Mapping and modelling mass movements and gullies in mountainous areas using remote sensing and GIS techniques

JAG Volume 3 - Issue 1 - 2001 l

Mapping and modelling mass movements and gullies in mountainous areas using remote sensing and GIS techniques J Alfred Dhruba

Zinckl, Jaime L6pez2, Graciela I Metternichtj, P Shresthal and Lorenzo Wzquez-Selem4

1 ITC, Division of Soil Sciences, PO Box 6, 7500 AA [email protected]) 2 Apartado

4 UNAM, lnstituto

de Geografla,

modelling,

gullies,

CP 04510,

mapping,

characteristics. In both cases, the soil material frequently conditions the initiation and development of the erosion processes and, at the same time, is affected by them. There are also dynamic relationships between gullies and mass movements themselves, since small landslides or earth slumps often convert into gully heads, while gullies frequently expand laterally through mass wasting. Mass movements and gullies are processes of multiple origin/causes and the initiation mechanisms might take place at the terrain surface or beneath. They are chaotic phenomena, triggered by sudden alteration of the environmental equilibrium and generating catastrophic damages Common factors, such as the complexity of the processes and interactions, the catastrophic character of the events, and the difficulty in predicting their spatial and temporal occurrence, contribute to making deterministic modelling cumbersome in both cases.

moni-

hazard assessment, remote sensing,

Natural as well as human-induced severe environmental

and monitoring.

areas is hampered

factors,

which

areas, among

other

and satellite

images

combinations

thereof,

and delineate remote

mass movements

cause

geometric

constraints. (visible,

But their effective

distortions

Nevertheless,

infrared

and gully

types.

movements

and

gullies

GIS modelling

information

approaches

to exploratory

of such chaotic highlight

ping, monitoring,

exploring

or

of mass

in combination The shortcomings

phenomena

relevance

and predictive

describes practical applications hazards of mass movements

the

bands),

used to discriminate

sensing data, is rapidly developing. modelling

and shadow

aerial photographs

and microwave

have been successfully

landslide

use

by cloud effects and relief-con-

and gullies, using ancillary

of deterministic

and gullies are

hazards. Remote sensing data offer promising

possibilities for identification in mountainous

with

e-mail:

D.F. Mexico

ABSTRACT

movements

phone: +31-53-4874-322;

GPO Box U1987, Perth, WA 6845, Australia Ciudad Universitaria,

mass movements,

KEYWORDS:

trolled

(fax: +31-53-4874-379;

Aereo 5057, Santa Marta, Colombia

3 Curtin University of Technology,

toring, GIS

Enschede, the Netherlands

as mass

of

GIS-assisted

modelling.

This paper

However, the development of modern earth observation techniques, in particular the availability of multitemporal remote sensing data, improves the mapping and monitoring possibilities. Similarly, GIS techniques facilitate the integration of multiple data layers and spatial simulation to explore cause-effect relationships. Such issues have been addressed by a number of authors from different perspectives, focussing on mass movements and/or gullies specifically, or environmental hazards in general. In this context, research with significant input of remote sensing and/or GIS has focused on a variety of aspects, including mapping and monitoring approaches [Kienholz et al, 1983; Pike, 1988; McKean et al, 1991; PalacioPrieto & Lopez-Blanco, 1994; Verstappen, 1995; Chowdhury & Flentje, 1996; lrigaray et al, 1996; Liener et a/, 1996; Rosenbaum & Popescu, 1996; Duan & Grant, 20001, the scale at which hazard maps are prepared for regional, local and site planning purposes [Rengers et a/, 1992; van Westen, 1993; Leroi, 19961, the synergy obtained from merging different kinds of remote sensing

of remote sensing and GIS for map-

cause-effect

relationships

and assessing

and gullies in hilly and mountainous

areas.

INTRODUCTION

Mass movements and gullies are severe environmental hazards in mountainous areas. Both erosion processes, especially mass movements, cause extensive material and human losses, which are often blamed in official statistics on primary causes such as earthquakes or hurricanes. Mass movements probably constitute the single most widespread hazard on the earth’s surface. For instance in New Zealand, one of the few countries that has a countrywide landslide map, 36% of the territory shows formerly or presently active mass movements [Eyles, 19831. Although the basic processes are fundamentally different, mass movements and gullies share some common

43

Mapping

data

and modelling

[Koopmans

Singhroy, al,

19981,

1998;

& Forero,

1995;

of

et a/, &

[Bocco

1996;

et a/, 1996;

Leroi

Chung,

1999;

1995;

Singhroy

sensing

1998;

modelling Wilson,

Zinck,

Shrestha

1998;

GIS

Metternicht

& pre-

1996;

1999;

1999;

pheric

conditions

RELIEF-INDUCED

remote

sensing

Further,

some

for

mapping

it describes

approaches

issues related

and

using

examples

oped

at ITC by advanced

monitoring

illustrates

to modelling

ards,

and

to the

Strong straints

to the efficient

tainous

areas. Topography

use of

drawn

from

research

to change

haz-

work

which

are

areas, among

GIS-assisted

gully and mass movement

relief variations

devel-

of the

ridges,

formation,

[Buchroithner,

students.

of remote the

including

movements

and gullies,

tral and temporal resolution bands

portions related

processes. incoming

to

This solar

[Lunetta,

at which

mass

energy,

gully

Changes

fea-

create

of the and

propagated

elec-

analysis

and mapping.

which

eate

smallest

the

selected

of the sensor determines

may be useful makes

The concept

ground

ture

area, is an important spatial

nises that

data

remotely

map unit,

being mapped.

sensed

characteristic

remote

GIS and

sensing,

niques,

which

allow

geneity

of the

should

of the

digital

and Shroder

application

of

temporal

ing cycle

of the

mass movement ral resolution

resolution

is determined Ideally,

or gully

mapping

than

elevation

and multi-direc-

interferometry

for relief,

cor-

and GPS,

scale, shadow,

weather

1). parameters, with

such as slope gradi-

respect to sun elevation,

which

cause the reflectance

of the training

and induces fully

frequent

slopes.

