-clicking on the gure will open the tool's context
menu rather than the desired gure's menu.
ˆ Matlab 2016b (and later), on Windows: Some users reported that, when
using the functions related to model and inversion of surface waves (but-
tons Model SW and START inversion (SW)), Matlab fails in nd-
ing the executable microtrem.exe which is launched inside the function
call_Albarello_2011.m. Various precautions have been taken to avoid
the problem. However, since I do not have Matlab 2016b, the solutions
adopted have not been fully tested.
CONTENTS 6

Figure 1: Matlab's graphical toolbar

If any problem occurs, please do not hesitate to contact me at the following

email: [email protected] .

0.0.3 Load an example

We provide three examples which can be found in OpenHVSR/EXAMPLES
folder. Examples 1 and 2 correspond to the same examples published in [3]. To
load the examples,

1. Go to the menu les and select  Load HVI Project 

2. Navigate to the example you want to open and select the project.m le,
then click Open

Example1_Simulated_data_Figures_6and7 (2D mode): This is a sim-

ulated data example, i.e. HVSR curves presented here were simulated.
In this example nine dierent HVSR curves are assumed to be obtained along
a linear prole and spaced at 100 m intervals.
The project-le project.m instructs OpenHVSR to load nine data les and
nine les describing the subsurface.
In this case the provided subsurface model is ready for starting the inversion.

Example2_real_data_Figures_9and10 (2D mode): In this example

eight dierent HVSR curves obtained along a linear prole and spaced at 1 km
intervals are provided.
The project-le project.m instructs OpenHVSR to load the eight data les
and eight les describing the subsurface.
In this case the provided subsurface model is the best model we found opti-
mizing for the frequency band 0.4-6.0 Hz.

Example3_Simulated_data_for_3D (3D mode): In this example the

same HVSR curves of example 1 are used, but the survey geometry (and conse-
quently the project-le), was changed to illustrate a 3D subsurface with topog-
raphy.
CONTENTS 7

0.0.4 Create your own project

0.0.4.1 The project-le
Input to OpenHVSR is providede by a project-le which requires a basic under-
standing of Matlab language.
Such project-le species the locations of measurements, data lenames
and an initial subsurface model under the measurement points. All the ex-
amples in the OpenHVSR/EXAMPLES/ folder contain project-les which
should be self explanatory, however, a minimal project-le structure is reported
in table 1, which describes a survey where three HVSR-curves (hvsr_1.txt,
hvsr_2.txt,hvsr_3.txt ), obtained from recordings performed along a line and
100 m spaced are located in the DATA folder (lines 1 to 9). The initial
subsurface assumed under hvsr_1.txt and hvsr_2.txt is described in the le
 subsurface_A.txt  (lines 12,13), while the initial subsurface assumed under
hvsr_3.txt is described in the le  subsurface_B.txt  (lines 15,16).

1 % Data S e c t i o n :
2 SURVEYS{ 1 , 1 } = [ 0 0 0 ] ; % Location [ x , y , z ] .
3 SURVEYS{ 1 , 2 } = 'DATA/ hvsr_1 . txt ' ; % HVSR c u r v e .
4
5 SURVEYS{ 2 , 1 } = [ 1 0 0 0 5 ] ;
6 SURVEYS{ 2 , 2 } = 'DATA/ hvsr_2 . txt ' ;
7
8 SURVEYS{ 3 , 1 } = [ 2 0 0 0 5 ] ;
9 SURVEYS{ 3 , 2 } = 'DATA/ hvsr_3 . txt ' ;
10
11 % I n i t i a l Subsurface Section
12 MODELS{ 1 , 1 } = 'MODELS/ subsurface_A . txt ' ;
13 MODELS{ 1 , 2 } = [ 1 2 ] ;
14
15 MODELS{ 2 , 1 } = 'MODELS/ subsurface_B . txt ' ;
16 MODELS{ 2 , 2 } = [ 3 ] ;
17
18 % optional section
19 % datafile_columns = [2 3 6 ] ;
20 % d a t a f i l e _ s e p a r a t o r = ' data_begins_here ' ;

Table 1: Example of a minimal input project-le

A project-le is a regular ASCII le, except that in order to be used as

Matlab script it must have the .m le extension.
Two main variables must be dened here:
ˆ SURVEYS
ˆ MODELS
CONTENTS 8

These two variables use the cell feature of Matlab which creates a container
that has the same structure of a matrix but it can contain dierent sort of data,
for example,

SURVEYS{ 3 , 2 }

is the entry of the SURVEYS cell at line 3 and column 2.

