DCW Data Quality 1992

1
Research Note
Data Quality
Data Quality
Statements For
Statements For
Spatial
Spatial
Databases
Databases1
R.J.Williams
Information Technology Division
Electronics Research Laboratory
Defence Science and Technology Organisation
July 1992
APPROVED FOR PUBLIC RELEASE
This paper overviews contemporary issues in incorporating data quality statements into
spatial databases. The paper includes discussion of two approaches; one emanating from the
Digital Chart of the World Project and one through a working group within the International
Cartographic association.
1 The text of this paper has not been edited at all from the original research note. However, some of the
illustrations have been redrawn; in some cases to include contemporary issues.

1
INTRODUCTION
Digital processing of spatial data brings immense benefits in the form of rapid, precise and
sophisticated analysis, but reveals weaknesses which may not otherwise be apparent.
Computers are very precise machines, and errors and uncertainties in data can lead to
serious problems, not only in the form of inaccurate results but in the consequences of
decisions made on the basis of poor data. Capabilities that excite enthusiasm among
potential users are the ability to change scale and the ability to overlay different themes of
information at random. These capabilities are indeed exceedingly useful; they constitute
much of the comparative advantage geographic information system technology (commonly
referred to as GIS) holds over spatial analysis based on analogue maps (Goodchild, 1991;
Abler, 1987).
These capabilities, however, can also mislead decision makers who are unaware of the
imprecision inherent in all cartography and who are untutored in the ways errors compound
when map scales are changed or when maps are merged. Burrough (1986) observes "a false
lure in the attractive, high quality cartographic products that cartographers, and now
computer graphics specialists, provide for their colleagues in environmental survey and
resource analysis. ... Many scientists and geographers know from field experience that
carefully drawn boundaries and contour lines on maps are elegant misrepresentations of
changes that are often gradual, vague or fuzzy".
Goodchild (1991) warns that "if the burgeoning GIS industry is indeed driven by false
perceptions of data accuracy, then the truth will be devastating: even the simplest products
will be suspect. The best insurance at this point is surely to sensitise the GIS user to the
accuracy issue, and to develop tools which allow spatial data handling systems to be used in
ways which are sensitive to error". That is, systems that use digital geographic information
require a method to maintain and manage their contents and processes over the long term.
Up until just a few years ago, the description of data quality and associated issues have been
neglected topics. Fortunately, however, the topic is now being recognised as one of
importance and the issue of the description of data quality is being addressed by a number of
research organisations and professional bodies throughout the world. The catalyst for this
work is because of incomplete coverage, variable accuracy, inconsistencies in standards and
inadequate sources.
Two approaches are worthy of assessment. One approach emanates from the Digital Chart
of the World (DCW) project while the other emanates from the Scientific Advisory Board of
the International Cartographic Association.
Digital Chart of the World (DCW)
The DCW Project is a United States Defense Mapping Agency research and development
effort (to which Australia, via the Royal Australian Survey Corps, is a cooperative partner),
whose ultimate objective is the promulgation of standards for the exchange of digital spatial
information and the development and distribution of a global topographic database on
compact disk (CD-ROM) (DMA, 1991).
DCW will be a new product of the Defense Mapping Agency (DMA). It will provide worldwide
coverage using a topologically based vector data structure to digitally represent the earth's
land surface information on a microcomputer accessible storage media. The 1:1000000 scale
Operational Navigation Chart (ONC) series will provide the majority of the information to
produce the DCW. The Jet Navigation Chart (JNC) series will provide the information over
the Antarctica. Features will be collected and stored along with their attributes at the level of
detail provided on the ONCs.
The purpose of the project is twofold:
1. To develop, refine, and establish a suite of standards that enable the exchange
and utility of spatial information; and

