Vous êtes sur la page 1sur 40

Accepted Manuscript

Geomorphometric comparative analysis of Latin-American volcanoes

Sergio Camiz, Maurizio Poscolieri, Matteo Roverato

PII: S0895-9811(17)30068-8
DOI: 10.1016/j.jsames.2017.02.011
Reference: SAMES 1662

To appear in: Journal of South American Earth Sciences

Received Date: 15 March 2016


Revised Date: 20 January 2017
Accepted Date: 20 February 2017

Please cite this article as: Camiz, S., Poscolieri, M., Roverato, M., Geomorphometric comparative
analysis of Latin-American volcanoes, Journal of South American Earth Sciences (2017), doi: 10.1016/
j.jsames.2017.02.011.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT

Geomorphometric comparative analysis of Latin-American Volcanoes


Sergio Camiz1,2, Maurizio Poscolieri2*, Matteo Roverato3

1
Dipartimento di Matematica “Guido Castelnuovo”, Sapienza Università di Roma - Italy
2
CNR - Istituto di Acustica e Sensoristica “O.M. Corbino”, Roma - Italy

PT
3
IGC Instituto de Geociências, Universidade de São Paulo - Brazil

RI
Keywords: DEM, geomorphometric classification, topographic gradients, volcanoes

*e-mail corresponding author: maurizio.poscolieri@idasc.cnr.it

SC
Abstract

U
The geomorphometric classifications of three groups of volcanoes situated in the Andes
Cordillera, Central America, and Mexico are performed and compared. Input data are
AN
eight local topographic gradients (i.e. elevation differences) obtained by processing each
volcano raster ASTER-GDEM data. The pixels of each volcano DEM have been
M

classified into 17 classes through a K-means clustering procedure following principal


component analysis of the gradients. The spatial distribution of the classes, representing
D

homogeneous terrain units, is shown on thematic colour maps, where colours are
TE

assigned according to mean slope and aspect class values. The interpretation of the
geomorphometric classification of the volcanoes is based on the statistics of both
gradients and morphometric parameters (slope, aspect and elevation). The latter were
EP

used for a comparison of the volcanoes, performed through classes’ slope/aspect


scatterplots and multidimensional methods. In this paper, we apply the mentioned
C

methodology on 21 volcanoes, randomly chosen from Mexico to Patagonia, to show


AC

how it may contribute to detect geomorphological similarities and differences among


them. As such, both its descriptive and graphical abilities may be a useful complement
to future volcanological studies.

Keywords: Geomorphometry, ASTER-DEM, Local topographic gradients,


Classification, Comparison, Volcanoes, Andes, Central-America, Mexico.

1
ACCEPTED MANUSCRIPT

1. Introduction

The study of landforms can take advantage from the study of terrain structures through
Digital Elevation Models (DEM) (Székely and Karátson, 2004; Wilson, 2012). With
their advent and their constant improvement and reliability, geomorphometry as
quantitative land-surface analysis is continuously developing; thus, it is now considered

PT
a scientific discipline on its own (Pike et al., 2009) and it is applied at large in several
fields, such as geomorphology, hydrology, soil science, vegetation science, and

RI
meteorology (Hengl and Reuter, 2009). In this framework, the classification of terrain
units as obtained through DEM analysis deserves a specific interest, in order to describe

SC
the structure of the study area and to characterize its homogeneous parts. Moreover, a
classification of the DEM elements (i.e. the pixels in the raster image), according to
geomorphometric parameters, may be taken into account to explain the behaviour of

U
other characters observed at pixel level (Janssen et al., 1990; Florinsky and Kuryakova,
AN
1996; Höersch, 2002; Garcia-Aguirre et al., 2007; Efiong et al., 2016), this way
contributing to a better understanding of the specific problem at hand. We find in
literature several approaches that take into account different characters; those which aim
M

at defining borders of homogeneous terrain units (Pike, 2002), are based on landform
attributes that include slope, aspect, and elevation. These morphometric parameters may
D

be easily derived from the DEM, as well as their derivatives, such as horizontal, vertical,
TE

and mean curvatures, etc. (Olaya, 2009; Florinski, 2012; Evans, 2013, Pedersen, 2016).

With respect to these, a methodology was proposed by Camiz and Poscolieri (2015) to
EP

classify into homogeneous terrain units the DEM pixels according to their local
topographic gradients. The conceptual approach was sketched in both Parcharidis et al.
C

(2001) and Poscolieri (2010) and some elements were applied already in some
occasions (see, e.g., Adediran et al. 2004; Abbate et al., 2006; Poscolieri et al., 2011;
AC

Camiz et al., 2013a) in which other items had to be studied in comparison with
geomorphometric classes. As aforementioned, this methodology classifies the pixels
simply according to eight neighbours' local gradients, oriented as the rose diagrams;
from these both the pixel's slope and aspect may be computed. The methodology results
in an exploratory Tandem Analysis (Arabie and Huber, 1994), based on Principal
Component Analysis of the eight topographic gradients, followed by a mixed clustering

2
ACCEPTED MANUSCRIPT

procedure aiming at building a nearly-optimal classification (Camiz and Poscolieri,


2015).

In this paper we show how the same methodology may be used to describe volcanoes
morpho-structures. For this, we show how a set of volcanoes of Mexico, Central, and
South America may be compared after the identification of their geomorphometric

PT
classes. Indeed, volcanology may take advantage from the use of DEMs (Székely and
Karátson, 2004) and its potential has not been totally exploited so far, albeit several

RI
works have already been carried out (Grosse et al., 2012; 2014). In the case of
volcanoes, shape and size are the result of complex evolutions involving several

SC
processes. Thus, the study of volcano morphology can give valuable insights into these
processes and their underlying causes, i.e. tectonic/structural setting, magma
composition and flux, eruptive style and climate (Cotton, 1944; Francis, 1993; Thouret,

U
1999; Davidson and De Silva, 2000; Grosse et al., 2012). In particular, Grosse et al.
AN
(2012) describe a systematic methodology to extract quantitative morphometric
parameters of volcanoes from DEMs, “to define a set of morphometric parameters that
comprehensively and objectively characterize the size and shape of most types of
M

volcanic edifices”. Indeed, they built a program, MORVOLC, that is able to extract from
a volcano DEM 44 parameters, to be used for these purposes: this was used together
D

with SRTM DEM in Grosse et al. (2014) to provide a global database of volcano
TE

morphometry. Whereas in Grosse et al. (2012) the issue of delimitation of volcano


edifices was solved empirically by combining slope and profile curvature, Bohnenstiehl
EP

et al. (2012) propose an automatic method introducing an adjustment of the base


elevation along topographic profiles by the morphometric index cross-sectional area to
outlining perimeter. The approach is considered still sensitive to user-defined
C

parameters including the contour interval (Pedersen, 2016). In Grosse et al. (2014)
AC

NETVOLC (Euillades et al, 2013) was applied, a program able to delimit the volcano
contour according to either minimum cost flow networks (using a DEM and its first-
and second-order derivatives) or three alternative functions, that take into account also
slope, elevation, and/or radial distance. In order to classify terrain units for
volcanological purposes, Pedersen (2016) proposes an object-based method, able to
partition the DEM into segments, rather than pixels, and works at segments level. This
way, the contouring of both volcano edifice and its substructures may be solved by the

3
ACCEPTED MANUSCRIPT

same procedure. We stem from a different point of view, namely to check to what
extent a methodology, initially developed for geomorphometric classification of any
DEM, is able to give consistent non-trivial results for volcano edifices and may
integrate other techniques devoted to describe the morphological settings of volcanoes,
according to specifically suited parameters as those used by Grosse et al. (2012; 2014).
Indeed, to cope with the requirements of the volcanoes morphological comparison, the

PT
classification methodology described in Camiz and Poscolieri (2015) was slightly
modified, as it will be discussed in Section 2.1. In practice, considering the ideal

RI
pseudo-conical shape of a volcanic edifice (Euillades et al. 2013) as a starting point, we
decided to start from 17 pre-defined classes centroids, corresponding to a “flat” class

SC
plus two sets of eight classes, initially oriented toward the main compass directions and
centred on two average slope values. The transformation of the initial centroids in the

U
final classes depends on each volcano structure. Thus, the resulting classifications for
AN
all volcanoes could be directly compared, contributing to the understanding of the
spatial distribution of homogeneous volcanic landforms.
The choice of the volcanoes was essentially driven by the purpose of analyzing edifices
M

of different size and morphological settings in order to test the proposed methodology
abilities. These settings range from nearly conical to complex edifices, which result
D

from emissions of various products: complex edifices commonly result from changes in
TE

vent location (i.e. migration of the locus of activity) and/or successive collapses that
occurred during their volcanological history.
EP

In this paper, after the description of the methodology and the comparison of the results
of its application to the chosen volcanoes, we shall show how these results may be used
to eventually classify the volcanoes, based on geomorphological similarities.
C
AC

2. Materials and methods

2.1 The analysis of geomorphometric structures

For the geomorphometric analysis, Camiz and Poscolieri (2015) implemented


GEOMORFO, a methodology based on the ideas described in Parcharidis et al. (2001),
Adediran et al. (2004), and Poscolieri (2010). The procedure is summarized in the
following.