multispectral

recognition,

slope gradient

In a case study Khola

a mini-

intensity

or feature

of land

watershed,

Nepal,

normalisation

assessment.

biased sampling

tech-

normalisation Figure

by Duan &

is

caus-

to approach

thus

normality.

use classification

in the

Likhu

Shrestha

[2001]

used

& Zinck data

to remove

for land degradation

solved the problem

of the training in the

latter

topography,

and this requires

constraints

This approach

The

and aspect variations

of multispectral

the topography-induced

trend

the

algorithms,

of rugged

differences,

fully

assumes that the

distributed.

case in areas

illumination

either

Furthermore,

classification

be normally

not the

in the fea-

a bias towards

shaded pattern

samples

feature

caused by

pixels. The effect space

plots

of data

is shown

in

1.

& Bishop [19981.

sensor.

higher

digital

belts.

haze and atmospheric

data,

variations

samples

LANDSLIDE Finally,

fea-

elevation

including

some kind of data transformation

local hetero-

such as discussed

or

ing large

estab-

analysis

to consider

data,

(Table

space plots

illuminated

with a

based on available

terrain

to correct

illumination

generally

[ 19991 recogsupport

process

sensing

based on statistical

delinover

by the current

approaches

researchers

landscape,

when

Lunetta

data

His view is supported explicit

interest

consideration

requirements.

of using spatially

[2000]

of

and

elongated

multi-seasonal

in topographic

training

or

of a minimum

to consistently

features

and mass

by such geometric

bioclimatic

solutions

duces an elongation

the scale

for mass movement

it possible

of gullies

tones

values from the same surface cover type to vary. This pro-

such as in the case of radar sensors

resolution the data

humidity, grey

to the slopes over long dis-

ent, aspect and orientation

formation

characteristics

or artificially

and

parallel

radar and stereo

is available

temperature,

affected

models,

and snow effects

of

cover

conditions

to the summit

to pro-

surface

understanding

ecosystem

radiation

spectrum or

a good

ability

remote

differ-

and shadow

are usually

of technical

rections,

of spectral

amongst

movements

requires

map unit,

Grant

tional

data

19991.

The spatial

mum

spec-

Spectral

but also to the

separability

between

tromagnetic

lishing

by the spatial,

of the electromagnetic

spectral

interactions

gully

A variety

and illumination

height

as they

develop

mass

of the data.

by the sensor,

enough

tures

is controlled

resolutions

on

snow

and may cross several

stud-

monitoring

does not refer only to the number

offered

specific vide

and

distor-

of the valleys

effects

con-

in moun-

produce

in anomalous

19951. The delineation

of

sensing to environmental

mapping

and climatic sensing

They also cause climatic

with

restrictions,

tances

REQUIREMENTS

ies,

of atmosevents.

relief displacements

is particularly

tures, which

MAPPING The application

by

resolu-

USE OF REMOTE

and elevation

the bottom

cloud

climatic RESOLUTION

ON THE

use of remote

reflected

others.

from

movements

DATA

conducted

debris flow

cause geometric

ences, scale differences,

purposes.

several

CONSTRAINTS

&

tions, addresses

study

SENSING DATA

ZOOO]. Our paper

3 - Issue 1 - 2001

early warning

likely to initiate

et a/,

Fabbri

Kniveton

a recent

tion data can be used to provide

et al,

Fujita

For instance,

Volume

l

et al [ZOO01 shows that high temporal

Kniveton

et al,

and hazard

1997,

& Zinck,

phenomena.

et

and

Garcia-Melendez

Fermont,

et a/, 1990;

JAG

Buchroithner,

remote

1992;

19981, and simulation

diction

and gullies

& Zinck,

integration

[Leroi

Metternicht

Zinck,

1993;

Metternicht

the

techniques

mass movements

the

should

the changes

A variety

by the revisit-

data

acquired

ping,

for

have a tempo-

evidenced

by the

44

MAPPING of approaches

including

factor

overlay,

process

models

has been used for landslide

on-ground statistical [Duan

monitoring, models,

& Grant,

remote and

map-

sensing,

geotechnical

20001. Singhroy

[19951,

Mapping

and modelling

mass movements

TABLE 1: Problem-and-solution

JAG

and gullies

matrix regarding

relief-induced

factors influencing

l

remote sensing data use

Volume

3 - Issue 1 - 2001

in high-mountain regions

[Buchroithner, 19951. REMOTE

SENSING

CARTOGRAPHY

AND

HIGH

MOUNTAINS

PROBLEM

SOLUTION

REMARKS

Relief

High acquisition altitude Small FOV Stereo data Digital elevation models Shape from shading lnterferometry

Displacement, Reduction Displacement, Reduction Also multi-sensoral Also from remote sensing. Stereo data SAR application Shadow problem. Remains most accurate relief information

GPS Scale

effects

Shadow

Weather,

Stereo data Digital elevation GPS

Intrinsic scale changes Permits mono-plotting,

models

Also multi-sensoral Illumination differences Base for illumination models CPU-intensive integration Simple. Not for completely dark slopes Shadow minimisation

Multi-seasonal data Multi-directional radar Digital elevation models Digital illumination models Grey value ratioing Solar noon satellites Haze, Clouds

Snow cover

geocoding

Multi-seasonal data (high temporal Haze correction Atmospheric correction Radar

resolution)

Also multi-sensoral Simple approach Problem of data availability Cloud penetration

Multi-seasonal data (high temporal Passive microwaves Radar

resolution)