2
2. To perform the necessary research and development steps to produce the DCW
in compliance with these standards.
In order to insure the suite of standards will be compatible with the international community,
as well as the US Department of Defense; allied partners, namely Canada, United Kingdom
and Australia, are participants in the overall research and development.
Standards to be developed for the DCW include format standards, media standards, a DCW
product specification, and a data directory standard to include tiling, coverage index, thematic
index, gazetteer index, and spatial query index.
The data structure used for DCW is a topologically structured vector structure in a relational
model and is known as Vector Product Format (VPF). VPF also contains data quality
information so that users may evaluate the utility of the data for a particular application.
International Cartographic Association initiative
Technological issues such as those concerned with digital data quality are also receiving
attention from working groups within professional organisations. Perhaps the lead
professional body in the disciplinary area concerned with spatial data is the International
Cartographic Association (ICA). The ICA has within its organisational structure a number of
commissions and working groups whose terms of reference, amongst other things, includes
"undertaking efforts on critical topics of research". The Scientific Advisory Board of the
International Cartographic Association has produced a set of guidelines as its contribution to
a clear and consistent approach to the assessment of data quality. These are presented in
ICA Data Quality Proposal.
VPF DATA QUALITY STATEMENT
The data structure used for DCW is a topologically structured vector structure in a relational
model and is known as Vector Product Format (VPF). VPF is a generic geographic data
model designed to be used with any digital geographical data in vector format that can be
represented using nodes, edges, and faces. VPF is based upon the georelational model,
combinatorial topology and set theory. VPF also contains data quality information so that
users may evaluate the utility of the data for a particular application.
VPF contains data quality information at a number of different levels within the database with
the detailed description being modified from the Spatial Data Quality section (Section 4)
of NCDCDS Report #7 (Moellering, 1986).
Data quality hierarchy
The VPF model is a hierarchical one with information held at database, library, coverage,
feature and primitive levels. Data quality information at the database level applies to all
libraries of the database, except where those libraries contain their own data quality
information of the same kind. Similarly, data quality information at the library level (which may
have been inherited from the database) applies to all coverages within the library, except
those that contain their own data quality information of the same sort. Coverage level data
quality information applies in the same manner to features. Feature level data quality
information in turn likewise applies to both spatial primitives and attributes that compose
them.
Data quality encoding
Data quality information is represented as attributes or as a coverage. If as attributes, it may
be either added to an existing VPF table, or as an independent table residing at the
appropriate level. If a coverage, it shall be a coverage whose area or complex features

3
designate areas with uniform data quality information of specified types. Figure 1 depicts the
attribute and coverage locations of data quality information through the database.
Types of data quality information
There are seven types of data quality information:
1. Source. Source describes the origin or derivation of a single feature, primitive or
attribute. This includes any processing techniques applied to the data, as well as
the data source.
2. Positional accuracy. Positional accuracy provides an upper bound on the
deviation of coordinates in VPF from the position of the real world entity being
modelled. Positional accuracy must be specified without relation to scale and
shall contain all errors introduced by source documents, data capture, and
processing.
3. Attribute accuracy. Attribute accuracy describes the accuracy or reliability of
attribute data.
4. Currency. Currency represents the date at which the data was introduced or
modified in the database. This date of entry is used as a proof of modification for
a single data element, permitting statistical interpretation of groups of data
elements.
5. Logical consistency. Logical consistency describes the fidelity of relationships
encoded in a VPF data set. Logical consistency requires that all topological
foreign keys match the appropriate primitive, that all attribute foreign keys match
the appropriate primitives or features, and that all tables described in feature
class scheme tables do indeed have the relationships described.
6. Feature completeness. Feature completeness indicates the degree to which all
features of a type for the area of the data set have been included.
7. Attribute completeness. Attribute completeness indicates the degree to which
all attributes of a feature have been included for that feature. Actually, since this
information can be derived from the feature itself, simply by counting null values,
this particular form of data quality information should not need to be explicitly
included.
Level Quality Attributes Quality Coverages
Database In a table within the database directory Within database directory
Library In a table within library directory, or
within the library attribute table or a table
related to it
Within the library directory
Coverage In a table within the library directory, or
within the library attribute table or a table
related to it
Within the same library directory as
the coverage to which the quality
information applies
Feature In a table within the coverage directory,
or within the feature’s feature table
Not applicable
Primitive In a table within the coverage directory,
or within the feature’s feature table
Not applicable
Figure 1: VPF Data Quality Information
These types of information above are VPF's standard types of quality data. Product
specifications, such as the Digital Chart of the World, call for additional types of data quality
information as well.

