l were used as 
lie significant 
at sequential 
regular basis. 
or) 
ollected in a 
bout storage, 
te data often 
versus vector 
lescribing the 
atial entities 
ordinates, or 
lents have been 
:er and vector 
its and format 
ae system user 
;lf (Haralick. 
appropriate point, line or areal symbolization. 
These data include the basic biological and physical 
land characteristics, man made features and land 
use. Certain social and economic information is 
also included, as it is often useful for planning 
and managem^n*- of the land base. This information 
usually hat, a geographic basis at both the 
collection and utilization stage. The phenomena 
that secondary data represent are often more dynamic 
than those of primary data. As a result of these 
changes periodic assessments for currency are 
required 
The third type (DEM) could be classed as primary 
data but it has some unique storage, retrieval and 
utilization characteristics that require special 
handling in a database. Therefore, it has to be 
--‘••'Mlshed as a separate data type. 
3.3 Digital Elevation Model (DEM) Data 
DEM data can be collected in one of three ways: 1) 
field observations, 2) photogrammetrically (tradi 
tional airphotos), or 3) utilizing digital stereo 
correlation of remotely sensed imagery. For the 
prototype LRIS, the DEM data were derived by the 
second and third methods. A digital elevation model 
was produced from 1:40 000 scale photography of a 
test area on a Wild AC-1 analytical plotter. An 
area covering approximately 5.6 by 5.6 km was 
chosen, and elevations were measured on the same 50 
m UTM grid that the Landsat data utilized. The area 
chosen has a good range of terrain types, with 
elevations ranging from 1650 m to 2550 m, and 
various surface cover conditions for testing the 
algorithms. There is a rugged area of mountainous 
terrain, some flat and rolling terrain and two 
lakes. 
described by a 
also described 
(Peucker and 
1 points are 
ates (relative 
r by relative 
>oint) (Baxter, 
les and points 
of an overlain 
s, which cover 
) may be full, 
ed, or sparse, 
scan line are 
). The usual 
ative merits of 
;ctors are more 
and rasters are 
n time (Barber, 
;r format is an 
dealing with an 
d lines are to 
an important 
considerations 
3) (Mepham and 
te sensing data 
d to a level of 
more efficient 
tering process 
al filter which 
Mepham, 1986). 
ted by polygons 
ique points. 
totype LRIS was 
broad range of 
ree basic data 
a (positional), 
id 3) digital 
andard and have 
elated informa- 
: 1) geographic 
th a range of 
ams established 
land registries 
les information 
can loosely be 
1980). It is 
a map with the 
' r -’hle 1. Data types for input into the LRIS. 
Classes 
Subclasses 
Source 
Survey 
Contri 
(Primary; 
Cadastral, 
Boundaries, 
Control Data 
digitized 1:50 000 NTS 
digitized RT series 
digitized 1:50 000 NTS 
Hydrography 
(Secondary) 
Lakes, Rivers, 
Swamps, Ice 
digitized 1:50 000 NTS 
Geology 
'’Secondary) 
Surficial, 
Bedrock 
digitized 1:50 000 RB 
digitized 1:1000 000GSC 
Transport 
ation 
(Secondary) 
Roads, 
Powerlines, 
Pipelines 
digitized 1:50 000 NTS 
Surface 
Cover 
(Secondary) 
Forest, Grass, 
Rock/Soil, 
Farm, Urban 
derived from digital 
Landsat data 
Digital 
Elevation 
Models 
(DEM) 
Airphoto, 
Digital Stereo 
1:40 000 airphotos 
psuedostereo test 
data 
With these differences in mind there were six 
classes of the different data types selected for 
processing in the study (Table 1). These classes 
have many of the varied subclass types found in land 
information. The data sources are also varied, with 
some data collected by digitizing existing maps, 
some derived from digital remote sensing imagery, 
and some from published descriptions. 
3.2 Primary and Secondary Data 
Survey control, hydrography, geology and 
transportation data were digitized directly from 
maps. The existing map data were digitized with a 
six parameter transformation to transfer the 
positional data to a UTM coordinate system. There 
is a header for each file, which contains the 
mounting information, composed of the control points 
utilized and their input coordinates and adjusted 
residuals. There is also a record containing the 
adjusted scale factors, offsets, rotation and 
non-perpendicularity parameters. The data files 
follow and contain an ID code, a character string 
description and the list of point coordinates 
describing the feature. The digitized features were 
displayed on a vector graphics screen as they were 
collected to allow verification of the data. Any 
blunders or corrections were flagged at the time of 
data capture and the data files were later edited to 
remove the errors. 
These traditionally derived data were then used to 
generate a test digital stereo data set to evaluate 
the third possible digital elevation model input 
method. The Landsat image was assumed to be an 
orthoimage and the grid reference systems for both 
the image and the digital elevation model were 
assumed to be the same. A parallactic angle of 30 
degrees was used, as this produces a significant 
elevation change from a horizontal displacement. 
This angle and the flying height of the satellite 
produce a typical rate change for all the X, Y, Z 
coordinates (ie a 50 m change in X produces a 50 m 
change in Z). Various methods of digital 
correlation and processing were evaluated and the 
resulting digital elevation models were compared to 
the airphoto derived model. The results indicate 
accuracies of subpixel range and suggest this as a 
viable method of obtaining DEM data from satellite 
systems, such as SPOT, which have stereo imagery and 
appropriate resolution. 
There are two methods for the storage and 
representation of elevation data. The data can be 
stored in the format in which it is collected, which 
will typically be either in the form of a regular 
grid of spot heights or an irregular grid of spot 
heights (a triangulated irregular network). It can 
also be stored in the form of contours derived from 
the collected data. While contours are simpler to 
interpret visually, they are also less accurate than 
the original data. In addition, the storage of 
contour data will generally require much more disk 
space than the storage of the original digital 
terrain model. This is particularly true in areas 
of relatively smooth (not flat) terrain, where only 
a limited number of spot heights is required to 
describe the surface. Also, the one advantage that 
contours have over spot heights (their simple visual 
interpretation) is rapidly decreasing in importance 
due to the increasing use of digital processing over 
manual processing and interpretation. 
4. DATA STORAGE 
4.1 Conceptual Design 
The Kananaskis database is being designed as a 
prototype of a database suitable for much larger 
application, such as the entire province of Alberta, 
Canada. With this in mind, some initial design 
criteria were established to guide its development. 
The first criterion examined was the question of 
response time (speed). It was decided that 
"ordinary" requests must be answered quickly in an 
online system. Such requests might be "Find all 
occupied dwellings within 5 km of a certain gas 
well" or "Find all forest that is primarily pine and 
is on property owned or leased by the Cut-Em Forest 
Company". In addition, there will also be other 
"strange" requests, such as "Find all lots larger