Fwd: Re: [GRASS-dev] VTK export for lat-long locations

[Sören Gebbert]

Hi,

Glynn Clements wrote:
> "Sören Gebbert" wrote:
> 
>> please dont use the vtk export modules with lat/lon projections.
>> The vtk export modules in grass are not 
>> usable with lat/lon projections right now.
>> I think i should put this info to the html docs.
>>
>> The lat/lon coordiantes are used to define the coordinates for each point.
>> VTK interpret these coordinates as planimetric and not as lat/lon.
>> This is not that bad for x,y coordinates, but for the height (z coordinate).
>> Because if x and y lat/lon coordiantes are of type [54.001/34.002] and the
>> height varies from 0m - 500m you will get a senseless result.
>>
>> I still dont know exactly how to implement a meaningful 
>> lat/lon ->planimetric coordinate transformation to get rid 
>> fo this behaviour.
>>
>> Any suggestions are welcome.
> 
> In increasing order of complexity and accuracy:
> 
> 1. Convert the lat/lon coordinates to radians then multiply by the
> radius of the earth (in whatever units the z values use).
> 
> E.g. if the Z values are in metres, using the radius of 6378137m for
> the WGS ellipsoid gives a scale factor of pi/180*6378137 ~ 111320
> 
> [I've noticed Hamish' comment, which essentially suggests the inverse,
> i.e. scale the Z values by 1/111320 = 9e-6, so that everything is in
> degrees.]
> 
> For a whole-earth map, this should be adequate; you will still get
> horizontal stretching as you move towards the poles, but no more so
> than is inherent with lat/lon.

I have implemented this approach in
r.out.vtk, r3.out.vtk and v.out.vtk.
The code is available in CVS.

Soeren

> 
> 2. As above, but scale the y coordinates by 111320 and the x
> coordinates by 111320*cos(lat), where lat is the mean latitude of your
> region. The main issue here is that, if the range of latitudes is
> significant, the choice of centre for the x scaling will make a
> difference to the shape of the projected data.
> 
> 3. Pick an "appropriate" geographic projection and transform the
> coordinates into that projection. The choice of projection needs to be
> done manually, based upon the size, shape and location of your region,
> and the relative importance of various properties (equidistant,
> equal-area, conformal, etc).
>