Mass Movement Classification Using Morphometric Parameters (Puebla, Mexico)

Abstract

The Sierra Norte de Puebla (Mexico) is a tropical mountain frequently affected by torrential rains induced by depressions and hurricanes. In October 1999 and 2005, intense precipitation triggered hundreds of landslides and caused heavy human losses and material damages. All the landslides surveyed in the field and their traces extracted from the satellite images have been analyzed by using morphometric parameters in order to characterize and to classify them. The spatial distribution of observed landslides types is not random but responds to both regional structural features and material nature involved in the movement. It is particularly important to establish these close relationships in order to assess landslide hazards by using different factors and methods and the approach proposed in this paper seems to be very useful to do that.

Full text

Proceedings of the Second World Landslide Forum – 3-7 October 2011, Rome
Ochoa-Tejeda Veronica
(1)
, Parrot Jean-François
(2)
, Fort Monique
(1)
Mass movement classification using morphometric
parameters (Puebla, Mexico).
(1) Université Paris Diderot, Sorbonne-Paris-Cité, UMR PRODIG 8586 – CNRS, France
(2) LAGE, Instituto de Geografía, UNAM, Mexico
Abstract The Sierra Norte de Puebla (Mexico) is a
tropical mountain frequently affected by torrential rains
induced by depressions and hurricanes. In October 1999
and 2005, intense precipitation triggered hundreds of
landslides and caused heavy human losses and material
damages. All the landslides surveyed in the field and their
traces extracted from the satellite images have been
analyzed by using morphometric parameters in order to
characterize and to classify them. The spatial distribution
of observed landslides types is not random but responds
to both regional structural features and material nature
involved in the movement. It is particularly important to
establish these close relationships in order to assess
landslide hazards by using different factors and methods
and the approach proposed in this paper seems to be very
useful to do that.
Keywords Morphometric parameters, Landslide
classification, Sierra Norte de Puebla (Mexico).
Introduction
Landslide hazard assessment is based on a landslide
susceptibility evaluation, that is to say the research of the
location of potential landslides that can occur in a given
region and remobilize specific terrain material (Guinau et
al., 2007). It is commonly assumed that the future
landslides triggering obey to the same conditioning
parameters than those observed in the past and the
present (Varnes, 1984; Carrara et al., 1995, Jaddah et al.,
2009). Such an analysis requires estimating the terrain
failure susceptibility and the behavior of the remobilized
material (Glade et al., 2005; Hürlimann et al., 2006) as
well as the land cover changes particularly when studying
the shallow landslides (Collinson et al., 1995; Fannin et
al., 1996; Garcia-Ruiz et al., 2010).
Independently of the classifications established by
different authors (i.e. Hutchinson, 1988; Cruden and
Varnes, 1996), we assume that the predominant types of
landslides are essentially related to the nature of the
material involved in the process (soils, colluvial deposits,
talus deposits and talus breccias, uncovered rock zones).
In the case of rock sliding and rock avalanches, eventually
coarse debris flows, the structural geologic factor plays a
predominant role (Ochoa-Tejeda, 2004), since the
presence of faults, fractures, as well as planes of weakness
(schistosity, joints, etc.), work like zones that supply the
gravitational movements.
The purpose of this paper is to analyze, in a context
of dense vegetation, all the landslides surveyed in the
field as well as their traces extracted from satellite
images. This analysis is mainly based on morphometric
parameters, and has been developped in order to
characterize and to classify the mass movements. High
resolution satellite images and high resolution Digital
Elevation Model (DEM) generated by a multidirectional
interpolation (Parrot and Ochoa-Tejeda, 2005) have been
used with such a goal. Actually IKONOS images give not
only information about land cover, presence of bare and
water surfaces, influence and extent of human
settlements, vegetation cover and area of fragmentation,
but can also be used to identify the different types of
landslides that occurred in the study region by means of
photo interpretation (Nichol et al., 2006) or using the
semi-automated method developed by Ochoa-Tejeda and
Parrot (2007).
We assume that it is possible to define geographical
unstable zones in a simple way and then to characterize
the type of events susceptible to be expected in a more or
less short time according to the type of geological
formation that would be involved in the mass movement.
Studied area
The Sierra Norte de Puebla (Mexico) is a tropical
mountain frequently affected by torrential rains inducing
regularly hundreds of landslides that affect steep
hillslopes mainly formed by intensively folded and
fractured metamorphic rocks and weathered sedimentary
formations both covered by slope deposits of variable
thickness.
This mountain is situated in the province of Puebla
located within the transition of the Trans-Mexican
Volcanic Belt (TMVB) and the Sierra Madre Oriental
mountains mainly formed by Mesozoic sedimentary
rocks. The TMVB consists of Late Tertiary and
Quaternary volcanic formations essentially of calc-
alkaline type (Alva-Valdivia et al., 2000).
The studied sub-scene is centered on the 135 km
2
region of La Soledad Lake (Fig. 2), where 98 landslides
recorded by the Regional Civil Protection Unit of Puebla,
actually 144 observed on the satellite image, have been
triggered by torrential rainfalls in October 1999.

