You are on page 1of 16

Engineering Geology, 7(1973) 99-114

Elsevier Scientific Publishing Company, Amsterdam - Printed in The Netherlands


A NUMERI CAL CLASSI FI CATI ON OF SELECTED LANDSLI DES OF
THE DEBRI S SLI DE- AVALANCHE- FLOW TYPE
R. J. BLONG
School of Earth Sciences, Macquarie University, North Ryde, N.S. 14I. [Australia)
(Accepted for publication April 5, 1973)
ABSTRACT
Blong, R. J., 1973. A numerical classification of selected landslides of the d~bris slide-avalanche flow
type. Eng. Geol., 7: 99-114.
Numerous classifications of landslides have been proposed based on a variety of classificatory
criteria. Several writers have mentioned the difficulties of distinguishing accurately between landslides
classed by Sharpe (1938) and Varnes (1958) as d~bris slides, d~bris avalanches, and d~bris flows. A
sample of 92 such landslides from the greywacke hill country of the North Island of New Zealand is
classified on the basis of as many as 19 numerical and 43 disordered multistate attributes. The results
of the agglomerative polythetic classifications do not help to distinguish these landslide phenomena
clearly. Until some distinctive criteria characterizing landslides of this type are identified the use of
unsatisfactory simple classifications is recommended.
INTRODUCTION
Landslides have been classified in a number of ways by workers in a variety of discip-
lines. In the present study, 92 landslides, mainly of the d6bris slide-d6bris aval anche-
d6bris flow type, are grouped according to a number of simple classifications. Subse-
quently, a large number of quantitative and qualitative morphometric attributes are
defined and the techniques of numerical t axonomy are employed in an at t empt to
identify the most suitable morphological parameters for classification purposes.
All 92 landslides examined were located in the upper Mangawhara catchment, an area
of deeply dissected, red-weathered, and ash-mantled greywacke hill count ry in the North
Island of New Zealand. Nearly all of the landslides occurred during two high-intensity
rainstorms on February 28th 1966, and February 2nd 1967.
LANDSLIDE CLASSIFICATION
A survey of the literature reveals that a variety of criteria has been used to classify
landslides.
100 R.J. BLONG
Bases of classification include: the lithology of the shear plane (Ladd, 1935; Zairuba
and Mencl, 1969), the mechanics of slope failure (Terzaghi, 1950; Yatsu, 1967), the
form in particular the geomet ry of the failure relative to the thickness of the moving
mass - (Ward, 1945; Skempt on, 1953), and the t ype of material involved, together with
the shape of the surface of rupture, and the arrangement of the d6bris (Sharpe, 1938:
Varnes, 1958).
Classification of landslides according to the parent material in which failure occurs
may not be very successful. Ladd' s (1935) classification, using this criterion, indicates
that similar types of failure occur in a wide variety of lithologies. Similarly, without the
prolonged investigations usually only conduct ed by soil engineering laboratories, classific-
ations of landslides according to causes of failure are not successful. Without such investi-
gations, the cause can frequently only be established by elimination of some possibilities.
Furt hermore, many landslides are at t ri but ed to several causes, although perhaps only to
one trigger mechanism (Sowers and Sowers, 1961, p.228).
Skempt on and Hutchinson (1969) and Hutchinson (1968) suggest t hat mass move-
ments exhibit such great variety t hat rigorous classification is hardly possible. The general
classifications, such as those of Sharpe and Varnes, have received the most at t ent i on in
geomorphic literature. However, many fieldworkers have experienced difficulty in fitting
particular movement s i nt o the individual categories of these classifications. Ward (1945,
p.172), Irwin-Hunt (1960, p.36), Bailey and Rice (1969, p.172), and Rice et al. (1969,
p.647) report lack of agreement on terminology and problems with classification.
Cumberland (1944, p.83) found difficulty in relating the many combinations of slippage
and flowage forms found in inland Taranaki, New Zealand, to the categories elaborated
by Sharpe. Yatsu (1966) is highly critical of Sharpe' s classification, considering Varnes' s
grouping to be superior.
TABLE I
The 92 landslides grouped according to
Varnes (1958)
Classification Number
Slump 1
D6bris slide 11
D~bris avalanche 2
Complex 78
Using Varnes' s classification, 78 of the 92 landslides investigated in the present study