4
ACCEPTED MANUSCRIPT

The DEM files are processed by computing, for each pixel, its difference in elevation
with respect to its eight closest neighbours, this way obtaining 8 topographic gradients:
the difference is negative if the central pixel is lower than its neighbour, positive if it is
higher. In Remote Sensing applications this corresponds to an 8-layer data set that can
be analysed by classification techniques as a multi-band imagery (Parcharidis et al.,
2001). As well, such data sets may be seen as data tables with all pixels as row units and

PT
the 8 gradients as column characters. As such, they can be submitted to
multidimensional analyses.

RI
The choice of partitioning the set of pixels of a DEM according to the gradients, instead

SC
of using slope, elevation, and aspect, is advantageous since it greatly simplifies
computations. In fact, both slope and aspect are angles that may be calculated in
different ways, thus leading to different results and, in addition, the very small

U
difference between 1º and 359º aspect angles may not be handled by the usual
AN
algorithms. Thus, the use of gradients is a need to overcome this problem.

The classification of the pixels is performed by first applying Principal Component


M

Analysis (PCA, Jolliffe, 2002) on the data table containing the eight gradients computed
for all pixels and then calculating the Euclidean distance between pairs of pixels on the
D

subspace spanned by the first two or three extracted principal factors. The use of PCA
before the classification is a current practice in exploratory data analysis (Lebart et al.,
TE

2006), when no intrinsic partition seems detectable in the data set. It is considered
useful to reduce the non-relevant information that is usually relegated to the least
EP

components. Our choice of using the coordinates of pixels of the first two principal
components only is in agreement with Camiz et al. (2013a, 2013b, 2014), in which they
C

were always accounted for over 90% of total variation. As for the interpretation of the
principal components, in the plane spanned by the first two the pixels are set in the
AC

different quadrants according to their slope orientation, with their distance from the
origin corresponding to their slope. Thus, the principal components represent the nearly-
orthogonal orientations in which the maximum dispersion of the local gradients occur,
like the main slopes. In the plane spanned by them, the eight gradients mimic the wind
rose, rotated according to fit the principal components. Their angles between the
gradients are skewed, depending upon their correlation, with those around the first

5
ACCEPTED MANUSCRIPT

principal component usually smaller than those around the second (see Camiz et al.,
2013b: fig 9).

The classification algorithm uses the Euclidean distances to measure the similarity
between pixels and is composed by a sequence of K-means, hierarchical (Gordon,
1999), and again K-means methods. The number of classes is then chosen by the user

PT
according to some intermediate results. For each resulting class mean and standard
deviations are calculated for the eight gradients, and for pixels' elevation, slope, and

RI
aspect, the last two calculated according to the gradients. It must be pointed out that, for
aspect, both parameters have been computed according to circular statistics (Fisher,

SC
1993), in order to take into account the identity of the extreme values of the scale (0º =
360º). Both slope and aspect means and standard deviations may be represented on rose
diagrams, in which the points are situated according to mean slope and aspect of each

U
class and each point whisker is proportional to the corresponding parameter standard
AN
deviation.

Thematic maps of the study area may be created, showing the pixels belonging to the
M

resulting classes with different colours. For each class, we considered the mean angles
of both aspect and slope to pinpoint the appropriate Hue and Saturation respective
D

values, according to the Hue-Saturation-Value (HSV) colour modelling (Agoston,


2005). Concerning the illumination, a mean shadow value was chosen. A flow chart of
TE

GEOMORFO is reported in the left column of Table 1 (see Camiz and Poscolieri, 2015,
for further details concerning the methodology).
EP

2.2 The methodology adjusted for volcanoes analysis

The described method was used by Camiz et al. (2013b) to study three Andean
C

volcanoes and the number of classes was a posteriori suggested by the results of the
AC

hierarchical classification procedure. Now, the present need of comparing volcanoes


forced us to slightly modify the methodology, to ensure that the results depend only on
morphological differences and not on methodological or practical choices. Indeed, the
loss of an objective method to decide a suitable “best” partition for each volcano,
prevents the use of the described method as such. Thus, we had rather use an identical
automatic procedure for all volcanoes and consider its results depending only on the
specific volcano structures. In particular, the distinction between specific volcanoes

6
ACCEPTED MANUSCRIPT

morphological features may be carried out by evaluating how each volcano structure
deviates from an ideal pseudo-conic shape, a result that could not be obtained from the
methodology described by Camiz and Poscolieri (2015). For this purpose, we slightly
modified the procedure to get a new one, GEOVOLCANOES, in which, after PCA, the
classification was simply performed by a K-means algorithm, initialized by a priori
defining the number of wished classes and their initial centroids. K-means builds the

PT
classes by iteratively reallocating each pixel into the class with the closest centroid and
re-computing the centroids themselves. Thus, the shift of each centroid from its original

RI
position to the final one depends on the particular morphological structure of each
volcano. For these reasons, the final positions of corresponding classes in the different

SC
volcanoes may be consistently used for comparison purposes.

To mimic the pseudo-conic shape (Euillades et al. 2013), we fixed as a priori start-

U
points two slopes along the eight wind rose directions, plus the origin, summing up to
AN
17 starting centroids. Thus, the centroid class 17’s slope and orientation were set to 0º,
in order to collect nearly flat areas. The other 16 classes were defined by fixing the
centroids on two points along each of the eight main compass directions (270º, 315º, 0º,
M

45º, ..., 225º), with two slope values, 12º and 28º respectively. The choice of these two
values is essentially based on the search of intermediate positions between some
D

maximum real slope and the flat area. For this task, we referred to the slope distribution
TE

discussed by Voight and Elsworth (1997: Figure 2): here, indeed, 12º and 28º represent
two modal values of an apparent bimodal distribution.
EP

Next, for each volcano, these 17 points were projected on the first two principal
components to be used as starting centroids of the K-means algorithm. The method
minimizes the within-class inertia, defined as the sum of squared distances of each pixel
C

to the centroid of the class of belonging. A test in a pilot study on some South-central
AC

American volcanoes (Camiz et al., 2014) proved effective in showing similarities and
differences among them. As the iterative procedure computes the new centroid of each
formed class and redistributes each pixel to the class with the closest centroid, until
convergence, the final position of the 17 centroids and the size of each class depend on
the particular volcano morpho-structure only. The flow chart of GEOVOLCANOES is
reported in the right column of Table 1.

7
ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M

Table 1. Flow-chart of the original methodology of geomorphometric classification


(GEOMORFO), on the left, and the modified version (GEOVOLCANOES) adapted for the
present study of volcanic structures, on the right.
D
TE

Furthermore, just as in the original procedure, means and standard deviations of the
eight local topographic gradients, elevation, slope, and aspect were computed for the 17
EP

classes of each volcano.

The classes were graphically represented by plotting rose diagrams, in which the mean
C

slope and aspect of each class are polar coordinates of the points-classes. This allows a
graphical direct comparison of different volcanoes' corresponding classes.
AC

2.3 The comparison of all volcanoes classifications

Each studied volcano was described by a series of statistics concerning mean slope and
aspect of each of the built classes: note that the 17th one gathered all pixels of reduced
slope, a kind of residual class with respect to the main structures. For this reason, we
decided to drop it from subsequent processing.

