Académique Documents
Professionnel Documents
Culture Documents
REFERENCES
Linked references are available on JSTOR for this article:
http://www.jstor.org/stable/633076?seq=1&cid=pdf-reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
is collaborating with JSTOR to digitize, preserve and extend access to The Geographical Journal
This content downloaded from 188.247.93.242 on Tue, 21 Feb 2017 08:10:54 UTC
All use subject to http://about.jstor.org/terms
AN ANALYSIS OF TERRAIN CLASSIFICATION
FOR LONG-RANGE PREDICTION OF
CONDITIONS IN DESERTS
People who plan to use land in ways that are different from those current usually
require information about the soil and other terrain properties. They may mount
special surveys to obtain the information, or use whatever happens to be available.
In either event the information will pertain to specific sites, which may be auger
bores, profile pits or trial plots far apart from one another. The planner, however,
is usually interested in continuous tracts of country, and therefore wishes to
predict in a spatial sense the nature ofthe terrain at many unvisited or unrecorded
sites from relatively few observations at specific sites.
Prediction may be based on the assumption that nearby sites are likely to be
similar or that sites are likely to be similar with respect to the property of current
interest if they are also alike in other respects, or both. The second is usually
assumed in practice, especially when sampling has been sparse: land is classified,
and conditions at unfamiliar sites are predicted from the information previously
obtained at any sites in the same class.
In earlier work we examined the validity of this second assumption, studying
the effectiveness of terrain classification in both temperate and hot desert climates
(Webster and Beckett, 1964; Beckett and Webster, 1965a, b; Beckett et al, 1972;
Perrin and Mitchell, 1969). Here we introduce an element ofthe first assumption,
for we are particularly concerned with the influence of distance in terrain classi?
fication. Thus, given that properties of a new site can be predicted well from those
This content downloaded from 188.247.93.242 on Tue, 21 Feb 2017 08:10:54 UTC
All use subject to http://about.jstor.org/terms
C. W. MITCHELL AND OTHERS 73
This content downloaded from 188.247.93.242 on Tue, 21 Feb 2017 08:10:54 UTC
All use subject to http://about.jstor.org/terms
74 TERRAIN CLASSIFICATION FOR
TABLEI
Land region One or more land systems that are 1:1M to 1:5M
adjacent or in close proximity and are
similar in lithology and surface form.
1:5M to 1:15M
Land province One or more land regions, adjacent or in
close proximity in the same large
lithological association or second order
structure.
Land zone A major climatic zone consisting of one or less than 1 :15M
more land divisions.
TABLE II
Scheme of classification for desert terrain and the numbers of groups recognized
This content downloaded from 188.247.93.242 on Tue, 21 Feb 2017 08:10:54 UTC
All use subject to http://about.jstor.org/terms
PREDICTION OF CONDITIONS IN DESERTS 75
related to one another and the relations between abstract classes and their local
forms.
With this hierarchical framework to guide them Perrin and Mitchell (1969)
set about classifying the terrain of the hot deserts. Information was gleaned from
over 1200 published accounts and from a systematic survey of available air photo?
graphy. Thirty-three sample areas, each covering approximately 1? square, were
examined in more detail, and field traverses were made to obtain round data. The
terrain was classified at the five levels of the hierarchy, and the numbers of classes
recognized in each category are given in Table II. Over 100 abstract land facets
were recognized in the areas examined. It was also possible to identify analogous
land systems, made up of the same patterns of recurring facets, in different parts
of the desert zone. For example, the fans from the Ibis and Mohawk Mountains of
California, the Hon Tertiary fan and Senonian hill outwash of north-western
Libya, the central plain of Abd al Kuri and the fans from the Jebel Dhanna Dome
in Arabia are all local forms of a single abstract land system?fans and bajadas.
Likewise the dissected limestone tablelands in the Jol district of Arabia have close
analogues in east Jordan, northern Algeria and Morocco that constitute an abstract
land system. At the level of land region and above, few, if any, classes obviously
Semi-arid
or non-arid areas
Land province
boundaries
Jabal Nafusah
Hammada al Hamra'
Sirte Lowlands
Hun-Bu Nujaym Trough
Jabal al Soda
Erg Ubari
Hammadat Marzuq
Erg Marzuq
TIBESTI
MOUNTAINS
This content downloaded from 188.247.93.242 on Tue, 21 Feb 2017 08:10:54 UTC
All use subject to http://about.jstor.org/terms
76 TERRAIN CLASSIFICATION FOR
recurred. Each class occupied a single specific tract of country or several sepa
tracts in close proximity, and it was not possible to group them into abstract cla
on their similarity. In practice therefore all classes at region level and above a
local forms.
