phical
sofa
icture-
neces-
posed
1 Ltd.
rt and
results
5 The
IS pos-
(to be
ores of
indow-
ind the
not af-
iccount
project
> front-
g envi-
1el der
1 somit
lierung
en wer-
urunab-
isch ge-
eidung-
n sind,
ind das
bald das
sch alle
ung der
n VGIS
kelt, die
s relativ
g ist ein
gserfah-
Itmodel-
rences
las
1 INTRODUCTION
The handling of geographical information systems (GIS) be-
comes increasingly difficult for the user. The complexity and
sheer amount of operations offered 1s growing faster than the
quality of user interfaces which are by far less comfortable
than has become standard for other software packages such
as word processors or spreadsheets. Many GIS still do not
have an adequate graphical user interface, often commands
still have to be entered at the systems level with a large
number of parameters. An additional problem is the combi-
nation of raster and vector data. Many users are over-
whelmed by the need to understand the principles underly-
ing transformations between different data models.
Current solutions to these problems focus on fancy icon-
based graphical user interfaces that depict the cryptic pro-
prietary commands. While the GIS modules usually retlect
the task flow from data input, via storage, manipulation,
analysis to data output, the individual operations are data-
centered. It is here that the project "Virtual Geographical In-
formation System" (VGIS) derives its justification. The pur-
pose of VGIS is to create a universal graphical user interface
(GUI) that can be used with any current GIS. VGIS' graphi-
cal front-end builds upon a limited set of universal data
structure-independent analytical GIS operations that allow
the user to build processing models — transformations be-
tween different data structures are to be executed automati-
cally. The graphical user interface works like a general pur-
pose flow charting tool that depicts the workflow consisting
of a number of processing steps to be applied to the input
data. This processing plan can be created and edited inter-
actively using a mouse. The plans can be saved just like
macros, to be used later on with other data. The visual pro-
gramming character facilitates easy adaptations to changing
needs. A first implementation of VGIS runs on the public
domain GIS GRASS™ coordinated by OGI at USACERL.
Current development concentrates on writing an interpreter
for the vector GIS ARC/INFO® from ESRI, Redlands, (CA),
while the basic research focuses on a more universal inter-
face to the Open Geodata Interoperability Specification
(OGF 1993).
The theoretical background for the universal analytical GIS
operations presented in section 2 discusses issues of data
model independence, and the interfacing with the Open GIS
specification. Section 3 introduces the general VGIS concept
while the more technical implementation aspects are given
in section 4. The second part. First applications of process-
ing models are presented in the 5th section with a practical
example given in section 6. Section 7, finally, provides an
outlook on the direction of future research conducted at the
Institute for Spatial Analysis and Planning (ISPA) of the
University of Vechta.
2 THE UNIVERSAL GIS OPERATIONS
Almost all GIS-related research is data-centered. Even hu-
man factor studies usually take the dichotomy of raster-
versus vector-GIS for granted and analyze GIS usage either
on a key-stroke or an a task level (Turk 1992, Medyckyj-
Scott and Hearnshaw 1993). In neither case the original pur-
pose for using a GIS in the first place is taken into account.
If a GIS is employed for assignments that go beyond mere
data repository chores, then the operations on data become
at least as important as the data themselves. Investigation of
these operations in a processing model context has been ne-
79
glected until very recently (Albrecht 1994, Voisard 1995,
Hamilton and Worboys 1996).
Many GIS users possess expert-level knowledge in the ap-
plication field in which the GIS is to be utilized, but have
neither the time nor desire to learn the technical intricacies
of a specific system (Albrecht 1994). The user's overall goal
should not be the mastery of a new system but more produc-
tive interaction with geographic information. An obvious re-
sponse is to provide a user interface which alleviates the
need for specialized training. This user interface should aim
at enhancing user interaction with geographic information
and with geographic problem solving rather than with sys-
tems. Much of the user interface problem is therefore not a
programming problem but a conceptual problem (Mark and
Gould 1991). Frank (1993) illustrates the crucial importance
of the user interface for the usability of a GIS with the fol-
lowing lines: "The user interface is the part of the system
with which the user interacts. It is the only part directly seen
and thus 'is' the system for the user."
As knowledge has been gained on the behavior of spatial al-
gorithms, a functional categorization has emerged (Burrough
1986. Tomlin 1990). However, it gave way to the basic ar-
chitecture of GIS that every student learns (needs to learn)
in the first introductionary course to GIS: data input, data-
base management, analysis and output, hereby violating the
functional continuation concept. Within each category, dif-
ferent means have been developed to handle the user inter-
face. Menu and command driven operations have become
the main way of interaction. Some are action driven
(producing immediate feedback) while others are environ-
ment driven, having a cumulative effect that ends in an ac-
tion driven function. A functional continuos GIS would be
mostly environment driven.
Modern GIS software is a multidisciplinary tool that must
allow for interdisciplinary support and is expected to be able
to integrate a variety of different data sources. These data
sources will be used in many ways and under a wide range
of decision support situations. To meet these demands, a
user interface is required whose generic functional model
consists of a small set of universal GIS operations that allow
for the automatic construction of a domain or task specific
derived model. It would act as a shell based on a high level
language consisting of spatial operators that have definable
hierarchical constructs. These spatial operators can be or-
ganized following a programmable schema that allow them
to generate the derived model. The core of such a shell,
however, would be the generic functional model. Section 3
will introduce the Virtual GIS (VGIS) project which aims at
implementing the above mentioned ideas.
The VGIS front-end to GRASS™ builds upon the 20 ana-
lytical GIS operations presented in (Albrecht 1996) and
summarized in the following. Based on a questionnaire
(Albrecht 1995) drawn from GIS operation taxonomies in
the relevant literature (Aronoff 1991, Burrough 1992, Good-
child 1992, de Man 1988, Rhind and Green 1988, Unwin
1990) the functionality of GIS is analyzed from a user's pint
of view.
The analysis of current user interfaces provides a good op-
portunity to study different approaches to the categorization
of GIS functionality. Again, most operations serve non-ana-
lytical means and are therefore not of concern to this study.
All systems offering special operations for particular appli-
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B2. Vienna 1996