are grouped as compl ex slope failures (Table I). Most of these 78 landslides involve com-
binations of t wo of the three t ypes of slope failure defined by Varnes as d~bris slide,
d~bris flow, d~bris avalanche. However, five of the 78 landslides do not clearly fit into the
compl ex category, as t hey do not involve combinations of materials or combi nat i ons of
NUMERICAL CLASSIFICATION OF SELECTED LANDSLIDES 101
TABLE II
The 92 landslides grouped according to a modified
Varnes's classification
Classification Number
Slump 1
D~bris slide 11
D~bris avalanche 2
D~bris slide-d~bris avalanche 24
D~bris avalanche-d~bris flow 49
Underthrust slide 3
Incipient underthrust slide 2
types of movement. Three of these five landslides have been called underthrust slides
(Blong, 1971). Two other movements exhibit only incipient underthrusting. The name
"underthrust slide" is given to a type of mass movement characterized by a series of
underthrust shear surfaces in coeval buried soils, subparallel transverse ridges, a marked
bulging toe, and overlapping erosional and depositional zones (Blong, 1971).
In Table II, the transitional types of failure have been removed from the complex
landslide category. This modified Varnes's classification was used in the field identific-
ation of landslide type. The transitional types d~bris slide-d6bris avalanche and d~bris
avalanche-d6bris flow were recognized in the field mainly on the basis of the degree of
cohesion retained in the moving mass. Where rafted blocks occur on the shear plane, but
where much of the moving mass has lost cohesion, a d~bris slide-avalanche is defined.
Where the landslide has characteristics of a d~bris avalanche, but where incoherent flow
deposits reach the base of the hillslope or beyond, the transitional type d~bris avalanche-
d~bris flow was recognized. Categorization of individual occurrences proved somewhat
subjective; continual reference was made to type examples.
With Varnes's original classification, the major flaw is that most landslides fall into the
complex category, a grouping that provides little information about the characteristics of
an individual slope failure. With the modified classification, the problem becomes one of
accurately placing a landslide in the correct category when so many categories are closely
related and transitional with one another. Both classificationsseem to be unsatisfactory.
Landslides can also be grouped according to the dominant mode of movement. Sowers
and Sowers (1961) recognized three types: rotational slides, linear shear slides, and flow
slides. For the sample of 92 landslides, 55 were classed as predominantly flowage pheno-
mena, 15 as translational movements, 2 exhibited rotation, 1 was classed as flow-fall, and
the remaining 20 slope failures were considered to be transitional between flowage and
translational phenomena.
102 R~ J. BLONG
In a further attempt at grouping the landslides according to their field characteristics,
50 of the 92 slope failures were considered to be simple, in that only one failure had
occurred and only one type of movement was involved (although that one type could be
flow-translation). Compound landslides exhibit more than one slope failure, tile hater
failure probably resulting from loss of support when tire first failure occurred, and closely
following the first in time. Compound failures also exhibit only one movement mode (i.e..
rotation, translation, flow, flow-translation). Twenty-four landslides were distinguished in
the sample of 92. Complex landslides exhibit more than one slope failure and more than
one movement mode. Eighteen landslides of this type were identified in the upper
Mangawhara catchment following careful examination of shear-plane configuration and
deposition morphology.
Landslides can also be grouped according to the number of phases of movement. The
distribution and nature of deposits on the shear plane and the shape of the shear plane
itself frequently provide evidence of more than one phase of failure. On the basis of such
evidence, 51 of the 92 landslides suffered only one phase of movement, 34 had failed
twice in rapid succession, and 7 landslides were considered to exhibit 3 phases o1 failure.
However, these last-described groupings of landslides are no less subjective than the
decision to place a landslide in one of the three categories d6bris slide, d6bris avalanche,
d6bris slide--d6bris avalanche. They may also be less informative.
The landslide depth/length ratio defined by Skempton (1953) has been used success-
fully by several authors to characterize landslide form. A single criterion can be used to
