tT n €———
cations, assist the user with headings such as ‘terrain
analysis or neighborhood.
In an analysis of the categorization schemes used in 'Map
Algebra' and related classifications, Schenkelaars (1994)
scrutinized the kind of queries that are necessary to build a
spatial analytical query language. His approach can be used
to differentiate between a user's and a developer's view of
GIS functionality.
A large number of operations (such as sliver line removal or
coordinate thinning) do not make much sense to the average
user, or are regarded as a nuisance. Others, (such as 'line-of-
sight' and 'viewshed analysis') are either synonymous or at
least part of another and therefore confuse an occasional
user. This does not mean that those with more experience
should have no access to their functionality, but rather that
such operations are hidden from an entry-level menu and
that the system has default values for the results of each of
these operations.
Generalization is a task that is closely related to zoom' and
scale change operations. As such, they are auxiliary and
users may expect them to be performed automatically. This
is not a trivial requirement and needs further research
(Timpf and Frank 1995), however, the results should be
hidden from the user of a GIS. In a similar vein, abstraction
procedures, e.g. the reduction of an area to its centroid, or
the regionalization resulting from a Dirichlet
(Voronoi/Thiessen) tessellation , are tasks that keep being
mentioned in the GIS literature (Aronoff 1991, Burrough
1986, Egenhofer and Frank 1992, Laurini and Thompson
1992). However, they were never mentioned by a single user
of the survey. Based on the information, provided in section
1, it is fair to assume that the modularized future GIS will
provide more user-friendly data input and management fa-
cilities and thereby render these functions obsolete.
Operations such as 'clump/labeling' are relics of the underly-
ing data structure (i.e. raster-based tessellation's) and there-
fore need to be eliminated from the list of truly universal
GIS operations. Purely geometric operations like line inter-
section' or 'point-in-polygon' are typical for the way that GIS
functionality is currently implemented. These operations are
too far beyond the non-technical horizon of the average users
as to support them in solving tasks.
Table 1 represents a first approximation of what is left of
the myriads of operations of one eliminates all those that
have no direct analytical purpose. Auxiliary functions such
as 'clump' / "labeling in the raster domain or 'topology
building! in the vector world have been discarded, yet it 1s
exactly this group of operations that make up to 80% of all
GIS operations in a regular session (Yuan and Albrecht
1995).
A most critical case represents all those operations that can
be subsumed under interpolation and surface generation.
With 'Search' and its subsidiary (re-)classification', already
one functional group was introduced that has definitely no
analytic character, but is such an important predecessor to
all analytic operations that it needs to be included. Likewise,
interpolation is of crucial importance in some applications,
while users in other domains ardently expect a GIS to per-
form all necessary interpolations by itself.
A number of operations carry different names in various
domains, despite being essentially the same; 'cost, diffusion,
spread is an example for such polymorphism. Two of the
three 'Network' operations listed in Table 1 can be repre-
sented by other operations, i.e. the 'flow-between-regions is
easily implemented by the 'Thiessen/Voronoi operation,
while the 'shortest path' is a repeated 'nearest neighbor' op-
eration. Figure 1 represents a conclusive list of user-ori-
ented, analytical, universal GIS operations; mind though,
that this list is not meant to be the only one that could be
conceived of. Rather it mirrors the expectations of the ran-
domly selected group of people who answered the above
mentioned questionnaire (students in Austria and Germany
as well as colleagues at a number of international confer-
ences in 1993 and 1994; most of them being ARC/INFO?
users).
Table 1. A first compilation of user-oriented GIS operations
Search
Thematic Search
Search by Region
(Re-)Classification
Location Analysis
Buffer /Thiessen
Corridor
Overlay
Terrain Analysis
Slope and Aspect
Catchment/Basins
Drainage Network
Viewshed Analysis
Flow Analysis and Network
Connectivity
Shortest Path
Flow between Regions
Distribution / Network
Costs, Diffusion, Spread, Gravity Modeling
Change Detection
Proximity
Nearest Neighbor
Spatial Analysis / Statistics
Pattern
Centrality
Complexity/Variation
Dispersion Measures
Frequency
Indices of Similarity/Diversity
Topology, Hole Description
Topology; Upstream Elements
3lobal Surface Fit (Trend/Fourier)
Multivariate Analysis
Regression
Autocorrelation
Measurements
Number of Items
Distance
Direction (Calculate Bearing)
Perimeter, Acreage, Height, Volume
Surface
Shape
Fractal Dimension
Adjacency, Contiguity
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B2. Vienna 1996
Searc
Locat
Terra
Distri
Neigl
Spati:
Meas
A numb
cussion ¢
analytic
guable; 1
the field
makes a
separate
‘cost’ op
‘spread’
priate; al
In the fc
different
whereas
Search «
and sear
search c
analysis
analytic
the 'sear
in other
a user-d
shaped r
classific
however
tial detei
Algebra'
of meas
expresse
number
The Lo
tions the
operatio;
could be
takes tw
only be :
