On => Precision in Geographic Information Systems . .
Do you want to use GIS as a scientific tool? Here are some challenges:
- Many folks develop their spatial (GIS) data bases by digitizing maps.
- This imposes many approximations and distortions from the source materials into the digital data,
- We don't want or expect such distortions in our source data.
- We often don't notice such distortions, but they are (1) often easy to detect with the proper methods and (2) potentially damaging to our work.
- Many folks acquire digital data from suppliers who develop their own source data from cartographic output,
- Such data have similar problems to those noted just above..
- Many software developers and GIS consultants gloss over precision as a concern in spatial data base development,
- This may be because there are huge numbers of topics to discuss - and precision is often sacrificed for lack of time to discuss it.
- It might also be because the topic rarely comes up, so consultants downgrade its importance in their own minds..
- Documentation on locational and thematic data precision is in its infancy
- Documentation often seems like a diary, or a secondary school chemistry laboratory report: we're justifying our work to our bosses.
- Many new users may be outside our profession. They want informative documentation that:
- Provides some awareness of the "strange, new data."
- Is written in English (or some other language..), NOT in professional jargon.
- Is copiously backed up with case histories and other citations.
- Instead of helping new users, much current documentation only confuses them.
- The result is an uninformed new user, who either:
- Abandons the data out of frustration, or
- Inadvertently misuses the data, because the documentation doesn't help prevent intelligent misinterpretation of data characteristics.
- In short: it's easy to overlook precision as an issue in GIS,
How do we take something used primarily for making maps, and use it for Science?
We often think of GIS as a framework for creating maps.
- But maps are simplifications (however effective or detailed) of the real world.
- Somewhat like editorial cartoons, good maps remove "extraneous details" that, in the author's eye, don't help make the author's point.
More accurately, GIS is an environment for managing highly detailed spatial data.
- Simplified maps are only one possible output from such a spatial data base.
- KNOWLEDGE is the goal. This may NOT be in the form of a map..
If the input data are imprecise (simplified, or stylized as are our output maps) we're limited to little more than simple mapmaking in our GIS.
- Thus, our GIS is little more than a cartographic output device.
- Our GIS is limited as a tool for scientific study.
If our GIS contains high-precision field measurements (like geodetic control points, analyses of soil and water chemistry, etc.), our ability to perform scientific study with GIS greatly increases.
- Yet many GISs, and people dealing with GIS, remain "mired" in the idea that GIS and high precision are unrelated.
- Some people have it even worse, and believe that rough data sets digitized from simplified maps (and even 1:5000 scale maps are greatly simplified!) are of wonderful quality.
So: do we
- accept low-precision data?
- accept that we must often use low-precision data at present, but hope for better precision in data (and software data-handling) in the future?
- work to improve the precision of our data, and the precision handling of our software?
This page supports the last option!
We often imbue our data bases with unnecessary errors, and fail to hold our data and software suppliers, as well as ourselves, to adequate standards of precision. We often ignore the importance of precision, rather than accept that some data sources are imprecise. If we accept that
some data sources are imprecise, we want to eventually have better data. If we accept the inadequacies of current low-precision data, we remain stuck with our low standards.
Here's an example of very high precision geodetic control data (shown as red spots) for the Island (and county) of Oahu, Hawaii, plotted with a moderate-resolution (about 1:80,000 scale) NOAA coastline. The geodetic control is in the North American Datum 83 (horizontal), with the following selection of vertical datums:
- "NGVD 29": 783 control points.
- "NAVD 88": 9 control points.
- "N/A": 291 control points.
The Arc Data version of the Digital Chart of the World is in an undocumented projection. (Data support specialists had various
answers to our question of "what datum do the data use?" The National imgry and Mapping Agency (formerly the Defense Mapping Agency) notes that DCW uses World Geodetic System 84 (The Pacific Command maintains another useful site discussing WGS-84)
for horizontal datum, and Mean Sea Level for vertical datum. Note the registration problems:
- Survey points in the northwestern part of the island, many of which run along the shoreline (including the coastal road) are plotted inland from the mapped "coastline."
- Many survey points in the southeastern coastal part of the island appear to be offshore.
NW Oahu geodetic control vs. coastline
|
Index map of Oahu, Hawaii
|
Northern Oahu geodetic control vs. coastline
|
Waikiki geodetic control vs. coastline
|
Is the shoreline mis-digitized? Are the geodetic control data actually so far off-the-mark horizontally? Probably not.
Are we having difficulty in rectifying (reprojecting) latitude-longitude data between these combinations of vertical and horizontal datums? Perhaps! We're told that the inaccurate noting of datums can result in 300 meter horizontal displacements, when the data are projected and plotted.
How often do we see full description of vertical and horizontal datums in a spatial data set? How well do our software packages handle projections and datums? Are we sure?
NOAA's National Geophysical Data Center, in cooperation with NOAA's National Geodetic Survey and the U. S. Geological Survey, has been conducting research in spatial representations of data. Initial observations are that we (as a scientific community) are FAR(!) from doing an adequate job in this area.
Treatment of this project
( presently under construction), will enhance this discussion of Precision in GIS.
Submit your favorite definition of On-Demand Mapping or related WWWeb-based mapping systems! Send us your favorite link to discussion on On-Demand/Interactive Mapping! Comments are welcome:
David Hastings
World Data Center-A, National Geophysical Data Center
303-497-6729 or
dhastings@ngdc.noaa.gov
URL: http://www.ngdc.noaa.gov/seg/globsys/gis/ondemand.shtml
Revised: 21 December 1997