separate various types of slope failure which are related to conditions broadly classed as
rotation, translation, or flow. However, as the length measurement is made from the
landslide crown to the depositional toe, where the moving mass is truncated by a stream,
accurate ratios cannot be established. For the present sample of 92 landslides, depth/
length ratios could be accurately measured in only 47 cases.
It is evident that the simple methods, used to differentiate between the closely related
landslide types investigated here, are inadequate. Very simple classifications, with only
three or four divisions, fail to distinguish among landslides that have different morpho-
logical features. As noted by previous authors, more detailed classifications can only be
applied subjectively; while individual landslides share some features in common, morpho-
logical diversity is still either apparent or subsumed under ' complex' groupings. The
present study investigates, therefore, numerous aspects of landslide morphology in an
attempt to identify criteria that usefully distinguish between d6bris flows, d6bris slides,
and d6bris avalanches.
DEFINITION OF LANDSLIDE FORM ATTRIBUTES
All types of landslide observed in the upper Mangawhara catchment are included in the
sample. Within small catchments, selected non-randomly, all landslides unaffected by
farm roads and/or other obvious human activity were included in the sample. Eighteen
out of 110 landslides sampled were subsequently rejected for the above reasons or because
NUMERICAL CLASSIFICATION OF SELECTED LANDSLIDES 103
data were incomplete. Field measurements were made using a simple slope surveying
device (Blong, 1972), with a 26-inch (66-cm) diameter bicycle wheel calibrated to record
distances, or with percentage visual estimation charts (Folk, 1951, p.33).
In the absence of prior information concerning the value of individual morphological
properties, a wide range of attributes was estimated. A total of 19 numerical attributes,
concerned mainly with size, shape, gradient, and locational properties of the erosional
zone and 43 disordered multistate attributes assessing general morphological character-
istics 1 of erosional, transportational, and depositional zones of the 92 landslides were
measured.
Although several computational procedures were tested (as detailed below), only the
results of one classification (Class II) are reported here in detail. Consequently, only the
seven numerical and the nine multistate attributes used are defined here. (Details of the
attributes used in Class I are available from the author.)
ELEN erosional slope length, the groundsurface length from the landslide headwall
to the foot of the shear plane measured in the direction of maximum slope.
HEAD - the height of the landslide headwall (mean of 3 measurements).
WID - erosional zone width measured at the foot of the shear plane.
EROS - the slope of the straight line joining the landslide crown (top of the headwall)
to the foot of the shear plane.
DO - a visual estimate of the degree of overlap between erosional and depositional
z one s .
PECO - the groundsurface distance from the hillslope crest to the base of the shear
plane expressed as a percentage of total hillslope groundsurface length.
CR - the ratio of landslide erosional zone area to the area of a circle having the same
perimeter as the landslide (cf. Miller, 1953, p.8).
The nine multistate attributes are listed in Table III.
COMPUTATIONAL PROCEDURES
The numerical attributes were first subjected to simple and multiple correlation and
regression analyses. These analyses enabled the interrelationships among attributes to be
identified (see Blong, 1973). In order to isolate those morphological attributes most valu-
able for landslide classification purposes, an agglomerative polythetic grouping was
performed using all 62 attributes. The general principles of numerical taxonomy have
been outlined by Sokal and Sneath (1963). More specific information concerning the
selection of proven methods for specific problems is contained in Lance and Williams
(1967a, b). Following the advice of Mr. P. W. Milne, Division of Computing Research,
CSIRO, Canberra, an agglomerative polythetic grouping procedure using the CSIRO pro-
These attributes express characteristics such as relationship of the landslide to topography, surface
roughness, parent material, cross-sectional shape, symmetry, type and location of depositional
material, and the number of phases of movement.
104 R. J. BLONG
TABLE Ii1
