[PDF]Landslide inventory for Moldavian Plateau, Romania

International Conference


Analysis and Management of Changing Risks for Natural Hazards
18-19 November 2014 | Padua, Italy


Abstract code: AP3
Landslide inventory for Moldavian Plateau, Romania
M, Niculita', MC, Margarint'


' Geography Department, Geography and Geology Faculty, Alexandru loan Cuza University,
lasi, Romania


Corresponding author details:
Geography Department, Geography and Geology Faculty, Alexandru loan Cuza University,
Carol |, 20A, 700505, lasi, Romania; mihai.niculita@uaic.ro


Keywords
Landslide inventory, satellite imagery, LIDAR DEM, Moldavian Plateau, Romania


Extended Abstract
INTRODUCTION


Landslides, as mass movement geomorphologic processes, are characteristic for any
climatic zone, and occur at any spatial and temporal slice (past, present, future). The causes
of landslides triggering are various, and usually, predisposing and preconditioning factors are
needed. The variety of mechanisms (Varnes, 1978, Cruden and Varnes, 1996, Hungr et al.,
2014) and magnitudes is a characteristic of these processes and shape. Clear clustering
properties are seen at temporal (Witt et al., 2010) and spatial level.

The cartographic representations of the extant slope failures, the landslide inventories, are
the most basic elements in analysis of controls on the spatial and temporal patterns of mass
movements and the environmental, human or geomorphic consequences of slides (McKean
and Roering, 2004, Van Westen et al., 2008, Guzzetti et al., 2012). Due to permanent
development of new remote sensing data and techniques, nowadays there is a general
increase of accuracy of landslides inventories, with more and more theoretical and applied
outcomes in landslide management and mitigation. The high quality of these images allow a
better understanding of local conditions for each landslide, and consequently a better
typological classification, through deciphering and interpretation of a large range of
morphological features (Tarolli et al., 2012).

In this context, we have created a polygon based landslide inventory containing 24,263
landslides for the entire area of Moldavian Plateau, NE Romania which cover 24,803 km.
The landslide inventory can be used to study and analyse the spatial patterns of the
landslides, both as processes and as morphologies created. The present approach can
constitute a baseline for the regional assessment of landslide susceptibility, and further for
hazard and risk evaluation (Guzzetti et al., 1999).


STUDY AREA

The Moldavian Plateau is one of the most representative unit plateau of Romanian territory
Despite many limits of the Moldavian Plateau cited in geographical literature (Bacauanu et
al., 1980), this approach will relate to the area comprised between Prut Valley (to the east),
Siret-Moldova valleys (to the west) and the northern border of the country (Figure 1). The
border between plateau unit and Carpathian Mountains, in north-western part of the study
region was drawn according the geological map of Romania at scale 1:200,000. The south-
west extreme part, dominated by fluvial terraces and holocene alluvial deposits, known as
Tecuci Plain, was not included in the studied area, being considered a part of Romanian


AP3 -1


International Conference


Analysis and Management of Changing Risks for Natural Hazards
18-19 November 2014 | Padua, Italy


Stereographic 1970 oS
projection G
(EPSG:3844)


(a)


DOOSONARAD
SOSSIOSi
SOS°SS 885:


PONS OOS mma
3


°
oO
vt


——— km
0 80 160 240
27°30° 10 O 10 20 30 40km
!


i=]
—vTt


10


=
oO


20 miles


720
640


Altitude [m]


Altitude [m]
global


landslided
= areas


s
>
oO
Ge.
oO
|
o
£
o
2
=
io
oa
“


Relative frequency [%]


Relative frequency [%]


100 150 200 250


= MuN
aon


Relative frequency [%]


o-=-
Sumo


Figure 1 Geographical position, physiographic subdivisions, elevation distribution and descriptive
statistics graphics of the Moldavian Plateau, based on SRTM data; on the left side there are
represented the distributions of altitude (a), slope (c), slope height (d) and slope aspect (e,f); for the
map (a) the letters indicate the geomorphological regions: a — Jijia Hills, Barlad Plateau composed
from b — Central Moldavian Plateau, c — Tutova Hills, d — Fdlciu Hills, e — Covurlui Plateau, f — Husi-
Sdrata-Elan-Horincea Hills, Suceava Plateau composed from g — Siret Hills, h — Falticeni Plateau, i —
Dragomirnei Plateau


