#385 Mercator projection problem and possible corrections
Submited By Kasemsan Manomaiphiboon
Air Quality Research Group, GeorgiaTech
March, 2005
++ Surrogate processing by MIMSSP (current version)
is not working on MERCATOR (GRID TYPE = 3 or
MERGRD3).
++ The problem is the program could not find the
overlapping area of a specified grid domain and a
shapefile and then stop. After examining the problem
in detail, I found that MIMSSP uses PROJ4 library
where, for MERCATOR, MIMSSP imports only lat_0,
lan_0, and lat_ts from GRIDDESC (as seen in Lines 138-
143 of src/mims_spatl_proj.c)
++ In fact, there is one critical variable for MERCATOR
that should be included. It is +y_0 (false northing).
Since +y_0 is not imported, PROJ4 assumes it to equal
0.0 automatically. Note that +x_0 (false easting) is not
important in this problem and PROJ4 also assumes it to
equal 0.0 (which is fine).
++ It should be noted that for some projections, the
assumed +x_0 and +y_0 equal to 0.0 does not cause
such a problem, as in Lambert Conformal Conic.
++ The effect of having +y_0 = 0 is crucial in
MERCATOR because it translates the center of the grid
domain (in GRIDDESC) by REARTH * LN(TAN(PI / 4 +
lat_0 / 2)) in north-south direction whereas we want to
keep the central (X,Y) = (0,0). So, an easy chance of
not finding an overlapping area happens and causes the
problem.
++ Correction options:
1) Include importing +y_0 in mims_spatl_proj.c (and in
generate script) or
2) What I did (just a temporary solution) is adjust the
value of YORIG in GRIDDESC to cancel out the effect of
translation from +y_0=0.0, which makes the program
run without any problem.
By Kasemsan Manomaiphiboon
Air Quality Research Group
GeorgiaTech
