nc2csv

A fast, specialized netCDF to CSV converter.

Build

This program builds with C++11, and depends on:

• libnetcdf
• libnetcdf_c++4
• pybind11

...so these should be installed in a standard location.

The use of bybind11 is to leverage some very useful features of the netCDF4 python library, so this python package should also be installed globally (though this is a run-time dependency, not a build-time one.)

This is a CMake project, so if the dependencies are met, you can build with:

cmake .
make

Using cmake -DCMAKE_BUILD_TYPE=Debug . will compile with debug symbols and no optimization - otherwise optimization at -O3 will be used.

Docker

To avoid dependency hell, nc2csv can also be run through Docker. A pre-built docker image exists at breadcrumbs2270/nc2csv, but you can also build an image yourself from this repo's Dockerfile.

nc2csv's docker image simply drops you into a shell with a compiled executable at /nc2csv/nc2csv. To expose files to docker, you can use a volume, as in:

docker run -v "$PWD:/ncfiles" -it breadcrumbs2270/nc2csv

This will create a directory named /ncfiles that has the contents of $PWD. You can then run nc2csv /ncfiles/* to convert the files, which will appear inside of $PWD.

Usage

Supply every file you want to convert as an argument, e.g.

./nc2csv netcdffile1.nc netcdffile2.nc ...

Each one will be converted to a file with the same name, having .nc removed and replaced with .csv. This means they will appear in the same directory as the original files.

What will be output is a CSV file with the following header:

time,lat,lon,<another variable>

... where "another variable" is the first variable encountered in the file that is NOT either "time", "lat", "long", "crs", "time_bnds", "lat_bnds", or "lon_bnds". You cannot manually specify a variable to convert, so currently this is only useful for netCDF files that have one variable, e.g. the TerraClimate datasets.

Options

• -c n | -concurrency n specifies the amount of threads that should run. Each thread will convert exactly one of the .nc files passed to the program. Many threads will consume a lot of memory but will convert multiple files significantly faster. Defaults to 3.
• -d dimension1,dimension2,dimension3... switches the program to dimension mode, where it will not output variables but instead simply the values of each of the dimensions provided. This is useful, for example, to get a list of the unique lat/long points or timestamps in a dataset. Note that the argument to -d is a comma-separated list of dimensions.
• -t dimension specifies the name of the time dimension. Necessary on datasets like Gridmet that deviate from the standard "time" name. Defaults to "time".
• -a dimension specifies the name of the latitude dimension. Defaults to "lat".
• -g dimension specifies the name of the longitude dimension. Defaults to "lon".

Options must be placed before any files passed to the program.