8
ACCEPTED MANUSCRIPT

To directly compare the means in the same class of two volcanoes, a statistical test, like
Student t or ANOVA, revealed useless, due to the too high number of units in each class.
With these class figures, even very small mean differences result statistically
significant. Just as an example, consider that the difference of 0.02 degrees of the slope
in class 2 between Tata Sabaya and Fuego is not significant, whereas the difference of
0.1 degrees of the slope of class 1 between Lanin and Socompa is. It is evident that no

PT
geological meaning may be attributed to both differences. In addition, the too high
number of comparisons would raise too much the risk of statistical error. Thus, we had

RI
better outlined, for all volcanoes, which classes diverged from their respective overall
means within each of the four groups of volcanoes. This was achieved by plotting the

SC
aspect vs. slope divergences of all classes of all volcanoes on a scatter diagram. The
diagram grid steps were fixed to 15º for aspect and 5º for slope divergences that we

U
considered suitable to outline the most significant ones.
AN
Both slope and aspect means of the remaining 16 classes of each volcano formed a
dataset composed of two subtables and were further used to classify the volcanoes
accordingly. As these two parameters might be structurally very different, to avoid
M

either group structure to overcome the other, we ran a Multiple Factor Analysis (MFA,
Escofier and Pagès, 2008) that provided us with a factorial representation, analogous to
D

PCA. MFA is a PCA in which groups of characters are differently weighed to


TE

counterbalance their possible structural differences. Technically speaking, this is


achieved by dividing each groups' data by the square root of the first eigenvalue of its
EP

corresponding PCA. Eventually, to perform the classification, a hierarchical ascendant


procedure was run on all MFA factor coordinates.
C

2.4 The Data


AC

The DEM files of the 21 examined volcanic complexes were obtained by using the
ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer) GDEM
(Global Digital Elevation Model) data, which are thoroughly described in Camiz et al.
(2013b). The used version, 2., overcame mostly the problem of the occurrence of
artificial sinks / holes or pixels lacking data present in the previous one (Li et al., 2012).
The DEM files were set up by picking out, within an ASTER GDEM, tile rectangular
subsets centred on the analysed volcanic edifices. Among the analysed volcanoes Santa

9
ACCEPTED MANUSCRIPT

Ana, Pico de Orizaba and Colima are the largest, while Tata Sabaya and Arenal are the
smallest ones.

PT
RI
U SC
AN
M
D
TE
EP

Table 2. The studied volcanoes: their position, height of the summit, elevation lower limit at
C

which the DEM was masked, and covered surface.


AC

Before applying GEOVOLCANOES, the border of the area to be studied ought to be


defined by masking the pixels outside. This step is not actually included in our
methodology: for this task, we could not use MBOA package (Bohnenstiehl et al, 2012),
because of lack of Matlab; we aimed at applying NETVOLC (Euillades et al., 2013), but
we could not, because i) it works on metric coordinates while we used geographic ones,
and ii) its implementation in our computers failed. Therefore, we manually proceeded
by visually defining the contour that best encompasses the volcanic morphology. This

10
ACCEPTED MANUSCRIPT

was carried out by overlapping on an appropriate shaded relief, covering all volcanic
units of the structure, the contour lines with 100m step. In order to choose the minimum
contour line that encompasses all landforms of the volcanic edifice, two N-S and E-W
profiles crossing the volcano summit were used. In some cases, to include/exclude some
particular morphological features, the contour has been manually adjusted by drawing
poly-lines on the DEM image on the screen. Then, all pixels outside the contour were

PT
masked and only the inner ones were taken into account in the following
geomorphometric analysis. In the last two columns of Table 2 the masking elevation

RI
limit of each volcano and the covered surface are reported.

SC
3. The volcanoes

21 volcanoes in South America, Central America, and Mexico were randomly selected.

U
In the selection process we gave preference to those edifices that strongly differ in their
AN
morphological features (e.g. shape, elevation, volume, etc). Both their geographical
coordinates, summit elevation, and covered area are reported in Table 2.
M

The South American volcanic arc (Fig. 1) includes over 200 potentially active
Quaternary volcanoes, and at least 12 giant caldera/ignimbrite systems, occurring in
D

four separate segments, referred to as the Northern (NVZ), Central (CVZ), Southern
TE

(SVZ), and Austral Volcanic Zones (AVZ) (Stern, 2004). Volcanism results from
subduction of both Nazca and Antarctic oceanic plates below South America. The
Andean Cordillera forms a mountain range, morphologically continuous along the
EP

western margin of South America, from the Caribbean Coast (Northern extreme) to
Cape Horn (Southern), more than 7,500 km long. For this study we focussed only on the
C

CVZ and SVZ and we selected four volcanoes of the CVZ geotectonic region:
AC

Parinacota, Tata Sabaya, Ollagüe, and Socompa; and four for the SVZ: Peteroa, Llama,
Villarica, and Lanin.

11
ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE
EP

Fig. 1. Tectonic setting of the Andes and distribution of volcanoes along the Cordillera (open
triangles). Moreover, the location of the eight South-American volcanoes (closed triangles),
C

object of this study, is shown.


AC

As concerns the Central America arc (from Guatemala to Panama), volcanism is the
result of the Cocos plate North-Eastward subduction beneath the Caribbean plate along
the Middle America trench (Fig. 2). The convergence rate between the Cocos and
Caribbean plates increases towards the South-East, from ~60 mm/year off Southern
Guatemala to ~90 mm/year off Southern Costa Rica/Panama (DeMets, 2001). This
volcanism is associated with Caribbean blocks of different crustal origins: the North-
Western Chortis block, and the South-Eastern Chorotega block. The Chorotega block is

12
ACCEPTED MANUSCRIPT

underlain by the Caribbean Large Igneous Province (CLIP), which was settled during
the Cretaceous. Although there is no consensus as to the location of the boundary
between Chortis and Chorotega blocks, the important observation is that the Chortis
block is underlain by old evolved continental crust (~100 Ma), whereas the Chorotega
block is not. For this study, we selected six volcanoes: Fuego, Atitlán, Santa Ana, San
Cristóbal, Momotombo, and Arenal whose positions are reported in Fig. 2.

PT
RI
U SC
AN
M
D
TE

Fig. 2. Tectonic setting of Northern-Central America showing, volcanoes (open triangles),


Middle America Trench (MAT), and the Chortis and Chorotega blocks. The boundary between
these blocks is not well defined. Letters indicate respective countries in Central America.
EP

Moreover, the location of the Mexican and Central-American volcanoes(closed triangles),


object of this study, is shown.
C

The Trans-Mexican Volcanic Belt (TMVB) is the largest Neogene volcanic arc in North
AC

America, with an area of 160,000 km2 and a length of almost 1000 km. The TMVB was
built upon Cretaceous and Cenozoic magmatic provinces and a heterogeneous basement
made of tectonostratigraphic terranes of different age and lithology (Ferrari et al.,
2012). The TMVB represents the most recent episode of a longer continental magmatic
activity that, since the Jurassic, produced a series of partially overlapping arcs as a result
of the eastward subduction on the Farallon plate beneath western Mexico (Ferrari et al.,
2012). The TMVB consists of a wide variety of volcanic rocks that include extensive

13
ACCEPTED MANUSCRIPT

lava plateaus, scoria cone fields, domes, stratovolcanoes, and ignimbrites produced by
caldera formation (Macías, 2007 and references therein). Numerous volcanoes in the
TMVB, such as Volcán de Colima, Nevado de Colima, Nevado de Toluca, Jocotitlan,
Popocatépetl, Pico de Orizaba and others have undergone major collapses during their
eruptive history (Siebe et al., 1995; Capra et al., 2002; Roverato et al., 2011). In this
area we selected seven volcanoes (see Fig. 2) that are: Ceboruco, Volcan de Colima,

PT
Jocotitlán, Nevado de Toluca, Popocatéptl La Malinche, and Pico de Orizaba.