Verónica Ochoa-Tejeda, Jean-François Parrot, Monique Fort – Mass movement classification using morphometric parameters.
2
Figure 1. Localization of the studied zone.
In this zone located between 19°53’ and 20°00’
(North latitudes) and 97°25’ and 97°30’ (W longitudes),
the main geological units are ranging from Palaeozoic to
Quaternary ages (Angeles-Moreno and Sánchez-
Martínez, 2002).
Figure 2. Geological map of the study zone. Qa-Ix = Xaltipan
formation (Quaternary Ignimbrites); Qa-Tb = Quaternary
andesitic tuffs; Plio-Tz = Teziutlan formation (andesitic flows
and tuffs); JS-Tp = Taman-Pimienta formation (marls and
limestone); JM-Tx = Tenexcate formation (sandstones and
conglomerates); JI-Hy = Huayacocotla formation (sandstones
and shale); Xucayucan mylonitic complex: Mi = El Mirador
(metamorphosed basaltic rocks); Cz = Cozolexco
(metabasaltes); Cc = Chicuaco (schist); La Soledad mylonitic
complex [Cs] = gneisses; = intrusive rhyolites.
The Palaeozoic formations are represented by
mylonites and strongly deformed phyllonites originally
defined as Xucayucan schists. Mesozoic folded
sedimentary rocks overlay discordantly the Palaeozoic
deposits and comprise limestones, sandstones, lutites,
limolites and conglomerates. Tertiary deposits from the
Pliocene are formed by andesitic lava flows up to 300 m
thick intercalated with volcanic tuffs. Quaternary
materials are composed of dacitic and rhyolitic
pyroclastic deposits derived from the activity of the Los
Humeros caldera, which is situated 35 km southwards of
the studied area. These materials overlay both Mesozoic
and Tertiary deposits.
Moreover, this zone is characterized by the
development of different superficial formations (regolith,
old landslide deposits) that are easily remobilized by the
frequent torrential rains affecting this tropical region.
The behavior of different types of material can therefore
be studied in this very limited zone.
The reinterpreted geologic map (Ochoa-Tejeda,
2009; 2010) takes into account field observation and the
information provided by INEGI (1994), Mooser (2000)
and Angeles-Moreno and Sanchez-Martinez (2002). We
reported (Fig. 2) landslide locations as the traces
extracted from the Ikonos image for the 1999 event, and
as points for the more recent mass movements).
Methodological approach
All the landslides surveyed in the field and their traces
extracted as previously described from the satellite
images, have been analyzed by using morphometric
parameters in order to characterize and to classify them.
Some parameters are directly related with the shape of
studied landslides: for instance, the surface S, the
perimeter P, the ratio between these two parameters, as
well as the presence of holes that allows defining a
porosity index.
The simplest method used to measure the surface
consists to calculate the total number of pixels N
bp
belonging to the connected component; but according to
Pratt (1978), another approach consists to consider the
surface as

 2
PS
PPS
, where P
S
are the pixels that
strictly belong to the surface and P
P
the pixels describing
the perimeter. The surface in m
2
or km
2
is obtained by
multiplying S by the value of the pixel surface. There are
some more accurate methods that take into account the
configuration of P
p
and the neighboring pixels in order to
know exactly the portion of the pixel P
p
that belongs to
the surface (Parrot, 2007). The perimeter P corresponds
either to the total number of pixels (N
p
) describing the
perimeter or to the length (L
p
) of this perimeter
measured by using the length of the segment linking two
successive pixel centers.
Taking into account the values of the former
parameters, it is possible to define various ratios such as
for example the ratio perimeter/surface [in
pixels]
 