AP3 - 2


International Conference


Analysis and Management of Changing Risks for Natural Hazards
18-19 November 2014 | Padua, Italy


Plain. Toward east, the Moldavian Plateau continues on the territory of the Republic of
Moldova. In between these limits, the study area covers 24,803 km’, with altitudes ranging
between 5 to 794 m as.l. The relief of the plateau was modelled on a monoclinal
sedimentary structure with a mean inclination of 7-8 m/km from north-west to south-east,
contrasting with crystalline basement of East-European Platform which sinks toward west,
under Carpathian Orogene (lonesi et al. 2005). This monoclinic structure of the sedimentary
coverture has controlled and designed a repetitive pattern of landforms (Niculita, 2011) with
obvious influences for many other environmental and human life characteristics
(hydrographic network, land use pattern, the main transport corridors, etc.).

The age of surface lithological formations is comprised between Cretaceous (which outcrop
in north-east, along Prut river banks) and Quaternary deposits (alluviums of floodplains and
terraces of the main rivers, and loess in southern part), with a large extension of Buglovian to
Romanian formations. Due to unequally and earlier uplifting of the northern part, which
progressively transformed actual plateau into an continental area, the sedimentary deposits
outcrop, in order of age, from north to south (lonesi et al. 2005). The lithology is represented
by alternances of consolidated rocks (sandstones, limestones, tuffs and fewer micro-
conglomerates) which mainly occur on the ridges and unconsolidated rocks, like clays, silts,
and sands, which occur beneath. The erosion has detached many morphological alignments,
supported by these consolidated rocks, some of these alignments being impressive, with a
remarkable and repetitive occurrence of deep seated landslides (Margarint and Niculita,
2014). These morphological alignments have delineated three main subunits of the
Moldavian Plateau (the Central Moldavian Plateau, the Suceava Plateau and the Jijia Hills —
Figure 1).


METHODS

As source for landslide affected areas and delineation, a wide range of remote sensing
imagery were used. These include optical satellite imagery available in the Google Earth®
and Bing Maps archives (SPOT, Formosat, WorldView, GeoEye, Pleiades, QuickBird,
KOMPSAT, IKONOS), national ortorectified imagery and where available, LIDAR DEM data
(for approx. 40% of the studied area). The digitization was performed in Google Earth® and
Quantum GIS software. Shading maps computed from LIDAR DEM data with 0,5 m spatial
resolution was used for the delineation of the landslides. Geomorphometric analysis of the
entire studied area and subsequently for the landslide inventory was performed by using
SRTM data (SRTM v.2), resampled at 30 m resolution with a bilinear interpolator and SAGA
GIS v. 2.12. Slope and aspect were computed with the maximum gradient method.

Landslide width and length was computed by considering the direction of the mass flow. In
this context, there is a special situation, with the landslides which develop on cuesta scarp
slopes. These landslides, usually have widths larger than the lengths, and requested special
processing for the separation from the landslides which have lengths larger than widths. The
processing involved the computation of the mass flow direction using the difference of height
for the lengths and widths of the bounding box rectangle for each landslide. The landslides
with widths larger than the lengths (Figure 2e), have smaller height difference on the long
side of the bounding box, while the landslides with lengths larger than the widths (Figure 2a)
have smaller height difference on the short side of the bounding box. Landslide height of
each landslide was computed as the difference between the minimum and maximum slope
height for that landslide. The slope height is computed for every pixel, as the difference
between the altitude of that pixel and the local drainage path. The slope and aspect angles,
for each landslide were considered as the mean values of all the DEM pixels corresponding
to the landslide extent. Beside the mean slope angle of the landslide affected topography,
the general slope given by the ratio between slope height and landslide length was
computed. Structural landforms (cuesta scarp slope and dip slope) classification was
described by Niculita, 2011.


AP3 - 3


International Conference


Analysis and Management of Changing Risks for Natural Hazards
18-19 November 2014 | Padua, Italy


Stereographic 1970
projection
(EPSG:3844)


Besegsse5


+,
0 80 160 240


27° ‘aad


oO
~&10 0 10 20 30 40k


se


20 mi


Figure 2. The spatial distribution of landslide inventory (the insets represent 3D perspectives of LIDAR
shading for typical landslides): a, e — rotational, b — lateral spread, c — translational, d — translational
and flow like, f- flow.