The SURVEYS variable in the project-le contains a description of one
measurement for each row; column 1 contains the coordinates (x,y,z) of a mea-
surement location (organized as a 1x3 vector and enclosed in squared brackets)
and column 2 contains the path to the corresponding hvsr le. For example

2 SURVEYS{ 1 , 1 } = [ 0 0 0 ] ; % Location [ x , y , z ] .
3 SURVEYS{ 1 , 2 } = 'DATA/ hvsr_1 . txt ' ; % HVSR c u r v e .

means that the curve contained in le DATA/hvsr_1.txt, was obtained

from the recordings performed at location (0,0,0).
Following the same strategy, the initial subsurface model is inputted through
the MODELS variable. Every row of this variable describes a locally 1D
subsurface model. The column 1 contains the path to the le describing the
subsurface, while column 2 contains a list of IDs (organized as a row vector)
which correspond to the rows of the SURVEYS variable that are associated with
the subsurface model. For example

12 MODELS{ 1 , 1 } = 'MODELS/ subsurface_A . txt ' ;

13 MODELS{ 1 , 2 } = [ 1 2 ] ;

means that the subsurface model contained in le MODELS/subsurface_A.txt

must be associated to the measurements described at rows 1 and 2 of the
variable SURVEYS. In other words, the curves in DATA/hvsr_1.txt and
DATA/hvsr_2.txt les, corresponding to locations (0,0,0) and (100 0 5) re-
spectively, are both associated to the same initial subsurface model described
in le MODELS/subsurface_A.txt.
Finally, variables

ˆ datale_columns

ˆ datale_separator

are optional and can be used to load dierent data-le structures, provided that
data-les are in ASCII format. By default, the program reads two-columns data
les, with frequency values and corresponding HVSR curve amplitudes stored
in the rst and second column respectively. If the input les include an header
CONTENTS 9

(footer is not allowed), followed by a table of data, these can be read by specing
the structure of the le, either from the interface (through the "Files/HVSR
les format" menu) or by uncommenting the lines 19,20 in the project-le. The
datale_separator variable can be any string. The user must insert this string in
every data-le, between the header and the data sections, in order to indicate to
the program where data begin. Finally, the datale_columns variable species
the column ID's corresponding to the frequecy scale, to the HVSR curve and to
the standard deviation. For example

19 datafile_columns = [2 3 6 ] ;
20 d a t a f i l e _ s e p a r a t o r = ' data_begins_here ' ;

means that OpenHVSR will look for the string data_begins_here in order
to nd the data-section of the le. Further, It will read the data section as a
matrix and it will use the data in the column 2 as frequency scale, column 3 as
the HVSR curve amplitude and nally, colum 6 as standad deviation.

0.0.4.2 Data-les
Almost any data le structure is loadable provided it is an ASCII / UTF-8 le.
Files containing a header can be loaded with minimum eort by placing any
string between the header and the data section of the le, for example in the
following data-le