100
bpp
NN

, the ratio perimeter versus
surface [in meters and square meters]
 
100 SL
p

or
the circularity index
 
100
2
 SP

.
The notion of porosity characterizes topographically
heterogeneous ensembles as encountered in some
landslides, especially in the case of rotational landslides
(graben-like structures). The calculation is as follow:

Proceedings of the Second World Landslide Forum – 3-7 October 2011, Rome
3
100
2





















 

p
sh
h
P
PP
P

where P
s
and P
p
are all the pixels describing the
shape (surface and perimeter) and P
h
the pixels
corresponding to the holes.
Another way to describe the shape is to compare
this shape and some plain shapes such as the rectangle,
the circle or the smaller convex zone that circumscribes
the studied item. It is then possible to measure the length
and the width of the studied shape, to define the relation
existing between these two values, etc… This approach
requires to research the gravity center (GC) and to define
the principal axis (PA) passing through the gravity center.
The gravity center GC with coordinates X
c
, Y
c
and the
second order moments

xx
,

yy
and

xy
are equal to:



Nbp
i
ic
X
Nbp
X
1
1



Nbp
i
ic
Y
Nbp
Y
1
1
 



Nbp
i
cixx
XX
1
2

 



Nbp
i
ciyy
YY
1
2

  



Nb p
i
cicixy
YYXX
1

N
bp
is the total number of pixels and X
i
Y
i
the
coordinates of a pixel i.
The principal axis PA that passes through GC is
equal to:
 
 
xxyyxy
tg

 22
if
 
0
xxyy

. When the
difference between
xx

and
yy

is equal to
0
, the
connected component does not present any orientation.
Otherwise, the PA orientation is calculated clockwise in
degrees (from 0° to 180°, 0° corresponding to the north).
It is also possible to compute all the distances between
GC and all the perimeter pixels; in this case, PA
corresponds to the perpendicular to the straight line that
links GC and the closest pixel belonging to the perimeter.
It is then possible to measure the length and width of the
rectangle circumscribing the studied landslide surface
and calculate its ratio. As the principal axis PA intersects
the shape perimeter in two points (D
1
and D
2
), the
direction D of the landslide can be computed (Ochoa-
Tejeda and Parrot, 2007). The relative hypsometric values
of D
1
and D
2
permit to define a minimum (A
min
) and a
maximum (A
max
). The direction D of the landslide
corresponds to the straight line that links GC and A
min
.
The measurement of the direction, especially in the case
of shallow (hypersaturated) debris slides allows defining
zones that present an uniform slope gradient related to
the triggering of such superficial mass movements.
There are many different procedures that allow
defining a convex zone. Among them, the Jarvis’s march
such as it is described by Akl and Toussaint (1978) is a
simple algorithm that efficiently draws a convex contour
(Fig. 3).
The surface of the convex zone S
c
(or the total
number of pixels N
tc
of this zone) and its perimeter P
c
calculated in pixels or in meters allows calculating several
parameters: the ratio S/S
c
or the relation N
bp
/N
tc
that
both correspond to two convexity indices based on the
surface, the ratio P
c
/P
p
(perimeter convexity or external
roughness index) or P
h
/N
tc
(porosity inside the convex
zone).
Figure 3. The Jarvis’s march procedure.
Results
The morphometric parameters as well as the
relationships between each other allow defining
accurately the characteristics of the different extracted
traces produced in the landscape by slope movements.
Only two diagrams are reported here in order to illustrate
some of the characteristics of these traces. The graph
width versus length (Fig. 4a) reveals that, except for two
items, small landslides essentially form a dense group of
linear type; the relation width/length ranges from 0.2 to
0.8 with an average of 0.55; 40% of the traces display a
relation equal or inferior to 0.5 and 24% of them are
inferior or equal to 0.4.
Figure 4. Two examples of parameters correlation.
Two bigger landslides are present. The first one
(symbol A in the figure 6) corresponds to a rotational
landslide in the region of San Jose Chagchaltzin; it has an
almost circular form (relation width/length of 0.82) with
a width size of 238 meters and a length of 289 meters.
The second landslide (symbol B in the figure 6) is located
northwards the El Dos village; its length is equal to 480
meters and the width to 90 meters; the ratio
width/length of this linear structure is equal to 0.18.
Figure 4b establishes the existing relation between
the external roughness index P
c
/P
p
and the surface S. The
more the index value decreases, the more irregular will be
the studied shape. The majority of the studied landslides
are grouped in the zone corresponding to small
landslides having an almost regular form; the external
roughness index of the cluster gravity center is equal to
75. Even if the large debris flow B is more elongated, it
has the same P
c
/P
p
value. For the large rotational
landslide A, even if it is almost circular, the great
irregularity of its contour related to the heterogeneous