Also multi-sensoral Snow type mapping Penetration to ground Snow type mapping

among others, has shown that mass movement features in a mountainous environment can be identified from remote sensing data, including the discrimination of different types of landslide (eg, rock slumps, block slides, debris lobes and slide scars) and the detection of conditioning factors such as faults and rupture lines. The use of airborne SAR data made it possible to recognise geomorphic patterns, while the combination of SAR and TM data added additional information on the vegetation cover. The comparative advantages of different kinds of remote sensing data and their integration to extract selective information on landslide characteristics (eg, distribution and classification) and factors (eg, slope, lithology, geostructure, neotectonics and land use/land cover) are also discussed by Singhroy [1995]. Overall, airphotos are judged to be the best data source for the identification and mapping of a large number of landslide-related features, while integrated satellite remote sensing data, combining optical and microwave ranges, are superior to the use of individual bands (Table 2). Using study examples from Canada, Singhroy et a/ [I9981 assessed the usefulness of integrated SAR/TM images and SAR interferometric techniques for landslide inventory and characterisation. They reached the following conclusions applicable to mountainous terrains: (1) RADARSAT SAR incidence angles varying from 40° to 59” provide the most suitable imagery to map landslides; (2) the interferometric SAR technique allows easy recognition of landslide

(b)

Feature space plots before (a) and after (b) data normalisation (units in reflectance values in eight-bit format), Likhu Khola watershed, Nepal [Shrestha & Zinck, 2001]. FIGURE 1:

45

Mapping

and modelling

mass movements

and gullies

JAG

TABLE 2: Remote sensing guideline for geohazard LANDSLIDES

A= Average;

features

on steep valley slopes;

airborne

SAR and TM images

retrogressive

AlRBORNE

SAR

SAR

P

P

E A

E E

E A

P P A

P P A

A

E

A

E

E

E

A

A

E A E

and (3) a combination is appropriate

of

the gullies:

to monitor

low

slope failures.

land

difficult

to

are elongated,

identify

gullies

which

are easier to map from

case study

develop

developed

into

eastern

Bolivian

gullied

badlands

data

small

ramified

remote

[Metternicht,

that are

scales. badland

sensing

Andes, from

the other

was assessed using Landsat

19981. Six information to terrain

features

1996;

possibility

visible

of

larities

land

and

study, near

effect

gullies

(2) faleroded

eroded

(mainly

gully

(ephemeral

caused

of the gullies,

reflectance

Zinck,

areas

erosion),

riverbeds

and

which lower

to

and surface

irregu-

traps the incoming

because

of

the

After

merging

Landsat

than

surface TM

that

and

a

light and infrared,

of the other

roughness

by the gullied

in the

be attributed

(Figure 2). In the middle

remained

backscattered

low reflectance

This can

by the depth

tures energy

with

showed

infrared.

reduces the reflectance

referring

intermingled

(4) moderately

(5) badlands

miscellaneous

shrubs),

(3) slightly

pavements).

shadow

TM and JERS-1

considered,

spatially

(mainly

dry season),

sheet erosion),

In the above

In a

kinds of sur&

vegetation

during

More

basin of

Metternicht

classes were

surface components

(TM)

E

rill erosion),

(6)

stone

areas,

data.

in the intra-mountainous

Cochabamba,

SAR

and

large

discriminating face features

narrow

at medium

often,

SARIVIR

(1) natural (bare

areas (mainly and

gullies

AIRPHOTOS

P = Poor

MAPPING

Individual

/NTEGRAT/ON

V/R

(mainly GULLY

3 - Issue 1 - 2001

assessment [Singhroy, 19951

SPACEBORNE

Distribution Classification Factors controlling slope stability Geomorphology of slopes Lithology Structure Neotectonics Landusellandcoverlinfrastructure E= Excellent;

Volume

l

fea-

component.

JERS-1 SAR data, areas remained

the rela-

50

n Badlands n Moderately

45

eroded areas

Slightly eroded areas 40

0

Miscellaneous

0

Fallow land

land

n Vegetation

35

1

2

3

4 LANDSAT

TM BANDS

5

6

(1 to 7)

FIGURE 2: Digital numbers of selected surface components from Landsat TM bands, Sacaba valley, Bolivia [Metternicht, 19961.The X-axis represents the visible (1, 2, 3). near infrared (4), middle infrared (5, 7) and thermal (6) bands of the Landsat TM sensor.

46

Mapping

and modelling

mass movements

JAG

and gullies

l

Volume

3 - Issue 1 - 2001

intricate distribution of branching gullies generates impure pixels, whose spectral reflectances are mixtures of the reflectances of the individual components. Thus the issue is not one of spectral confusion, but one of spectral mixing of the surface components within a single pixel. To solve this problem, a linear mixture model including five pure end-members (n-l), one of them being pure gullies, was applied to a 6-band Landsat TM data set (the thermal range being excluded). The procedure is described in Metternicht & Fermont [1998]. This made it possible to generate an error map from spectral unmixing. The error between the original map and the root-mean-square image representing the best-fit output proportion map was lower than 1 percent in 36 percent of the classified area, and between 1 and 10 percent in 63 percent of the area. Only 1 percent of the classified area had an error greater than 10 percent. These results were further used to improve the image classification.

tively low, but became more variable so that the separability of the gullies from the other erosion features and surface components slightly improved (Figure 3). The effect of data fusion on spectral class separability was assessed using transformed divergence (TDij), a measure to select the best band combination for pixel labelling. Output values are usually scaled between 0 and 2000. For TDij>l600, good separability between classes i and j can be expected [Richards, 19931. In the study case [Metternicht, 19961, best between-class separability was obtained when combining JERS-1 SAR (L-band) and TM bands 1, 3, 4, 5, 6 and 7, causing all transformed divergence values to cross the threshold figure of 1600. This is particularly the case between badlands and miscellaneous land, as well as between badlands and moderately eroded areas. The improvement of class separability is also reflected in the spatial distribution of the gullies before and after data merging. With Landsat data alone, gullies appear as large undifferentiated areas. Comparatively, Landsat and JERS-1 SAR data together produce a sharper spatial segmentation of the gullies.