RI
4. Results

SC
For each analyzed volcano, the results consist of detailed statistics of geomorphometric
parameters for each of the 17 classes of the obtained classification, together with a

U
thematic colour classification map. Here, as an example both are reported for the Lanin
volcano in Table 3 and Fig. 3 right, respectively. In Table 3, the first column shows the
AN
sequential number of the class, the percentage of pixels belonging to the class, the area
covered by the class in Km2, and the colour given to the class in the classification map.
M

The second column provides a morphological interpretation of the class; the third
column reports mean elevation, slope, and aspect. The fourth column reports the
D

average of the eight gradients with respect to the central pixel, coloured in blue the
TE

positive (pointing out a descent toward the outer pixels) and in red the negative ones
(pointing out a descent toward the central pixel). The arrow at the centre of the gradients
matrix points to the steepest descending slope. In Fig. 3 right, the thematic map is
EP

obtained by colouring each pixel, as aforementioned, according to its corresponding


class. It is worth noting a kind of three-dimensional effect that results this way. The
C

interpretation of this map may be the following: three ridges are recognized NE of the
AC

summit due to the alternation of red and green colours (within the white ellipse in the
figure), corresponding to N and ESE aspects. Close to the summit, in the S-SW sector, a
horse-shoe shape scarp is visible, highlighted by three green spots SE and a blue-red
NW oriented slope (white arrows in the figure). Another E-W ridge appears in the lower
part of the same sector, as indicated by the red spot, once again N oriented. Eventually,
on the NW side, another crest results at the end of the selected area. This crest could be
not correlated with the main recent volcanic features, therefore it could be questionable
whether to take it into account as we did or to mask it out. Nevertheless, we think that

14
ACCEPTED MANUSCRIPT

the obtained representation contributes to understand the problem and suggests further
inspection.

PT
RI
U SC
AN
M
D
TE
C EP
AC

Table 3. Descriptive statistics of the 17 classes obtained by processing the topographic gradients
of the Lanin volcano. Each class, of which the covered surface is pointed out, is provided with
a morphologic interpretation and the mean values of both main landform attributes and
topographic gradients with respect to the central pixel of a 3x3 kernel (colored in red for
negative and in blue for positive).

15
ACCEPTED MANUSCRIPT

PT
RI
Fig. 3. The volcano Lanin. Left: example of delimited volcanic structure prior to apply

SC
geomorphometric analysis. Right: the thematic classification map resulting from the analysis,
arrows and ellipse point out the interpreted volcanic structures.

U
Another example comes from the Colima volcano (see next Fig. 6 bottom-left), a well
know stratocone that displays complex geomorphological features and two different
AN
vents, the Nevado (dormant) and the Fuego (active). Both edifices show horse-shoe
shape scarps southward facing as highlighted by the green and blue colours SE, S and
M

SW oriented. E-W important fault scarps, S and N facing, are clearly indicated by green
colour SE oriented, blue colours S and SW oriented and red colour N oriented,
D

respectively. These structures are interpreted as the result of the interaction between the
TE

regional Colima rift and lateral spreading of the volcanic complex toward south (Norini
et al., 2010) that could explain the main direction of several sector collapses that
characterized the Colima complex.
EP

The volcanoes thematic maps are reported in Figs. 4, 5, and 6 for South-, Central-
America, and Mexico, respectively. The sizes of the maps are proportional, thus
C

comparable. As well, the pixels' colours represent both the aspect (through hue) and the
AC

slope (through saturation). Indeed, looking at the cited figures, the structure of the
edifices is detectable. The overall averages of slope and aspects of all classes of all
volcanoes are shown in Table 4, where also standard deviations, minima and maxima
are reported. In Figs. 7, 8, 9, and 10, the polar diagrams show mean and standard
deviation of both slope and aspect of each class in the eight South-American, the six
Central-American, and the seven Mexican volcanoes, respectively. Their inspection
allows to appreciate similarities or differences between the corresponding classes of the
analysed volcanoes. As it may be a complicated task, due to the overlapping of the

16
ACCEPTED MANUSCRIPT

PT
RI
SC
Fig. 4. Thematic maps of the eight South-American volcanoes.

U
AN
drawn volcanoes parameters, especially for low mean slope classes, we plotted the
divergences of the corresponding classes' mean slope and aspect from their average
M

values within the American volcanoes studied sector, considering steps of 15º for aspect
and 5º for slope (see Figs. 11 to 15). This way, we may identify larger deviations that
D

may be interpreted as some specific feature of a volcano.


TE

In the volcanoes Llaima, Villarica, Lanin, and Peteroa (Fig. 7), located in the SVZ
Andes sector, the classes oriented 330°, N, and 30° (10, 11, and 12) appear clustered,
the first two groups mostly in terms of aspect, while the third is more clustered as slope.
EP

Moreover, Villarica class 9, facing 315°, is far from the same class of the other
volcanoes, while Lanin and Peteroa show some particular anomalous pattern, including
C

the steepest classes, 14 of Lanin and 14 and 15 of Peteroa, Southward oriented (45° to
AC

50° slope ranges). Indeed, in Fig. 11, in which the deviations from the corresponding
classes means of both aspect and slope are represented, both classes 14 and 15 (S
oriented) result for Peteroa more than 10º steeper than the mean, whereas for Villarica
they are gentler of more than 10º. As for slope, class 8 of Lanin is over 10º steeper than
the mean; dealing with aspect, classes 4 and 5 of Villarica appear more than 25º shifted
clockwise with respect to the mean, whereas classes 1 and 8 of Llaima are over 25º
shifted counterclockwise.

17
ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
Fig. 5. Thematic maps of the six Central-American volcanoes.
M
D
TE
C EP
AC

Fig. 6. Thematic maps of the seven Mexican volcanoes.

18
ACCEPTED MANUSCRIPT

Item Mean St.dev. Minim um Maxim um


SM1 17.71 4.00 12.40 30.41
SM2 10.60 3.19 4.27 16.43
SM3 18.29 4.95 12.19 33.32
SM4 10.85 4.07 4.69 20.14
SM5 17.27 3.30 10.27 23.64
SM6 12.32 4.15 4.43 23.60

PT
SM7 20.24 6.03 11.98 33.20
SM8 11.21 5.30 4.34 24.75

RI
SM9 31.53 3.35 26.55 39.84
SM10 27.99 4.01 18.88 34.45
SM11 33.69 4.03 26.90 44.31

SC
SM12 27.97 4.18 17.21 36.20
SM13 31.63 4.64 23.65 42.93
SM14 31.26 4.85 23.66 47.28

U
SM15 36.60 6.60 26.93 56.34
SM16 28.27 5.66 14.81 35.64
AN
AM1 264.05 14.96 235.84 293.17
AM2 309.89 16.83 270.07 336.28
AM3 -2.00 11.28 -22.31 16.32
M

AM4 49.38 21.81 10.03 104.22


AM5 92.60 19.94 51.70 131.82
D

AM6 138.78 15.95 88.37 161.64


AM7 179.17 15.79 141.48 204.49
TE

AM8 219.44 15.34 190.66 253.65


AM9 271.13 21.08 222.42 313.47
AM10 319.37 10.25 297.61 336.01
EP

AM11 0.12 6.98 -16.07 13.81


AM12 40.48 7.53 23.14 51.69
AM13 90.15 12.07 58.90 123.07
C

AM14 141.66 10.52 121.76 163.90


AC

AM15 184.14 6.95 174.37 201.41


AM16 220.66 9.14 190.24 236.76

Table 4. Mean, standard deviation, minimum, and maximum of the average values of slopes
(SMxx) and aspects (AMxx) of the 16 most significant geomorphometric classes, computed on
all the 21 volcanoes.

19
ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
Fig. 7. Polar chart of the four South-American volcanoes Llaima, Villarica, Lanin, and Peteroa.
All classes are centred to their means; the error bars correspond to the standard deviations of
M

both slope and aspect.


D
TE
C EP
AC

Fig. 8. Polar chart of the four South-American volcanoes Ollague, Parinacota, Tata Sabaya, and
Socompa. All classes are centred to their means; the error bars correspond to the standard
deviations of both slope and aspect.

20
ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
Fig. 9. Polar chart of the six Central American volcanoes. All classes are centered to their
means; the error bars correspond to the standard deviations of both slope and aspect.
M
D
TE
C EP
AC

Fig. 10. Polar chart of the seven Mexican volcanoes. All classes are centred to their means; the
error bars correspond to the standard deviations of both slope and aspect.