4
DCW metadata
The DCW is one database with two libraries. The database level includes three tables: a
database header table, a database description table and a library description table. The
database header table contains metadata pertaining to the DCW data and includes
information on security and release information.
The DCW library is a directory containing VPF tables, coverages and index tables. One table,
known as the library header table, identifies the data set, sources, extent, projection, security,
and data quality information in the library (Figure 2).
As the Digital Chart of the World is available for public release from February 1992, the
schema will be the first containing a 'data quality statement' that will be supported as a
'standard'. Therefore, future defence data (in vector format) should include, as a minimum,
that information as shown in Figure 22.
It seems unfortunate that, although 'data quality statements' have been identified as being
important, the implementation in VPF (and therefore in DCW) is somewhat simplistic and
poorly described in accompanying documentation. This component of VPF (and DCW) is
clearly one needing further development and enhancement.
VPF Column Name DCW Column Name Record Entry
Undefined ID 1
Product Type PRODUCT_TYPE DCW
Name LIBRARY_NAME DCW
Data Structure Code DATA_STRUCT_CODE 1, 2, and 6
Series SOURCE_SERIES ONC
Source Identification SOURCE_ID Complete ONC Series
Edition SOURCE_EDITION Varies with source map
sheet
Source Name SOURCE_NAME Operational Nav Charts, Jet
Nav Charts
Source Date SOURCE_DATE 1989
Ellipsoid Name ELLIPSOID_NAME WGS
Ellipsoid Code ELLIPSOID_CODE None
Vertical Reference Name VERT_REF_NAME Mean Sea Level
Vertical Reference Code VERT_REF_CODE MSL
Vertical Datum Code VERT_DATUM_CODE Unknown
Geodetic Datum Name GEOD_DATUM_NAME Unknown
Geodetic Datum Code GEOD_DATUM_CODE Unknown
Longitude of SW Corner LON_SW_MBR 0 Longitude
Latitude of SW Corner LAT_SW_MBR 90 South Latitude
Longitude of NE Corner LON_NE_MBR 0 Longitude
Latitude of NE Corner LAT_NE_MBR 90 North Latitude
Longitude LON_BOUND_FACE +- 180 degrees
Latitude LAT_BOUND_FACE +- 90 degrees
Projection Name PROJECTION_NAME Decimal degrees
(Unprojected)
Projection Code PROJECTION_CODE Unknown
Security Classification SECURITY_CLASS U
Downgrading DOWNGRADING No
2 It is apparent that the developers of the 'data quality' module of DCW lacked experience
and/or knowledge in cartography and surveying. In the draft documentation there are errors and
uncertainties. Firstly, the projection is noted to be 'Unprojected' with decimal degrees but horizontal
unit of measure is given as 'Meters' (possibly should be expressed in arc units). Secondly, the vertical
unit is expressed as 'Meters' but the source material was an aeronautical chart with elevation in 'Feet'.
The absolute horizontal accuracy was given as +-2040 meters (perhaps +-2 KM might have been more
commensurate with the source scale).

5
Date DOWNGRADING_DAT N/A
Releasability RELEASABILITY Unrestricted
Feature Completeness FEATURE_COMPLETE 100% of ONC
Attribute Completeness ATTRIBUTE_COMPL 100% of ONC
Consistency LOGICAL_CONSIST TBD
Edition Number DATASET_ED_NO 1
Creation Date CREATION_DATE TBD
Revision Date DATASET_REV_DATE TBD
Recompilation Date RECOMP_COUNT 0
Revision Count REVISION_COUNT 0
Specification ID PRODUCT_SPEC_ID MIL-D-89009
Date SPEC_DATE April 29, 1991
Amendment SPEC_AMENDMENT N/A
Earliest Source EARLIEST_SOURCE 1971
Latest Source LATEST_SOURCE 1989
Quantitative Attribute QUANT_ATTRIBUTE Unknown
Qualitative Attribute QUAL_ATTRIBUTE TBD
Collection Criteria COLLECTION_SPEC ONC Specs & DCW Design
Absolute Horizontal Accuracy ABS_HORIZ_ACC +- 2040 meters
Unit of Measure HORIZ_UNITS Meters
Absolute Vertical Accuracy ABS_VERT_ACC +- 610 meters
Unit of Measure VERTICAL_UNITS Unknown
Relative Horizontal Accuracy PT_PT_HORIZ_ACC N/A
Relative Vertical Accuracy PT_PT_VERT_ACC N/A
Comments COMMENTS Source map editions from
1971 to 1989
Figure 2: Schema for DCW Library Header Table
The second initiative referred to earlier, that by the International Cartographic Association,
offers an approach to improve on the weakness in the VPF 'data quality statement'.
ICA DATA QUALITY PROPOSAL
The International Cartographic Association (ICA) through its Scientific Advisory Board, has
developed a set of guidelines as its contribution to a clear and consistent approach to the
assessment of data quality.
The guidelines are intended to satisfy certain basic requirements:
1. Defensible. Qualitative rating schemes like 'high', 'medium' and 'low' would
be difficult to defend because of subjectivity, in the form of inconsistency
between assessors, and confusion over what the terms mean. The
guidelines emphasise objective measurement, with summaries as simple,
unambiguous choices.
2. Informative. The purpose of a rating should be to give the user the greatest
possible amount of useful information. If ratings are to be designed by a
testing scheme, they should be designed to pass as many detailed results of
testing as possible on to the user. They should reflect likely uses by
anticipating what the user will be doing with the data.
3. Definitive. It is important that the differences between ratings be as definitive
as possible, and not based on subjective scales of assessment.
Rather than attempt to assess quality in an absolute sense, the guidelines emphasise the
quality of data relative to user needs and anticipated uses, by comparing reality to likely
expectations. In many cases spatial databases are assembled from well known and widely
distributed sources, so an important measure of quality is the degree to which the information
content of the source has been captured accurately in the database: this relative measure
may be more useful to the potential user than an absolute measure of quality.