Disordered multistate landslide attributes
Qualitative
attribute number
State description* Number of
landslides
in state
M9 Shear-plane shape
M10 Cross-section shape
of shear plane
M16 Shear-plane shape is
formed on
M26 Depositional outline
M28 Depositional
material consists of
M33 Depositional area
has a toe which is
1. rectilinear
2. convex
3. concave
4. sigmoid
5. multiconcave
6. multirectilinear
7. convex-rect i l i near
8. concave--rectilinear
1. concave
2. slightly concave
3. rectilinear
4. slightly convex
5. convex
6. rect i l i near-convex
8. sigmoid
1. ill situ red weathered greywacke
2. colluvial greywacke
3. Hamilton ash beds
6. colluvial mixture
7. yellow weathered and shattered
greywacke in situ
1. semicircular
2. lobate
3. elongated lobate
4. regular (other t han 1, 2, 3 above)
5. irregular
6. very irregular
7. obliterated
1. rafted blocks
2. flow material
3. veneer deposition
4. combi nat i on 1 and 2
5. combi nat i on 2 and 3
8. no evidence
9. one large block
1. absent
2. 0- 20 cm
3. 20- 50 cm
4. 50- 100 cm
5. 100- 200 cm
6. 200- 300 cm
7. > 300 cm
46
t3
18
1
l
I
8
4
41
25
22
l
1
1
1
38
3
5
5
41
10
23
35
11
1
2
10
6
36
2
35
2
7
4
56
1
12
11
9
2
1
NUMERICAL CLASSIFICATION OF SELECTED LANDSLIDES
TABLE III (continued)
Disordered multistate landslide attributes
105
Qualitative State description*
attribute number
Number of
landslides
in state
M34 Majority of 1. on hillslope 52
deposition occurs 2. on valley floor 31
3. beyond valley floor at foot of
hillslope 9
M38 Failure 1. simple 50
2. compound 24
3. complex 18
M40 Dominant movement 1. rotation 1
2. translation 15
3. flow 55
6. flow-translation 20
7. fl ow-fal l 1
*Where no landslides exist in a particular state, the state is not listed but the original numbering is
retained.
grams MULTBET, GROUPER, MAXGOWER, and GOWECOR were sel ect ed f or use in
t he pr esent s t udy. A cent r oi d sort i ng st r at egy was e mpl oye d t oget her wi t h a Shannon-
t ype i nf or mat i on st at i st i c ( Lance and Williams, 1967a). The f our pr ogr ams pr ovi de: an
hi er ar chi cal s t r uct ur i ng o f t he i ndi vi dual l andl i des, an anal ysi s o f t he si gni fi cance o f each
at t r i but e in f or ci ng each maj or gr oup f or mat i on, a pr i nci pl e co- or di nat e axes anal ysi s
( Gower , 1966) , and a s ummar y of t he cor r el at i on bet ween each l at ent r oot and each
at t r i but e ( Lance and Wi l l i ams, 1967a, b; Lance et al . , 1968). These resul t s al l ow deci si ons
t o be made a bout t he val ue o f each at t r i but e in t he cl assi f i cat i on.
RESULTS
The fi rst cl assi f i cat i on (Class I) e mpl oye d all 19 numer i cal and 43 di s or der ed mul t i -
st at e at t r i but es . Fi g. 1 i l l ust r at es t he hi er ar chi cal st r uct ur e of t he cl assi fi cat i on. Onl y t he
l ast t en fusi ons are shown. The hi gh val ue of t he Eucl i dean met r i c ( 452. 82) for t he fi nal
f usi on i ndi cat es t he essent i al di ssi mi l ar i t y bet ween t he t wo maj or gr oups, A and B.
The fi rst cl assi f i cat i on is t oo l engt hy and c ompl e x t o be s ummar i zed in t abl e f or m.
Perusal of t he r esul t s i ndi cat es t hat t her e is no cr i t er i on mut ual t o all gr oups in an hi er-
archi cal level. No single l andsl i de f or m at t r i but e consi der ed her e can be used as a classi-
f i cat or y cr i t er i on f or t he sampl e of 92 l andsl i des. Fur t he r mor e , f r om t he resul t s, i t seems
doubt f ul t hat t he use of several at t r i but es t oget her pr ovi des a mor e decisive cl assi f i cat i on.
In view of t he di f f i cul t i es expr essed by vari ous aut hor s in cat egor i zi ng i ndi vi dual l and-
slides, t hese concl usi ons are not surpri si ng.
106 R.J. BLONG
S O 0 0
A
38
G
23 I l S
M
8 15
8
C
20
E
K L
11 11
4000
N u rn d e rs r e f e r t o n u m b e r o f l a n d s l i d e s i n e a c l l g r o u p
I
300"0
200.0
Fig.1. Landslide classification structure for the last 10 fusions using 19 numerical attributes and 43
disordered multistate attributes.
However, the agglomerative polythetic classification illustrated in Fig.1 can also be
used to examine the validity of the classifications presented earlier. The simple classifi-
cations, such as the modified Varnes's classification and the Sowers and Sowers division of
landslide phenomena, are based on the general appearance of slope failures. A wide variety
of individual aspects of landslides is considered and assumptions are made regarding the
genesis of the landslides. Similarly, the grouping of landslides according to the number of
