ESMF gen weights problem

Hello,

I'm having a problem with ESMF regridding where ESMF_regrid_gen_weights
appears to succeed but the var n_s = 0 and thus row, col and S are all
empty. Not surprisingly, this leads to a segfault if I try to do a
regrid_with_weights with this file. The weird part is that the target
grid will work when going from one source grid but not another, as
explained below.

I have two potential input files/source grids:
(1) an unmodified WRF file (technically it's been postprocessed but I
haven't done any spatial subsetting yet). This grid is 357 x 381.
(2) a spatial subset of (1) created by selecting points in the original
file that lie within particular lat/lon boundaries (the "subdomain"
file). This grid is 119 x 167.

I have two potential target grids:
(3) every other point from (1), 179 x 191.
(4) every other point from (2), 60 x 84.

Case 1. Input grid is lat/lon from (1) , output grid is (3). Result: OK
Case 2. Input grid is lat/lon from (1) , output grid is (4). Result: OK
Case 3. Input grid is lat/lon from (2) , output grid is (4). Result:
n_s = 0, segfault on regrid.

The clues so far point to a problem with relating source grid (2) to
target grid (4), but this target grid does actually work with source
grid (1).

There are no messages in the PET0* log file and no obvious debug output
suggesting errors.

This is NCL 6.1 on OS X 10.8.2. I've put the script, input WRF files
and ESMF files in file ESMF_Problem_Reusch.zip in the incoming directory
at ftp.cgd.ucar.edu. For the ESMF files, prefix "Dmn" indicates case 3,
prefix "Full" = case 2. Likewise, the WRF files have suffix "domain"
and "full".

Thanks,
Dave

-- 
Associate Research Professor of Climatology
Dept of Earth and Environmental Science
MSEC 304; 801 Leroy Place
New Mexico Tech
Socorro, NM 87801

_______________________________________________
ncl-talk mailing list
List instructions, subscriber options, unsubscribe:
http://mailman.ucar.edu/mailman/listinfo/ncl-talk
Received on Tue Nov 13 14:00:42 2012