FOSS4G 2017 - Geonotebook: an extension to the jupyter notebook for exploratory geospatial analysis (foss4g 2017)
- 1. GeoNotebook An extension to the Jupyter Notebook for exploratory geospatial analysis Christopher Kotfila, Aashish Chaudhary - Kitware Inc. Petr Votava - NASA Ames FOSS4G 2017
- 2. What is GeoNotebook? ● Python & Javascript extensions to Jupyter Notebook ○ Interactive map connected to python execution environment via RPC. ○ Integrated tile server for rendering geospatial data onto the map. ● Supports interactive exploration and subsetting of geospatial data (especially raster data).
- 4. Development Context ● NASA ESTO - AIST 2014 Grant ○ Public cloud tools for extending analysis capabilities. ○ Bring analysis to the data, rather than data to the analyst ● NASA Earth Exchange (NEX) @ NASA Ames ○ Portal for collaborative sharing network for researchers ○ Large scale data processing pipelines for enabling ‘science-quality’ satellite data products.
- 6. Technical Objectives ● No assumptions about relationship between data access and data visualisation. ○ Data access always at native resolution ● Support custom visualisation/data sources via third party python packages ● Sensible defaults; target “traditional” Jupyter user.
- 7. Technical Stack ● GDAL - file I/O with custom VRT generation ● Mapnik - styling and orchestration of downsampling ● KTile - TileStache fork for serving tiles ● Jupyter Notebook - plugins + custom python kernel ● GeoJS/OpenLayers - interactive presentation of map
- 8. Current Limitations ● Complex deployment of tile rendering stack (Mapnik/GDAL); docker image available ● Re-rendering transformations of data subsets (e.g rendering in-memory data). ● Interactive elements break some of the implicit promises about reproducibility of notebooks.
- 9. Call to Action ● Originally developed as a demonstration of what was possible for Q/A on NEX data processing pipelines. ● Still early days ○ Apache 2.0 licensed software seeks interested community of users & contributors. ○ Cathedral and the Bazaar - “plausible promise”
- 10. Many Thanks! Contributing projects: GDAL, Mapnik, TileStache, Jupyter Notebook, NumPy, Fiona, Shapely, Rasterio, Pytest and many others. Contributing humans: Mike Little†, Raman Nemani†, Petr Votava, Andrew Michaelis†, Jon Beazley‡, Doruk Ozturk‡, Dan Lamanna‡ † NASA, ‡ Kitware Inc.
- 11. Questions! Github: OpenGeoscience/geonotebook Gitter: gitter.im/OpenGeoscience/geonotebook RTD: geonotebook.readthedocs.io Email: chris.kotfila@kitware.com docker run --net=host -v /path/to/notebooks:/notebooks geonotebook/geonotebook:latest
- 15. Nascent Geotrellis integration Geotrellis & GeoPySpark Deep Learning for Semantic Segmentation of Aerial Imagery