The efficiency of the different portions of the spectrum (ie, visible, infrared, thermal and microwave) to properly separate gullies from the other selected surface components was assessed using a procedure based on the percentage of spectral confusion, derived from the transformed divergence analysis, in relation to the total number of training samples [Metternicht, 19961. Applying this criterion to the study case, the different regions of the electromagnetic spectrum performed relatively well when separating gullies from fallow land and from slight-

However, in any case of band combination or data fusion, the class accuracy for gullies remained low (54 percent), when compared to the other surface components included in the study. This is because gullies are heterogeneous areas including variable mixtures of natural vegetation, stone pavements and eroded soils. The

n Badlands 180

w Moderatelyeroded areas H Slightlyeroded areas 1

160

0

Miscellaneousland

c] Fallow land

1

n Vegetation

140 -

120 -

3

4

5

LANDSAT TM SANDS (1 to 7’) and JEW-1

6

7

8

SAR (8) DATA

Digital numbers of selected surface components from Landsat TM and JERS-1SAR bands, Sacaba valley, Bolivia [Metternicht, 19961.The X-axis represents merged Landsat TM and JERS-1SAR data (1 to 7), and JERS-1SAR data alone (8). FIGURE 3:

47

Mapping and modelling mass movements and gullies

ly and moderately tral

eroded

confusions

covered

with

between

spectral

gullies

shows

band

algorithm

ranges

contextual spectral

regardless

knowledge

relationship features,

image

gullies,

their

Mass movements,

4).

librium

sensor,

quake.

between-class about

the

cover,

geological

structure

surface

exceeds

Terlien, which the

19961. aims allow

the

mass movements different

Ideal of the

one

to

properties

thus from

would

erosion

predict

gullies

[van

of the material

geomorphic

response

micro-fabric

be solifluction. solid

consistence

flocculated

the predicted

An

aggregated

would

micro-fabric

45

ministic

mass movement

correlation

between

of the soil cover,

builds

the

consistence

would

deterministic

In contrast,

understanding

semi-

would

INFRARED

fabric

a

of the

relationships

of the soil material

must still be achieved.

pro-

mass

rhexistatic

framework

movements, between

and the geomorphic

As quite

some progress

the

is needed

353025 20 15 10 5 0

Fa: Fallow land

SI: Slightly eroded areas

Ba: Gullied badlands

Rf: Rock fragments

eroded areas

FIGURE4: Spectral confusion among terrain surface features, Sacaba valley, Bolivia [Metternicht,

48

and stone pavements

19963.

micro-

response

40-

MO: Moderately

for better

q THERMAL q MICROWAVE

Ve: Natural vegetation

simple

and moisture

To improve

the basic conceptual

deter-

mass move-

elaboration,

can be used as a basis for designing of

criti-

19961. While

profiles

modelling

of areas

up during

and predict

further

graphi-

identification

1986,

able to explain

mantle

or consistence

models.

sliding.

VISIBLE

graphical

contents

potential

require

soil

the Atterberg

be

and

and solid consistence

models

surfaces.

exists. A simple

allows

of the year [Zinck,

hazards

of the and/or

and a

would

micro-fabric

promote

ment

thus in

a dispersed

plastic consistence,

material, when

series,

its

creates

shear

water-potential

a mass movement

response

is deflocculated,

geomorphic

as time

deter-

through

horizons

content

But

which

The rheological

as

capacity

and the real moisture

cal periods

For instance,

mudflow.

with

where

modelling

liquid state, the expected Similarly,

profiles

soil

hazard

between

of the

behaviour. mainly

functioning

holding

The

or an earth-

properties

properties.

moisture

equi-

geomorphodynamics,

consecutive

a mass movement

preferably

through

of the soil or substratum susceptibility.

limits,

cal comparison

&

be deterministic,

Mechanistic

the

have

Westen

phenomena

processes.

its intrinsic

the micro-fabric

and

perspectives

modelling

at explaining

mechanics

would from

model from

actual

the water

include

to mass wasting,

to

phenom-

change.

rainfall

of the soil material

and hydrological

the

suddenly

and hydrogeological

between

When

and

factors

nature

susceptible

BOUNDARY CONDITIONS to

[Zinck,

are chaotic

topography,

mines its propensity mechanical

MODELLING

made

stability

of a meta-stable

be abnormal

The conditioning

vegetation

planes

been

might

it is the intrinsic

is needed.

the terms

drastically

factor

contrast

Attempts

geomorphic

as well as gullies,

situation

activating

requires

characteristic

positions

for

ena. They occur when

classification

the

conditions

espe-

(Figure

knowledge

the

Volume 3 - Issue 1 - 2001

l

19961.

the vis-

of gullies

improve

vide

in discrimi-

cover

detection

to

and geomorphic

spec-

surfaces

of the satellite

In particular,

between

poorly

and/or

accurate

separability.

and

Similarly,

performed

combinations

applied,

ranges.

and vegetation

that

major

gullies

and stone pavements,

and infrared

ible and microwave nating

between

rock fragments

cially in the visible

This study

areas. In contrast,

occurred

JAG

Mapping

and modelling

mass movements

before deterministic modelling gullies can be fully undertaken,

and gullies

JAG

l

Volume

3 - Issue 1 - 2001

of mass movements and the current shortcomings

can be partially compensated by G&assisted modelling, including exploratory and predictive approaches [Zinck, 1997, 19991. EXPLORATORY

APPROACH

Exploratory models attempt to identify non-explicit cause-effect relationships between environmental hazard types (eg, soil erosion processes) and affected soil types (ie, soil map units) in order to predict, from these relationships, soilscape areas potentially exposed to degradation. Because it relies on relatively simple GIS operations, cartographic modelling is an exploratory mode frequently implemented to this end [Bocco et al, 19901. The overlay of information layers, usually represented by series of thematic single-attribute maps, allows highlighting of areas of coincidence between factors presumably controlling erosion processes and features resulting from these processes.

