[netcdfgroup] netCDF Operators NCO version 4.9.3 have arrived

• To: NCO Announce Mail List <nco-announce@xxxxxxxxxxxxxxxxxxxxx>, netcdfgroup@xxxxxxxxxxxxxxxx
• Subject: [netcdfgroup] netCDF Operators NCO version 4.9.3 have arrived
• From: Charlie Zender <zender@xxxxxxx>
• Date: Sun, 31 May 2020 15:26:00 -0700

The netCDF Operators NCO version 4.9.3 have arrived.

http://nco.sf.net (Homepage, Mailing lists, Help)
http://github.com/nco (Source Code, Issues, Releases, Developers)

What's new?
Version 4.9.3 contains many regridding fixes and features:
This version updates adds robustness to the NCO weight-generator
for PG2 grids, updates default options for TempestRemap, implements
a high frequency mode for the splitter, and fixes some issues with
ncremap, ncclimo, and ncrcat and works-around a backwards
incompatibility with the latest netCDF library, 4.7.4.

Work on NCO 4.9.4 has commenced and will improve NCO weight-generator
accuracy, reduce vertical interpolation memory use, and supply more
accurate weight generation options for rectangular lat-lon grids.

Work on NCO 5.0.0 has commenced "under the hood". A key leap in that
release will be support for netCDF4 user-defined types. Printing of
netCDF4 user-defined types ENUM and VLEN is ready now (though
unsupported) with the --udt flag. 5.0.0 will contain the finished
version of that, and include options for invoking mbtempest in place
of tempest.

Enjoy,
Charlie

NEW FEATURES (full details always in ChangeLog):

A. ncatted prints additional diagnostics when referenced attributes
   are not found.
   http://nco.sf.net/nco.html#ncatted

B. ncclimo adds a new clm_md='hf' type for splitting high-frequency
   (hgh_frq) datasets. Currently these datasets must be daily means.
   http://nco.sf.net/nco.html#ncclimo

C. ncremap now automatically parallelizes the loading of data into
   KD-trees during regridding when OpenMP is available.
   This significantly improves weight-generation throughput.
   http://nco.sf.net/nco.html#ncremap

D. The ncks --frac_b_nrm and --area_wgt options both now automatically
   imply --chk_map.
   http://nco.sf.net/nco.html#chk_map

E. Options for TempestRemap have been updated to current guidance:
   No TR algorithms invoke the TR --volumetric option anymore.
   Many TR algorithms now add the TR  --correct_areas option.
   http://nco.sf.net/nco.html#tempest

F. ncremap's grid intersection algorithms are improved.
   In particular, "physics grids" (aka PG2 grids) intersections
   are now more robust for all resolutions in both directions
   up to ne1024.
   http://nco.sf.net/nco.html#ncremap

G. ncks accepts a new option --cell_area_quad which implements the
   TempestRemap quadrature algorithm instead of the NCO piecewise
   constant algorithm to compute cell areas. The TR algorithm often
   produces slight improved map statistics, though sometimes produces
   worse statistics. We are working to understand this better before
   deciding whether to adopt it as the default area algorithm.
   http://nco.sf.net/nco.html#ncks

H. ncks includes a workaround to the backwards incompatible changes
   to nc_inq_var_filter() inadvertently introduced in netCDF 4.7.4.
   Without this workaround, NCO may quit when printing files that
   store results compressed with non-DEFLATE filters.
   http://nco.sf.net/nco.html#ncks

I. ncap2 eliminated an old and little-used custom correlation
   procedure that worked for data with missing values yet that ran
   afoul of the NCO's new BSD license.
   http://nco.sf.net/nco.html#ncap2

J. The Users Guide now fully documents the ncremap and ncks options
   chk_map, job_nbr, par_typ, SGS, unq_sfx.

K. Add --frac_b_nrm option to ncks map-checker
   http://nco.sf.net/nco.html#chk_map

BUG FIXES:

A. ncremap once again passes non-default regridding options to
   the NCO weight-generator in map-only mode. This appears to
   have been inadvertently broken in 4.9.2. The solution is to
   avoid 4.9.2.

B. ncclimo in splitter mode now correctly globs all files in a
   directory. Previsously, the globbing option to ncclimo was
   broken for the splitter. Thanks to Jill Zhang for reporting.
   The solution is to upgrade.

C. ncrcat once again re-bases time coordinates stored as integers.
   It lost that capability in an unknown but recent NCO release.
   The solution is to upgrade.

D. Curvilinear regional NCO grids that work with ESMF now work
   with the NCO weight generator.

Full release statement at http://nco.sf.net/ANNOUNCE

--
Charlie Zender, Earth System Sci. & Computer Sci.
University of California, Irvine 949-891-2429 )'(