[OSRS-PROJ] PROJ and ellipsoid shifts

[Frank Warmerdam]

Jean-Paul Berroir wrote:

>  
> Hello all
> 
> being unfamiliar with cartographic projections, I hope I won't bother 
> you with too naive questions.
> 
> I have to process together data of different nature and sources:
>     * satellite (landsat ETM) data, geocorrected (level 1G) , delivered 
> in UTM projection and unspecified ellipsoid -in which case I assume WGS84
>     * data given in latitude/longitude for WGS84
>     * data from ground based stations, delivered in the local 
> co-ordinate system, in my case (former East Germany):
>                              Projection: Gauss Krueger in 6° stripes (or 
> Transverse Mercator)
>                                  projection parameters: central meridian 
> 15E false northing 3,500,000m
>                               Ellipsoid: Krassowsky with Pulkovo Datum 
> (dX=28, dY=-130 dZ=-95)
> 
> then come my questions:
> -up to version 4.3, proj was restricted to projections only, within the 
> same ellipsoid/datum system. It seems that the newer version -and 
> program cs2cs- is able to do the job. Am I right, or is cs2cs also 
> restricted to one ellipsoid at a time?
> 
> -to check my results I was provided with some checked remarkable points, 
> which co-ordinates are known in lat/long and in the local system.
> I was never able to find these results with proj/cs2cs
> Example:
> "ground truth": geographical, WGS84: 52.4172632N 13.6254409E becomes 
> y=3406605m x=5810804m on GK/Krassowsky etc
> 
> The command
> cs2cs  +ellps=WGS84 +proj=latlong +to +ellps=krass +proj=tmerc +lon_0=15 
> +x_0=3500000
> gives me approximately the same figures, yet slighty different:
> y=3406478.97   x=5810764.06 (difference is 40m in x and 126m in y)
> 
> -although this difference might be sufficiently small for my application 
> I would be glad to know if it is normal or much too large, in comparison 
> to cartography standards
> -the difference in x and y looks very much the same as the Pulkovo 
> datum. I'm not sure of the meaning of this datum: is it a shift from 
> Krassovsky ellipsoid or from WGS84? In the first case, how can I tell it 
> to cs2cs, and in the second case, why the same cs2cs command with datum 
> specified as +towgs84=28,-130,-95
> gives me absurd results (e.g. x=3500000.00  y=10002137.50)
> 
> -last but not least, wanting to check my command before bothering you, 
> I've asked for more info with option -v, and look at what is said about 
> ellipsoid:
> (with above cs2cs command, option -v added)
> # ---- From Coordinate System ----
> #Lat/long (Geodetic)
> #
> # +ellps=WGS84 +proj=latlong
> # ---- To Coordinate System ----
> #Transverse Mercator
> #       Cyl, Sph&Ell
> # +ellps=krass +proj=tmerc +lon_0=15 +x_0=3500000
> #--- following specified but NOT used
> # +ellps=WGS84
> 
> how do I have to understand the last comment?
> 
> Many thanks for your help
> 
> Jean-Paul Berroir


Jean-Paul,

I get the same problem, and I have found it is a botched change I made
to the Convert_Geodetic_To_Geocentric() routine in September.  As far as
I can tell this will always screw up any conversion from geodetic to
geocentric.   I have correct the geocent.c file and now I get values close
to your reference position, but still off by 12 meters.  I am assuming this
is due to the +towgs84 transform being only an approximation of the correct
transformation.

I have committed the fixed geocent.c, and also place it at:

   ftp://gdal.velocet.ca/pub/outgoing/geocent.c

I would appreciate your trying the fix and verifying things are OK for you.
I think I had better issue a PROJ.4.4.5 soon since this is a very serious
flaw in the datum shifting support.

Best regards,

  

-- 
---------------------------------------+--------------------------------------
I set the clouds in motion - turn up   | Frank Warmerdam, warmerdam at pobox.com
light and sound - activate the windows | http://pobox.com/~warmerdam
and watch the world go round - Rush    | Geospatial Programmer for Rent


----------------------------------------
PROJ.4 Discussion List
See http://www.remotesensing.org/proj for subscription, unsubscription
and other information.