SLOPE IN % _

KW”NrrAtlE4

___________ ?hOFTOTAL GUUJED *REpI

Observed gullied areas per slope gradient units at Huasca de Ocampo, Mexico [Vazquez-Selem & Zinck, 19941.

FIGURE 5:

60

Erosion features caused by gullies or mass movements are not randomly distributed on the landscape. They develop in response to a combination of controlling factors The simple overlay of a gully distribution map on top of maps representing environmental factors, such as geoforms, slope gradients, soils, lithologic units and land use types, highlights the degree of spatial coincidence between gully occurrence and specific factors. Such cartographic coincidence helps identify the most favourable combination of conditions and reveals underlying causeeffect relationships.

50 iis % 9 3 2 0 rp

40

30

20

10

0 1

Cartographic modelling through thematic map overlay was performed in a volcanic area located about one hundred kilometres northeast of Mexico City, in Huasca de Ocampo County [Vdzquez-Selem & Zinck, 19941. Frequency graphs were established for each forming factor to highlight the conditions most favourable to the occurrence of gullies. For example, nearly half of the total gullied area corresponds to a narrow slope gradient range of 4-7 percent (Figure 5). Thus, gully erosion does not increase proportionally to slope gradient and, in this sense, substantially deviates from the principles governing rill and sheet erosion.

2

3

4

5

6

7

6

i

lb

11

DOMINANT SOIL UNITS m

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)

% OF UNIT AREA

% OF TOTAL GULLIED

AREA

Ustifluvents-Fluvaquents Ustorthents (frequently Lithic) Ustorthents-Ustropepts Ustropepts (shallow: Typic and Lithic) Ustropepts (deep: Typic and Fluventic) Typic or Lithic Ustropepts-Lithic Ustorthents-Typic Haplustalfs Typic Ustropepts-Typic Haplustalfs Typic Haplustalfs Typic Paleustalfs-Typic Haplustalfs Typic Argiustolls-Typic Haplustolls Fluventic Ustropepts-Typic Argiustolls-Typic Haplustalfs

Observed gullied areas per dominant soil units at Huasca de Ocampo, Mexico [Vazquez-Selem & Zinck, 19941.

FIGURE 6:

Similarly, the overlay of the gully map and the soil map shows that most gullies develop on Alfisols (Figure 6). In fact, about 90 percent of the gullied surface within the study area concentrates on deep Paleustalfs, although these soils account for only 39 percent of the total area. Alfisols in this area are composed of two main layers: a colluvial cover material, 40 to 60 cm thick, which lies on a buried Bt horizon belonging to a truncated subsoil. The surface of discontinuity between the two materials is marked by ancient human artefacts (2400 years BP). The

difference of permeability between the well-structured colluvial cover and the highly clay-dispersible subsoil favours horizontal water flow along percolines, from which gullies initiate. In contrast, the areas covered with Mollisols are virtually free of gullies, although these soils occur in conditions of slope and land use similar to those of the Alfisols.

49

Mapping

and modelling

PREDICTIVE

mass movements

JAG

and gullies

tionships from the spatial coincidence between observed erosion features (eg, gullies, landslides) and landscape factors. Such models are not able to take into consideration, for the purpose of prediction, the role played by the activating factors (ie, exceptional rainfalls, earthquakes) in triggering the processes. The modelling rules are mainly based on the conditioning factors of the environment (eg, slope, vegetation cover, rock substratum) and a few soil properties. This type of model is able to (1) reproduce the spatial distribution of existing gullies for validation purposes and (2) predict the potential occurrence of gullies in areas with favourable conditions.

1

3 - Issue 1 - 2001

and 6). Using these rules, six models were established with decreasing boundary conditions (Table 3). The first model, for instance, takes into account only the thematic map units with a high percentage of observed gullied area (more than 10 percent or more than 15 percent according to the environmental factor considered). Such a combination of rules is highly selective, since only a few units satisfy the requirements. As a consequence, the gullied area estimated by model 1 is small. But, at the same time, the model is efficient because a large proportion of the calculated gullied area corresponds to existing gullies. The other models operate with less restrictive rules in decreasing order.

To illustrate the above considerations, a set of nested models was developed to confirm observed gullies and assess the hazard of potential gullies in the same Huasca de Ocampo area of central Mexico, where the exploratory models were developed [Vazquez-Selem & Zinck, 19941. The assessment rules mobilise only the area percentages covered by observed gullies in the units of the thematic maps representing environmental factors. Six environmental factors were selected on the assumption that they contribute, in one way or another, to gully formation: geoforms, lithologic units, slope gradients, slope shapes, dominant soils and land uses. Two criteria were implemented to account for the area percentages: (1) the percentage of area with observed gullies in each thematic map unit, and (2) the percentage that the observed gullied area in each thematic map unit represents in rela-

Model

Volume

tion to the total extent of gullies (562 ha) within the whole study area (8009 ha). The first criterion was considered more diagnostic than the second one, because it is independent of the total gullied area and thus reflects better the intrinsic susceptibility of each thematic map unit to gully formation. To establish class limits, critical threshold values of area percentages were determined by iteration from the graphics showing the frequency of gullies per classes of environmental factors (eg, Figures 5

APPROACH

Predictive models usually implemented in GIS are based on rules and expert knowledge. Such models lack deterministic capability, because they neither simulate nor explain the mechanisms involved in gully formation or mass wasting. They are built on the results of the exploratory analysis, which identifies cause-effect rela-

TABLE 3:

l

The relative efficiency of the various models in corroborating observed gullies is presented in Figure 7. The total gullied surface area of 562 ha is equivalent to 7 percent of the study area. An ideal model would confirm 100 percent of the gullied area with only 7 percent of the study area. Thus, an efficient model is one which approximates this optimum performance. To assess the efficiency of the models, the gullied area calculated by each model was compared to the area percentage of existing gullies properly confirmed. Accordingly, models Zb, 3a, 3b and 4a are the best predictors.