6
The guidelines use certain terms which require definition:
1. Reality: independently verifiable ground truth; an item of information that can
be verified by visiting the appropriate place on the earth's surface and
making a measurement or observation;
2. Source: the documents (often maps) from which the database was built. The
source is assumed to be available for assessment of the quality of the
database;
3. Database: the product being tested; a set of digital records organised in
some appropriate structure. The assessment of quality extends not only to
the records themselves, but also to information that can be deduced from the
records by simple processes. For example, a user may wish to know the
accuracy of the length of a digital line, whether length is stored explicitly in
the database or computed from the line's coordinates;
4. Source errors: inaccuracies apparent in the source when its contents are
compared to reality. These may include the uncertainties due to different
interpretations of ground truth;
5. Processing errors: inaccuracies introduced by digital processing (including
digitising) and thus apparent in the database when its contents are compared
to the source.
The guidelines describe two distinct approaches, and each has two levels: overall summary
rating, and detailed assessment. In the latter area sections of the guidelines have been
adapted and modified from the Spatial Data Quality section (Section 3) of the proposed US
National Spatial Data Transfer Standard (SDTS). This standard will be the basis of the
proposed Australian Spatial Data Transfer Standard (ASDTS) (Moellering, 1986).
The intent of the guidelines is they be used to assemble an informative Data Quality
Statement to accompany the database.
A summary rating is assessed using one of two methods, depending on whether accuracy is
determined with respect to source document or ground truth.
Overall Summary Rating - Method 1
Method 1 is used to assess databases with respect to source documents, but also must
address the quality of the source document itself, usually by reference to independent
reports. A Method 1 rating has two parts, e.g. A1 C, denoting a database that captures
accurately the entire contents of a source document of unknown quality. These parts are:
1. A measure of the relationship of the digital database to its source;
2. A statement of the quality of the source.
Refer to Figures 3 and 4 for measure of rating for Method 1.
A The database is not significantly less accurate than the document from which it
was derived. Source map features are within 0.5mm of their source map
positions at the scale of the source map (e.g. the location of a feature in the
database is within 50m of its marked position on a 1:100 000 scale source map).
Features attributes are coded correctly in all cases tested. Details of the testing
procedure must be supplied (i.e. sampling design, numbers of features tested)
as all data cannot be guaranteed error-free.
A1 The database is a complete representation of the information on the
source map: all facts that could reasonably be deduced from the
source map can also be found from the database.
A2 The database is an incomplete representation of the information on the
source map. Certain classes of features are missing, or certain
relationships between features cannot be obtained from the database.
The data quality statement should specify the missing classes and