S i t e ID : DPC_INGV2014_MAE,
Instrument : EXT−
No . o f c h a n n e l s : 3
Sampling r a t e : 125 Hz
Start recording : 25/07/14 0 4 : 0 0 : 0 0
End r e c o r d i n g : 25/07/14 0 4 : 3 0 : 0 0
Trace l e n g t h : 0h50 ' 0 0 ' '
Channel l a b e l s : NORTH−SOUTH EAST−WEST UP−DOWN
f r e q . H/V Std . Dev . 95% c o n f . i n t . Vel . Ampl . Spec .
data_begins_here
0.01525 0.000 0.000 0.000 0.000
0.03051 0.000 0.000 0.000 0.000

the string data_begins_here was chosen and inserted as separator. The

user must then tell OpenHVSR that the data-les include a header, either from
the interface (thrugh the menu Files/HVSR les format) or by inserting the line
datale_separator = 'data_begins_here'  in the project-le.
OpenHVSR assumes that the frequency scale, the H/V curve and the stan-
dard deviation occupy the columns 1, 2 and 3 respectively, however, the user
CONTENTS 10

can customize the column order by modifying the datale_columns variable

as explained in section 0.0.4.1.
Starting September 27, 2018, the program accepts HVSR datasets which
curves are dened on dierent frequency scales.

0.0.4.3 Subsurface-les
The subsurface (see for e.g. table 2) is inputted using one or more ASCII les
where subsurface visco-elastic parameters are organized in a table. Each row
corresponds to a layer and the last one corresponds to the half-space. Columns
1 to 6, correspond to:

ˆ Vp (compressive P waves velocity)

ˆ Vs (shear S waves velocity)

ˆ Rho (density)

ˆ H (thickness)

ˆ Qp (P-waves daming)

ˆ Qs (P-waves daming)

Note that the values of H, Qp, Qs must be set to 999. for the Half-space.

230. 115. 1.80 2. 15. 5.

750. 250. 1.80 7. 30. 10.
2100. 915. 1.90 999. 999. 999.

Table 2: Example of local 1D subsurface model

0.1 Interface description and use

The algorithm is composed of several routines integrated into a main graphical

user interface (GUI), which is organized in tabs.

0.1.1 Tab 1. Main parameter settings and 2D prole op-

timization
Tab 1 (g. 2), is dedicated to the general settings for the simultaneous model-
ing and/or inversion of a set of HVSR curves. In case of inversion, the input
data consist of HVSR curves which are obtained after the elaboration of the
corresponding eld data by using any third party software (e.g. Grilla, Geopsy,
Jsesame, etc.).
CONTENTS 11

Figure 2: View of tab 1: modeling and inversion setup.

Setup of the forward modeling routine

The portion of the left panel (a), contains the control parameters necessary to
setup of the forward model routine (FWD). For each location of the survey, the
FWD calculates the theoretical transfer function of a layered subsurface model
based on Tsai's approach and enhanced to manage frequency-dependent atten-
uation and body-wave dispersion [6, 7, 4]. The modeling algorithm represents
the subsurface as a stack of viscoelastic homogeneous layers over a half space.
Each layer is described by its thickness, H, and its mechanical behavior dened
by the density (Rho), the compressive and shear wave velocities (Vp , Vs) and
the corresponding attenuation factors (Qp, Qs) which are frequency dependent
and follows

Q = Q0 f k (1)
where Q0 is the attenuation factor at 1 Hz and k is a constant which is
assumed to be xed for every location. Finally body waves dispersion, is con-
sidered trough the logarithmic low [1]
  
−1 f
V (f ) = V (fref ) 1 + (πQ0 ) ln (2)
fref
Text-boxes at (a) allow to set the constant k , the reference velocity V (fref ),
and the frequency band and frequency step to simulate the HVSR curves. The
stand-alone modeling can be started by simply pressing the MODEL button
(f).
CONTENTS 12

Setup of the inversion

The remaining controls of this tab are used to setup the inversion. The inversion
strategy is based on the guided Monte Carlo method, where at every iteration a
randomly perturbed version of the best tting model (i.e. the model which best
reproduces the data) is produced and used to compute a set of simulated curves
to be compared with the experimental HVSR curves. The generation of many
trial models allows exploring the parameters space while looking for a new and
better tting model. The controls placed on this tab allow the user to:

ˆ set the maximum percentage of change to be associated to each family of

parameters (e)

ˆ set the statistical distribution to be used to generate the perturbation;

either uniform or normal (d)

ˆ set the weights to be associated to dierent portions of the HVSR curve

(i.e the weights on data) in case the user wants to focus on a particular
portion of the HVSR curve (panel b). Right-clicking on this panel open a
context menu which allow to modify the Frequency Weighting Function.