Rule-basedspatial models for gully prediction at Huasca de Ocampo, Mexico [Nzquez-Selem & Zinck, 19941 Geopedologic map (7)

Lithologic

Slope gradient map

Slope shape map

Dominant

map

map

Landuse map

> 15 %

>15% unit area

>lO% unit area

> 15 % un’it area

> 15 % unit area

>lO% unit area >5% unit area

unit area

soil

Model

2

[>lO% unit area and >lO% total gullied area] or >15% unit area

>15% unit area or >30% total gullied area

>5% unit area or >5% total gullied area

>lO% unit area or ~20% total gullied area

~10%

unit area

Model

3

>5% unit area

>5% unit area and >5% total gullied area

>5% unit area or >5% total gullied area

>5% total gullied area

~5% unit area

~-3% unit area

Model

4

>2% unit area

>2% unit area

~2% unit area

~2% unit area

>2% unit area

>2% unit area

Model

5

SO% unit area

>O% unit area

>O% unit area

20%

>O% unit area

>O% unit area

Model

6

no gullies

no gullies

no gullies

no gullies

no gullies

no gullies

unit area

(1) Only geomorphic component of the geopedologic map taken into account Examples: “>I 5% unit area” means that only units in which observed gullres cover >15% of the area of the unit are considered rn the model’s calculations; “>5% total gullied area” means that only unrts that contain ~-5% of the total gullied area within the study area are considered In the model’s calculatrons.

50

Mapping

and modelling

mass movements

and gullies

JAG

l

Volume

3 - Issue 1 - 2001

% of total study area

= (total area calculated by the model/ total study area) x 100 % of total gullied area = (gullied area calculated by the model/ total gullied area) x 100

5a .5b

ideal model Type “a” models: Including all thematic maps except the geomorphic component of the geopedologic map (see Table 3) Type “b” models: based on the geomorphic component of the geopedologic map alone (see Table 3) Intersection model of 3-l 6% slopes and Alfisols Intersection model of 5-l 1% slopes and highly susceptible geoforms

2a

3 20 d la

FIGURE 7: Relative efficiency of rule-based models for gully prediction at Huasca de Ocampo, Mexico [Vazquez-Selem & Zinck, 19941.

10

I 0

10

I

I

20

I

30

I

40

I

50

I

60

I

70

I

80

I 100

90

% OF TOTAL STUDY AREA

to describe ard

cordillera E z

oped

40 30

9 3 2

classes.

of the Colombian

to identify

20

its

0

physico-mechanical

Andes,

dological

map

a model

1

2

3

4

5

6

areas (Figure

GULLY HAZARD UNITS m

% OFUNIT

AREA

homogeneity

% OF TOTAL GULLIEO

AREA

homogeneity

was

of

the

material and

model

distribution zoning,

of

a geope-

to extrapolate established

properties

of the soil units were

slope class-

on the basis of

to extrapolation,

soil

develash cover

used included

used as vector

9). Previous

haz-

central

was devel-

in a volcanic

For hazard

basin the causal

valley,

to the potential

properties,

mass movements.

the whole

0

1) 2) 3) 4) 5) 6)

river

classes of the

observed

and establish

Coello

areas favourable

es, susceptibility

3ully

factors

In the

opment of mass movements [L6pez & Zinck, 19911. Criteria

ifi %

the environmental

severity

over

in sample

the structural and

validated,

the

spatial

by testing

hazard severity classes Very high High Moderate Low Very low No hazard

Observed gullied areas per gully hazard units at Huasca de Ocampo, Mexico [Vazquez-Selem & Zinck, 19941.

FIGURE 8:

In addition also

be

to confirming used

favourable the

are organised

this sequence

hazard

prone

severity.

to future

decreasing Instead predictive

existing

identify

to the potential

models

straints, of

to

that

the models meet

development

also represents initiation.

rates (Figure

of operating

only

models

can directly

order

Models

areas

Since

of con-

a decreasing

1 identifies

can

conditions

of gullies.

per decreasing

Model

gully

hazard

gullies,

areas

scale

strongly

2 to 6 indicate

8).

on the

basis of area criteria,

use concrete

Mass movement hazard zones in the upper Coello river basin, Colombia [L6pez & Zinck, 19911.

FIGURE 9:

parameters

51

Mapping

the

and modelling

normal

using

mass movements

distribution

geostatistics

of the selected

to characterise

JAG

and gullies

parameters

their

spatial

variabili-

and

mass movements

because

they are difficult Remote ping

are severe

of the damages

environmental

they cause and because

sensing data substantially of landslide

tures,

but their

use is limited

which

cause geometric

contribute

to the map-

and gully

erosion

by relief-controlled

distortions

fea-

Koopmans, B.N. & R.G. Forero. 1993. Airborne SAR and Landsat MSS as complementary Information source for geological hazard mapping. Photogrammetry & Remote Sensing 48(6): 28-37

factors,

and atmospheric

con-

straints Best results are obtained ity

through

merging applying remove

the

visible,

The

techniques

complexity

and expert tionships tion

features

of

knowledge, and

potentially

exposed

favourable

formation

multi-source

in single

pixels.

to

presence

of

hampers

to explore

because

they

rela-

Lopez, J. & J.A. Zinck, 1991. GIS-assisted modelling of soil-induced mass movement hazards: a case study of the upper Coello river basin. ITC Journal 1991-4: 202-220.

areas meet

Lunetta, R., 1999. Applications, project formulation, and analytical approach. In: R. Lunetta & C. Elvidge (Eds.), Remote Sensing Change Detection: Environmental Monitoring Methods and Applications. Taylor & Francis, London, pp. 1-19.

conditions.

McKean, J., S. Buechel & L. Gaydos, 1991. Remote sensing and landslide hazard assessment. Photogrammetric Engineerrng & Remote Sensing 57(g): 1 185-l 193.