7
relationships as precisely as possible, since users may otherwise
assume the database to be a complete representation of the source.
B There is substantial loss of accuracy with respect to the source document: some
parts of features are more than 0.5mm from their correct positions, and feature
attribute coding errors are present. The data quality statement should include a
detailed assessment of the severity of the errors. In the case of certain types of
errors (e.g. unsnapped junctions) it is important to know whether these errors
could be removed by appropriate processing by the user. If so, the user should
be supplied with appropriate parameters (e.g. the size of the required snap
tolerance).
Figure 3: Summary Rating Method 1 – Relationship between database and source
A Known and documented. The source document has been tested in a rigorous
process of quality assessment, or has been certified by an identified group, and
a data quality statement is either present on the source document or available.
Note that the quality of the source document will have been assessed against
which the uses for which it was originally intended, e.g. visual communication of
information about geographic variation. Unfortunately this will not be sufficient to
ensure quality for many applications of the digital database. For example,
information about the quality of contour positions is not useful in determining
uncertainty in estimates of slope obtained from a contour database.
B Implied, or documented, but not met in fact. Testing has shown that the actual
quality of the source is not equal to reasonable expectations, or to the
information in the supplied data quality statement. Details of testing should be
provided if available.
C Unknown. No information is available about the quality of the source.
Figure 4: Summary Rating Method 1 – Quality of the source
Overall Summary Rating - Method 2
Method 2 assesses the quality of the database by direct reference to ground truth, and has
only one part: that being a statement of the relationship of the database to ground truth. Refer
to Figure 5.
DETAILED ASSESSMENT – ICA PROPOSAL
Spatial databases frequently contain multiple themes, often from different sources. A detailed
assessment of data quality must address each theme individually, particularly in comparisons
with ground truth. Detailed assessment is relevant in two cases:
1. In determining the accuracy of the database in relation to its source (Method
1 above); and
2. In determining the accuracy of the database in relation to ground truth
(Method 2).
There are significant differences in the approaches in the two cases.
Detailed Assessment - Method 1
Each assessment consists of five sections:
1. Lineage;

8
2. Positional accuracy;
3. Attribute accuracy;
4. Logical consistency; and
5. Completeness.
Lineage
The lineage portion of a quality report includes a description of the source material from
which the data were derived and the methods of derivation, including all transformations
involved in producing the final digital files. The description should include the dates of the
source material and the dates of ancillary information used for update. The date assigned to
a source should reflect the date that the information corresponds to the ground; however, if
this date is not known, then a date of publication may be used, if declared as such.
A The relationship to ground truth has been tested by a scientific body appropriate
to the data’s theme.
A formal program of testing was carried out using either:
1. A source of higher accuracy; or
2. Ground checks.
Accuracy is described in accordance with the procedures for detailed
assessment.
B The relationship to ground truth has not been tested directly. Information on data
quality must be inferred as for Method 1, i.e. the accuracy of the database with
respect to its source, and the quality of the source with respect to reality.
Figure 5: Summary Rating Method 2 – Quality assessment through ground-truthing
Any database created by merging information obtained from distinct sources should be
described in sufficient detail to identify the actual source for each element. In these cases,
either a lineage code on each element or a quality overlay (source data index, etc) should be
provided.
The lineage report should include information on all coordinate transformations applied to the
data, including changes of projections, and the parameters used in each transformation (e.g.
figures of the earth).
Positional accuracy
Descriptions of positional accuracy should consider the quality of the final product after all
transformations. The information on transformations forms a part of the lineage portion of the
quality report.
Measures of positional accuracy may be obtained by one of the following optional methods:
1. Deductive estimate: an estimate of positional accuracy based on knowledge of the
errors introduced in each production step. Any deductive statement should describe
the assumptions made concerning error propagation (e.g. independence);
2. Internal evidence: an estimate based on repeated measurements, e.g. by having
several operators digitise the same source material;
3. Comparison to source: an estimate based on graphic inspection of results and
comparison with source ("check plots"); and
4. Independent source of higher accuracy: the preferred test for positional accuracy
is a comparison to an independent source of higher accuracy. The number of test
points and sampling design should be reported.

