NetCDF: How to mask/filter out non-land values in global dataset, preferably using Python and/or NCO?

Question

I have a global data at 0.25 degree resolution that I'd like to mask so that it only contains data values over land.

The data covers full 360 degrees in the lon dimension and from -60 to 60 degrees in the lat dimension.

The file header, as well as summary lat and lon coordinate values, are listed below:

netcdf cmorph_global_daily {
dimensions:
    lat = UNLIMITED ; // (480 currently)
    lon = 1440 ;
    time = 7305 ;
variables:
    float lat(lat) ;
        lat:units = "degrees_north" ;
        lat:long_name = "Latitude" ;
    float lon(lon) ;
        lon:units = "degrees_east" ;
        lon:long_name = "Longitude" ;
    float prcp(lat, lon, time) ;
        prcp:_FillValue = NaNf ;
        prcp:units = "mm" ;
        prcp:standard_name = "precipitation" ;
        prcp:long_name = "Precipitation" ;
        prcp:description = "CMORPH Version 1.0BETA Version, daily precip from 00Z-24Z" ;
    int time(time) ;
        time:units = "days since 1900-01-01" ;
        time:long_name = "Time" ;
        time:calendar = "gregorian" ;

// global attributes:
        :history = "Mon Mar 26 10:44:42 2018: ncpdq -a lat,lon,time cmorph_adjusted_daily.nc latlontime/cmorph_adjusted_daily.nc\nThu Mar 15 10:21:10 2018: ncks -4 cmorph_adjusted_daily.nc cmorph_adjusted_daily.nc" ;
        :nco_openmp_thread_number = 1 ;
        :title = "CMORPH Version 1.0BETA Version, daily precip from 00Z-24Z" ;
        :NCO = "4.7.2" ;
data:

 lat = -59.875, -59.625, -59.375, -59.125, ..., 59.125, 59.375, 59.625, 59.875 ;
 lon = 0.125, 0.375, 0.625, 0.875, 1.125, ..., 359.125, 359.375, 359.625, 359.875 ;

I would prefer to do this using Python/numpy and/or NCO since that's my typical toolset.

score 5 · Accepted Answer · answered Mar 29 '18 at 12:01

5

Once you have a variable on the same grid to mask with, you can use ncap2 where, e.g.,

ncap2 -s 'where(LANDMASK != 1) prcp=prcp@_FillValue' in.nc out.nc

If your mask is on a different grid than your data, you can use (on Linux/Mac) the masking features of ncremap, e.g., to remap your data to your mask (or visa versa) with something like

ncremap --msk_dst=LANDMASK -d mask.nc prcp_in.nc prcp_out.nc

answered Mar 29 '18 at 12:01

Charlie Zender

5,929
14
19

1

Obi-Wan strikes again. Thanks, Charlie. – James Adams Mar 29 '18 at 20:22

ClimateUnboxed · Answer 2 · 2021-05-04T10:23:19.527

4

A solution with CDO

# make landseamask on 0.25 grid with 1 for land and missing for sea
cdo -P 8 -f nc2 setctomiss,0 -gtc,0 -remapcon,r1440x720 -topo seamask.nc

# not sure if you need to do this but lets put your data on same grid
cdo remapcon,r1440x720 yourdata.nc yourdataremap.nc

# now mask the data
cdo mul yourdataremap.nc seamask.nc yourdatamasked.nc

hope that works...

EDIT: You can also simply make the landsea mask directly on the grid of your data in this way:

# make landseamask on 0.25 grid with 1 for land and missing for sea
cdo -P 8 -f nc setctomiss,0 -gtc,0 -remapcon,yourdata.nc -topo seamask.nc

# now mask the data
cdo mul yourdataremap.nc seamask.nc yourdatamasked.nc

edited May 04 '21 at 10:23

answered Mar 29 '18 at 10:51

ClimateUnboxed

7,106
3
41
86

I used the first part of this to get the mask file, it worked like a charm. Because my file is so large the remap step took forever, so I followed Charlie's approach above for remapping and masking. Thanks a lot for your help, Adrian. – James Adams Mar 29 '18 at 20:22
You could always use remapbil for bilinear interpolation, much much faster than conservative remapping – ClimateUnboxed Jan 21 '20 at 12:22

NetCDF: How to mask/filter out non-land values in global dataset, preferably using Python and/or NCO?

2 Answers2

Linked