|
I have noticed a problem with the ncep routine w3fb11.f
which is supposed to be the the reverse of w3fb12.f) posted
on the narf.fsl.noaa.gov server, a problem related to the
accuracy of the machines on which the routines are run.
If I specify the grid origin xi,xj=(1,1) and first call
w3fb12 to convert that grid coordinate to longitude and
latitude and then call w3fb11 to convert that longitude
and latitude back to grid coorinates I do not get the
original (1,1) but instead (with my machine's accuracy)
1.000061,-6.10E-05). The problem seems to lie in the lines
at the end of w3fb11.f:
C IF COORDINATE LESS THAN 1
C COMPENSATE FOR ORIGIN AT (1,1)
IF(XI.LT.1.) XI = XI - 1.
IF(XJ.LT.1.) XJ = XJ - 1.
which in my case convert an "almost 1" XJ of 0.999939 to an
XJ of 0.000061.
I don't know how significant this problem might be for other
users of these routines on 20km MAPS test directory but wish
to alert them to the potential problem.
Jack
|