The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008
210
uncertainty is quite natural part of our life and the surrounding
world and the fact of uncertainty is very stimulating for the
research on the field of defining, measuring, modelling and
visualizing. Moreover, the uncertainty opens the space for
further questions and the answers to this question can help us to
do better decisions [7].
Data are not perfect from many reasons: incomplete data,
precision of measurements, discreet description of connective
phenomena or inherent part reflecting our understanding of
things [11]. To reduce uncertainty of data it is mainly the
question of the proof of recognized quality assurance. Users
often take the pragmatic approach to the cost versus accuracy.
Sometimes, without the relevance testing, the resolution of data
is used for the whole set of different task. Then the problem of
over-defined and under-defined objects brings the difficulties
[8].
2.2 Uncertainty Modelling
Especially uncertainty of a geographic object can be modelled
through uncertainty of its geo-spatial, temporal and thematic
attributes. Uncertainty of relations takes into consideration
spatial, temporal and spatial-temporal relations. To add suitable
attribute or to spread the net of relations reduce the uncertainty
of the object.
The special case is to model objects uncertainty using spatial-
temporal approach to the objects and incorporate spatial-
temporal relationships. The dynamics of object is very powerful
tool to obtain exact results about the object and phenomenon
behaviour to support further decision [16].
The decision making process is always associated with some
level of uncertainty which can rise from: definition of the
problem, data used, sequence of operations used to obtain result
and understanding of result.
In case when the concept is not precisely defined, the
characteristics are not precisely measured and insufficient
estimation of parameters is in disposal than the techniques of
fuzzy logic may be applicable. This way allows computing and
evaluating results when the object or phenomenon state can be
sufficiently described using linguistic terms that can be
consequently quantified in some fashion.
Temporal aspects
GIS packages mostly lack the ability to perform temporal
analysis of spatial data in general. The temporal analysis is
applied for selected applications with specially arranged
conditions. The fact is that decision support procedure is highly
dependent on the ability to represent and analyse temporal
processes in a GIS. Moreover, the widespread application of
GIS in urban, regional and environmental planning, in
marketing and logistics requires the integration of temporal data
in order to perform analyses or to deal with temporally variable
data.
Another motivation for the development of temporal GIS
techniques is that the great amounts of data have been collected
the research projects have been solved and important results
have been achieved, which cannot be revaluated and further
reused. The development of appropriate methods and tools as
well as their implementation is affected by the lack of standards
for the temporal data representation and analysis. It means the
set of tasks where no single criterion is sufficient to arrive at a
decision with the comfortable level of certainty.
2.3 Multi-criteria decision problems
In many cases only the combination of many information
sources and eventually to combine expert knowledge can
produce requested information. Very similar situation except
GIS is also in medical tasks connected with diagnosis
determination.
Parameters estimation
The use of modelling functions strongly depends up exact
definition of parameters as follows:
• specification of one or more target locations,
• specification of the neighbourhood that surrounds the
targets,
• specification of the way spatial elements are
interconnected,
• set of rules that specify the allowed movement along
these interconnections,
• specification of a function to be performed on the
"things" found surrounding the targets in the specified
neighbourhoods,
• set of resources (such as a finished manufactured
product),
• one or more locations where the resources reside
(warehouse or factory location),
• objective to deliver the resources to a set of destinations
(customers),
• set of constraints that places limits on how the objective
can be met (speed of travel, time spent delivering the
products, etc.)
The different requirements can be described by the set of factors
and coefficients, but these factors are often connected to the
critical characteristics coming from the selected area and
surrounding objects that can influence the estimation quality in
the frame of classification procedure. It is possible to define
priority function PF with the aid of priority functions of the
selected set of factors and weighting coefficients as follows:
n
F = '^ j W i fi, where = 1 (1)
/ = 1 i
The great part of parameters can be successfully put more
precisely with the aid of expert knowledge. GIS applications are
frequently used in producing new information by combining
information from different sources, by spatial analysis of
existing data and by implementation of additional information
coming from previously processing and analysis, expert
knowledge, objects dynamics and trends.
Usually the objective in applications involving contextual
modelling is to locate the area or areas where the given criteria
apply and eventually calculate the measure of exposure to
hazard in case of infections, diseases and pests, find the optimal
routes and produce different complex scenarios. The powerful
tool is the way of buffering where the expert knowledge can
help us to set the ranges and find the areas with defined ways of
protection.
The evaluation procedure consists of the two steps: to set up
parameters and determine their importance at first and provide
the sensitivity analysis to demonstrate the effect of selected
