[Talk-ca] Your new coastline

Andrew MacKinnon andrewpmk at gmail.com
Fri Sep 2 20:08:47 BST 2011

On Thu, Sep 1, 2011 at 11:15 PM, James A. Treacy <treacy at debian.org> wrote:
> On Thu, Sep 01, 2011 at 10:30:34PM -0400, Andrew MacKinnon wrote:
>> Given that this sort of work is time consuming it will take a while to
>> finish. However, 99% of the work that requires importing coastlines
>> from CanVec is done, and realigning coastlines using Bing is a lot
>> less disruptive and less error-prone.
> I'm curious why you would trust the Bing imagery more than canvec. In
> addition to not being very high resolution, I would think that Bing
> would suffer from problems with registration (alignment of images to
> lat/lon) which would have to be checked against ground readings. Of
> course canvec should also be checked for accuracy with local readings.
> Further, my understanding is that much of the canvec data is generated
> from local surveying, which uses high end GPS which are extremely
> accurate.
> Locally (Kitchener-Waterloo) I have found that the canvec data is very
> accurate and most imagery less so.

I have been using the Bing imagery where high resolution imagery is
available and Canvec where high resolution Bing imagery is not
available. My impression is that for coastlines, tracing from Bing
imagery is more accurate than the Canvec data.

Keep in mind that some of the Canvec data is VERY out of date. While
the road data in Canvec is fairly up to date, the rest of the Canvec
data seems to be old (1990s, 1980s even?) Canvec data shows woods,
buildings etc. which clearly haven't existed for many years, for
example it often shows forests in areas where new subdivisions have
been built recently, old industrial buildings which were torn down 10
years ago and replaced with housing, long-ago abandoned rail spurs to
industrial areas and long-demolished agricultural buildings. I would
not trust anything except the roads layer in Canvec to be up to date.

More information about the Talk-ca mailing list