[PROJ] PROJ 6.1.0RC1

[Bas Couwenberg]

On 2019-05-06 15:05, Even Rouault wrote:
>> Removing the cs2cs args and +init= prefix result in the gtx being 
>> used,
>> but the magin is still too large.
>> 
>> See the attachment.
> 
> ok, so this is on
> 
> Test:   10 edge_rd
> Exec:   cs2cs EPSG:28992+5709 +to EPSG:4937 -f %.9f
> Input:  50000.4500 335999.6700
> [...]
> Output: 51.003976841	3.891248343 44.739810857
> Expect: 51.003976532	3.891247830
> Latitude exceeds 1e-08 degrees: 3.08999993592352e-07
> Longitude exceeds 1e-08 degrees: 5.12999999813246e-07
> 
> With proj 5.2.0:
> $ echo "50000.4500 335999.6700" | ~/proj/install-proj-5.2.0/bin/cs2cs
> +init=rdnap:rdnap +to +init=epsg:4258 -f %.9f
> 3.891248343	51.003976842 44.740002097
> 
> And with proj master/6.1
> $ echo "50000.4500 335999.6700" | PROJ_LIB=$PWD:data src/cs2cs 
> EPSG:28992+5709
> +to EPSG:4937 -f %.9f
> 51.003976841	3.891248343 44.739810857
> 
> So for lat,long I get almost the same values with both PROJ versions 
> (1e-9 deg
> difference on latitude). Perhaps the test expected result should be 
> updated ?

Tests 07 - 10 are already handled differently, the "Use of RDTRANS2008 
and NAPTRANS2008.pdf" [0] documents the following:

"
    Points 07 - 10 are outside the region where interpolation between
    either the NLGEO2004 geoid or the RD correction grid points is
    possible. RD is defined only within the region enclosed by the
    following points (in RD), outside this region RD coordinates can
    be computed, but the output should be handled with care.

    Corners of the validity region for RD:
    ┌────────┬────────┐
    │  x (m) │  y (m) │
    ├────────┼────────┤
    │ 140000 │ 630000 │
    │ 100000 │ 600000 │
    │  80000 │ 500000 │
    │  -8000 │ 390000 │
    │  -8000 │ 335000 │
    │ 100000 │ 335000 │
    │ 160000 │ 288000 │
    │ 220000 │ 288000 │
    │ 301000 │ 450000 │
    │ 301000 │ 615000 │
    │ 260000 │ 630000 │
    └────────┴────────┘
"

The other tests (01 - 06) should all pass with values within the allowed 
margin of those expected.

Appendix 2 of the PDF documents the following:

"
  The transformation between ETRS89 and RD/NAP procedure should be tested 
in both directions. The
  rdtrans2008.gsb consists of one parent grid and one child grid, the 
naptrans2008.gtx consists of one
  grid, the bounding boxes of the grids, in Bessel-1841 geographic 
coordinates, are given by:

                      NTv2-parent grid  NTv2-child grid  VDatum-grid
  Latitude Minimum    50° 30’ 00.000”   50° 30’ 00.000”  50° 30’ 00.000”
  Latitude Maximum    55° 50’ 00.000”   54° 00’ 00.000”  55° 50’ 00.000”
  Latitude Interval   00° 05’ 00.000”   00° 00’ 30.000”  00° 00’ 30.000”
  Longitude Minimum   02° 30’ 00.000”   03° 00’ 00.000”  02° 30’ 00.000”
  Longitude Maximum   07° 40’ 00.000”   07° 40’ 00.000”  07° 40’ 00.000”
  Longitude Interval  00° 05’ 00.000”   00° 00’ 30.000”  00° 01’ 00.000”

  The coordinates below are computed with the RDNAPTRANSTM2008. the 
differences when using the
  rdtrans2008.gsb and naptrans2008.gtx grids should not exceed, except 
for point 10 (see limitation 2)
  of the rdtrans2008 NTv2-grid):

  RD x and y coordinates:                      0.001 meters
  NAP heights and ETRS89 ellipsoidal heights:  0.001 meters
  ETRS89 latitude and longitude:               0.00000001 degrees
"

This margin was not exceeded with PROJ 5.2.0 (for tests 01 - 06).

[0] 
https://salsa.debian.org/debian-gis-team/proj-rdnap/blob/debian/2008-8/Use%20of%20RDTRANS2008%20and%20NAPTRANS2008.pdf

Kind Regards,

Bas