Table II summarizes the classification achieved for desert terrain. Note that a
particular local form of a land facet can occur only within one local land system,
and a local form of a land system only within one land region. As above however, the
abstract facet to which its local forms belong can occur in many local land systems,
This content downloaded from 188.247.93.242 on Tue, 21 Feb 2017 08:10:54 UTC
All use subject to http://about.jstor.org/terms
PREDICTION OF CONDITIONS IN DESERTS 77
PACIFIC
OCEAN
p Semi-arid
ijj or non-arid areas
_ Land province boundaries
0 Km 500
18 Colorado Plateau,
15 Texas High Plains (Pecos Valley)
16 Mexican Meseta (Altoplanice) (Colorado River Valley)
17 Sonora (incl. Salton Trough) 19 Great Basin
and hence regions, divisions and provinces, and it may occur in more than one
abstract land system. Perrin and Mitchell (1969) have described in full their survey
procedure and the classes of terrain recognized.
This content downloaded from 188.247.93.242 on Tue, 21 Feb 2017 08:10:54 UTC
All use subject to http://about.jstor.org/terms
78 TERRAIN CLASSIFICATION FOR
Sampling
The desert zone was sampled by choosing four main sample areas (Fig. 4),
corresponding to four land divisions within it, and choosing sites within them to
represent a minimum of four local forms of each of twenty-four facets. This
ensured adequate replication of classes in all categories down to local forms of land
facets. Expeditions were mounted to visit the chosen sites in Libya, Socotra and
Abd al Kuri, and Trucial Oman (now United Arab Emirates) and Bahrain, while
data had been obtained for south-west USA by visiting sites during the earlier
classification stage (Mitchell, 1969). Nearly 800 sites were visited in all. Table II
shows how they were distributed among the classes of the hierarchy. At each site
Horn South-west
of LAND DIVISIONS
Africa
USA
Fig. 4. Fragment of the hierarchy of local classes for desert terrain. In most instances
each class at every level above that of the facet consists of several classes in the level
immediately below. In a few instances, e.g. the Abd al Kuri land province, it contains
only one. Facets and, in some instances, land systems are grouped into abstract classes
without regard to their place in the hierarchy. Thus the upper fans of the Abd al Kuri
Central Plain and those of the Mohawk Mountains Outwash are regarded as the same
abstract facet. Likewise the two facets labelled 'Colluvial Flats3 belong to the same
abstract. The Abd al Kuri Central Plain and the Mohawk Mountains Outwash are
analogues and belong to the same abstract land system while the Newberry Area does not
This content downloaded from 188.247.93.242 on Tue, 21 Feb 2017 08:10:54 UTC
All use subject to http://about.jstor.org/terms
PREDICTION OF CONDITIONS IN DESERTS 79
This content downloaded from 188.247.93.242 on Tue, 21 Feb 2017 08:10:54 UTC
All use subject to http://about.jstor.org/terms
80 TERRAIN CLASSIFICATION FOR
60 <J
"3 S S'Ss
J3
i> s^
21
<u
?S .s ?-
"? ^ ??
^3 1 o J
?.? I S 2
11 I ?M.J s
S5 T? ? . n Ji e 5 -2 ?P 8 c - c-......
5o
o.
This content downloaded from 188.247.93.242 on Tue, 21 Feb 2017 08:10:54 UTC
All use subject to http://about.jstor.org/terms
PREDICTION OF CONDITIONS IN DESERTS 81
^00
3 ? ^
S as^9SY<N<^oocO(N^fi>|1as|-HuPvov9vov?-"|i>v9
QQ ? s WwB^2w w ? w w w C ^
^ g
1 ? g-?ilS7S^RS3:*5SS??pSSS5?8?278?
-5.3 H-1 * ^2a?^s-g-.S6^S^^.S?26S?^CM
co
QQ
C K
8| ? 9 8 8 ? ? s * ?
Q> .? ^ co w^ ea ^h g B O ^ O ? O ? ? ^
s
2 ^?.2 ? .2 ?~^ "> "^
This content downloaded from 188.247.93.242 on Tue, 21 Feb 2017 08:10:54 UTC
All use subject to http://about.jstor.org/terms
82 TERRAIN CLASSIFICATION FOR
This content downloaded from 188.247.93.242 on Tue, 21 Feb 2017 08:10:54 UTC
All use subject to http://about.jstor.org/terms
PREDICTION OF CONDITIONS IN DESERTS 83
2 8
-*r*Q O^O^^OCNvOCM^iOcO
"5 ? S
2 ?5 &:?