Verónica Ochoa-Tejeda, Jean-François Parrot, Monique Fort – Mass movement classification using morphometric parameters.
4
topography during the rotational sliding implies a low
value of the external roughness index. It should be noted
that the porosity parameter of this item is the highest
obtained in the studied zone.
All the possible combinations between the various
parameters calculated in such a way have been tested;
finally, because these indices present a weak correlation
creating the strongest dispersion of all the points, the
graph between the perimeter convexity index P
c
/P
p
and
the ratio perimeter versus surface θ allows classifying
better the different traces extracted from the Ikonos
satellite image (see Fig. 5).
It is then possible to define six classes taking into
account ranges of common values of these two
parameters.
The parameter  in fact reflects the globularity of a
shape. Regardless of its length a line has a  value equal
to 100. This value decreases when this line thickens. Thus
for a circle the size of which is greater than 150 pixels, the
value  is lower or equal to 10 and this value decreases
when the circle is growing.
Figure 5. Landslide classification using the diagram perimeter
convexity P
c
/P
p
versus the ratio perimeter/surface
 
100 SL
p

.
Dividing the  axis in three groups, high, medium
and low values, it may be noted that shallow debris slides
correspond to θ values > 60, debris flow developed in
slope deposits have θ values comprised between 40 and
60, and many other features have a value < 40. Among
these latter, a few items (large debris flows and complex
movements) have a perimeter convexity index value < 50.
Figure 6. Repartition of the mass movement families in the studied region.

Proceedings of the Second World Landslide Forum – 3-7 October 2011, Rome
The last mentioned group corresponds to a very
thick line, reason why its θ value is lower than 40, and the
low value of the perimeter convexity index Pc/Pp
indicates that the shape corresponds to a broken line
(constrained by thalweg).
The following group (Pc/Pp between 50 and 80)
corresponds to thick lines which present a slightly curved
contour; all the debris avalanches are comprised in this
group.
Finally convex forms have a P
c
/P
p
value greater than
80. Actually, according to field observations, it is possible
to introduce a division in this last group.
We obtain then a class where are located the
entirely convex shape of high globularity (θ > 18) and
another one where convex shape not so globular (18 > θ >
40); this distinction allows to sort out rotational from
translational landslides.
A label and a color are attributed to the different
traces according to the landslide class they belong to. As
shown by the figure 6, the six classes are not randomly
distributed and different groups are perfectly
distinguishable.
As presented in this figure, the shallow debris slides
(1) are located on thin colluviums that especially cover
the metamorphic rocks, and the debris slides (2) occur on
thicker colluviums that correspond to a mixture of soils
and broken metamorphic and sedimentary elements. The
translational and rotational landslides (respectively 3 and
4) are essentially related to older slope deposits. Debris
avalanches (5) are mainly observed in the folded and
fractured metamorphic formations and the complex
movements (6) are related to uncovered rock zones.
Discussion
According to the classification provided by using
morphometric parameters, it is obvious that the
distribution of the different mass movement types is not
random, but corresponds to the material that can be
involved in the movement. This is particularly clear for
the linear traces remobilizing slope deposits. Thus it was
possible to define geographical unstable zones in a simple
way and to characterize the type of events susceptible to
be expected in a more or less short time according to the
type of material that would be involved in the mass
movement. For instance, on the main road linking
Tlatlauquitepec and Atotocoyan (Fig. 7), shallow slides
and superficial debris flows triggered by 2005 rainfall
took place exactly at the location considered after the
1999 event as an unstable hillslope covered by thick
unconsolidated colluviums.
The soil behaviour depends on its texture, porosity
and mineralogical composition. In the present case, the
colluvial formation reported in figure 7 comes from
weathering, erosion and degradation of the upper part of
the metamorphic formation. It is a very heterometric,
easily broken material with a high porosity, and it can be
rapidly saturated as it was the case in 1999 and 2005.
Parrot and Ochoa-Tejeda (2009) and Ochoa-Tejeda
and Fort (2011) have recently shown the relationship
existing between heavy rain periods (tropical hurricane)
and the hillslope behaviour, hence mass movement types.
For instance, as it occurred in 1999 (see Fig. 8), when a
sequence of three days of intense rainfall was followed by
a pause of two-three days, then by another short period
of intense showers, the material remobilized by the mass
movements affected not only the superficial slope
deposits, but also uncovered rock zones. When the
rainfall interruption is shorter as it occurred in 2005, the
landslides affect mainly superficial formations, and may
reactivate previous landslide zones.
Figure 7. Position of the October 1999 mass movements
(extracted from the satellite image) and the landslides detected
at the end of 2005.
The method presented in this paper appears quite
efficient when extracting and characterizing shallow
slides, debris flows and even debris avalanches, i.e. more
or less linear structures. In contrast the sorting out of
“globular”, compact and convex shapes appears more
difficult and claims a perfect understanding of
parameters significance. In a tropical mountainous area
with relatively dense vegetation cover these traces are
easier to extract than linear structures yet are more
difficult to classify. Even if the method needs some
improvements in the definition of adaptive parameters to
avoid this weakness, the algorithm is easy to implement
and the output is obtained rapidly enough so that it can
directly be used and implemented in crisis management
by the authorities when infrastructures and populations
are severely impacted by landslides occurrences.
Figure 8. Cumulative pluviometric curves during 7 years (1998
untill 2005).