phases of movement or by the recognition of the landslide as simple, compound, or
complex, concentrates on shear-plane phenomena; a variety of criteria is examined and
judgements (albeit subjective) are made concerning the nature of the failure. The com-
puter classification considers a wide variety of landslide features, some requiring a general
impression of the landslide, some relating to specific morphological aspects of the failure
zone, the depositional zone, and the landslide environment. The classification procedure
subsequently selects those landslide attributes that promote the distinctiveness of groups.
Within these limitations, it seems reasonable to expect some correspondence between the
computer-produced groupings and those intuitively divined in other classifications.
Table IV illustrates the relationships between the agglomerative grouping achieved here
and four of the classifications discussed earlier. The computer classification can also be
NUMERICAL CLASSIFICATION OF SELECTED LANDSLIDES
TABLE IV
Comparison of various classifications
107
Group name: Total A B C D E F G H I J K L M N
Number ingroup: 92 38 54 20 34 24 10 23 15 14 6 11 13 8 15
M37
Modified Varnes's classification:
1. Slump 1
2. D6bris slide 11
5. D6bris avalanche 2
11. D~bris slide-avalanche 24
9. D~bris avalanche-flow 49
3. Underthrust slide 3
12. Incipient underthrust slide 2
M40
Dominant movement is:
1. Rotation
2. Translation
3. Flow
6. Flow-translation
7. Flow-fall
M38
Failure is:
1. Simple
2. Compound
3. Complex
1 1 1 -
3 8 6 2 2 - 2 1 5 1 - 2 2 -
1 1 1 1 1
1 2 3 8 1 5 5 1 0 1 - 8 - 3 2 - 1
32 17 - 17 19 - 19 13 8 9 5 14
- 3 3 3
- 2 2 2 . . . .
1 1 1 1 -
1 5 2 13 10 3 3 - 1 1 4 6 - 3 1 -
55 33 22 1 21 19 2 20 13 1 - 10 9 6 I4
20 1 19 9 10 2 8 1 - 9 - 1 1 - 1
1 1 1
50 25 25 16 9 9 - 17 12 10 6 8 1 7 6
24 12 12 - 12 11 1 5 3 2 9 1 8
18 1 17 4 13 4 9 1 - 4 - 1 3 - 1
M39
No. of phases of movement:
1. 51
2. 34
3. 7
25 26 16 10 10 - 17 12 10 6 8 2 7 6
11 23 4 19 12 7 4 3 4 3 9 1 7
2 5 - 5 2 3 2 2 - 2
compar ed wi t h an unmodi f i ed Var nes' s cl assi fi cat i on of t he 92 l andsl i des by r egr oupi ng
d6bris s l i de- aval anche, d6bris aval anche- f l ow, under t hr ust slide, and i nci pi ent under-
t hr ust slides as ' compl ex' l andsl i des.
It is evi dent t hat at no level in t he hi er ar chy is t her e a close cor r es pondence bet ween
t he cl assi fi cat i ons l i st ed and t he comput er groupi ng pr esent ed. The lack of cor r es pondence
does less t o i nval i dat e t he comput er cl assi fi cat i on t han it does t o emphasi ze t he failings of
mor e t r adi t i onal l y- or i ent ed groupi ngs, even where t hese cl assi fi cat i ons have been modi f i ed
by exper i ence rel evant t o t he pr esent sampl e.
To t est t he hypot hes i s t hat t her e is no superi or way of di st i ngui shi ng unequi vocal l y
bet ween l andsl i des br oadl y classed as d6bris sl i des- d6br i s f l ows - d6br i s aval anches, a
second comput er cl assi fi cat i on was per f or med (Class II). In Class I t oo many at t r i but es
are empl oyed, many of whi ch do little except conf use. Some of t he numeri cal at t r i but es
108 R..I. BLONG
1 6 0 , 0 -
P
el
0
41
U V
11 12
I 1
Numbers r l f e r to number of landslides In each group.
R
/ , 0
T
t l
0
_o 120,0
X
d
,.=,
8 0 - 0
Fig.2. Landslide classification structure for the last ten fusions using seven numerical and nine
disordered multistate attributes.
are logically and mat hemat i cal l y correl at ed, t hus rei nforci ng the aggl omerat i on of parti-
cular groups within t he classification. Fur t her mor e, some of t he di sordered mul t i st at e
at t ri but es are ext r emel y subjective.
Fol l owi ng exami nat i on of t he relationships among the numeri cal at t ri but es (Blong,
1973), the seven at t ri but es defi ned above were sel ect ed as satisfactorily represent i ng
quant i t at i ve aspects of landslide mor phol ogy. The ot her twelve numeri cal at t ri but es are
either moder at el y hi ghl y correl at ed wi t h those listed above or cont r i but ed little t o the
precedi ng analysis. In an at t empt t o reduce t he compl exi t i es of t he earlier classification,
34 of t he 43 di sordered mul t i st at e at t ri but es have been eliminated. The remai ni ng nine
at t ri but es (Table I l l ) were selected as coveri ng t he maj or charact eri st i cs of t he sample
landslides.
The classification st ruct ure for t he last t en fusions is shown in Fig.2. Three maj or
groups of landslides are evi dent f r om the classification st ruct ure. A gr oup of eleven land-
