In general, MetPy tries to support minor versions of dependencies released within the last two years. For Python itself, that means supporting the last two minor releases.
matplotlib >= 2.1.0
numpy >= 1.16.0
scipy >= 1.0.0
pint >= 0.10.1
pandas >= 0.22.0
xarray >= 0.13.0
traitlets >= 4.3.0
pooch >= 0.1
The easiest way to install MetPy is through
pip install metpy
conda install -c conda-forge metpy
The source code can also be grabbed from GitHub. From the base of the source directory, run:
pip install .
This will build and install MetPy into your current Python installation.