REFERENCES Bocco, G., J Palacio & C.R. Valenzuela, 1990. lrng using GIS and geomorphic knowledge. 253-261.

E., 0. Rouzeau, J.-Y. Scanvic, C.C. Weber & C.G. Vargas, 1992. Remote sensing and GIS technology in landslide hazard mapping in the Colombran Andes. Episodes 15(l): 32-33.

Liener, S., H. Kienholz, M. Liniger 8 B. Krummenacher, 1996. SLIDISP - A procedure to locate landslide prone areas. In: K. Senneset (Ed.), Landslides. Balkema. Rotterdam, pp. 279-284.

distribu-

and identify

scales: (Ed.),

E., F. Pontarollo, J.D. Gascuel & M.P. Gascuel, 1996. Development of a 3-D model for block trajectories, based on synthetic imagery and stress vs. deformation laws. In: K. Senneset (Ed.), Landslides. Balkema, Rotterdam, pp. 271-278.

Leroi,

and spatial

at different K. Senneset

Leroi,

based on rules

it possible

factors hazards

unmixing

evolution

mass movements to

(eg,

and mass movements.

approaches,

make

forming

spectral

and

data

data) and when

by the

of gullies

Leroi, E., 1996. Landslide hazard - Risk maps objectives, tools and developments. In: Landslides. Balkema, Rotterdam, pp. 35-51.

class separabil-

caused

formation

G&assisted

between

of gullies

of

such as linear

modelling

empirical

improving

and microwave

confusions

surface

deterministic More

integration infrared

spectral

contrasting

when

1983. Maps area).

Knrveton, D.R., P.J. de Graff, K. Granica & R.J. Hardy, 2000. The development of a remote sensing based technique to predict debris flow triggering conditions in the French Alps. International Journal of Remote Sensing 2 l(3): 429-434.

to control.

and monitoring

3 - Issue 1 - 2001

Kienholz, H., H. Hafner, G. Schneider 8 R. Tamrakar, Mountarn hazards mapping in Nepal’s Middle Mountains, of land use and geomorphic damages (Kathmandu-Kakani Mountain Research and Development 3(3): 195-220.

CONCLUSIONS hazards

Volume

Ingaray, C., T. Ferndndez & J. Chacon, 1996. Inventory and analysis of determining factors by a GIS in the northern edge of the Granada Basin (Spain). In: K. Senneset (Ed.), Landslides. Balkema, Rotterdam, pp. 1915-I 921.

and

ty.

Gullies

l

Gully erosron modelITC Journal 1990-3:

Buchroithner, M.F., 1995. Problems of mountain using spaceborne remote sensrng techniques. Research 15(11): 57-66.

Metternicht, G.I., 1996. Detecting and monitoring land degradation features and processes in the Cochabamba valleys, Bolivia. A synergistic approach. ITC Publication 36. ITC, Enschede, 390 pp.

hazard mapping Advanced Space

Metternicht, G.I. & A. Fermont, features by linear mrxture Environment 64: 254-265.

Chowdhury, R.N. & P.N. Flentje, 1996. Geological and land instability mapping using a GIS package as a burlding block for the development of a risk assessment procedure. In: K. Senneset (Ed.), Landslides. Balkema, Rotterdam, pp. 177-182.

Metternicht, G.I. & J.A. Zinck, 1998. Evaluating the information content of JERS-1 SAR and Landsat TM data for discrimination of soil erosion features. Photogrammetry & Remote Sensing 53: 143-l 53.

Duan, J. & G. Grant, 2000. Shallow landslide delineation for steep forest watersheds based on topographic attributes and probability analysis. In: J. Wilson & J. Gallant (Eds.), Terrain Analysis: Principles and Applicatrons. John Wiley & Sons, New York, NY, pp. 31 l-329.

Palacio-Prieto, J.L. & J. Lopez-Blanco, tive remote sensing tool for Journal 1994-3: 233-237.

1994. Videography: an alternamonitoring gully erosion. ITC

Pike, R.J., 1988. The geometric signature: quantifying landslrde-terrain types from digital elevation models. Mathematical Geology 20(5): 491-511.

Eyles, G.O., 1983. The distribution and severity of present soil erosion in New Zealand. New Zealand Geographer 39(l): 12-28. Fabbri, A.G. & Ch.-J.F. Chung, 1999. Favourability functions for spatial prediction of resources, hazards and environmental impacts. Proceedings, International Conference on Geoinformatics for Natural Resource Assessment, Monitoring and Management, 911 March 1999, IIRS, Dehradun. Indian Institute of Remote Sensing, National Remote Sensing Agency, Dehradun, pp. 359. 367.

Rengers, N., R. Soeters & C.J. van Westen, 1992. and GIS applied to mountarn hazard mapping. 36-45.

Remote sensing Episodes 15(l):

Richards, J.A., 1993. Remote Sensing Digital Image introduction. Springer Verlag, Berlin, 340 pp.

Analysis:

an

Rosenbaum, M.S. & M.E. Popescu, 1996. Using a geographical information system to record and assess landslide-related risks in Romania. In: K. Senneset (Ed.), Landslides. Balkema, Rotterdam, pp. 363-370.

FuJita, K., S. Obayashi & K. Kasa, 1996. Some aspects of landslide prediction using satellite remote sensing data. In: K. Senneset (Ed.), Landslides. Balkema, Rotterdam, pp. 1545-l 550. Garcia-Melendez, E., I. Molrna, M. Ferre-Julia & J. Aguirre, Multisensor data integratron and GIS analysis for natural mapping in a semiarid area (Southeast Spain). Advanced Research 2 l(3): 493-499.