slides, labelled O in Fig.2, is qui t e distinctive, being f or med at a l ow level o f fusi on and
remai ni ng distinctive until all landslides are j oi ned t oget her in t he one group. Ten of t he
eleven movement s are t ransl at i onal ; most o f t he deposi t ed mat eri al consists o f raft ed
bl ocks or has remai ned as a single bl ock. Deposi t i onal out l i nes are semicircular or l obat e.
The deposi t i onal overlap (DO) for t he gr oup o f eleven averages 74%. All t he landslides in
this gr oup have a deposi t i onal t oe and all are simple failures in t hat onl y one phase of
movement has been identified. Erosi onal slopes are relatively gentle wi t h a mean gradient
of 24 . The failure zones have an average wi dt h of 17 m.
On t he ot her hand, Gr oup P wi t h 81 landslides is charact eri zed by fl ow (68%) and
fl ow-t ransl at i on (25%) movement s. The deposi t i onal zones are el ongat ed l obat e or l obat e,
NUMERICAL CLASSIFICATION OF SELECTED LANDSLIDES 109
and the average depositional overlap is only 22%. Of these failures 69% have no obvious
depositional toe, and the mean erosional gradient is 34 . Shear planes average only 11 m
in width. More than half of these landslides are either compound or complex slope
failures.
In terms of the modified Varnes's classification (Table II), the group with eleven land-
slides includes all the underthrust and incipient underthrust slides, the single slump, and
five of the twelve d~bris slides. The taxonomic classification, then, confirms the subjec-
tive grouping to some extent, but does not accurately duplicate the field classification.
As indicated in Fig.2, Group P is composed of two sub-groups of nearly equal size.
Group Q, with 41 members, is compared with Group R (40 members) in Table V in order
of increasing similarity of attributes between groups. The biggest difference between
these groups is in terms of location on the slope. Where landslides occur very low on the
hillslope (Group R), nearly one quarter of the depositional outlines have been obliterated.
Otherwise, the differences between the two groups are not large, and do not form the
basis of a rational classification of landslides.
TABLE V
Comparison of two major landslide groups
Group Q (41 members) Group R (40 members)
PECO (mean) 55% 80%
ELEN (mean) 17.0 m 16.7 m
DO (mean) 26% 18%
M26 depositional outline 24% elongated lobate 40% elongated lobate
39% lobate 12.5% lobate
0 obliterated 23% obliterated
15% regular 12.5% regular
0 lobate 12.5% obate
M33 height of 61% no toe 78% no toe
depositional toe 22% 20-50 cm 0 0-50 cm
0 100-200 cm 12.5% 100-200 cm
10% 50-100 cm 0 50-100 cm
Group R is itself composed of two distinct groups (Fig.2). Group T, with fifteen
members, has a mean value of PECO of 95%. Consequently, all of the landslides in this
group have had the majority of tile depositional material removed from the hillslope.
Group S, with 25 members, has a mean value of PECO of 70%.
Group Q is also composed of two smaller groups, although these are less distinct than
in the case of Group R (Fig.2). The sub-group with eighteen members (Group U) is
dominated by both flow and translational movements, while Group V is dominated by
flow-translational and flow movements. Other differences are relatively minor.
In general, at the higher levels of classification, no actual landslide characteristics seem
capable of distinguishing adequately between the types of landslides recognized in the
field. However, it should be remembered that the computer classification was undertaken
110 R. J. BLONG
because of t he di ffi cul t i es of di f f er ent i at i ng obj ect i vel y bet ween closely rel at ed sl ope-
failure t ypes. PECO, a l andsl i de l ocat i on at t r i but e, is t he most effect i ve measur e used her e
in gr oupi ng l andsl i des. Shear pl anes l ocat ed near t he base of hi l l sl opes, wi t h hi gh values of
PECO, t end t o have t he deposi t i onal phase of t he move me nt obl i t er at ed, deposi t i on pre-
domi nant l y be yond t he f oot of t he hi l l sl ope, and shear pl anes f or med in colluvial mat eri al
or i n-si t u yel l ow- br own weat her ed gr eywacke. All t hese l at t er charact eri st i cs are, m fact ,
t he result of shear-pl ane l ocat i on near t he base of t he hi l l sl ope.
Table VI i ndi cat es t he poor cor r es pondence bet ween t he second comput er cl assi fi cat i on
and t he f our t r adi t i onal groupi ngs of t he 92 l andsl i des. Al t hough t he under t hr ust slides
