[mapserver-users] Re: Layer scale problem with Mercator (EPSG:900913)

[Daniel Morissette]

paalkr wrote:
> Hi!
> 
> This issue is quite important to us, does anybody feel like contracting the
> task of fixing this? Please contact me by email if interested,
> pal.kristensen at statkart.no
> 

(For reference, the original post is available at
http://n2.nabble.com/Layer-scale-problem-with-Mercator-EPSG-900913-td2575585.html)

You get this kind of effect because MapServer bases its scale 
calculation on the units of the output projection, and not on the 
location of your map on the sphere... it basically trusts the fact that 
if your projection says "meter" then we can trust the units to really be 
meters on the ground and base the scale on that. The problem in your 
example is that for the area of your map, one unit/meter is not the same 
size in UTM33 and in Mercator... at the equator one unit of the Mercator 
projection is about one meter on the ground, but as you move up North, 
this is no longer true, and one unit covers more like 2, 3 or more 
meters on the ground.

So I'd be tempted to say that this is not a MapServer problem, it's a 
poor projection choice problem, since your projection is lying to you on 
the size of a unit/meter for the area of the world you are looking at.

It seems that to work around this we'd need an option in MapServer to 
enable the concept of "true scale", based on the center of the map on 
the sphere and not on the size of the projection units at the equator 
(or wherever the projection's units are true), basically bypassing the 
projection's definition of what a meter is for your part of the world. 
Note that this would work around the current issue but would come with 
its own side-effects: if you draw tiles using "true scale" in Mercator 
or EPSG:4326 in OpenLayers, then the "true scale" of each tile will not 
be the same, and the tiles in the North part of the map may not show the 
same layers as the ones at the equator... so "true scale" could be 
useful for single tile mapping only and not for tiled mapping interfaces.

All that being said, I played a few years ago with a similar idea to 
have a more meaningful scale when you use lat,lon coordinates in 
degrees, basically basing the scale on the latitude of the center of the 
current map view, but I abandoned the idea when I saw the side-effects 
(layers turning on and off) as you pan North/South. There is still code 
in MapServer's msInchesPerUnit() function (mapproject.c) that is 
currently #ifdef'd that would work for UNITS DD if you enable it.

I'd suggest you try enabling this code (recompile with 
-DENABLE_VARIABLE_INCHES_PER_DEGREE and then retest with EPSG:4326. If 
that fixes the issue for the EPSG:4326 case (I think that will), then we 
could talk about making this an official option in MapServer, and making 
it work for projected coordinate systems in addition to DD.

Daniel
-
Daniel Morissette
http://www.mapgears.com/