V" PEL TE cC X
I bmp -. e YW NN NP NV
« N
et =a UU OO N
6095740
A geometric spatial reliability diagram should indicate the sources
from which the final thematic map was compiled and which parts
of the data can be considered reliable based on an established
accuracy standard (e.g. National Map Accuracy Standards). For
example, Figure 4 depicts a geometric reliability diagram where
a thematic map has been compiled from SPOT panchromatic data,
from USGS digital line graph (DLG) transportation data, and a
USGS digital elevation model (DEM) containing "good" and "bad"
data. It is evident that the geometric reliability of such data
sources is clearly stated in the legend. The legend also identifies
that the DLG vector data were converted to raster format and
resampled to 10 x 10 m. Additional information such as the root
mean square error (RMSE) associated with the resampling
procedures of each data set can also be included. This type of
annotation helps readers identify portions of the final thematic
map which have reduced geometric reliability and can be useful
for improved decision making. It need not be present on the map,
but should be easily accessible on the system by the user.
Most modern mapping applications utilize thematic data obtained
on different dates and/or at different minimum mapping units.
Although a final map may look uniform in its accuracy, it is
actually an assemblage of thematic information from diverse
sources which vary in accuracy. Newcomer and Szajgin (1984)
and Walsh et al. (1987) suggest that the highest accuracy of any
GIS output product is only as accurate as the least accurate file
used in its creation. It is important for the reader to know the
source of the error by depicting them in a thematic reliability
diagram. The thematic reliability diagram shown in Figure 5
identifies two sources of data used in a supervised classification
of wetlands and the location of in situ samples used to assess map
accuracy. Scientists who map wetlands might be concerned that
only DLG wetland data were used. Also, the diagram reveals that
the in situ sampling was spatially biased toward locations which
were accessible only by boat. These two facts can help a reader
to determine the value of a thematic map product derived from the
application of various techniques.
When developing digital geometric and thematic reliability
diagrams, there is a need to standardize their design and function.
The most common questions pertain to the information content
and the amount of detail presented on such diagrams. First, the
diagrams should contain information on the data source (e.g.
USGS 1:24,000 topographic quad). Second, the date of the
original compilation of source data and the dates of subsequent
updates should be included. Third, details on the spatial resolution
to which the data may have been resampled (e.g. 10 x 10 m
resampling of a Landsat TM scene) should be clearly stated.
Fourth, the reliability diagrams must indicate the areas which
would be considered "bad" data, or more specifically data that do
not conform to some accepted accuracy standards. Fifth, if in situ
data is used, then the bias or limitations in the acquisition of such
measurements, such as the number of sample points used or the
restricted access to parts of a study area should be shown.
By including this information in geometric and thematic reliability
diagrams a reader is made aware of the overall accuracy of the
final map. It will also limit the liability of the producers of such
maps, and increase the public confidence in the integrity of
products from the remote sensing and GIS community.
LINEAGE (GENEALOGY) OF THEMATIC MAPS
AND IMAGE MAPS
It is important to identify the difference between lineage and
spatial reliability diagrams. Lineage documentation records the
entire history of all analytical operations performed on a dataset,
and its resultant products. For example, Chrisman (1983) defined
"lineage" as the documentation of data sources and transformations
127
[iterations] applied to them. Conversely, the spatial reliability
diagrams previously discussed provide details on the sources used
in the compilation of the ‘final product’.
Remote sensing and GIS final products are produced from basic
source materials. Manual "book-keeping" of the processes used
for deriving the final product is cumbersome and rarely performed.
There are systems which provide automated methods such as
‘history’ or ‘audit’ files to keep track of the iterations and
operations performed. However, none of these methods are
capable of fulfilling the informational requirements of a true
‘lineage’ report which itemizes the characteristics of image and
cartographic sources, the topological relationships between source,
intermediate and final product layers, and the transformations
applied to sources to derive the output products (Lanter, 1990).
The National Committee for Digital Cartographic Data Standards
(NCDCDS) proposed that lineage information be included in every
*quality report' of a digital cartographic product (NCDCDS, 1988).
The committee specified five requirements for the lineage criteria,
including:
a) source material from which the data were derived;
b) methods of derivation, including transformations
applied;
c) if data from different distinct sources are used, such
sources must be identified;
d) include reference to specific control information used,
e.g. National Geodetic Reference System or if other
points are used then sufficient detail must be
provided to allow recovery; and
e) description of the mathematical transformations of
coordinates used in each step from source material
to final product.
Lanter (1991) categorized geographic data layers into source
layers, intermediate layers, and product layers. ^ Lineage
information on source layers should include the NCDCDS digital
cartographic data standards, while intermediate layers require
documentation on the nature of the transformations used in their
derivation. Final product layers must be associated with
information concerning their use, such as the users' role in
decision making, release dates, and those responsible for product-
layer maintenance (Lanter, 1990).
Lineage or genealogical documentation should, therefore, form an
integral part of the annotation of remote sensing or GIS products.
Software designed to document lineage must have the following
components: 1) lineage tracing, 2) maintain data quality
information, 3) automatic error detection, 4) rule building (i.e.
flexibility to users on building their own rules into a knowledge
base about how their GIS data should be handled), 5) data-driven
user interface, and 6) project management (such as keeping track
of times, dates, and user names to show who did what to the
database and when) (Lanter, 1989). This will resolve data
management problems by maintaining an automated, dynamic
model of the database. In addition, the user will have information
on cartographic materials used and a chronicle of the remote
sensing or GIS transformations applied to derive the final
products. In most cases it may only be necessary to explicitly
state in a textual legend a) the name of the lineage file, e.g.
Jensen.21092, and b) the cognizant scientist (and his/her address)
who was responsible for creating the final product. The lineage
file must then accompany the final product file.
CONCLUSION
Remote sensing and GIS products will be cartographically
enhanced by adopting the five types of annotation discussed in this
paper. It is also important for the image processing and GIS