complement (logical NOT) respectively. The 
results can be formed into a new quadtree 
file or sent directly to the printer. 
Alternatively, there is a menu driven 
interface which allows the detailed study 
of one quadtree. This can be used to find 
the colour of a point, traverse a line 
across the image or find a neighbour to a 
point. 
Work is now being done to integrate the 
GIS into a relational data base management 
system for storing regions and objects. A 
B-tree memory management system is being 
developed to store very large quadtrees 
(Abel, 1984). 
LAND EVALUATION IN THE PEAK DISTRICT 
To illustrate the potential of the GIS 
based on linear quadtrees for regional 
evaluation of land resources, a data base 
was created by digitising geographic data 
from maps and aerial photographs. Landsat 
MSS imagery was classified to provide land 
cover at a much coarser level and is being 
included in the GIS. 
Description of study area 
A study area was selected close to 
Matlock, Derbyshire. This was chosen 
because of the variety of geology, relief, 
soils and land use in a relatively small 
area. The region lies on a boundary of 
Carboniferous limestone and Millstone 
grit. There is a range of relief from 
valleys below 100 m to summits over 350 m. 
An interesting change in land use occurs 
from the relatively fertile alluvial 
valleys which support intensive use 
(orchards, arable, settlements., etc), to 
the upland areas characterised by 
extensive uses, such as rough grazing and 
forestry. Soils change from alluvium on 
the valley floor, through surface water 
gleys to brown earths on valley sides, 
with podsols occurring on the highest 
ground. The region therefore represents a 
relatively rich source of geographic 
variation and contains many identifiable 
sub-regions suitable for illustrating land 
resources evaluation by GIS. 
The area of approximately 25 sq km was 
chosen for compatibility with existing 
mapsheets, and so that the resolution 
after digitising would be sufficient for 
the type of queries received. The square 
region selected corresponded to O.S. 
mapsheet SK 26 SE at 1:10560 scale. This 
was one of the older O.S. maps produced in 
1971 so heights were recorded in feet. The 
resulting raster grid of 256 x 256 pixels 
lead to a ground resolution of about 20 x 
20 m, whilst the LANDSAT MSS data was 
resampled to give a ground resolution of 
80 x 80 m pixels. 
Data Capture 
Two different methods were used to build 
arrays of spatial data for quadtree 
encoding: 
1. Manual Digitising 
A GTCO digitiser linked to an IBM PC/XT 
was used to digitise selected features 
from the O.S. map at a scale of 1:10560. A 
series of functions for polygon to raster 
conversion form part of the GIS. These 
allow vector data to be mapped on to an 
array of a specified size. Functions for 
filling regions then produce binary array 
images. A set of 10 binary array images of 
size 256 x 256 were made for images of 
geology, soils, land use and climatic 
data. 
2. Video digitising. 
A video digitising system was used to 
produce 256 x 256 8 bit images of contours 
from the O.S. map. The contours were first 
traced onto transparent overlays and each 
contour interval was coloured by hand to 
produce a grey scale image. This was then 
captured by the video digitiser and frame 
grabber then transformed to fit a square 
256 x 256 array. The result was 10 binary 
array images of relief. 
All the binary array images were then 
quadtree encoded to produce a database for 
the area for testing various operations 
on quadtrees and land evaluation. 
RESULTS 
Figures 3-8 show selected results from a 
land resources evaluation for the Matlock 
area. Parameters considerd were geology, 
elevation, soils and land use. Although 
several more parameters are to be included 
into the database, this example indicates 
how a GIS approach could be used to 
identify areas most suitable for 
agriculture. 
Fig (3) shows those areas where there is 
grit parent material. This large area 
highlights the way in which the quadtree 
representation saves space by storing 
maximal blocks at different levels within 
the tree. The northern three quarters of 
Fig (3) is stored as only 6 quadtree nodes 
for the 6 maximal blocks. The complex 
boundary of the grit area requires 
considerably more nodes to be accurately 
represented. A characteristic feature of 
quadtree encoded images is the tendency 
for smaller blocks at the edges of 
regions, and larger ones towards the 
centre. 
Fig (4) shows land below 800 ft above sea 
level. This map was produced by the union 
of all height maps for intervals above 800 
ft, to produce the map 'height_above800'; 
this was then complemented to produce 
areas not above 800 ft. Following each 
union operation, maximal blocks are 
formed, since four small blocks with a 
common ancestor may be produced. Because 
the combined image often contains more 
space-saving larger blocks, the union 
image of Fig (4) need not necessarily 
contain more nodes than any single image 
from which it is derived. Fig (4) 
contains many larger blocks, showing how 
the quadtree stores large regions 
compactly, while being accurate enough to 
closely approximate the contours which 
bound the region. 
Fig (5) indicates areas of acid brown 
earth soils, which occur in two regions on 
the valley sides. Note that four smaller 
blocks are not always merged into a single