358
M.A. Chapman et al.
that this detail will constitute a complete data quality report that gives sufficient
information to make users able to assess the data’s fitness for use.
Different aspects of the source data and their respective applications are
important for communicating error information. The data lineage describes source and
update material (with dates), methods of derivation, transformations, and other
elements relating to the processing history. Positional accuracy deals with how closely
the locational data (encoded coordinate values) represent true locations. Attribute
accuracy is similarly concerned with non-locational descriptive data. Logical
consistency refers to the fidelity of encoded relationships in the structure of spatial data
(e.g. where appropriate, the degree to which topological relationships have been
verified should be reported). The completeness part of the quality report includes
information about the geographic area and subject matter coverage (e.g., selection
criteria and other mapping rules that are relevant completeness indicators).
Indeed, uncertainty can be conveyed in different formats such as statistical
indicators and visual aids. Some useful mechanisms for displaying this information are
given in subsequent sections.
ERROR ABSORPTION
The next stage of the decision strategy deals with deciding whether product quality
satisfies the requirements of the current task and recognizing what can and cannot be
modified. Indeed, if the quality is found to be acceptable, then the issue of uncertainty
has effectively been dealt with. However if the quality is not acceptable, then further
action is required. Regardless of the efforts put on the error reduction techniques, it
should be realized that there will always remain some residual uncertainty which users
must decide to either absorb (i.e. accept) if they wish to use the data, or else reject it.
This stems from the fact that all measurements are of limited accuracy and no model
will ever perfectly reflect the real world. The amount of uncertainty absorbed can be
considered to be the risk associated with using the data or product.
DATA VERSIONS AND ERROR REDUCTION
The purpose of this stage is to decrease the magnitude of uncertainty to an acceptable
level. If the uncertainty of the product does not meet the application requirements,
versions of data that may satisfy the objectives should be tested. Different versions of
data may already be created using techniques such as image pyramids or multiresolution methods, and they are stored in the database. Since a top-to-bottom
approach is the usual theme in GIS - users look at the global picture first then zoom to
view finer detail - to find the specific data needed, accurate data can be found in the
lower level of the pyramid. However if the object shows fuzzy characteristics, upper
level data is needed.
If no version of data meets the users need (i.e. the user reaches the lowest
level of the pyramid) then error reduction techniques should be applied. For example,
in the case of parcel-based systems, Bedard (1987, p. 181) recognized that actions such
M.A. Chapman et al.
that this detail will constitute a complete data quality report that gives sufficient
information to make users able to assess the data’s fitness for use.
Different aspects of the source data and their respective applications are
important for communicating error information. The data lineage describes source and
update material (with dates), methods of derivation, transformations, and other
elements relating to the processing history. Positional accuracy deals with how closely
the locational data (encoded coordinate values) represent true locations. Attribute
accuracy is similarly concerned with non-locational descriptive data. Logical
consistency refers to the fidelity of encoded relationships in the structure of spatial data
(e.g. where appropriate, the degree to which topological relationships have been
verified should be reported). The completeness part of the quality report includes
information about the geographic area and subject matter coverage (e.g., selection
criteria and other mapping rules that are relevant completeness indicators).
Indeed, uncertainty can be conveyed in different formats such as statistical
indicators and visual aids. Some useful mechanisms for displaying this information are
given in subsequent sections.
ERROR ABSORPTION
The next stage of the decision strategy deals with deciding whether product quality
satisfies the requirements of the current task and recognizing what can and cannot be
modified. Indeed, if the quality is found to be acceptable, then the issue of uncertainty
has effectively been dealt with. However if the quality is not acceptable, then further
action is required. Regardless of the efforts put on the error reduction techniques, it
should be realized that there will always remain some residual uncertainty which users
must decide to either absorb (i.e. accept) if they wish to use the data, or else reject it.
This stems from the fact that all measurements are of limited accuracy and no model
will ever perfectly reflect the real world. The amount of uncertainty absorbed can be
considered to be the risk associated with using the data or product.
DATA VERSIONS AND ERROR REDUCTION
The purpose of this stage is to decrease the magnitude of uncertainty to an acceptable
level. If the uncertainty of the product does not meet the application requirements,
versions of data that may satisfy the objectives should be tested. Different versions of
data may already be created using techniques such as image pyramids or multiresolution methods, and they are stored in the database. Since a top-to-bottom
approach is the usual theme in GIS - users look at the global picture first then zoom to
view finer detail - to find the specific data needed, accurate data can be found in the
lower level of the pyramid. However if the object shows fuzzy characteristics, upper
level data is needed.
If no version of data meets the users need (i.e. the user reaches the lowest
level of the pyramid) then error reduction techniques should be applied. For example,
in the case of parcel-based systems, Bedard (1987, p. 181) recognized that actions such