21
ACCEPTED MANUSCRIPT

The volcanoes Ollagüe, Parinacota, Tata Sabaya, and Socompa (Fig. 8), located in the
CVZ Andes sector, at about 25°-40º slope range exhibit similar classes 13 to 16, facing
E, SE, S, SW, respectively, and, partially, class 9, facing W (concerning slope only). The
classes 10, 11, and 12, facing NW-N-NE, respectively, are not so clearly grouped as the
ones facing the South quadrant. This may result from the special behaviour of Socompa,
in which the classes 3, 4, and 11 are far from the corresponding ones of the other

PT
volcanoes, as their mean slopes appear steeper than the overall means (see Fig. 12). In
particular, the class 3, usually exhibiting gentle slope, is about 12º steeper than the

RI
mean, at the same level of the other large slope classes, whereas the class 4 is shifted
counterclockwise by over 20º with respect to the mean. In addition, comparing Figs. 7

SC
and 8, it might be outlined that in CVZ Andes sector the groups of gentle and steep
slope classes appear more clearly separated with respect to the volcanoes located in the

U
SVZ Andes sector. It is worth noting that in Fig. 12 the pattern of deviations in the CVZ
AN
sector is mostly within 5º slope and 15º aspect, thus far more concentrated around the
mean than that of SVZ, as represented in Fig. 11, a sign of larger homogeneity of the
examined volcanoes in this sector.
M

As regards the Central-American volcanoes (Fig. 9), the classes with teep slope appear
reasonably clustered around the rose diagram, more markedly the 9, 10, 11, and 12, W,
D

NW, N, and NE oriented, respectively. Somehow exceptional appears Fuego volcano, in


TE

which class 7, usually exhibiting gentle slope, is located at the same level of the large
slope ones, while class 8 behaves similarly, but with lower deviation. This causes for
Fuego a counterclockwwise shift in the 11 to 15 class range with respect to the other
EP

volcanoes. It must be noted that the gentle slope classes appear more clustered than
those of volcanoes located in the CVZ Andean sector, but not as much as those located
C

in SVZ; in particular, in the Central American sector, the gentle slope classes appear
AC

better separated than in the latter from the steepest ones, especially those facing W.
Looking at Fig. 13, where aspect vs. slope divergences are shown, it results that all
classes of Santa Ana are well clustered around both mean slope and aspect; conversely,
the other volcanoes exhibit some peculiar classes deviations: indeed, both Atitlan and
Fuego's classes are in general steeper than the mean - in particular Atitlan's 10 and
Fuego's 7 and 8. Moreover, the latter two are also shifted counterclockwise, as well as
Atitlan's 1 and 2, whereas Atitlan's 4 and 5 and Fuego's 4 are highly shifted clockwise.

22
ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
Fig. 11. Slope angle divergence from class mean vs. respective aspect divergence for the four
volcanoes of the SVZ Andes area.
M
D
TE
C EP
AC

Fig. 12. Slope angle divergence from class mean vs. respective aspect divergence for the four
volcanoes of the CVZ Andes area.

23
ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
Fig. 13. Slope angle divergence from class mean vs. respective aspect divergence for the six
M

Central-American volcanoes.
D
TE
C EP
AC

Fig. 14. Slope angle divergence from class mean vs. respective aspect divergence for the seven
Mexican volcanoes.

24
ACCEPTED MANUSCRIPT

On the opposite Momotombo and San Cristobal classes exhibit gentler slope than the
mean, in particular Momotombo's class 16. Classes 4 and 8 of San Cristobál appear
highly rotated (counter- and clockwise, respectively) with respect to the mean aspect:
this may be due to the presence of main ridges along this orientation that favoured the
observed aspect divergences.
The classes of Mexican volcanoes (Fig. 10) appear quite less clustered than the other

PT
groups of volcanoes. Indeed, some homogeneity appears for the steepest classes 11, 12,
13, and 14, facing N, NE, E, and SE, respectively, with the exception of Colima, as

RI
regards N and NE facing classes 11 and 12, and Jocotitlán classes 13 and 14, E and SE
facing. Moreover, Colima exhibits another anomalous trend towards NW (class 9) and

SC
Jocotitlán towards S and SW (classes 14 and 15). It is interesting to note that in this case
the classes 1 facing W are sufficiently well clustered. The pattern of deviations of

U
Mexican volcanoes, depicted in Fig. 14, shows that Pico de Orizaba (except for class
AN
2), Nevado de Toluca (except for class 2) and partially Jocotitlán (except for classes 2,
13, and 15), Ceboruco (except for classes 4, 7, and 15) and Popocatetepl (except for
classes 1 and 9, the most divergent as for slope and aspect) appear quite clustered
M

around the class means, while Colima and Le Malinche are more widespread, especially
in terms of aspect counterclockwise divergences.
D

In general, whereas the steepest classes show some clustering among different
TE

volcanoes, the gently sloping ones appear more confused. This may depend on the
irregularities of the low-lying terrain units, due to the balance between emplacement
EP

(lava flows, pyroclastic activity, etc.) and erosional processes.

In Fig. 15 the deviations of the means of the four volcanoes sectors are represented with
C

respect to the overall means of each class. It is possible to observe that the SVZ sector
exhibits the largest deviations both in terms of aspect and slope divergences, while
AC

Mexican volcanoes sector deviations appear slightly steeper and counterclockwise


rotated than the mean.

25
ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
Fig. 15. Slope angle divergence from group class mean vs. respective aspect divergence for the
four volcanoes sectors.
M

5. Discussion
D

The volcanoes studied in this work may reflect contrasting examples of continental
TE

margin arcs: the Central American Volcanic Front is developed on thin to thick crust,
contains many young and historically active volcanoes, and is characterized by a
humid, erosive climate; Central Andes Volcanic Zones are on thick crust, most
EP

volcanoes are dormant or extinct, and they exhibit a very arid, low-erosion climate,
while in Southern Andes climate is humid, and volcanoes are more active (Grosse et al.,
C

2009). Compared with Central American arc and the Andean setting, the TMVB has
AC

much more variable width and climate and displays another peculiarity with respect to
most volcanic arcs: the main volcanic edifices are aligned oblique to the general
orientation of the arc (Ferrari et al., 2012).

In order to identify possible morphological analogies or differences among the 21


volcano structures, based on the discussed classification of the terrain units, we attempt
here a classification, based on principles analogous to those aforementioned: this time

26
ACCEPTED MANUSCRIPT

the classification was preceded by a MFA, described in section 2.3, in order to avoid
unbalances within the two groups of parameters, slopes and aspects.

PT
RI
U SC
AN
M
D

Fig. 16. The pattern of the 21 volcanoes on the first factor plane issued by the Multiple Factor
TE

Analysis, performed on both mean slope and aspect of the 16 geomorphometric classes of pixels
of each volcano. This plane summarizes around 40% of the total squared deviation of both slope
and aspect of the 16 classes of each volcano from the mean overall values of both parameters
reported in Table 4. The lines link the compromise position of each volcano with the
EP

corresponding two partial representation of slopes (S) and aspects (A).

In Fig. 16 the volcanoes are represented on the first factor plane of MFA. The two
C

factors are accounted for 20.64% and 19.16% of the total inertia respectively, so that the
AC

plane summarizes around 40% of it. Considering the correlations between factors and
original characters, it results that the first factor is mostly correlated with the slopes and
the second mostly with the aspects: this allows to interpret the pattern of the volcanoes
according to the divergences of their classes with respect to the average. It is evident the
opposition between Momotombo and Peteroa along the first axis, due to the fact that
slopes are respectively smoother and steeper than average, whereas on the second one
La Malinche, Colima, and Fuego are opposed to Socompa, Ollagüe, and San Cristobal,

27
ACCEPTED MANUSCRIPT

due to the rotation of the aspect with respect to the average, counterclockwise for the
first group and clockwise for the second one. The position of each volcano is a
compromise between its partial positions according to the classes' slope and that
according to classes' aspect only. In the Figure, both partial positions, indicated by S and
A, respectively, are linked by a line to the corresponding volcano's compromise
position. Looking at the thematic maps of the volcanoes, one may see that the extreme

PT
positions along the first axis are occupied by volcanoes whose shape is far from the
typical conic one, in particular because of the variously-shaped structures in the ring-

RI
plane. More difficult is to understand at first sight the differences along the second axis.
Thus, considering the opposition along the first axis, Momotombo NE side was enriched

SC
by lava-flows and deposits, something missing in the other sides. Indeed, despite the
presence of the lake, that may have blurred ancient deposits, it is evident that all recent

