http://nco.sf.net (Homepage, Mailing lists, Help)
http://github.com/nco/nco (Source Code, Issues, Releases)
What's new?
Version 5.3.0 significantly improves the unpacking of ELM/CLM/CTSM
history and restart datasets, and seems complete in this regard.
ncrcat has more friendly behavior when appending, and ncclimo now
outputs more metrics (such as valid area per gridcell) when in
regional statistics mode. This release many be skipped if these
changes do not interest you.
Work on 5.3.1 is underway and includes improved NC_STRING support
for autoconverting from netCDF4->netCDF3 for common workflows.
Enjoy,
Charlie
NEW FEATURES (full details always in ChangeLog):
A. Once again, ncks leveled-up how it unpacks ELM/CLM history/restart
files. New features include automatic writing of fractional landunit
and fractional column variables frc_landunit and frc_column,
respectivly. frc_landunit is equivalent to the variable PCT_LANDUNIT
sometimes placed in history files, but is expressed as a decimal
fraction (not a percent) and is derived from restart files.
The landunit dimension of frc_landunit is of size ten: Index zero
contains the sum (should be 1.0) of standard landunit types/indexes
1-9 soil, crop, ..., urban medium density).
frc_column is a bespoke version of the same, only for column (not
landunit) types. For datasets with Multiple Elevation Classes (MECs):
index = 0 is soil column, index = 1 is MEC == 1, indexes 2..10
are remaining MEC columns, index 11 is sub-total of MEC columns, index
= 12 is deep lake column, index = 13 is wetland column, and index = 14
is grand total of natural columns. For non-MEC datasets, index = 0 is
soil column, index = 1 is glaciated column, index = 2 is deep lake
column, index = 3 is wetland column, and index = 4 is grand total of
natural columns.
ncks --s1d --hrz=history.nc restart.nc s1d_r05.nc # Restart file
ncks -m -v frc.? s1d_r05.nc
netcdf s1d_r05 {
dimensions:
column = 15 ;
landunit = 10 ;
lat = 360 ;
lon = 720 ;
variables:
float frc_column(column,lat,lon) ;
float frc_landunit(landunit,lat,lon) ;
}
http://nco.sf.net/nco.html#s1d
http://nco.sf.net/nco.html#unpack_sparse
B. ncks can now grid snow-related variables into a top-down
(ocean-like) vertical grid that many think is more intuitive.
By default ELM/CLM/CTSM store the negative of the number of active
snow layers in the variable SNLSNO. Restart files for these models
store the active snow layer butted-up against the lowest layers in
the level dimension (so that they are continguous with soil layers
to simplify hydrologic calculations). This makes snow variables in
restart files hard to visualize. By default S1D now uses SNLSNO, if
present, to unpack snow variables in a first layer downwards order,
increasing with depth. Inactive layers at the bottom (i.e., where they
reside physically). The resulting snow variables appear like ocean
state variables over uneven bathymetry, with missing value
underneath. We call this "snow-ocean" ordering to contrast it with
the on-disk storage order of snow variables.
http://nco.sf.net/nco.html#s1d
http://nco.sf.net/nco.html#snw_ocn
http://nco.sf.net/nco.html#snw_rdr
http://nco.sf.net/nco.html#no_snw_ocn
C. ncks S1D now accepts the "--rgr lut_out=$lut_out" option to specify
which column type(s) to place in each gridcell in the output file.
The argument lut_out is the standard landunit type of the column,
with (thus far) two additional values that will output area-weighted
averages of multiple landunit types:
lut_out Output will be value of column(s) in this Landunit
1 Vegetated or bare soil
2 Crop
3 Landice (plain, no MEC)
4 Landice multiple elevation classes
5 Deep lake
6 Wetland
7 Urban tall building district
8 Urban high density
9 Urban medium density
10 Area-weighted average of all landunit types except MEC glaciers
13 Area-weighted average of soil+(non-MEC) glacier
For example,
ncks --s1d --rgr lut_out=1 --hrz=hst.nc rst.nc s1d.nc # Output Soil LUT
ncks --s1d --rgr lut_out=13 --hrz=hst.nc rst.nc s1d.nc # Avg Soil+Glacier
http://nco.sf.net/nco.html#s1d
http://nco.sf.net/nco.html#lut_out
D. ncremap always warns when it regrids land model files that contain
a "landfrac" variable and SGS renormalization is not requested.
This is because ELM/CLM/CTSM use "landfrac" for sub-gridscale
fraction, and ignoring that causes errors along coastlines.
EAMxx also archives that same "landfrac" but every EAMxx variable
is valid throughout every gridcell so renormalization would be
superfluous at best. Now ncremap no longer emits the "landfrac"
WARNING for EAMxx files.
http://nco.sf.net/nco.html#eamxx
E. As of NCO version 5.3.0 (December, 2025) ncclimo automatically
outputs additional metrics with global statistics. The output files
containing the global timeseries also contain the variable
valid_area_per_gridcell. This field is equivalent to the product of
the area variable and the sgs_frc variable (if any). Thus for
ELM/CLM/CTSM, this field equals area times landfrac, while for EAM/CAM
this variable simply equals area. The output files also contain the
area and sgs_frc variables separately. The presence of these variables
in output allows downstream processors (e.g., zppy) to generate
additional masks and weights for rescaling the statistics. For
example, these fields can be used to rescale global sums into any
units desired.
ncclimo -P elm --split --rgn_stt=sum -c v3.LR.piControl -s 460 -e 461 \
--drc_in=drc_in --drc_out=drc_out
http://nco.sf.net/nco.html#glb_stt
http://nco.sf.net/nco.html#rgn_stt
BUG FIXES:
A. ncrcat ensures that the last time value and monotonicity are
both defined before the monotonicity check is performed in REC_APN
mode. This fixes overly noisy false-negative WARNINGs about
monotonicity. Thanks to Seth McGinnis (NCAR) for reporting this
problem.
Full release statement at http://nco.sf.net/ANNOUNCE
--
Charlie Zender, Dept. of Earth System Science
University of California, Irvine 949-891-2429
