[postgis-users] Confusion about projections and Spatial Data

[Rick Zoolker]

Thanks for your help.  It's been very very very helpful.
I do have a few followup questions for you or the group.
I guess I understand all of what you said except for your #2 where you
state that all the functions work with lat/long except for distance
calculations.  As far as I can tell any perimeter calculation as well
as area calculations will almost certainly be wrong when dealing with
lat/long coordinates.  Also, polygon operations (e.g. intersection,
union, etc) will certainly be wrong if they cross any "boundaries"
(i.e. +/- 180 deg).  I'd have to think about if they actually return
the correct result when you don't cross boundaries, but I'm thinking
that the fact that each degree of longitude is not a consistent length
(i.e. it varies by latitude), plus the fact that projected straight
lines become curves will mess things up in these operations.  You all
probably know this off the top of your heads, or perhaps the
distortions are not really enough that I should be concerned about
them.?
Thanks again for your help
  -Rick

On 8/10/07, Obe, Regina <robe.dnd at cityofboston.gov> wrote:
>
> Very good questions.  I may be wrong on some of my assumptions, but this is
> my understanding
>
> 1) Everyone ships in lat long format because it covers the widest area of
> space and can be easily transformed into other SRIDs or out of all SRIDs.
> Its not even entirely true that everyone does.  Most state municipalities
> for example often ship their data in a coordinate system specific to thier
> area or state.
>
> 2) As far as I know most functions will work with lat long except distance
> gives you funky results because it is measuring around circles for lat long.
>
> 3) Why do databases have this whole cartesian thingy and not work straight
> with lat long?  Because screens are flat, printed maps are flat and when you
> look at isolated areas you can make them look flat and not skew measurements
> and layout.   So I guess its a hack to make your problem simpler.  If you
> look at the upcoming SQL Server Katmai which is supposed to have Geography -
> it will challenge that and try to deal with geodetic data as a "first class"
> reference system.
>
> 4) Not everyone stores their data projected, but often will transform on the
> fly if they do not project.  I do when I know I am only dealing with a very
> isolated region like a single state or city municipality.
>
> 5) Its not true that tools like ArcView and MapInfo want lat long.  Actually
> MapInfo is pretty sophisticated I think when it comes to projections so its
> almost transparent to the user.  I'm not quite sure how it does it, but
> encoded in each tab file is the projection info and you can lay on a bunch
> of layers in different projections and it will magically pick the most
> suitable one and reproject all the layers to that.  My knowledge of ArcView
> is less, but I think ArcView tends to be more picky about projections so
> requires more knowledge on the user's side to know how to deal with multiple
> layers in different projections.
>
>
> Hope that helps,
> Regina
>
>
> ________________________________
> From: postgis-users-bounces at postgis.refractions.net on
> behalf of Rick Zoolker
> Sent: Fri 8/10/2007 1:34 PM
> To: postgis-users at postgis.refractions.net
> Subject: [postgis-users] Confusion about projections and Spatial Data
>
>
>
> OK, I feel like I've got a fundamental block about how spatial data is
> actually used in the real world and how spatially aware databases
> actually make use of that data.  This is probably quite obvious for
> everyone so please bear with me.
>
> I've been playing with a bunch of spatial databases (Oracle, Postgis,
> DB2) and from what I can tell, all tend to roughly follow the OGC SFA
> spec (Oracle using different terminology).  Now it also seems that the
> SFA spec deals EXCLUSIVELY with Cartesian coordinates.  Now here's my
> first confusion.  If they only deal with Cartesian coordinates, why do
> they allow you to specify SRIDs that are lat/long based.  Will any of
> the SFA functions work with lat/long (other than ST_Transform(), which
> is actually a SQL/MM and not even included in SFA)??
> My next area of confusion is that I've looked at a bunch of data from
> people such as Navteq, teleatlas, and a bunch of individual companies
> databases, and it seems that every single one ships and stores their
> data in lat/long.  So what the heck?  If the tools can't deal with
> lat/long, why does everyone ship and store their data in lat/long?
> Does everyone project this data before actually loading it into the
> db?  Help!?
>
> Now, it appears that Oracle in particular can do some functions with
> lat/long directly (the distance calculation comes to mind), and I know
> Postgis has some specialized distance functions for geodetic
> coordinates, but I'm not sure if that extends to the whole function
> set (i.e. can I do an Intersection of two polygons specified in
> lat/long coordinates??).  Why doesn't the OGC SFA spec and libraries
> support all the functions with geodetic coordinates directly rather
> than require projected coordinates?
>
> Finally, how do the analysis and visualization tools that people put
> on top of these databases deal with these questions? (e.g. MapInfo,
> ArcView, etc.)  They seem to want lat/long data as well, but if you
> stored project coordinates in the database isn't there a problem??
>
> HELP!
>
>   -Rick
> _______________________________________________
> postgis-users mailing list
> postgis-users at postgis.refractions.net
> http://postgis.refractions.net/mailman/listinfo/postgis-users
> OK, I feel like I've got a fundamental block about how spatial data is
> actually used in the real world and how spatially aware databases
> actually make use of that data.  This is probably quite obvious for
> everyone so please bear with me.
>
> I've been playing with a bunch of spatial databases (Oracle, Postgis,
> DB2) and from what I can tell, all tend to roughly follow the OGC SFA
> spec (Oracle using different terminology).  Now it also seems that the
> SFA spec deals EXCLUSIVELY with Cartesian coordinates.  Now here's my
> first confusion.  If they only deal with Cartesian coordinates, why do
> they allow you to specify SRIDs that are lat/long based.  Will any of
> the SFA functions work with lat/long (other than ST_Transform(), which
> is actually a SQL/MM and not even included in SFA)??
> My next area of confusion is that I've looked at a bunch of data from
> people such as Navteq, teleatlas, and a bunch of individual companies
> databases, and it seems that every single one ships and stores their
> data in lat/long.  So what the heck?  If the tools can't deal with
> lat/long, why does everyone ship and store their data in lat/long?
> Does everyone project this data before actually loading it into the
> db?  Help!?
>
> Now, it appears that Oracle in particular can do some functions with
> lat/long directly (the distance calculation comes to mind), and I know
> Postgis has some specialized distance functions for geodetic
> coordinates, but I'm not sure if that extends to the whole function
> set (i.e. can I do an Intersection of two polygons specified in
> lat/long coordinates??).  Why doesn't the OGC SFA spec and libraries
> support all the functions with geodetic coordinates directly rather
> than require projected coordinates?
>
> Finally, how do the analysis and visualization tools that people put
> on top of these databases deal with these questions? (e.g. MapInfo,
> ArcView, etc.)  They seem to want lat/long data as well, but if you
> stored project coordinates in the database isn't there a problem??
>
> HELP!
>
>   -Rick
> _______________________________________________
> postgis-users mailing list
> postgis-users at postgis.refractions.net
> http://postgis.refractions.net/mailman/listinfo/postgis-users
>
>
>
>
> ________________________________
>
>
>
>
> The substance of this message, including any attachments, may be
> confidential, legally privileged and/or exempt from disclosure pursuant to
> Massachusetts law. It is intended solely for the addressee. If you received
> this in error, please contact the sender and delete the material from any
> computer.
> _______________________________________________
> postgis-users mailing list
> postgis-users at postgis.refractions.net
> http://postgis.refractions.net/mailman/listinfo/postgis-users
>
>