Verónica Ochoa-Tejeda, Jean-François Parrot, Monique Fort – Mass movement classification using morphometric parameters.
6
Conclusion
We propose here a new method for extracting landslide
traces from satellite images. The study area is a tropical
mountain frequently subjected to torrential rains, and
characterized by relatively dense vegetation cover and a
variety of geological and superficial formations. The
landslide traces were analyzed in using a set of
morphometric parameters, including perimeter, surface,
convexity, porosity, and defining some specific ratios.
The method developed here allows characterizing and
classifying easily different landslide types. It appears
quite robust for identifying linear landslide types such as
shallow debris slides, triggered by intense hurricane rains
and causing most damages to the population and
infrastructure. This method appears as fairly simple and
rapid to implement, and its results are robust enough, so
that it can be applied in other tropical regions as a useful
tool in crisis management following landslide
occurrences in areas of difficult access.
References
Akl S G, Toussaint G (1978) Efficient Convex Hull Algorithms for
Pattern Recognition Applications. Proc. 4th Int. Joint Conf. on
Pattern Recognition, Kyoto, Japan. p. 483-487.
Alva-Valdivia L M, Goguitchaichvili A, Urrutia-Fucugauchi J, Ferrari L,
Rosas-Elguera J, Zamorano-Orozco J J (2000) Paleomagnetic data
from the Trans-Mexican Volcanic Belt: Implications for tectonics
and volcanic stratigraphy. Earth, Planets and Space. 52(7): 467-
478.
Angeles-Moreno E, Sánchez-Martínez S (2002) Geología, geoquímica
y geología estructural de las rocas del basamiento del macizo de
Teziutlan, estado de Puebla. Profesional Thesis, Facultad de
Ingenieria, UNAM, México, 105 p.
Carrara A, Cardinali M, Guzzetti F, Reichenbach P (1995) GIS
technology in mapping landslide hazard, in: Geographical
Information Systems in Assessing Natural Hazards, edited by:
Carrara A and Guzzetti F, Academic Pub., Dordrecht, the
Netherlands, 360 p.
Collinson A J C, Anderson M G, Lloyd D M (1995) Impact of
vegetation on slope stability in a humid tropical environment: a
modelling approach. Proc. of the Institution of Civil Engineers.
Water, Maritime and Energy. 112: 168–175.
Cruden D M, Varnes D J (1996) Landslides types and processes. In:
Turner A K, Schuster R L (Eds.), Landslides: Investigation and
Mitigation. Spec. Rep.-Transp. Res. Board, 247: 36– 75.
Fannin R J, Wise M P, Wilkinson J M T, Rollerson T P (1996) Landslide
initiation and runout on clearcut hillslopes. Proc. 7th Internat.
Symposium on Landslides, Trondheim, Norway, 195–199.
García-Ruiz J M, Beguería S, Alatorre L C, Puigdefábregas J (2010)
Land cover changes and shallow landsliding in the flysch sector of
the Spanish Pyrenees. Geomorphology. 124:250–259.
Glade T, Anderson M, Crozier M J (Eds.): Landslide hazard and risk,
Wiley, 803 pp., 2005.
Guinau M, Vilajosana I, Vilaplana J M (2007) GIS-based debris flow
source and runout susceptibility assessment from DEM data – a