AP3 -4


International Conference


Analysis and Management of Changing Risks for Natural Hazards
18-19 November 2014 | Padua, Italy


For the Moldavian Plateau, a geomorphological historical landslide inventory (Malamud et al.,
2004, Guzzetti et al., 2012) was prepared by using a large series of remote sensing imagery
and data, and numerous field campaigns. This was realized during more than 18 months, by
the two authors of this article. This experience confirm us the fact that the landslide
inventories are essentially factual in nature (Fell et al., 2008). Firstly, a number of five
obvious elements of mass displacement, were used as main criteria, and selected to identify,
locate and to de-lineate the occurrence and spatial extension of landslides: (i) crowns
(undisplaced material, still in place), (ii) scarps (main and/or minor), (iii) roughness of the
mass displaced, (iv) flanks, and (v) toes. Subsequently, using the line editors of specialized
software, the surfaces affected by landslides were drawn individually as polygons features. In
the next step, especially in the cases when an obvious spatial continuity was reported, the
following operations of mapping generalization have been applied: simplification,
amalgamation and refinement.

This landslide inventory is classified as a geomorphological and historical one in the sense
that the optical imagery used as data source for a large area of the Moldavian Plateau,
doesn’t always allowed the delineation of event based landslides, but instead requested their
inclusion in the entire area affected by landslide processes. The same approach was applied
to the areas where LIDAR DEM shading was used as source for landslide delineation (data
available only for 40% of the study area) for getting consistency, although this data source
allowed the delineation of event based landslides. Only the small event based landslides,
recognizable both on remote sensing images and LIDAR DEM shading were considered as
individual landslides, attributable to a certain event. This is consistent with methodology
applied by other authors (Ardizzone et al, 2007, Van den Eeckhaut et al, 2005).

Lastly, through geomorphometric analysis of polygons, environmental settings of landslide
locations, field experience and many verifications, the polygons were attributed according to
Varnes (1978) and Cruden and Varnes (1996) systems.


RESULTS

Overall, the landslide inventory of Moldavian Plateau contains 24,263 landslide polygons,
which correspond to a landslide density of 1.02 landslides per km’. The total area of the
landslides is 4534.7 km? (the overlapping landslides were merged before computing the total
area) that represent 18.3% from the total area of the plateau, corresponding to a landslide
density of 0.19 km? of landslide area for every km’. The following types of landslides have
been identified: 1 — rotational, 2 — translational, 3 — lateral spread, 4.1 — rotational-
translational complexes, and 4.2 — rotational-translational-flow complexes. The absolute
frequency of these types for the physiographic regions of the Moldavian Plateau (Figure 1) is
presented in Table 1.

The most frequent landslide area is 2500 to 3500 m? (1044 counts), because the majority of
landslides have small areas (Figure 3b). Meanwhile the small landslides cover a small
proportion of the area affected by landslides: landslides with areas under 10,000 m?
represent 22.2% from the total number of landslides and 0.6% of the total landslide area.
Those with areas between 10,000m? and 100,000 m? represent 44.5% from the total number
of landslides and 9.5%of the total landslide area.

Large landslides although represent a small proportions of the total number of landslides,
cover the biggest area. Landslides with area comprise between 100,000 m* and 1 km?
represent 29.8% from the total number of landslide sand 49.4% of the total landslide area.
Landslides exceeding 1 km? (3.5%from the total number of landslides) cover 40.5% from the
total landslide area.

The spatial distribution of landslide index (Fig. 3a), defined (Trigila et al, 2010) as the
proportion of landslide surface for a grid of 1x1 km, show the hotspots of landslide density.
The distribution of the landslide index values show an exponential distribution, so we can
state that spatially, the areas prone to landslide processes are occupied in a large proportion.


AP3-5


International Conference


Analysis and Management of Changing Risks for Natural Hazards
18-19 November 2014 | Padua, Italy


Stereographic 1970
projection a
(EPSG:3844) ai


0 80 160 20"
27°30'140 OQ 10 20 30 40
oO
a


47° 30'=
o | Area [km]


b


Landslide


200


Landslide index [%]


PP


‘an a of total


40 60 ~=—,80
Landslide Area (m )


10°


10°


10°


Area [km’] TT
= = a=


Frequency Density f(km~)


10!