r emai n t oget her as a gr oup wi t h some of t he d6bris slides, as t hey di d in t he first com-
put er cl assi fi cat i on (Tabl e IV), it is evi dent t hat t he comput er groupi ngs cut across t he
ot her cl assi fi cat i ons. However , t he cor r es pondence achi eved bet ween t he cl assi fi cat i ons is
TABLE VI
Comparison of classifications with computer Classification 2
Group Name: Total O P Q R S T U V
Number in group: 92 11 81 41 40 25 15 25 15
M37
Modified Varnes' s classification:
1. Slump 1 1 . . . . . . . . . . .
2. D~bris slide 11 5 6 5 1 1 - 3 2
5. Dgbris avalanche 2 - 2 1 1 - 1 1 -
11. D~bris slide-avalanche 24 - 24 18 6 3 3 1 17
9. Ddbris avalanche-flow 49 - 49 17 32 21 11 13 4
3. Underthrust slide 3 3 . . . . . . . . . .
12. Incipient underthrust slide 2 2 . . . . . . .
M40
Dominant movement is:
1. Rotation 1 1 . . . . . . . . . .
2. Translation 15 10 5 4 1 1 2 2
3. Flow 55 55 21 34 23 11 16 5
6. Flow-translation 20 20 16 4 1 3 16
7. Flow-fall 1 - 1 - 1 - 1 -
M38
Failure is:
1. Simple 50 11 39 17 22 12 10 11 6
2. Compound 24 - 24 12 12 10 2 7 5
3. Complex 18 - 18 12 6 3 3 -- 12
M39
Number ofphases of movement:
1. 51 11 40 18 22 12 10 12 6
2. 34 - 34 20 14 11 3 4 16
3. 7 - 7 3 4 2 2 2 1
NUMERICAL CLASSIFICATION OF SELECTED LANDSLIDES 111
greater than t hat resulting from the first comput er grouping. Nevertheless, the poorest
correspondence occurs in the case of those landslides broadly classed as d~bris sl i des-
d6bris avalanches-d6bris flows.
A comparison of the two comput er classifications is difficult because the major land-
slide characteristics responsible for the agglomeration of each group are different in each
case. In Classification 1 six groups lettered E, F, G, H, I, and J occur at the second major
hierarchical level. Similarly, in the second comput er classification, groups O, S, T, U, and
V form the second hierarchical level. Table VII indicates the group i nt o which each land-
slide falls for each of the comput er classifications as well as for the four simple classifi-
cations based on disordered multistate variables 37, 38, 39, and 40, and as presented i n
Tables III, IV, and V1.
It is apparent from Table VII that the removal of a large number of variables from
Classification 1 has produced many changes in the groupings achieved in Classification 2.
The classifications cannot, therefore, be regarded as stable. The data presented in Table
VII also show t hat landslides grouped together in even four out of six of the classifications
can be widely separated in the other two classifications.
SUMMARY AND CONCLUSIONS
Neither of the two comput er classifications of landslides presented here reveals any
single landslide form at t ri but e or group of attributes that can be used to distinguish,
clearly, between such closely related types of slope failures as d6bris slides, d6bris flows,
and d6bris avalanches. Furt hermore, the format i on of individual groups within the com-
puter classifications is dependent upon a wide variety of landslide characteristics. The
groups, once formed, frequently bear no resemblance to landslide categories defined in
more traditional classifications; this is not an indictment of the comput er groupings. The
lack of agreement between the comput er classifications tested indicates that stability has
not been achieved, and that other characteristics will have to be investigated i f a rational
descriptive classification of landslide types is to be constructed.
No clear basis can be identified for differentiating among the landslides grouped by
Varnes (1958) as complex, but to group so many landslides of possibly diverse forms as
complex remains unsatisfactory. The removal of deposits by stream action prevents the
use of Skempt on' s (1953) depth/length ratio in a large number of cases in the upper
Mangawhara catchment. Although landslide location on the hillslope proved a useful
criterion in the second comput er classification, because some other local landslide
characteristics are partly controlled by distance from the base of the hill, this attribute is
hardly likely to provide the basis of a universally applicable classification.
As no one of the classifications of landslides at t empt ed here provides any rational basis
for comparison of slope failures from area to area, until a realistic scheme is forthcoming
it seems sensible to rely on a simple division of landslides according to the nature of the
dominant movement . However, it seems t hat the three basic t ypes of failure - rotation,
translation, and flowage - must be supplemented by a transitional fourt h t ype, that
f

