Managing scientific projects often requires balancing system-level dependencies (GDAL, PROJ, HDF5) with Python package reproducibility. Conda and uv serve complementary roles:

• Conda excels at installing compiled libraries and cross-language dependencies.

• uv provides fast, modern dependency resolution and reproducible lockfiles for Python packages.

Why Hybrid?

• Geospatial and climate libraries (e.g., GDAL, rasterio) are best installed via Conda-forge.

• Python-only packages (numpy, pandas, scikit-learn) are faster and more reproducible with uv.

• Combining both ensures stability and speed.

Workflow Template

• Create Conda Environment

conda create -n hydro-env python=3.11 gdal proj 

or

conda env create -f environment.yml

Activate the new environment:

conda activate hydro-env

• Install uv Inside Conda

pip install uv

or

conda install conda-forge::uv

simply, installing uv with the virtual environment creating.

conda create -n hydro-env -c conda-forge python=3.11 gdal proj uv

• Initialize uv Project

Navigating into the project directory, and initializing it:

uv init
uv add numpy pandas rasterio

This creates pyproject.toml and uv.lock inside the project, aligned with Conda’s Python. Or synchronizing the environment when setting up on a new machine with uv.lock

uv sync

uv sync possibly upgrade newer versions of dependencies and updates uv.lock. For a production or research cases to guarantee reproducibility and prevent accidental upgrades, using:

uv sync --frozen

It will read uv.lock only and enforce exact versions.

• Run Scripts

uv run script.py

This ensures dependencies from uv.lock are resolved while Conda provides system libraries.

• Jupyter Integration

uv add jupyter

then:

uv run jupyter notebook or uv run jupyter lab

Best Practices

• Keep environment.yml for Conda (system deps).

• Keep pyproject.toml + uv.lock for uv (Python deps).

• Document both so collaborators can choose Conda or uv depending on their workflow.

• Align .python-version with Conda’s Python to avoid mismatches.

Conclusion

The hybrid approach leverages Conda’s strength in compiled libraries and uv’s speed in Python dependency management. For geo-related projects, this ensures reproducibility, portability, and efficiency across diverse research teams.

Related commands

• Export the Conda environment to a YAML file using this command:

conda env export > environment.yml

• Check the available Jupyter kernels:

uv run jupyter kernelspec list

• Remove a virtual environment managed by uv, simply delete the directory .venv.

• Remove a Conda environment:

conda deactivate
conda env remove --name hydro-env

References

Working on projects - uv Astral Docs

Python tools

Related

Colon, double colon and ellipsis python

Ports of python geospatial analysis tools python

Add text or annotation into the axes in matplotlib python

Setup font size and other groups in matplotlib python

Share on:

---

« PREVIOUS: Access home assistant remotely using Tailscal Funnel