ˆ set the perturbation of the parameters as depth-dependent. This is done

by modifying the Depth Weighting Function of panel (c). For each depth
z the eective amount of perturbation will be the value of this function at
z multiplied by the maximum percentage of change. Right-clicking on this
panel open a context menu which allow to modify the Depth Weighting
Function.

ˆ Specify which degree of freedom are locked (i.e. kept xed) during inver-
sion. The Show Lock Table button opens the Lock Parameters dialog
window in gure 3. Here the locked parameters are shown on the right
top table as unchecked. The right bottom table on the other hand can
be used to relax the constrain on the maximum ratio Vp /Vs , so enabling
higher Vp values in the selected layers in order to account for the water
table.

Finally, since this inversion is focused on retrieving a 2D or 3D subsurface,

we need a criterion to decide which set of simulated curves best reproduces the
experimental data. This is achieved by calculating the value of an objective
function (eq. 3) which is a positive, real value function of the subsurface pa-
rameters Vp, Vs, Ro, Qp, Qs, and H (called collectively m ), and seeking its
global minimum. Indeed, the objective function can be expressed as
5
X
E(m) = aM (m) + bS(m) + αj Rj (mj ) (3)
j=1

where the rst term on the r.h.s. represents the mist between the data and
the simulated curve. The second term introduces into the objective function a
CONTENTS 13

Figure 3: Lock Table

forcing condition on the slope of the curves, and the third term implements a
smooth lateral variation constraint on a particular family of parameters mj = Vp
, Vs , Rho, Qp , Qs , for j = 1 to 5 respectively. For details on equation 3, please
refer to [3].
Parameters a and b can be set using the menu  Settings/Objective function 
(see section 0.2.2) while the parameters αj , which highlight the importance of
the regularizing term for the dierent subsets of parameters are usually chosen
by the user, even though an automatic selection strategy is available by pressing
the Automatic weighting button. When pressed, a dialog window opens which
asks the user to set the amount of lateral variation expected for each family of
parameters.
As a nal note, it is important to remember that the settings of tab 1 will
also aect all other tabs.

0.1.2 Tab 2. Local 1D model optimization

Tab 2 (g. 4), is devoted to the independent optimization of the local 1D
subsurface at the site of the specic HVSR curve at hand.
CONTENTS 14

Figure 4: View of tab 2: Independent optimization of a specic (local) site.

Tab 2 shows:
ˆ an aerial view (a) of all measurement sites (i.e. the location associated to
each HVSR curve). The selected location is highlighted with a red circle.
Further the user can navigate through the sites by using the buttons with
the left/right arrows (b). Sites are normally indicated in blue but the
user is allowed to lock any site in order to prevent further modications
or in order to introduce an abrupt discontinuity when it is known to exist.
To lock/unlock a location use the Lock and Unlock buttons (b); the
locked sites will turn red. Finally, a curve can be discarded (i.e. not
considered) by using the Disable and Enable buttons. As a nal note,
when using the program in 3D mode (see 0.0.1), the user is allowed to
dene a customized linear prole by using the dene prole option of
the context menu which is opened by right-clicking on panel (a)
ˆ a set of controls (b and c) along with the name of the visualized data-le
ˆ the central panel (d) shows the data space. Here the experimental curve
(black) is shown along with its experimental uncertainty curves (gray, if
available). On the same plot, the stand-alone modeled curve (green) which
is obtained by pressing the MODEL button of Tab 1 and the simulated
curves corresponding to the best tting model (red curve) and to the last
model generated during the Monte Carlo inversion (blue curve) are shown
ˆ the table describing the current subsurface (e)
ˆ the 1D prole of one media property (f), which can be selected by pressing
the buttons on top.
ˆ the mist-vs-number of generated Monte Carlo trials plot (panel g), where
the vaue of the mist (i.e. aM + bS , see eq. 3), expressed as percentage
CONTENTS 15

Figure 5: Model Manager

of the initial mist, is plotted against the number of trial model generated
(black line). The red and green lines, plot the behavior of terms aM and
bS respectively.