U
effusive activity was NE-oriented. This may explain the gentle slope of the classes in
AN
the N section. On the opposite, Peteroa has a structure very far from the conic one, as it
is a complex of three NS aligned volcanoes: this is coherent with the steepest classes, N
and S oriented. Considering the second axis, the opposition between La Malinche and
M

both Socompa and Ollagüe may be understood considering the conical shape of all, but
with relevant semi-circular scarps, with different orientations.
D

The inspection of the hierarchical classification process of the volcanoes, computed on


TE

all extracted factors, thus considering all inertia, suggests to consider partitions in 4 and
9 classes: thus, we try to discuss both. It is interesting to observe that in the 9-classes
partition a big group of 8 volcanoes results (Llaima, Parinacota, Tata Sabaya, Arenal,
EP

Santa Ana, Ceboruco, Nevado de Toluca, and Pico de Orizaba), then a group of three
(Lanin, Fuego, and Colima), and other pairs or singletons. Aggregating progressively
C

from 9 to 4 classes, pairs and singletons, but Peteroa and La Malinche, merge into the
AC

main group. We can interpret these results by saying that most volcanoes share common
geomorphological characteristics, regardless their more or less complex structure. The
most peculiar volcano is undoubtedly Peteroa, that differs from all others by having the
slope of all classes larger than the overall volcanoes means. Indeed, this represents
pretty well the massive aspect of the complex. Considering that Peteroa activity does
not construct a central conic vent and that the whole complex is subject to erosion of
looser material (ashes), its shape may result from the remaining of the strongest

28
ACCEPTED MANUSCRIPT

material, such as lavas and welded deposits. Concerning La Malinche, it distinguishes


from the others mainly because of a counterclockwise rotation of most classes. The
group Lanin, Fuego, and Colima distinguishes from the others by the steeper slope of
their SSW facing classes. Even if very different among each other, these volcanoes share
some morphological structures oriented toward SSW, which appear as erosive gullies or
structural depressions: as an example, Colima displays very steep horse-shoe-shaped

PT
scarps, the main slopes of which are facing toward SSW. This may depend upon some
structural anomalies that influence the positioning of the classes centroids during the

RI
partitioning process. Moreover, it is interesting to point out a group of two volcanoes in
the partition in 9 classes, Villarica and San Cristóbal, which are characterized (as well

SC
as Momotombo) by all classes slopes lower than the average. This may be explained by
the growing of these volcanoes in unconfined landscapes, in which the volcanic edifices

U
may grow with lower angles. This is totally different from all others, in particular from
AN
Peteroa that built up in a confined setting.

6. Conclusion
M

The methodology proposed by Camiz and Poscolieri (2015) is an exploratory technique


D

able to identify the most significant features issued by a DEM. It may be used as a
background (or a complement) for more sophisticated studies, involving other aspects
TE

of an area under study. In this paper, this methodology was adapted to a set of 21
volcanoes, randomly chosen from Mexico to Patagonia, to contribute in detecting
EP

similarities and differences among volcanoes. As such, both its descriptive and
graphical abilities may be a useful complement to volcanological studies. In this
C

respect, its exploratory use may be a guide for a focused study improving its results. Far
from being exhaustive, we think that it may integrate comparative works (e.g. Grosse et
AC

al., 2012; 2014) on more specific volcanic aspects or, better, act as a guide for them.
These preliminary results, whose interpretation in volcanological terms was attempted
in the discussion, do not seem sufficient for a complete geomorphological
understanding of the volcanoes. It might be further validated by deeper analyses and
possibly linked with other methodologies, in order to better understand the suggested
relations among volcanoes. Among other volcanoes aspects, local and regional
geotectonic settings and type of volcanic activity should be taken into account.

29
ACCEPTED MANUSCRIPT

Acknowledgements

This paper was developed in the framework of the cooperation of the first author as
associated to the Institute of Acoustics and Sensors “O.M. Corbino” of the Italian
National Research Council (CNR). Matteo Roverato acknowledges the grant of the

PT
Brasilian CAPES/CNPq Programa Ciências sem Fronteiras, Atração de Jovem Talento
402564/2012-0.

RI
References

SC
Abbate, G., Cavalli, R.M., Pascucci, S., Pignatti, S., Poscolieri, M., 2006. Relations
between morphological settings and vegetation covers in a medium relief landscape

U
of Central Italy. Annals of Geophysics 49(1), 153-165.
AN
Adediran, A.O., Parcharidis, I., Poscolieri, M., Pavlopoulos, K., 2004. Computer-
assisted discrimination of morphological units on north-central Crete (Greece) by
M

applying multivariate statistics to local relief gradients. Geomorphology 58, 357-370.


Agoston, M.K., 2005. Computer Graphics and Geometric Modelling: Implementation
D

and Algorithms. Springer, London, 300-306.


Arabie, P., Hubert, L., 1994. Cluster analysis in marketing research, in R. J. Bagozzi
TE

(Ed.), Advanced methods of marketing research, London, Blackwell, 160-189.


Bohnenstiehl, D.R., Howell, J.K., White, S.M., Hey, R.N., 2012. A modified basal
EP

outlining algorithm for identifying topographic highs from gridded elevation data,
Part 1: Motivation and methods. Computers & Geosciences 49, 308-314.
C

Camiz, S., Poscolieri, M., 2015. Geomorfo: a program for the classification of terrain
units. In: Jasiewicz, J., Zwolinski, Z., Mitasova, H., Hengl, T. (Eds.),
AC

Geomorphometry for Geosciences, Poznán (PL), Bogucki Wydawnictwo Naukowe,


Adam Mickiewicz University in Poznan - Institute of Geoecology and
Geoinformation, 49-152.
Camiz, S., Papageorgiou, E., Poscolieri, M., Parcharidis, I., 2013a. Correlation between
Landforms and Ground Deformation at Nisyros Volcano (Greece). In: Proceedings
of the ESA Living Planet Symposium 2013, Edinburgh (UK), 9-13 September 2013,
ESA SP-722, 1-6.

30
ACCEPTED MANUSCRIPT

Camiz, S., Poscolieri, M., Roverato, M., 2013b. Comparison of three Andean volcanic
complexes through Multidimensional Analyses of geomorphometric data. In: G.
Bianchini et al., Proceed. of the E-ICES 8, Buenos Aires, Comisión Nacional de
Energía Atómica, 52-74.
Camiz, S., Poscolieri, M., Roverato, M., 2014. Morpho-Structural Comparison of Latin
American Volcanoes. In: C. Hucailuk, N. Núñez, E. Molina (Eds.), Proceed. of the

PT
E-ICES 9, Buenos Aires, Comisión Nacional de Energía Atómica (CD-ROM, ISBN
978-987-1323-36-4), 46-64.

RI
Capra, L., Macias, J.L., Scott, K.M., Abrams, M. and Garduño-Monroy, V.H., 2002.
Debris avalanches and debris flows transformed from collapses in the Trans-Mexican

SC
Volcanic Belt, Mexico - Behavior and implications for hazard assessment. Journal of
Volcanological and Geothermal Research 113, 81-110.

U
Cotton, C.A., 1944. Volcanoes as Landscape Forms. Christchurch, Whitcombe and
AN
Tombs Publishing.
Davidson J., de Silva S.L. (2000). Composite volcanoes. In: Sigurdsson, H., Houghton,
B., McNutt, S.R., Rymer, H., Stix, J. (Eds.), Encyclopedia of Volcanoes. Academic
M

Press, New York, 663-681.


DeMets, C., 2001. A new estimate for present-day Cocos–Caribbean Plate motion;
D

implications for slip along the Central American volcanic arc. Geophys. Res. Lett.
TE

28, 4043-4046.
Efiong, J., Digha, O.N., Asouzu, O.E., 2016. Spatial Analysis of Land Surface -
Vegetation Relationship in Mountainous Areas of the Tropics Using Srtm-3 Dem,
EP

Journal of Geography and Geology, 8(2), 59-75.


Escofier, B., Pagès, J., 2008. Analyses factorielles simples et multiples. Objectifs,
C

méthodes et interpretation. Dunod, Paris.


AC

Euillades, L.D., Grosse, P., Euillades, P.A., 2013. NETVOLC: an algorithm for
automatic delimitation of volcano edifice boundaries using DEMs. Computer &
Geosciences 56, 151-160.
Evans, I.S., 2013. Land urface derivatives: history, calculation and further development,
Proceedings of Geomorphometry 2013, Nanjing (China), 1-4.
http://geomorphometry.org/system/files/Evans2013geomorphometry.pdf
Ferrari, L., Orozco-Esquivel, M.T., Manea, V., Manea, M., 2012. The dynamic history

31
ACCEPTED MANUSCRIPT

of the Trans-Mexican Volcanic Belt and the Mexico subduction zone.


