GeoPandas depends for its spatial functionality on a large geospatial, open source stack of libraries (GEOS, GDAL, PROJ). See the Dependencies section below for more details. Those base C libraries can sometimes be a challenge to install. Therefore, we advise you to closely follow the recommendations below to avoid installation problems.

Installing with Anaconda / conda

To install GeoPandas and all its dependencies, we recommend to use the conda package manager. This can be obtained by installing the Anaconda Distribution (a free Python distribution for data science), or through miniconda (minimal distribution only containing Python and the conda package manager). See also the installation docs for more information on how to install Anaconda or miniconda locally.

The advantage of using the conda package manager is that it provides pre-built binaries for all the required and optional dependencies of GeoPandas for all platforms (Windows, Mac, Linux).

To install the latest version of GeoPandas, you can then do:

conda install geopandas

Using the conda-forge channel

conda-forge is a community effort that provides conda packages for a wide range of software. It provides the conda-forge package channel for conda from which packages can be installed, in addition to the “defaults” channel provided by Anaconda. Depending on what other packages you are working with, the defaults channel or conda-forge channel may be better for your needs (e.g. some packages are available on conda-forge and not on defaults).

GeoPandas and all its dependencies are available on the conda-forge channel, and can be installed as:

conda install --channel conda-forge geopandas


We strongly recommend to either install everything from the defaults channel, or everything from the conda-forge channel. Ending up with a mixture of packages from both channels for the dependencies of GeoPandas can lead to import problems. See the conda-forge section on using multiple channels for more details.

Creating a new environment

Creating a new environment is not strictly necessary, but given that installing other geospatial packages from different channels may cause dependency conflicts (as mentioned in the note above), it can be good practice to install the geospatial stack in a clean environment starting fresh.

The following commands create a new environment with the name geo_env, configures it to install packages always from conda-forge, and installs GeoPandas in it:

conda create -n geo_env
conda activate geo_env
conda config --env --add channels conda-forge
conda config --env --set channel_priority strict
conda install python=3 geopandas

Installing with pip

GeoPandas can also be installed with pip, if all dependencies can be installed as well:

pip install geopandas


When using pip to install GeoPandas, you need to make sure that all dependencies are installed correctly.

  • fiona provides binary wheels with the dependencies included for Mac and Linux, but not for Windows.

  • pyproj, rtree, and shapely provide binary wheels with dependencies included for Mac, Linux, and Windows.

  • Windows wheels for shapely, fiona, pyproj and rtree can be found at Christopher Gohlke’s website.

Depending on your platform, you might need to compile and install their C dependencies manually. We refer to the individual packages for more details on installing those. Using conda (see above) avoids the need to compile the dependencies yourself.

Installing from source

You may install the latest development version by cloning the GitHub repository and using pip to install from the local directory:

git clone
cd geopandas
pip install .

It is also possible to install the latest development version directly from the GitHub repository with:

pip install git+git://

For installing GeoPandas from source, the same note on the need to have all dependencies correctly installed applies. But, those dependencies can also be installed independently with conda before installing GeoPandas from source:

conda install pandas fiona shapely pyproj rtree

See the section on conda above for more details on getting running with Anaconda.


Required dependencies:

Further, optional dependencies are:

  • rtree (optional; spatial index to improve performance and required for overlay operations; interface to libspatialindex)

  • psycopg2 (optional; for PostGIS connection)

  • GeoAlchemy2 (optional; for writing to PostGIS)

  • geopy (optional; for geocoding)

For plotting, these additional packages may be used:

Using the optional PyGEOS dependency

Work is ongoing to improve the performance of GeoPandas. Currently, the fast implementations of basic spatial operations live in the PyGEOS package (but work is under way to contribute those improvements to Shapely). Starting with GeoPandas 0.8, it is possible to optionally use those experimental speedups by installing PyGEOS. This can be done with conda (using the conda-forge channel) or pip:

# conda
conda install pygeos --channel conda-forge
# pip
pip install pygeos

More specifically, whether the speedups are used or not is determined by:

  • If PyGEOS >= 0.8 is installed, it will be used by default (but installing GeoPandas will not yet automatically install PyGEOS as dependency, you need to do this manually).

  • You can still toggle the use of PyGEOS when it is available, by:

    • Setting an environment variable (USE_PYGEOS=0/1). Note this variable is only checked at first import of GeoPandas.

    • Setting an option: geopandas.options.use_pygeos = True/False. Note, although this variable can be set during an interactive session, it will only work if the GeoDataFrames you use are created (e.g. reading a file with read_file) after changing this value.


The use of PyGEOS is experimental! Although it is passing all tests, there might still be issues and not all functions of GeoPandas will already benefit from speedups (one known issue: the to_crs coordinate transformations lose the z coordinate). But trying this out is very welcome! Any issues you encounter (but also reports of successful usage are interesting!) can be reported at or