9
Attribute accuracy
Accuracy assessment for measures on a continuous scale (interval/ratio) should be
expressed in terms of a numerical estimate of expected discrepancies (standard or RMS
error). Accuracy for measures on a discrete scale (nominal) should be given as percent
correct, which could be expressed in the form of a misclassification matrix with summary
statistic for the classified attributes. Sampling design and sample size should be reported.
Logical consistency
A report on logical consistency should describe the fidelity of relationships encoded in the
data structure of the database. Tests for permissible values may be applied to any data
structure. Such a test can detect gross blunders, but does not ensure all aspects of logical
consistency. A data base containing lines may be subjected to general questions such as 'Do
lines intersect only where intended? Are any lines entered twice? Are all areas completely
described? Are there any overshoots or undershoots? Are any polygons too small, or any
lines too close?
For exhaustive areal coverage data transmitted as chains or derived from chains (see the
layer model discussion below), it is permissible to report logical consistency as 'topologically
clean' under the condition that an automated procedure has verified the following conditions:
1. All chains intersect at nodes;
2. Cycles of chains and nodes are consistent around polygons. Or, alternatively, cycles
of chains and polygons are consistent around nodes; and
3. Inner rings embed consistently in enclosing polygons.
Completeness
The quality report should include information about selection criteria, definitions used and
other relevant rules used to capture features from the source. For example, geometric
thresholds such as a minimum area or minimum width should be reported.
The report on completeness should describe the relationship between the objects
represented and the abstract universe of all such objects present in the source. In particular,
the report should describe the exhaustiveness of a set of features.
Detailed Assessment - Method 2
Two different strategies are acceptable, depending on the nature of the theme:
1. Layers; and
2. Objects.
Each assessment consists of five sections:
1. Lineage;
2. Positional accuracy;
3. Attribute accuracy;
4. Logical consistency; and
5. Completeness.
Layers
The theme represents a single variable with a value everywhere, e.g. a map of soil class,
land use, or elevation. The database will likely be expected to provide estimates of the value
of the variable at specific points, and the measure of accuracy should inform the user of the
uncertainty involved in determining such values.

10
Objects
The theme consists of a set of well-defined geographic features with associated attributes.
Features should be sufficiently well-defined to be identifiable on the ground, allowing a test of
positional accuracy to be made with respect to ground truth. Building footprints, shorelines,
rivers, mountain peaks, bridges and roads are examples of well-defined geographic features.
In cases where the object is highly interpreted and thus not suitable for ground truth (an
independent observer could not reasonably be expected to identify correctly whether an
arbitrarily chosen point was located inside the object or not), accuracy cannot be evaluated
(e.g. location of object 'The Top End' of the Northern Territory).
Accuracy should be assessed using the same five categories identified above (lineage,
positional and attribute accuracy, logical consistency and completeness). For the layer
model, positional accuracy should be omitted as it is not relevant, but attribute accuracy is
particularly important, and attention should also be paid to the data structure aspects of
logical consistency. For the object model positional accuracy is particularly important, but the
data structure will likely impose few logical consistency conditions.
Lineage
The lineage portion includes a description of the entire process of data handling from raw
ground observations through to the digital database, including all transformations involved in
producing the final digital files. The description should include the dates of raw observations,
and the dates of ancillary information used for interpretation or update.
Any database created by merging information obtained from distinct sources should be
described in sufficient detail to identify the actual source for each element. In these cases,
either a lineage code on each element or a quality overlay (source data index, etc.) should be
provided.
The lineage report should include information on all coordinate transformations applied to the
data, including changes of projections, and the parameters used in each transformation (e.g.
figures of the earth).
Positional accuracy (object model only)
Descriptions of positional accuracy should consider the quality of the final product after all
transformations. The information on transformations forms a part of the lineage portion of the
quality report.
Measures of positional accuracy may be obtained by one of the following optional methods:
1. Deductive estimate: an estimate of positional accuracy based on knowledge of the
errors introduced in each production step from raw observations to digital database.
Any deductive statement should describe the assumptions made concerning error
propagation (e.g. independence);
2. Internal evidence: an estimate based on repeated measurements, e.g. by having
several operators collect and process the same data; and
3. Comparison to ground truth: an estimate based on actual ground check of the
positions of objects, e.g. using GPS.
Attribute accuracy
Accuracy assessment for measures on a continuous scale (interval/ratio) should be
expressed in terms of a numerical estimate of expected discrepancies (standard or RMS
error). Accuracy for measures on a discrete scale (nominal) should be given as percent
correct, which should be expressed in the form of a misclassification matrix with summary
statistic for classified attributes. Sampling design and sample size should be reported.