Tectonophysics 522-523, 122-149, doi:10.1016/j.tecto.2011.09.018.
Fisher, N.I., 1993. Statistical Analysis of Circular Data. Cambridge University Press,
Cambridge (UK).
Florinski, I.V., 2012. Digital Terrain Analysis in Soil Science and Geology. London,
Elsevier.

PT
Florinsky, I.V., Kuryakova, G.A., 1996. Influence of topography on some vegetation
cover properties. Catena, 27, 123-141.

RI
http://dx.doi.org/10.1016/0341-8162(96)00005-7
Francis, P.W., 1993. Volcanoes: A Planetary Perspective. Oxford University Press,

SC
Oxford.
Freda, C., Palladino, D., Pignatti, S., Trigila, R., Onorati, G., Poscolieri, M., 1990.

U
Volcano-tectonic scenario of Mt. Vulsini (Central Italy) from Landsat-MSS Images
AN
and digital elevation data. ISPRS Jour. of Photogr. and Remote Sensing 45, 316-328.
Garcia-Aguirre, M.C., Ortiz, M.A., Zamorano, J.J., Reyes, Y., 2007. Vegetation and
landform relationships at Ajusco volcano Mexico using a geographic information
M

system (GIS), Forest Ecology and Management, 239, 1-12.


http://dx.doi.org/10.1016/j.foreco.2006.10.031
D

Gordon, A.D., 1999. Classification. Chapman and Hall, London.


TE

Grosse, P., van Wyk de Vries, B., Petrinovic, I.A., Euillades, P.A., Alvarado, G., 2009.
Morphometry and evolution of arc volcanoes. Geology 37, 651-654.
Grosse, P., van Wyk de Vries, B., Euillades, P.A., Kervyn, M., Petrinovic, I.A,. 2012.
EP

Systematic morphometric characterization of volcanic edifices using digital elevation


models. Geomorphology 136, 114-131.
C

Grosse, P., Euillades, P. A., Euillades, L.D., de Vries, B.V.W., 2014. A global database
AC

of composite volcano morphometry. Bulletin of Volcanology, 76(1), 1-16.


Hengl, T., Reuter, H.I., (2009). Geomorphometry: Concepts, Software Applications.
Developments in Soil Science 33. Amsterdam, Elsevier, 765.
Höersch, B., Braun, G., Schmidt, U., 2002. Relations between landforms and vegetation
in alpine regions of Wallis, Switzerland. A multiscale remote sensing and GIS
approach. Computers, Environment and Urban Systems, 26, 113-139.
http://dx.doi.org/10.1016/S0198-9715(01)00039-4

32
ACCEPTED MANUSCRIPT

Janssen, L.F., Jaarsma, J., van der Linder, E., 1990. Integrating topographic data with
remote sensing for land-cover classification. Photogrammetric Engineering &
Remote Sensing, 56(11),1503- 1506.
Jolliffe, I.T., 2002. Principal Component Analysis. Springer, New York, Springer Series
in Statistics, 2nd ed.
Lebart, L., Piron, M., Morineau, A., 2006. Statistique Exploratoire

PT
Multidimensionnelle, Visualisation et Inférence en Fouille de Données. Dunod,
Paris.

RI
Li, P., Shi, C., Li, Z., Müller, J., Drummond, J., Li, X., Li, T., Li, Y., Liu, J., 2012.
Evaluation of ASTER GDEM VER2 Using GPS Measurements and SRTM VER 4.1 in

SC
China. In: Proceed. of the XXII Congress of International Society of
Photogrammetry, Remote Sensing and Spatial Information Sciences, 25 August-1

U
September 2012, Melbourne, 181-186. http://dx.doi.org/10.5194/isprsannals-I-4-181-
AN
2012.
Macías, J.L., 2007. Geology and eruptive history of some active volcanoes of México.
In: Alaniz–Álvarez, S.A., Nieto–Samaniego, Á.F. (Eds.), Geology of México:
M

Celebrating the Centenary of the Geological Society of México. Geol. Soc. Amer.
Special Paper 422, 183-232.
D

Norini, G., Capra, L., Groppelli, G., Agliardi, F., Pola, A., Cortes, A., 2010. Structural
TE

architecture of the Colima Volcanic Complex. Journal of Geophysical Research:


Solid Earth, 115 (B112209), doi:10.1029/2010JB007649
Parcharidis, I., Pavlopoulos, A., Poscolieri, M., 2001. Geomorphometric analysis of the
EP

Vulcano and Nysiros island: clues of the definitions of their volcanic landforms. In:
Giovannelli, F. (Ed.), Proceedings of the International Workshop “The Bridge
C

Between Big Bang and Biology”, Stromboli (Messina, Italy), 13-17 September 1999.
AC

CNR President’s Bureau of the National Research Council pub., Special Volume,
310-320.
Pedersen, G.B.M., 2016. Semi-automatic classification of glaciovolcanic landforms: An
object-based mapping approach based on geomorphometry. Journal of Volcanology
and Geothermal Research 311, 29–40.
Pike, R.J., 2002. A bibliography of terrain modelling (geomorphometry), the
quantitative representation of topography. U.S. Geological Survey, Open-File Report

33
ACCEPTED MANUSCRIPT

02-465.
Pike, R.J., Evans, I.S., Hengl, T., 2009. Geomorphometry: a brief guide. In: Hengl, T.,
Reuter, H.I. (Eds.), Geomorphometry: Concepts, Software, Applications. Elsevier,
Developments in Soil Science, 33, 3-30.
Poscolieri, M., 2010. Geomorphometry as a value added to land analysis. In: Proceed.
of the E-ICES 5 Conference, Malargüe (Argentina), 24-27 November 2009.

PT
Comisión Nacional Energía Atómica, Buenos Aires, 236-244.
Poscolieri, M., Foumelis, M., Parcharidis, I., Camiz, S., 2011. Relationship between

RI
topographic relief and ground deformation at Cephalonia Island (Greece). In:
Kalogirou S. (Ed.), Proceed. of the ECQTG 2011 - 17th European Colloquium on

SC
Quantitative and Theoretical Geography, Athens (Greece). Greek Society for
Demographic Studies, 484-491.

U
Roverato, M., Capra, L., Sulpizio, R., Norini, G., 2011. Stratigraphic reconstruction of
AN
two debris avalanche deposits at Colima Volcano (Mexico): insights into pre-failure
conditions and climate influence. J. Volc. Geoth. Res. 207, 33-46.
Siebe, C., Abrams, M., Macìas, J.L., 1995. Derumber Gigantes, Depositos de Avalancha
M

de Escombros y Edad del Actual Cono del Volcán Popocatépetl. In: Volcán
Popocatépetl: Estudios Realizados Durante la Crisis de 1994-1995. CENAPRED-
D

UNAM, 195-220.
TE

Székely, B., Karátson, D., 2004. DEM-based morphometry as a tool for reconstructing
primary volcanic landforms: examples from the Börzsöny Mountains, Hungary.
Geomorphology, 63(1), 25-37.
EP

Tachikawa, T., Kaku, M., Iwasaki, A., Gesch, D.B., Oimoen, M.J., Zhang, Z.,
Danielson, J.J., Krieger, T., Curtis, B., Haase, J., Abrams, M., and Carabajal, C.,
C

2011. ASTER Global Digital Elevation Model Version 2 - Summary of Validation


AC

Results. Report, Federal Government Series, NASA, 27.


https://lpdaacaster.cr.usgs.gov/GDEM/Summary_GDEM2_validation_report_final.pdf
Thouret, J.C., 1999. Volcanic geomorphology - an overview. Earth-Science Reviews
47, 95-131.
Voight, B, Elsworth, D., 1997. Failure of volcano slopes. Géotechnique, 47(1), 1-31.
Wilson, J.P., 2012. Digital terrain modelling. Geomorphology 137, 107-121.