L
a
n
d
l
i
s
d
e

N
u
m
b
e
r

M
3
7

M
3
8

M
3
9

M
4
0

C
l
a
s
s

I

C
l
a
s
s

I
I

L
a
n
d
s
l
i
d
e

N
u
m
b
e
r

M
3
7

M
3
8

M
3
9

M
4
0

C
l
a
s
s

I

C
l
a
s
s

I
I

L
a
n
d
s
l
i
d
e

N
u
m
b
e
r

M
3
7

M
3
8

M
3
9

M
4
0

C
l
a
s
s

1

C
l
a
s
s

I
I

g
~

g
~

~
'
~
.

g

g

a

<


Z

NUMERICAL CLASSIFICATION OF SELECTED LANDSLIDES 113
dominated bot h by flow and by translational failure. Where possible, this four-fold
division should be supplemented by Skempton' s depth/length ratio.
ACKNOWLEDGEMENTS
The author is indebted to Mr. C. F. Pain (Australian National University) for his con-
siderable help and critical comment during field work. Mr. P. W. Milne (Division of Com-
puting Research, CSIRO, Canberra) provided invaluable advice and assistance with
computing procedures. Professor G. H. Dury (The University of Wisconsin), R. J. Wasson,
M. F. Clarke and M. A. J. Williams (Macquarie University) all made valuable criticisms of
a draft of the manuscript. Financial assistance during field work and for comput i ng was
provided by The University of Sydney and Macquarie University.
REFERENCES
Bailey, R. G. and Rice, R. M., 1969. Soil slippage, an indicator of slope instability on chaparral water-
sheds of southern California.Prof. Geogr., 21(3): 172-177.
Blong, R. J., 1971. The underthrust slide - an unusual type of mass movement. Geogr. Ann., 53A:
52-58.
Blong, R. J., 1972. Methods of slope profile measurement in the field. Aust. Geogr. Stud., 10: 182-
192.
Blong, R. J., 1973. Relationships between morphometric attributes of landslides. Z. GeomorphoL, in
press.
Cumberland, K. B., 1944. Contrasting regional morphology of soil erosion in New Zealand. Geogr.
Rev., 34: 77-95.
Folk, R. L., 1951. A comparison chart for visual percentage estimation. J. Sediment. Petrol., 21 : 32-
33.
Gower, J. C., 1966. Some distance properties of latent root and vector methods used in multivariate
analysis. Biometrika, 53: 325-338.
Hutchinson, J. N., 1968. Mass movement. In: R. W. Fairbridge (Editor),Encyclopedia of Geomor-
phology. Reinhold, New York, N.Y., pp.688-695.
Irwin-Hunt, J. R., 1960. Some Quantitative Aspects of Mass Movement. Thesis, Kings College,
University of London, London, 76 pp.
Ladd, G. E., 1935. Landslides, subsidences and rock-falls. Am. Railw. Eng. Assoc. BulL, 6: 1092-
1162.
Lance, G. N., Milne, P. W. and Williams, W. T., 1968. Mixed data classificatory programs. III. Diag-
nostic systems.Aust. Comput. J., 1(3): 178-181.
Lance, G. N. and Williams, W. T., 1967a. Mixed-data classificatory programs. I. Agglomerative systems.
Aust. Comput. J., 1(1): 15-20.
Lance, G. N. and Williams, W. T., 1967b. A general theory of classificatory sorting strategies. I. Hier-
archical systems. Comput. J., 9(4): 373-380.
Miller, V. C., 1953. A quantitative geomorphic study of drainage basin characteristics in the Clinch
Mountains areas, Virginia and Tennessee. Dept. Geol., Columbia Univ., ONR Proj. NR 389-420
Tech. Rep., 3: 30pp.
Rice, R. M., Corbett, E. S. and Bailey, R. G., 1969. Soil slips related to vegetation, topography, and
soil in southern California. WaterResour. Res., 5(3): 647-59.
Sharpe, C. F. S., 1938. Landslides and Related Phenomena. Pageant, N.J., 125 pp.
Skempton, A. W., 1953. Soil mechanics in relation to geology. Proc. Yorks. Geol. Soc., 28: 33-62.
Skempton, A. W. and Hutchinson, J. N., 1969. Stability of natural slopes and embankment found-
ations, lnt. Soil Mech. Conf.. Mexico, State-of-the-Art Rep., 291-340.
114 R. J. BLONG
Sokal, R. R. and Sneath, P. H. A., 1963. Principles ojNumerical Taxonomy. Freeman, San [:rancisco,
Calif., 359 pp.
Sowers, G. B. and Sowers, G. F., 1961. Introductory Soil Mechanics and Foundations. Macmillan, New
York, N.Y., 386 pp.
Terzaghi, K., 1950. Mechanism of landslides. Geol. Soc. Am., Geol. (Berkeley) Vol. : 83 - 123.
Varnes, D. J., 1958. Landslide types and processes. Highw. Res. Board. Spec. Rep., 29: 20-47.
Ward, W. H., 1945. The stability of natural slopes. Geogr. J., 105: 170-197.
Yatsu, E., 1966. Rock Control in Geomorphology. Sozosha, Tokyo, 135 pp.
Yatsu, E., 1967. Some problems on mass movements. Geogr. Ann., 49A: 396-401.
Zaruba, Q. and Mencl, V., 1969. Landslides and their Control. Elsevier, Amsterdam, 202 pp.

You might also like