NCToolkit will not run on Windows, because of system dependency limitations in its computational backend. Use WSL if you're on Windows.
This is the recommended route: it installs NCToolkit and all of its system dependencies (CDO, NCO) in one go, and it just works out of the box.
conda install -c conda-forge nctoolkit
NCToolkit is also on the Python Package Index. This installs the core dependencies only — you'll then need to install the system dependencies yourself.
pip install nctoolkit
For slightly better default plots, install with the complete extra, which pulls in optional plotting dependencies:
pip install nctoolkit[complete]
NCToolkit relies on two command-line tools to do its heavy lifting:
CDORequired
Climate Data Operators is the computational backend for almost every NCToolkit method. NCToolkit will not work without it.
NCOOptional
NetCDF Operators is used by a handful of methods that offer it as an alternative backend to CDO, via nco_command.
The easiest way to install either is with conda:
conda install -c conda-forge cdo conda install -c conda-forge nco
On Ubuntu, CDO is also available through apt:
sudo apt install cdo
If you installed the non-complete version from PyPI, install cartopy for nicer map plots. Cartopy has some extra system dependencies, so prefer installing it with conda — it resolves the environment far more reliably here than pip:
conda install -c conda-forge cartopy
See cartopy's own installation guide for platform-specific notes.