A CES
single source location or the complete stream network. This
includes the delineation of stream links, stream order, flow di-
rection and upstream elements. The ‘viewshed' operation is the
only other one that requires an input file containing height val-
ues. In addition to that a view point must be designated (this
could also be a route or an area) and the viewer's height above
ground needs to be specified. The search distance is an op-
tional parameter.
The "Distribution/Neighborhood' group of operations is
probably the most geographic of all. Non-statistical queries
about the relationship between spatial features are usually
answered with this set of functionalities. The
'cost/diffusion/spread' operation takes one arbitrarily dimen-
sioned feature and calculates the value of neighboring attrib-
utes according to some spread function. The spread can be in-
fluenced by barriers which simulate spatial impedance (the
'cost' character), accessibility, or the relative distance under
anisotropic conditions. The spreading function is usually ex-
pressed by an equation while the friction can be represented
either by an equation or by a special friction coverage. The
'shortest path' functionality described in the 'Network' section
above, can also be implemented by a 'spread' along a given
network. Especially, if the shortest path calculation is to be
based on relative distances, 'cost/diffusion/spread' might be
more appropriate than the ‘nearest neighbor’ operation.
'Proximity' is less of a singular operation than a functional
group of numerous technical operations that carry out the same
functionality. Proximity measures can be applied to all features
of an input file or to selected only. In case of multi-dimensional
features, the user needs to specify whether it should be meas-
ured from edge to edge or from center to center. Finally, a
maximal distance may be specified for what is considered to be
proximal.
Similarly, the nearest-neighbor’ operation uses a number of
different algorithms, depending on the mode, which usually is
(but does not have to be) conditional to the input data. Aside
from the common specification of input and output files, (the
input can be one or several features of any type), this operation
needs particulars about the unit of measurement (i.e. length or
number of nodes) and the mode (e.g. along a path or as-the-
crow-flies).
It could be argued that ‘proximity’ belongs into the Measure-
ment' group, while nearest-neighbor’ is a special case of the
'cost/diffusion/spread' operation, which in itself is nothing but
a complex "reclassification' This would render the whole
group obsolete. From a technical point of view, this argument
is valid, however, it does not correspond with the requirements
on the user's side and is therefore not adhered to here.
All statistical measures, that inhibit a certain degree of com-
plexity, are categorized as Spatial Analysis ' 'This includes the
landscape ecological ‘pattern and dispersion' measures, (such
as frequency, indices of similarity, relative richness, diversity,
dominance, fragmentation, density, Shannon index, and de-
grees of freedom) as well as ‘centrality or connectedness’,
‘shape measures’ (€.g. skewness, compactness), and the whole
set of tools for "multivariate analysis’. They all result in singu-
lar figures, which is why they could arguably be categorized as
Measurements’. Some of the computations, however, are so
complex that users would be confused if they were grouped
among measures like perimeter’ or ‘acreage’.
82
'Pattern' and 'dispersion' measures are possibly the most proto-
typical of all ‘Spatial Analysis’ operations, at least with respect
to descriptive statistics. ‘Centrality’ gives either the center ofa
point cluster or a measure of connectivity in a network. ‘Shape’
measures are used in a wide array of applications, e.g. in geo-
morphological, bio-geographical, political ("gerrymandering"),
or archeological practices. A number of basic parameters that
can be found in the Measurements’ group (‘acreage’, perime-
ter', 'centroid', etc.), are used here to describe elongation, ori-
entation, compactness, puncturedness or fragmentation.
The last operation in the ‘Spatial Analysis’ group is again a
header for a whole bag of secondary operations. Multivariate
analysis’ is comprised of a number of techniques to describe
the relationships and dependencies among the spatial objects in
scrutiny. Although these are definitely analytic in character, it
can be argued that their functionality is covered by such well-
established statistical software packages as SAS, SPSS or S-
plus and therefore, do not need to be classified as a universal
GIS operation. In the other hand, operations like regression’,
autocorrelation’ and 'cross-tabulation' are so often used in a
GIS context, that they are included here as well.
The Measurements’ group is virtually infinite. In it's core, it
consists of a number of simple geometric calculations
(‘distance', ‘direction’, ‘perimeter’, ‘acreage’, ’height', volume’,
‘surface’, fractal dimension’), these are then extended by sim-
plest statistics ('number', ‘histogram’, 'mean^, and finish with a
few topological measures, such as ‘adjacency’ and 'dough-
nuts/holes".
3 THE VGIS CONCEPT
VGIS is not intended to be yet another GIS. It is conceived as a
shell that can be draped over an existing GIS, using the func-
tion offered by that GIS (see Figure 2). The current implemen-
tation requires one interpreter for each existing GIS, but the
prospect of vendors adhering to the OGF's Open Geodata
Model alleviates this necessity.
The VGIS-Shell consists of four modules:
e the graphical user interface (GUI)
ean interpreter for the processing plans
ea tool to generate the processing plans
e the underlying of-the-shelf GIS
2.1 The Graphical User Interface
Compared to regular GIS user interfaces there are a number of
significantly different features that cause the innovative user
friendliness of VGIS:
e Complete reconfiguring of the user interface
The GIS manager has complete freedom to adjust the
user interface to the needs of his customers. Only those
operations that are useful to solve the users' task will be
offered. Their structure and presentation in sub-menus
are also fully configurable.
e Powerful, complex operations
The operations offered to the user are no isomorphic
mapping of the operations of the underlying GIS. Rather
they are custom-made to each application. These opera-
tions are independent of whatever data structure is used.
Necessary transformations are initiated by VGIS and per-
formed by the underlying GIS.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B2. Vienna 1996
