Hello,
grass 6.4.4 on Centos 6 (from EPEL).
Before I try to turn this into a bug I'd like to be sure that I am not
doing something stupid with projections (happened in the past...).
I have a Lambert equal area location and a lon-lat location, the lon-lat
location covers the whole earth, the Lambert does not. I would like to
project some points from the lon-lat location to the Lambert one,
selecting only those points in a specific region on the Lambert map
which is much smaller than the whole Lambert map.
To that aim, I first do a "grid" vector over the region of interest in
the Lambert location with "v.mkgrid map=region_countries_whole_grid
grid=40,40" (a rather loose grid such that it is easier to project on
different locations). I import that grid in the lon-lat location, use
it to select points and then, back in the Lambert location use v.proj to
import the points. Except that it gives me
v.proj --verbose input=lon_lat_demand_relocated location=whole_region_demand_map_asiamed_cities
Input Projection Parameters: +proj=longlat +a=6378137 +rf=298.257223563 +no_defs +towgs84=0.000,0.000,0.000
Input Unit Factor: 1
Output Projection Parameters: +proj=laea +lat_0=55 +lon_0=20 +x_0=0 +y_0=0 +no_defs +a=6370997 +b=6370997
Output Unit Factor: 1
Reprojecting primitives: WARNING: pj_transform() failed: latitude or longitude exceeded limits
ERROR: Error in pj_do_transform
Visually it seems that all the points are on the "grid" vector (which is
obviously not a regular grid once projected on the lon-lat location).
What is very strange is that projecting the "grid" vector back to the
Lambert location works, but not projecting the points back??
Am I doing something silly, or does it looks like a bug?
Additional info (similar report and lack of datum):
I found a possibly similar issue reported:
http://lists.osgeo.org/pipermail/grass-user/2012-June/065049.html
but no definitive answer.
In some answer to another thread that started with the same issue
erroneously,
http://lists.osgeo.org/pipermail/grass-user/2012-October/065805.html
Markus asked to do a g.region -b, but in fact this does not work in the
Lambert location for me, as there is no datum information. Lack of
datum may be a cause of the v.proj issue above, although I doubt it.
The Lambert location map comes from HYDRO1K (the eu basins):
http://edcftp.cr.usgs.gov/pub/data/gtopo30hydro/
https://lta.cr.usgs.gov/HYDRO1K
https://lta.cr.usgs.gov/HYDRO1KReadMe
I used the grass helper to setup the region and it didn't set a datum
(but it set an ellipsoid). In the .prj file that comes with vectors,
there is
PROJCS["Sphere_ARC_INFO_Lambert_Azimuthal_Equal_Area",GEOGCS["GCS_Sphere_ARC_INFO",DATUM["D_Sphere_ARC_INFO",SPHEROID["Sphere_ARC_INFO",6370997.0,0.0]],PRIMEM["Greenwich",0.0],UNIT["Degree",0.0174532925199433]],PROJECTION["Lambert_Azimuthal_Equal_Area"],PARAMETER["False_Easting",0.0],PARAMETER["False_Northing",0.0],PARAMETER["Central_Meridian",20.0],PARAMETER["Latitude_Of_Origin",55.0],UNIT["Meter",1.0]]
but I couldn't find the "D_Sphere_ARC_INFO" datum in the datum proposed
by grass. The other information from the .prj file were correctly
extracted by the grass helper, as far as I can tell.
--
Pat