case study in NW Nicaragua. Natural Hazards and Earth System
Sciences. 7: 703-716.
Hürlimann M, Copons R, Altimir J (2006) Detailed debris flow hazard
assessment in Andorra: A multidisciplinary approach.
Geomorphology. 78: 359-372.
Hutchinson J N (1988) General report morphological and
geotechnical parameters of landslides in relation to Geology and
Hydrology. In Bornnard C (ed), Proceedings of the Fifth
International Symposium on Landslides, A. A. Balkema,
Rotterdam, Netherland, 1: 3-35.
INEGI (1994) Carta Geologica de Poza Rica. Esc: 1:250 000. Mexico.
Jadda M, Shafri H Z M, Mansor S B, Sharifikia M, Pirasteh S (2009)
Landslide Susceptibility Evaluation and Factor Effect. Analysis
Using Probabilistic-Frequency Ratio Model. European J. of
Scientific Research. 33(4): 654-668.
Mooser F (2000) Mapa geológico de Laguna Verde, Mexico.
Comision Federal de Electricidad.
Nichol J, Shaker A, Wong M-S (2006) Application of high-resolution
stereo satellite images to detailed landslide hazard assessment.
Geomorphology. 76: 68-75.
Ochoa-Tejeda V (2004) Propuesta metodologica para el estudio de
inestabilidad a partir de los MDT y la Percepción Remota. Master
thesis, UNAM, 213 p.
Ochoa-Tejeda V (2009) Control estructural de la inestabilidad de
laderas: Modelación tridimensional de los rasgos geológicos.
Aplicación al estudio de los procesos de remoción en masa de La
Soledad, Sierra Norte de Puebla, México. (2009). PhDthesis,
Earth Sciences, Geophysical Institute, UNAM , 217 p.
Ochoa-Tejeda V (2010) Etude des facteurs favorables au
déclenchement des glissements de terrain dans les formations
superficielles et les affleurements rocheux de la Sierra Norte de
Puebla (Mexique). PhD thesis, Université Paris 7 Denis Diderot,
Paris France. 215 p.
Ochoa-Tejeda V, Fort M (2011) Relation entre la pluviométrie et le
déclenchement des glissements de terrain dans La Soledad,
Sierra Norte de Puebla, Mexique. Bulletin de l’Association des
Géographes Français. 1: 27-34.
Ochoa-Tejeda V, Parrot J-F (2007) Extracción automatizada de trazas
de los deslizamientos utilizando un modelo digital de terreno e
imágenes de satélite de alta resolución. Ejemplo de aplicación: La
Soledad, Sierra Norte, Puebla, México. Revista Mexicana de
Ciencias Geológicas. 24(3): 354-367.
Parrot J-F (2007) Tri-dimensional parameterization: an automated
treatment to study the evolution of volcanic cones.
Géomorphologie. 3: 37-47.
Parrot J-F, Ochoa-Tejeda V (2005) Generación de Modelos Digitales
de Terreno Raster. Método de digitalizacion. Geografía para el
Siglo XXI, UNAM, 31 p.
Parrot J-F, Ochoa-Tejeda V (2009) Auto-related fractal analysis of
triggering factors and landslide assessment. An example from the
Sierra Norte de Puebla, Mexico. 6
th
annual meeting AOGS.
Singapore, 11-15 august 2009.
Pratt W K (1978) Digital image processing. Ed Wiley, New York. 750
p.
Varnes D J (1984) Landslide hazard zonation: a review of principles
and practice. IAEG Commission on Landslides and Other Mass-
Movements, UNESCO Press, Paris, 63 p.