To run the inversion algorithm press the Optimize independently button (c).
When inversion is running, all tables and graphs can be set to either updated
every time a new best model is found or dynamically, using the toggle real-time
radio button (c). The inversion can be paused at any time and once stopped, the
user is allowed to: inspect the 1D graph window (f), where the N most tting
models (N is settable) are plotted along with the best match. Further when
a satisfactory t is reached for a given location the user is allowed to extend
the local subsurface model to the previous/sucessive location (using the spread
left/right buttons) or even to the entire prole (spread button). Furthermore,
the experienced user is also allowed to manually correct the subsurface model
both on the basis of their own knowledge about the local geology, or alternatively
after using the tools of tabs 4 and 5 (discussed later).
To open the Model Manager window (g. 5), right-click on the model table
(e) and select modify. Once open, the user is allowed to manually edit the
model table and test the eect of the correction prior to deciding whether to
keep or discard the introduced changes.
Finally, if for any reason the user is not satised with the evolution of the
inversion process, the Revert button will restore the initial subsurface model.
CONTENTS 16

0.1.3 Tab 3. Prole view

Tab 3, (Fig. 6) is a 2D/3D view of the best tting subsurface. For a 3D
subsurface, the volume is fully sliceable and navigable.

Figure 6: Tab 3: 2D/3D subsurface model view

0.1.4 Tab 4. Condence

Tab 4 (Fig. 7) allows for plotting the condence of the results.

Figure 7: View of tab 4. This shows the condence of the best model as a
function of two selected parameters using the function CONF_LIM [4, 2, 5].
When the two parameters pertain to dierent measurement locations, this tool
can be used to gain condence on lateral variations.
CONTENTS 17

This plot (shown in panel d) is based on theCONF_LIM routine [4] which,

in turn, is based on the F-distribution computation [5, 2]. Since the parameter
space is a multidimensional space, the condence plot allows the user to inspect
the distribution of mist values with respect to two chosen degrees of freedom
(d.o.f). The rst d.o.f is selected using controls on panel (a). The buttons with
the arrows allow to select which site to investigate, while the two drop-down
lists allow to select the visco-elastic parameter and the layer under investigation.
The second d.o.f is selected in the same way using panel (b), while graphical
properties of the plot are set using the controls of panel (c). When the mist
is plotted as function of two d.o.f. pertaining to the same site (i.e. to the
same HVSR curve and the same local 1D subsurface), the result of the gure
is exactly the same of Herak's. The consistent improvement demonstrated by
this method is that now the condence plot can be expressed as a function of
d.o.f. pertaining to dierent sites thus allowing the evaluation of the condence
regarding any found lateral heterogeneity.

0.1.5 Tab 5. Single parameter sensitivity

Finally, Tab 5 (Fig. 8) implements a novel strategy to investigate the condence
on the result. It allows for plotting the variation of mist (rst two term on
r.h.s of eq. 3) with respect to a chosen parameter m, retaining the dependency
on the frequency. This allows to judge how each portion of the HVSR curve
is aected by a particular parameter (plot on panel c). Further, positive and
negative variations, plotted in red and blue respectively, are associated to an
increase and decrease of the frequency dependent contributions to the mist.
CONTENTS 18

Figure 8: View of Tab 5. The frequency dependent variation of the mist with
respect to a single parameter (d.o.f) variation is shown (c). We refer to this
as single parameter sensitivity and it can be used to understand the correct
variation to be introduced into a model when improbable parameter values are
present

The d.o.f. under investigation is set using the controls (a) while the amount
of variation of the parameter is set using the controls (b).