34
ACCEPTED MANUSCRIPT
MEAN
GEOMORPHOMETRIC RELIEF TOPOGRAPHIC
ELEVATION/
CLASSES INTERPRETATION GRADIENTS (m)
SLOPE/ASPECT
1 (6.42% -> 8.93 Km²) 1909.40 m 4.65 -3.80 -12.19
Averagely sloping 18.43° 8.49 -8.46
areas facing SW
240.62° 12.39 3.95 -4.49
2 (8.43% -> 11.72 Km²) Gently sloping areas 1821.40 m 9.81 4.14 -1.50
facing NE 14.84° 5.78 -5.72
310.75° 1.76 -4.01 -9.70
3 (2.99% -> 9.02 Km²) 1878.10 m 11.79 11.21 10.71
Gently sloping areas 25.35° 0.68 -0.57
facing N
357.60° -10.55 -11.21 -11.79
4 (4.63% -> 13.99 Km²) 1602.09 m 1.79 3.62 5.45

PT
Quite gently sloping 10.15° -1.86 1.88
areas facing NE
25.39° -5.41 -3.60 -1.75
5 (2.92% -> 8.81 Km²) 1993.89 m -7.57 1.36 10.15
Gently sloping areas 17.61° -8.98 9.00
facing E

RI
79.82° -10.08 -1.19 7.82
6 (4.05% -> 12..2 Km²) 1898.19 m -9.42 -5.12 -0.84
Gently sloping areas 14.97° -4.29 4.29
facing SE
145.00° 0.91 5.17 9.45

SC
7 (1.94% -> 5.86 Km²) Moderately Steep 2057.89 m -19.22 -14.34 -9.01
sloping areas facing 31.81° -5.04 5.18
S-SE 164.42° 9.35 14.40 19.25
8 (3.00% -> 9.07 Km²) Averagely sloping 1886.64 m -8.08 -10.80 -13.23
areas 24.75° 2.67 -2.60
facing mostly S

U
190.66° 13.26 10.86 8.30
9 (1.94% -> 5.85 Km2) Moderately steep 1970.33 m 18.91 3.72 -11.60
sloping areas facing 28.27° 15.55 -15.54
AN
W-NW 286.55° 11.82 -3.52 -18.75
10 (2.06% -> 6.22 Km²) 2056.40 m 22.38 13.37 4.09
Quite steep sloping 32.37° 9.49 -9.43
areas facing NW
330.73° -3.81 -13.34 -22.41
11(1.59% -> 4.79 Km²) 1961.30 m 18.70 20.28 21.24
M

Steep sloping areas 40.02° -1.21 1.32


facing N
2.84° -21.02 -20.18 -18.50
12 (2.52% -> 7.61 Km2) Quite steep sloping 2018.15 m 3.93 12.03 20.04
areas 29.56° -8.34 8.39
D

facing NE 28.67° -19.97 -11.93 -3.60


13 (0.79% -> 2.37 Km2) 2174.30 m -14.33 3.94 21.66
Steep sloping areas 32.60° -18.74 18.81
facing E-NE
TE

76.74° -21.39 -3.48 14.98


14 (1.50% -> 4.53 Km2) 1929.06 m -24.41 -10.38 4.44
Steep sloping areas 32.69° -14.80 14.82
facing SE
131.11° -4.00 10.62 24.52
15 (0.55% -> 1.67 Km2) Very steep sloping 2328.06 m -24.23 -26.34 -27.39
EP

concave areas 47.11° 1.81 -1.42


facing S 183.00° 27.49 26.26 24.71
16 (1.45% -> 4.38 Km2) 1985.78 m 0.39 -12.43 -24.59
Steep sloping areas 33.63° 12.79 -12.69
facing SW
218.54° 24.76 12.66 0.12
C

17 (7.28% -> 21.99 Km2) Quite flat areas 1638.74 m -0.00 -2.02 -4.01
slightly facing SW 7.43° 2.03 -2.00
AC

224.09° 4.05 2.03 0.03


ACCEPTED MANUSCRIPT
Item Mean St.dev. Minimum Maximum
SM1 17.71 4.00 12.40 30.41
SM2 10.60 3.19 4.27 16.43
SM3 18.29 4.95 12.19 33.32
SM4 10.85 4.07 4.69 20.14
SM5 17.27 3.30 10.27 23.64
SM6 12.32 4.15 4.43 23.60
SM7 20.24 6.03 11.98 33.20

PT
SM8 11.21 5.30 4.34 24.75
SM9 31.53 3.35 26.55 39.84
SM10 27.99 4.01 18.88 34.45

RI
SM11 33.69 4.03 26.90 44.31
SM12 27.97 4.18 17.21 36.20

SC
SM13 31.63 4.64 23.65 42.93
SM14 31.26 4.85 23.66 47.28
SM15 36.60 6.60 26.93 56.34

U
SM16 28.27 5.66 14.81 35.64
AM1 264.05 14.96 235.84 293.17
AN
AM2 309.89 16.83 270.07 336.28
AM3 -2.00 11.28 -22.31 16.32
AM4 49.38 21.81 10.03 104.22
M

AM5 92.60 19.94 51.70 131.82


AM6 138.78 15.95 88.37 161.64
AM7 179.17 15.79 141.48 204.49
D

AM8 219.44 15.34 190.66 253.65


AM9 271.13 21.08 222.42 313.47
TE

AM10 319.37 10.25 297.61 336.01


AM11 0.12 6.98 -16.07 13.81
AM12 40.48 7.53 23.14 51.69
EP

AM13 90.15 12.07 58.90 123.07


AM14 141.66 10.52 121.76 163.90
AM15 184.14 6.95 174.37 201.41
C

AM16 220.66 9.14 190.24 236.76


AC
ACCEPTED MANUSCRIPT

Delimiting study area


Masking outside

PT
GEOMORFO GEOVOLCANOES

RI
INPUT DEM INPUT DEM

U SC
PCA (2 or 3 axes) PCA (2 axes)

AN
Define 100 classes at random Assign 17 centroids

M
KMEANS on

D
100 random classes
HIERARCHY building TE KMEANS on
EP

Choice of # of classes 17 centroids


C
AC

KMEANS on
chosen classes
Classification map Classification map
Statistics' output Statistics' output
ACCEPTED MANUSCRIPT
Summit Masking
Volcano Latitude Longitude Elevation Elevation Area (kmq)
(m) Limit (m)
South-America
Parinacota 18º09'46″S 69º08'34″W 6348 4700 129.57
Socompa 24°23′45″S 68°14′45″W 6051 3650 240.03
Tata Sabaya 19°08'07"S 68°31'27"W 5430 3900 59.57
Ollagüe 21º18'18″S 68º10'45″W 5868 3900 149.58

PT
Peteroa 35°14′24″S 70°34′12″W 4107 2600 159.06
Llaima 38º41'49″S 71°43′50″W 3125 300 563.95
Villarica 39°25′15″S 71°56′21″W 2860 1350 123.70

RI
Lanin 39°37′59″S 71°30′00″W 3747 1300 139.04
Central-America

SC
Atitlán 14°35′00″N 91º11'10″W 3535 70 240.44
Fuego 14°28′58″N 90°52′58″W 3763 50 360.58
Santa Ana 13º51'11″N 89°37′48″W 2381 300 1933.59

U
San Cristóbal 14°35′00″N 91°11′10″W 1745 200 257.04
Momotombo 12º25'18″N 86º32'26″W 1280 20 346.65
AN
Arenal 10º27'48″N 84°42′12″W 1670 500 39.79
Mexico
M

Ceboruco 21°07′30″N 104°30′30″W 2280 1100 96.86


Colima 19°30′46″N 103°37′02″W 3839 1200 1179.14
Jocotitlán 19°43′26″N 99°45′25″W 3900 2750 75.93
D

La Malinche 19°13′48″N 98°01′48″W 4461 2800 238.50


Nevado de Toluca 19°06′29″N 99°45′29″W 4690 3000 448.78
TE

Pico de Orizaba 19°01′01″N 97°16′12″W 5700 2800 744.06


Popocatepetl 19°01′23″N 98°37′19″W 5426 2300 577.53
C EP
AC
ACCEPTED MANUSCRIPT
• A geomorphometric comparison of 21 South- and Central-America volcanoes is proposed

• the volcanoes ASTER/DEMs are classified according to 8 local topographic gradients

• the volcanoes are compared according to the 17 obtained classes differences

PT
RI
U SC
AN
M
D
TE
C EP
AC

Vous aimerez peut-être aussi