1998. Estimating erosion surface modelling. Remote Sensing of

1998. hazard Space

Shrestha, D.P. & J.A. Zinck, 1999. Land degradation assessment using geographic informatron system: a case study in the Middle Mountain region of the Nepalese Himalaya. Proceedings, International Conference on Geoinformatics for Natural

52

Mapping

and modelling

mass movements

JAG

and gullies

l

Volume

3 - issue 1 - 2001

RESUME

Resource Assessment, Monitoring and Management, 9-l 1 March 1999, IIRS, Dehradun. Indian Institute of Remote Sensing, National Remote Sensing Agency, Dehradun, pp. 391409.

van Westen, C.J. & M.T.J. Terlien, 1996. An approach towards deterministic landslide hazard analysis in GIS. A case study from Manizales (Colombia). Earth Surface Processes and Landforms 2 1: 853-868.

Des mouvements en masse et des ravinements naturels ou provoques par les hommes sont des risques environnementaux graves. Des don&es de teledetection offrent des possibilites prometteuses pour I’identification et le suivi de I’evolution. Mais leur utilisation effective dans des zones montagneuses est g@nee par des nuages et des facteurs dependants du relief, qui entrainent des distorsions geometriques et des zones d’ombres, entre autres contraintes. Neanmoins, des photos aeriennes et des images satellite (visible, infrarouge et des bandes en hyperfrequence), ou des combinaisons resultantes, ont et@ utilisees avec succes pour differencier et delinier des types de glissements de terrain et des ravinements. Une modelisation SIG de mouvements en masse et de ravinements, utilisant une information auxiliaire en combinaison avec des don&es teledetection se developpe rapidement. Les defauts de modelisation deterministe de tels phenomenes chaotiques tels que des mouvements en masse et ravinements met en valeur I’importance d’approches assistees SIG pour une modelisation exploratoire et predictive. Cet article decrit des applications pratiques de teledetection et SIG pour cartographier, suivre I’evolution, explorer les relations de cause a effet et @valuer les risques de mouvements en masse et ravinements dans des zones accidentees et montagneuses.

Vazquez-Selem, L. & J.A. Zinck, 1994. Modelling gully distribution on volcanic terrains in the Huasca area, central Mexico. ITC Journal 1994-3: 238-251.

RESUMEN

Shrestha, D.P. & J.A. Zinck, 2001. Land use classification in a mountainous area of Nepal: integration of image processing, digital elevation data and field knowledge. JAG 2001-l : 78-85. Shroder Jr., J.F. & M.P. Bishop, 1998. Mass movement in the Himalaya: new insights and research directions. Geomorphology 26: 13-35. Singhroy, V., 1995. SAR integrated techniques for geohazard ment. Advanced Space Research lS(11): 67-78.

assess-

Singhroy, V., K.E. Mattar & A.L. Gray, 1998. Landslide characterisation in Canada using interferometric SAR and combined SAR and TM images. Advanced Space Research 21(3): 465476. van Westen, C.J., 1993. Application of geographic information systems to landslide hazard zonation. ITC Publication 15. ITC, Enschede, 245~~.

Los movimientos en masa y las cdrcavas, que Sean naturales 0 inducidos por actividades humanas, representan severos riesgos ambientales. Los datos obtenidos por teledeteccion ofrecen posibilidades prometedoras para la identification y el seguimiento.

Verstappen, H.Th., 1995. Aerospace technology and natural disaster reduction. Advanced Space Research 15(11): 3-l 5. Wilson, J.M., 1996. ROXIM: A computer program to simulate rockfall. In: K. Senneset (Ed.), Landslides. Balkema, Rotterdam, pp. 1643-1648.

Pero, su uso efectivo en areas montariosas queda limitado por efectos de nubes y factores controlados por el relieve, 10s cuales causan distorsiones geometrjcas y areas de sombra, entre otras limitaciones. Sin embargo, fotografias aereas e imagenes satelitdrias (en las bandas del visible, del infrarrojo y de las microondas), o combinaciones de estas, han sido utilizadas exitosamente para diferenciar y delinear tipos de deslitamiento y carcava. La modelizaci6n de movimientos en masa y carcavas en SIG, usando information contextual en combination con datos de teledeteccion, se esta desarrollando rapidamente. Las limitaciones de la modelizacion deterministica de fenomenos caoticos coma son movimientos en masa y carcavas subrayan la importancia de 10s enfoques basados en SIG para la modelizacion exploratoria y predictiva. Este articulo describe aplicaciones practicas de teledeteccion y SIG para mapear, monitorear, explorar relaciones de causa a efecto, y evaluar riesgos de movimientos en masa y c6rcavas en areas de colinas y montaiias.

Zinck, J.A., 1986. Propiedades y estabilidad mecdnicas de 10s suelos en ambiente de selva nublada. In: 0. Huber (Ed.), La selva nublada de Ranch0 Grande, Parque National Henri Pittier. Fondo Edit. Acta Cient. Venez. y Seguros Anauco C.A., Caracas, Venezuela, pp. 91-105. Zinck, J.A., 1996. La susceptibilidad de 10s suelos a la erosibn por movimientos en masa. Con referencia especial a las montarias tropicales htimedas. In: J. Aguilar R., A. Martinez R. y A. Rota R. (Eds.), Evaluation y Manejo de Suelos. Junta de Andalucia-SECSUniversidad de Granada, Espatia, pp. 25-48. Zinck, J.A., 1997. Riesgos ambientales y suelos. Enfoques para la modelizaci6n de la erosion por carcavas y movimientos en masa. Revista de la Sociedad Espariola de la Ciencia del Suelo, Edition Especial 50 Aniversario, Granada, Espaiia, pp. 283-297. Zinck, J.A.. 1999. GIS-assisted approaches to modelling soil-induced gully and mass movement hazards. Proceedings, International Geoinformatics for Natural Resource Conference on Assessment, Monitoring and Management, 9-11 March 1999, IIRS, Dehradun. Indian Institute of Remote Sensing, National Remote Sensing Agency, Dehradun, pp. 368-376.

53