[Talk-us] Characterizing Systematic Errors in Hawaii and US Pacific Island Data

Scott Atwood scott.roy.atwood at gmail.com
Fri Oct 2 19:38:20 BST 2009


Based on suggestions in an earlier thread, I have begun the task of
analyzing the errors in the existing data for the islands of Hawaii and
other US Pacific Islands.  I have begun with a few of the smaller, simpler
islands, and I'd like to start sharing my findings.
Kohoʻolawe:

This is an uninhabited island off the coast of Maui, so it has essentially
nothing other than natural features to map.  The existing OSM data consists
of the shoreline, a few labeled bays, and a political boundary that
presumably is supposed to correspond to the the shoreline.

The shoreline and bay data corresponds very well with the satellite images.
 The boundary data is lower resolution (fewer polygon points), and is
displaced approximately 1km to the east.  The boundary data appears to have
been imported data, and I don‘t think it has been subsequently modified.

Lanaʻi

This is another small island off the coast of Maui.

Both the coastline and road network appear to correspond closely to the
satellite imagery.  The road network has significant manual modification
history, which suggests if there were systematic daturm errors in the
initial import, they were all manual resolved in the relatively small road
network of this island.

The boundary data, much like Kohoʻolawe, is lower resolution, and is
displaced by the same or similar 1km to the east.


Next Steps for Kohoʻolawe and Lanaʻi?

These two islands appear to be in pretty good shape, except for the boundary
data.  Since the political boundary should presumably correspond exactly to
the physical boundary of these islands, what is the best way to proceed?  I
assume that the existing erroneous political boundary should be removed, and
then a new political boundary should be recreated based on the existing,
correct shoreline data.  Can the coordinates of the shoreline points and
ways be cloned and tagged as a political boundary?  Or can/should tags be
added to the existing shoreline data that designates them as both the
physical and political boundary simultaneously?


I'll continue analyzing data for additional islands as time permits.

-Scott


-- 
Scott Atwood

Cycle tracks will abound in Utopia.  ~H.G. Wells
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-us/attachments/20091002/22982a85/attachment.html>


More information about the Talk-us mailing list