10
Frequency Density f(m7)


10°
ie? ao* 16° ao* 40"


10" 40
Landslide Area (km?) T 1 2 3 41 42


Figure 3. Landslide area patterns in Moldavian Plateau: a — landslide index for 1x1 km (1 km’) grid; b
— histogram of landslide area in logarithmic scale; c — histogram of landslide index for 1x1 km grid; d
— frequency distribution in log-log scales; e - frequency distribution in linear scales; f — boxplots for
landslide area, of all landslides (T), and of landslide types (1, 2, 3, 41, 42).


AP3 - 6


International Conference


Analysis and Management of Changing Risks for Natural Hazards
18-19 November 2014 | Padua, Italy


Rotational slides comprise 2.5% of the inventaried landslides, translational 12.2%, river
induced 0.2%, rotational-translational complexes 62.8% and _ rotational-translational-flow
complexes 22.3%.

Considering the physiographic regions of the Moldavian Plateau, the proportion of landslide
area from the total landslide area are: Central Moldavian Plateau (30.4%), Jijia Plateau
(26.5%), Tutovei Hills (20.3%), Siret Hills (7%), Suceava Plateau (6.2%), Falciu Hills (4.1%),
Covurlui Plateau (2.3%), Husi-Sarata-Elan-Horincea Hills (2.3%). Regarding the density
(proportion of landslide area from total area) the rank changes: Central Moldavian Plateau
(38.2%), Sucevei Plateau (33.6%), Tutovei Hills (29.1%), Husi-Sarata-Elan-Horincea Hills
(26.5%), Falciu Hills (21.8%), Jijia Plateau (19.3%), Siret Hills (17.9%), Covurlui Plateau
(4.5%).


Table 1. The absolute frequency of landslide types from the physiographic units of Moldavian Plateau.


physiographic units 1 2* 3* 41* 42* Total**
Covurlui Hills 5 733 139 118 3 998
Falciu Hills 12 580 45 114 106 857
Tutova Hills 129 4179 288 1251 72 5919
Central Moldavian Plateau 124 2925 107 92 811 4059
Jijia Hills 92 4271 90 1427 660 6540
Siret Hills 61 950 64 479 23 1577
Suceava Plateau 89 1874 301 464 31 2759
Husi-Elan-Sarata-Horincea Hills 4 353 13 104 1 475
Total** 516 15865 1047 4049 1707 23184
physiographic units 1* 2* 3* 41* 42* Total**


* the columns 3 to 7 represent the landslide type code according to the text

** The inconsistency of the row and column total surface and proportion is due to the multiple
superposing of landslide polygons along regions borders, and the inclusion of some landslides to Siret,
Prut and Barlad rivers valleys (Figure 1)


CONCLUSIONS
Using a historic landslide inventory for the Moldavian Plateau, this article has analysed the
typology, causal factors and patterns of landslides. While the landslide inventory is not
complete, as the frequency density of area shows, the distribution shape is similar to the
theoretical one and we conclude that the acquired landslide information can be used to
assess the spatial patterns. Further expansion of the landslide inventory requires the
estimation of landslide age and the separation of event based landslides.
The present landslide inventory, prove that Moldavian Plateau is an important hotspot for
these gravitational processes (18.3% from the total area in 24 263 polygons), despite a
relative lifelessness recorded of the mass movement processes for the last decades.
In the Moldavian Plateau the following types of landslides have been identified: rotational,
translational, lateral spread, flow and complex landslides. The analysis allows us to consider
that geological conditions (monoclinic structure and _ clayey-silty lithology) and
morphostructure have a strong control in the spatial distribution of landslides. Landslides
occur on all lithological classes, mostly on clays, sands, limestones and sandstones. The
landslides occupy the entire range of slope positions (upper, middle, lower), with differences
depending on the type. Spatial and and geomorphometric analysis of the landslide polygons
variables: altitude, length, width, height, slope angle, aspect, have proved this link between
mass movements processes and the geologic, morphostructural and topographic setting.
This general pattern is completed by a large range of local preparatory factors.
Further extensions of the inventory by including age, event based and multi-temporal
attributes and locations will complete the general image of landslide types and patterns in
>>>