[gdal-dev] Re: Poor Aggrement Between Layers

[Kent Eschenberg]

This problem has been traced, I believe, to an error in the original relief map 
in the Lambert projection provided by the National Atlas of the US (NAUS). The 
same map in the Albers projection seems correct.

On 1/2/2008 Kent Eschenberg wrote:
 > The two map layers described at
 >
 >    nationalatlas.gov/mld/srld48i.html (relief)
 >    nationalatlas.gov/mld/roadtrl.html (roads)
 >
 > don't seem to agree. If I look at the Los Angeles area I must shift
 > the roads about 10,596 feet to the right and 1,849 down.

Further testing has been done with these 3 datasets:

   A: nationalatlas.gov/mld/roadtrl.html (roads, geographic)
   B: nationalatlas.gov/mld/srld48i.html (relief, Lambert)
   C: nationalatlas.gov/mld/srld48a.html (relief, Albers)

The comparisons were done by converting one map to the same projection as the 
other map as follows:

   Compare A to B->geographic: disagree
   Compare A to C->geographic: agree
   Compare A->Lambert to B: disagree
   Compare B->geographic to C->geographic: disagree
   Compare A to C->geographic->Lambert->geographic: agree

The transformations at the "->" were performed by GDAL. For example, 
B->geographic was performed with

    gdalwarp -t_srs WGS84 srld48i200l.tif srld48i200l_WGS84.tif

where the projection of the input file is read from the file and associated 
files in the same directory.

Note that both the first and the last comparison exercised the GDAL 
transformation from Lambert to geographic. It seems therefore that this 
transformation is being done correctly by GDAL. The only conclusion I can make 
is that B is in error.

The funny thing is that Jay at the NAUS sent an image from their online 
MapMaker utility that uses A and B, and the results agree. Maybe MapMaker 
doesn't use exactly the same version of B as is available to the public for 
downloading.

Kent
Pittsburgh Supercomputing Center