[Talk-ca] Canadian imports: good or bad?

Bégin, Daniel Daniel.Begin at RNCan-NRCan.gc.ca
Thu Apr 26 12:48:09 BST 2012


Bonjour Steve,

I'm pretty comfortable with your propositions and wording, as a contributor :-)

However, as data provider representative, my emails on this list aimed at providing information to help the community to better understand the product, not decide for them.

So I invite the rest of the community to comment on it!

Best regards,

Daniel

Ps: I'll write a little something about accuracy ( from home, as a contributor!-)





-----Original Message-----
From: Steve Singer [mailto:steve at ssinger.info] 
Sent: April 25, 2012 22:03
To: Bégin, Daniel
Cc: Paul Norman; talk-ca at openstreetmap.org
Subject: RE: [Talk-ca] Canadian imports: good or bad?

On Wed, 25 Apr 2012, Bégin, Daniel wrote:

> Steve, Paul,
>
> I was on the impression that the consensus was more about using Canvec 
> where it is the best available source and, when it is not, the data 
> could be imported, but only after an exhaustive verification using 
> available data/imagery.

'best available source' as a standard has appeal to me, and I think this varies by layer (ie your comments in the other email about older
hydrography) I think often people are importing all of the layers at once when without evaluating what they are importing.

>
> Whatever is the consensus, it should be documented in the wiki.
>

Agreed.  I think right now we have consensus on saying:

* The osm-ca community wants to import Canvec data
* The imports should be done carefully to avoid duplicating objects
* Coastlines and large lakes should only be imported by experienced users

(which is basically what the wiki already says)

Paul proposed two additional guidelines here:
http://lists.openstreetmap.org/pipermail/talk-ca/2012-April/004721.html

"1. The buildings data from CanVec should not be imported unless it can be verified against imagery, in which case you might as well trace the buildings from imagery."

Ie if the imagery (and there is no other source like a local mapper) isn't good enough to verify the buildings then don't import them. It seems, to me, that so many of the 20+ year old building data is no longer valid that we might want to discourage the use of this  layer without Do we have consensus on this point?


"2. There is not a consensus among the community that CanVec data can be imported without verifying the data for internal consistency and where possible against imagery."

I like the sentiment but I don't like the 'negative wording' it doesn't tell people what we DO have consensus on, so it doesn't tell them what they can 
import. Nor does it explicitly prevent any sort of import.   My wording from 
this morning apparently wasn't good either.

How about

* When importing Canvec data you should verify that the data you are importing is consistent with other data.  For example check that forests aren't sitting in lakes. Sometimes the different Canvec layers are not consistent because the data comes from different sources.  You should try to fix consistency issues as you import data.

(anyone should feel free to propose some better wording)

Is there something we can say in the guidelines to help people judge accuracy? (In most of the areas I map I've found the Canvec data lines up VERY well with Bing and my GPS traces)


Steve


> Furthermore, I think that "internal consistency/accuracy/existence" 
> should be defined...
>
> consistency: ?
>
> Accuracy: Bing imageries in urban areas are pretty good and easy to 
> correct, if necessary, using available GPS tracks. It is not the case 
> outside these areas.
>
> I suspect that Bing imageries are not always corrected using a good 
> digital elevation model. It means that in hilly areas, the image shows 
> an object somewhere on the ground while the object is actually 
> somewhere else, due to Z distortion.
>
> Existence: Again, outside urban areas, the resolution of Bing 
> imageries doesn't allow for detailed validation. You won't be able to 
> see small objects, even if they are there!
>
> Best regards,
> Daniel
>
> -----Original Message-----
> From: Steve Singer [mailto:steve at ssinger.info]
> Sent: April 25, 2012 07:12
> To: Paul Norman
> Cc: talk-ca at openstreetmap.org
> Subject: Re: [Talk-ca] Canadian imports: good or bad?
>
> On Wed, 25 Apr 2012, Paul Norman wrote:
>
>>> 2. There is not a consensus among the community that CanVec data can 
>>> be imported without verifying the data for internal consistency and 
>>> where possible against imagery.
>>
>> If no one disagrees with the fact there is not a consensus that 
>> importing CanVec without minimal verification is acceptable I'll go 
>> ahead and document on the Wiki, using Andrew Allison's examples.
>
> +1.
>
> Is there enough support to use the positive rather than the negative language, ie 'There is consensus among the community that Canvec data should only be imported when the data elements have been verified for internal consistency/accuracy/existence with the available imagery'
>
> Steve
>
>
>>
>>
>> _______________________________________________
>> Talk-ca mailing list
>> Talk-ca at openstreetmap.org
>> http://lists.openstreetmap.org/listinfo/talk-ca
>>
>
>
> _______________________________________________
> Talk-ca mailing list
> Talk-ca at openstreetmap.org
> http://lists.openstreetmap.org/listinfo/talk-ca
>


More information about the Talk-ca mailing list