[Proj] Dutch correction grid

[molnar at sas.elte.hu]

Dear Jan,

> I see. So this would be a floating point grid with two bands, the first
> for the x (lon) offset, the second for the y (lat) offset. All
> measurements in arcseconds.
>

I can share my experiences creating a grid shift binary (gsb) file for
Hungary (between "Hungarian Datum 1972 (HD72)" and WGS84). As NTv2
standard was invented on the Western Hemisphere, some parameters might
seem strange for those, who learned geodesy east from Greenwich.
1: You should create a four column table. First and second columns should
be the latitude and longitude shift on the first datum grid points,
starting from the bottom right (!) corner of the grid in westward
direction.
Third and fourth column should be the standard deviation of the latitude
and longitude shift value
2: As longitude grows westward for the standard, longitude shifts should
be opposite sign as calculated (not lon(2nd datum)-lon(1st datum) but the
opposit).
3: the delimitation character for this table in ascii format should not be
'tab' but 'white space'. (At least it worked for me...)
4: you should add an ascii header and a one line footer to your table. The
header should you like this: (see my comments after # signs)

NUM_OREC 11
NUM_SREC 11
NUM_FILE  1
GS_TYPE SECONDS
VERSION NTv2.0
DATUM_F HD72
DATUM_T WGS84
MAJOR_F 6378160.000
MINOR_F 6356774.516
MAJOR_T 6378137.000
MINOR_T 6356752.314
SUB_NAMEGGPSH95
PARENT  NONE
CREATED 04-10-10
UPDATED 04-10-10
S_LAT     164520.000000
N_LAT     174960.000000
E_LONG    -82800.000000
W_LONG    -57600.000000
LAT_INC      360.000000
LONG_INC     360.000000
GS_COUNT 2130



> It's not meant for a datum shift, I've got two targets:
>
> - maps in epgs:28992 (Dutch coordinates, with corrections on a regular
> grid  up to 25 cm)
>
> - maps in the 19th century Bonne-based projection, with corrections for
> about 1000 triangulation points (mostly church towers), up to 60 meters.
> These should have to be gridded.
>
> Am I correct that I can create NTv2 files this way and use them in the
> PROJ-parameter list as @nadgrids=...NTV2?
>
> Jan
>
>
>
>
>
>
> On 8-4-2011 18:29, Frank Warmerdam wrote:
>> On 11-04-08 11:28 AM, Jan Hartmann wrote:
>>> I don't quite understand, Frank. I have a spreadsheet with four
>>> columns: x, y,
>>> correction for x and correction for y. What kind of raster file do I
>>> need to
>>> create with these data?
>>
>> Jan,
>>
>> As input we would require an input raster with two "bands"
>> representing the
>> offset in longitude and latitude.   I'm not sure off hand what the units
>> are - possibly arcseconds?   The input file would need to be
>> georeferenced
>> in latitude and longitude.
>>
>> If your input points are sparse and irregular you could likely use
>> gdal_grid to interpolate a grid from the point data.
>>
>> My point was mainly that you don't necessarily have to write an NTv2
>> "writer" - it is sufficient to get the datum shift data into any GDAL
>> support grid format.
>>
>> Best regards,
>>
> _______________________________________________
> Proj mailing list
> Proj at lists.maptools.org
> http://lists.maptools.org/mailman/listinfo/proj
>