918 
results of human activity. Mitchell 
(1973) stresses that land evaluation is a 
broad term which encompasses analysis, 
classification and appraisal of 
information from a variety of sources for 
a potential land use. Analysis involves 
selecting characteristics which have 
importance for a particular application 
and compiling land characteristics. 
Classification relates to the organisation 
of characteristics which distinguish one 
area from another and which characterise 
each. Appraisal uses these 
characteristics, along with other 
properties, to assign a value to a piece 
of land, expressed either by a numerical 
value or by a judgement of its worth in 
qualitative terms. 
A land resources evaluation system has 
several basic requirements. Mitchell 
(1973) identifies three: 
1. a means of answering queries from 
users; 
2. a means of acquiring, storing, 
analysing and displaying information about 
the land and its potential uses; 
3. a means of retrieving and 
manipulating information; 
The traditional approach to fulfilling 
requirements for land resources evaluation 
has been by preparing manually various 
maps and transparent overlays showing 
features, such as slope, aspect, soils, 
drainage and other characterisitcs and by 
preparing statistical and textual reports. 
Visual comparison and interpretation of 
maps and reports leads to an evaluation of 
regional land resources for a particular 
application. The basic source of 
information for all these maps has 
usually been aerial photographs, though 
other forms of remote sensing are 
increasingly being used to aid sub 
division of the land. Computers are used 
increasingly to store, process and 
retrieve at least some of the data and 
GIS have been developed using a fixed 
cell size grid or polygon respresentation, 
but much geographic data is still stored 
in analogue maps because these have 
provided access much more quickly than 
existing GIS when large volumes of 
geographic data are involved. 
GEOGRAPHIC INFORMATION SYSTEMS 
Marble and Peuquet (1983) describe the 
development of GIS and observe that a GIS 
is designed to accept large volumes of 
spatial data, derived from a variety of 
sources including remote sensing, and to 
store, retrieve, manipulate, analyse and 
display these data. The development of 
intelligent GIS in which the concepts and 
techniques of artificial intelligence and 
database systems are integrated represents 
a major new field of research (Smith and 
Pazner, 1984a; 1984b; Smith and Peuquet, 
1985; McKeown et al. 1984). 
In designing a GIS, a critical decision is 
the choice of data model. This is the 
abstraction that is used to represent 
properties which are considerd to be 
relevant to the application in the 
computer. Peuquet(1974) reviews the 
different types of spatial data models 
that have been used in GIS and compares 
their performance. Geographic data have 
been represented using many different 
types of data models, but a basic 
difference is between vector and raster 
types. 
1 Vector type 
In this type of data model, the basic 
logical unit in a geographical context 
corresponds to a line on a map. It is 
recorded as a series of x-y coordinates 
with a heading describing the feature. 
Vector data is widely used in cartographic 
GIS and many other types which have been 
developed for specific projects. 
2 Raster type 
This type of data model uses a fixed-sized 
square cell or raster to represent 
geographic data in a binary array or grey 
scale .image. The development of data 
models based on raster has been largely 
driven by advances in the technology of 
remote sensing and computing over the past 
decade (Marble and Peuquet 1983). The use 
of MSS scanning systems in satellite 
remote sensing has been a major influence. 
At the same time, there have been 
significant advances in the technology of 
raster scan and video digitising systems. 
These have accelerated digitising maps and 
related documents. Because all these 
systems use a square cell or raster, it is 
generally agreed that this is the only 
practical tiling or tessellation. A number 
of other possibilities exist which may be 
theoretically better than the regular 
tessellation (Bell et. al., 1983). 
Peuquet (1984b) discussed the main 
advantages of raster type of data models. 
Apart from the practical benefits of being 
able to get massive sets of raster data 
from satellite remote sensing, and raster 
scanning of maps, it is compatible with 
array data structures and various hardware 
devices for input and output. Peuquet 
(1984b) and McKeown (1984) argue that 
existing vector and raster data models are 
limited however by two basic factors: 
1. the rigidity and narrowness in the 
range of applications and types of 
geographic data which can be accommodated; 
2. the unacceptibly low levels of 
efficiency for storage and response to 
queries for the current and anticipated 
volumes of geographic data. 
These factors restrict the potential of 
automated GIS based on the use of vector 
or raster data models to cope with the 
variety of different forms of geographic 
data and the massive volumes. For these 
reasons, attention has recently focused on 
another data model known as the quadtree. 
QUADTREE DATA MODEL 
A data model which has become increasingly 
important in recent years is the quadtree, 
which is based on the concept of recursive 
decomposition of a grid. The idea of the 
quadtree was formulated by Klinger (1971) 
but ha 
includi 
and 
(1980,1 
theory 
broaden 
They ha 
in com 
image p 
GIS. Sa 
review 
hierarc 
that th 
quadtre 
regions 
based 
decompo 
not all 
of in 
quadtre 
and f 
common 
chain c 
studied 
scienti 
Quadtre 
A quad 
binary 
image. 
in the 
that a 
array c 
represe 
dividin 
further 
occurs 
(possit 
homoger 
black 
level c 
process 
branche 
corresf 
pixels 
node c 
which i 
North E 
termina 
corres; 
which 
necessa 
represe 
Thus, 
single 
at le\ 
quadtre 
illusti 
region 
binary 
square 
lc and 
Initial 
carriee 
the t 
represe 
five i 
ancest« 
node. 1 
et.al.I 
search 
based 
storage 
Accord: 
each j 
the n< 
pointei 
space i