11
Attribute accuracy may be assessed by comparison to ground truth, internal evidence or
deductive estimates.
Logical consistency
A report on logical consistency should describe the fidelity of relationships encoded in the
data structure of the database. Tests for permissible values may be applied to any data
structure. Such a test can detect gross blunders, but does not ensure all aspects of logical
consistency. A data base containing lines may be subjected to general questions such as 'Do
lines intersect only where intended? Are any lines entered twice? Are all areas completely
described? Are there any overshoots or undershoots? Are any polygons too small, or any
lines too close?
For exhaustive areal coverage data transmitted as chains or derived from chains (see the
layer model discussion below), it is permissible to report logical consistency as 'topologically
clean' under the condition that an automated procedure has verified the following conditions:
1. All chains intersect at nodes;
2. Cycles of chains and nodes are consistent around polygons. Or, alternatively, cycles
of chains and polygons are consistent around nodes; and
3. Inner rings embed consistently in enclosing polygons.
Completeness
The quality report should include information about selection criteria, definitions used and
other relevant rules used to capture features from the source. For example, geometric
thresholds such as a minimum area or minimum width should be reported.
The report on completeness should describe the relationship between the objects
represented and the abstract universe of all such objects in reality. In particular, the report
should describe the exhaustiveness of a set of features.
IMPLEMENTATION STRATEGY
As digital geographic data has not usually contained details of data quality explicitly within its
structure or in associated documentation, there is a requirement to formulate an
implementation and management strategy to incorporate this form of information. Such a
strategy needs to take into consideration the diversity of forms and formats currently in
existence as well as the sheer magnitude of the task if fine detail is required immediately for
all data assets (not only from within Defence but also the wider community).
An implementation strategy is complex and involves knowledge of digital data requirements,
production and acquisition priorities, and coordination through a number of ADF
organisations. It is therefore the subject of another study. A strategy would include, however,
a number of steps:
1. Compilation of a register of digital data assets of defence and civilian agencies;
2. Assembling an overall summary rating of the data sets; and
3. Producing detailed descriptions for the data sets.
Any implementation plan, however, involves a 'cost'. But such a 'cost' should not only be
considered in terms of dollars and man-hours, it should also be evaluated against benefits to
Defence systems. As technology evolves, future weapons systems, navigation systems,
command and control, targeting, and intelligence systems will become 'smarter'; and the
'smarter' the systems become the more reliance there will be on the data on which they base
their 'decisions'. This means that the systems will require detailed knowledge of the 'quality'
and reliability of the data (similar to those discussed in the ICA Data Quality Proposal).
In the meantime, there are in excess of thirty separate projects (that need to access digital
geographic data in one form or another) being staffed in the Forces Executive, Navy, Army
and Air Force acquisition programs. It, therefore, seems appropriate to commence the

12
implementation process of applying 'data quality labels' to existing data sets and those in
current production and to guidelines compatible with our Defence partners. For example, a
number of systems (such as the F/A-18 Mission Data Planning Facility, Electronic Chart
Display and Information System, Mine Warfare Systems Centre Information System,
Australian Army Tactical Command Support System, and Operational Movements Planning
System) require digital feature data for a range of analyses, and it seems appropriate to
format these data and include 'quality statements' that are being introduced as MILITARY
STANDARDS by other ABCA organisations. As such, the VPF Data Quality Statement
should be used as Stage One of an implementation strategy.
REFERENCES
Abler, R.F.(1987). "The National Science Foundation National Center for Geographic
Information and Analysis", International Journal of Geographical Information Systems,
1,1987.
DMA (1990). Vector Product Format - Draft, Prepared for Defense Mapping Agency
Systems Center by Environmental Systems Research, Institute, Inc., dated November 1990,
[US] MIL-STD-600006
DMA (1991). DCW Product Specification - Interim, Prepared for Defense Mapping Agency
Systems Center by Environmental Systems Research, Institute, Inc., dated April 1991, [US]
MIL-D-89009
Goodchild, Michael F. (1991). "Keynote Address", Symposium on Spatial Database
Accuracy, Department of Surveying and Land Information, The University of Melbourne,
June 19-20, 1991.
ICA (1990). "Guidelines for Quality Evaluation of Spatial Databases", Draft report by
Scientific Advisory Board, International Cartographic Association dated December
19,1990.
Moellering, H. (ed) (1986). Issues in Digital Cartographic Standards, Report #7, National
Committee on Digital Cartographic Data Standards [US], The Ohio State University, OH.,
1986.

DCW Data Quality 1992

More Related Content

What's hot (20)

Similar to DCW Data Quality 1992 (20)

More from Robert (Bob) Williams (20)

Recently uploaded (20)

DCW Data Quality 1992