?3 # g
o '^^."?^S^><T)'^fr)'^-'cooo'^vfvfo
"is ^S^PSojc^t^co^mint^in^^^
o ?-> i? ?: s ^
^ i-^^'s' q^oooooooom^^ooin
*5> ^2 K c h CM^ini>iridNCNiri^23!S
? SlBi'-S, int^cM^^^Tfirim^cOTf
* ? "8 B
**"?? "^c-^.o op^^p^a^cpopcMOvcMqs
u ti *> e 'Si in^^vbob-^vbo^^^co
I ?S '?
Q Q 3 K u
K ?'?<)??> COt^^CN^CMCO^^O^iri
s ^ 5
wH ^ ^
J 2 ?^?
!S s ? ONob^coc^co^c^^cbco^
?? ??. S CM CM^h^hCMCO^hCMCMITi^h
g
05
u o o 8
2 a b'-'u^r-s
.3 S
CJ
This content downloaded from 188.247.93.242 on Tue, 21 Feb 2017 08:10:54 UTC
All use subject to http://about.jstor.org/terms
84 TERRAIN CLASSIFICATION FOR
TABLE VI
TABLE VII
l.The 62 properties included the following: gradient upslope and downslope; spaci
maximum, and average size of surface stones; vegetation cover and the height and spa
of dominant plants; soil strength at the surface and 7-5 cm depth; Munsell hue, value,
chroma of the surface and each of 3 soil layers; width, depth, granulometric fractio
Atterberg limits, electrical conductivity, pH, % CaC03, and ?0 CaS04 of each of 3
layers.
2. This represents that percentage of the total variance which is unaccounted for by the
classification.
This content downloaded from 188.247.93.242 on Tue, 21 Feb 2017 08:10:54 UTC
All use subject to http://about.jstor.org/terms
PREDICTION OF CONDITIONS IN DESERTS 85
A cknowledgements
The original research on which this paper is based w
Engineers, and we thank the Military Vehicles and
(formerly MEXE) for support and collaboration.
Environmental Research Council for funding the l
Mrs S. A. Bell for help with the computing.
References
Beckett, P. H. T., and Webster, R. 1965a Field trials of a terrain classification?organization
and methods. Report No. 873, Military Engineering Experimental Establishment,
Christchurch.
Beckett, P. H. T., and Webster, R. 1965b Field trials of a terrain classification?statistical
procedure. Report No. 874, Military Engineering Experimental Establishment, Christ?
church.
Beckett, P. H. T., Webster, R., McNeil, G. M., and Mitchell, C. W. 1972 Terrain evaluation
by means of a data bank. Geogrl. J. 138: 430-56.
Brink, A. B. A., Mabbutt, J. A., Webster, R., and Beckett, P. H. T. 1966 Report of the working
group on land classification and data storage. Report No. 940, Military Engineering
Establishment, Christchurch.
Gower, J. C. 1962. Variance component estimation for unbalanced hierarchical classification.
Biometrics. 18: 537-42.
Linton, D. L. 1951 The delimitation of morphological regions. In: L. D. Stamp and S. W.
Wooldridge (Editors), London Essays in Geography, pp. 199-217.
Mitchell, C. W. 1969 Terrain evaluation?Descriptions of visits made to USA, Libya and
Middle East. Report No. 1118, Military Engineering Experimental Establishment,
Christchurch (unpublished).
Perrin, R. M. S., and Mitchell, C. W. 1969 An appraisal of physiographie units for predicting
site conditions in arid areas. Report No. 1111, Military Engineering Experimental
Establishment, Christchurch.
Snedecor, G. W., and Cochran W. G. 1967 Statistical methods. 6th edition. Iowa State
University Press: Ames.
Webster, R., and Beckett, P. H. T. 1964 A study of the agronomic value of soil maps inter-
preted from air photographs. Transactions, 8th International Congress of Soil Science 5:
795-803.
Webster, R., and Beckett, P. H. T. 1968 Quality and usefulness of soil maps. Nature, 2
680-682.
This content downloaded from 188.247.93.242 on Tue, 21 Feb 2017 08:10:54 UTC
All use subject to http://about.jstor.org/terms