0.2 Menus

Main aspects of the program menus:

0.2.1 Menu Files

ˆ Load HVI Project : load a project-le from any point of the computer

ˆ HVSR les format : allow to specify a string (that the user must insert
beforehand into the HVSR data-les) to be used as header/data separator
and allow to specify the columns where frequency scale and HVSR curve
are stored. This potentially allows the program to read any le format

ˆ Save as new project : allows to save the current subsurface and data
as a completely new project that can be used independently. This is
particularly useful when adding layers to the subsurface in order to rene
the result

ˆ Save Elaboration : saves the status of the current elaboration into a .mat
(Matlab) le format for later use

ˆ Resume Elaboration : resume an elaboration stored into a .mat le

CONTENTS 19

Figure 9: Setup Manager

0.2.2 Settings
ˆ Setup : open the Setup Manager window (g. 9) to set the physical
constraints of the subsurface visco-elastic parameters and allows to specify
the target Earthquake

ˆ Objective function : allows to setup parameters a and b of equation 3.

0.2.3 view
ˆ HVSR: allows to switch between dierent visualizations of the HVSR curve

ˆ prole: set graphical options

ˆ coormap: allows to switch between colormaps

0.3 Suggestions

ˆ We suggest to perform Inversions using multiple steps and using the Save
as new project from one step to another. Start with a subsurface model
with few layers and limit the curve inversion to low frequencies. Suc-
cessively, when a good model is found, save as new project and edit the
subsurface les splitting the deeper layer in two or three sublayers. Use
the new project with thicknesses kept xed but widening the frequency
band to higher frequencies. Then, continue with the next, shallower layer,
and so on. Usually this strategy reveal to be more eective for curves with
the main peak at frequencies higher than 2 Hz.
CONTENTS 20

ˆ Changing set inversion parameters (such as weighting functions, frequency

step or frequency band), during inversion is highly discouraged. However,
this not aect the nal result if the change results with a lower value of
the mist function. We encourage to save your work as a new project
before changing any inversion parameter.

ˆ Look at the subsurface model in table (Tab 2, e). When a certain portion
of the HVSR curve is inverted by the Monte Carlo method, the elastic
parameters of each layer are randomly perturbed. When a good match
between the data and the simulated curves is found, the subsurface model
that produced the synthetics is accepted. Unfortunately, if some parame-
ter has a poor inuence on the shape of the simulated curve and in turn,
on the mist (i.e. has a low sensitivity), a good match can be found even
if an improbable value of such parameter is present. The tools of tabs 4
and 5 allow the user to judge if an improbable parameter value should be
changed in order to obtain a more plausible model.
Bibliography
[1] K. Aki and P. G. Richards. Quantitative seismology, Theory and methods.
University Sciences Books, Susalito, California, second edition, 2002.

[2] F. Bianco, E. Del Pezzo, M. Castellano, J. Ibanez, and F. Di Luccio. Separa-

tion of intrinsic and scattering seismic attenuation in the southern Apennine
zone, Italy. Geophysical Journal International, 150:1022, 2002.

[3] S. Bignardi, A. Mantovani, and N. Abu Zeid. OpenHVSR: imaging the

subsurface 2D/3D elastic properties through multiple hvsr modeling and
inversion. Computers & Geosciences, 1:110, 2016.

[4] M. Herak. ModelHVSR-A Matlab tool to model horizontal-to-vertical spec-

tral ratio of ambient noise. Computers & Geosciences, 34:15141526, 2008.

[5] K. Mayeda, S. Koyanagi, M. Hoshiba, K. Aki, and Y. Zeng. A comparative

study of scattering, intrinsic, and coda q-1 for hawaii, long valley and central
california between 1.5 and 15 Hz. Bulletin of the Seismological Society of
America, 97:66436659, 1992.

[6] N.C. Tsai. A note on the steady-state response of an elastic half-space.

Bulletin of the Seismological Society of America, 60:795808, 1970.

[7] N.C. Tsai and G.W. Housne. Calculation of surface motions of a layered
half-space. Bulletin of the Seismological Society of America, 60:16251651,
1970.

21