In the digital plotting space, there are many open source projects which seek to display and organize datasets. One such project is GeoViews, but unlike many of the other projects, it brings geographic plotting to the next level. This project is an impressive tool for anyone in the geographical, meteorological, or oceanographic data science world.

Open Source Directions recently interviewed Philipp Rudiger, who works on consulting projects for Anaconda Inc. and is a core developer and maintainer of various PyViz projects, including GeoViews. The idea for this project came as Philipp was working with a client from Met Office, the UK meteorological agency. Initially, they tried to take Iris gridded datasets, which were capable of showing gradient datasets for things like temperature or precipitation, and make it easy to explore that data in Matplotlib/Cartopy so that they could visualize the datasets. This process was arduous and the team needed an easier way to project different datasets and see the dataset evolve over time or over other variables. Their solution was GeoViews.

This project has continued to improve with time, and now they offer greater controls through the integration of Bokeh. In addition to these integrations, they have also been integrating Datashader and Dask so that the user can dynamically shade these datasets, and scale visualizations to datasets that would be impossible to view with classical plotting tools.

This project is particularly visual so the demo was a great opportunity to really see what GeoViews was able to do. If you want to see that demo, the video is on YouTube here, at timestamp 12:09.

Looking at where GeoViews is headed, there are many areas of planned improvements. Specifically, these are areas that the project could benefit from contributors of time or resources. The first issue they would like to tackle is scalability. Currently, it is difficult to use this on a massive scale because it is limited to your available memory resources, however, a contributor could help create a Dask backed rasterization of rectilinear, curvilinear, and unstructured meshes. Another big improvement area would be to expose more custom renderers, building on libraries such as CesiumJS and/or Deck.GL, which would allow for 3D rendering of your datasets. These are just two examples of how the project is planned to develop, but it needs support from users and community members. To see the complete roadmap, visit the Quansight page here.

We enjoyed having Philipp on the webinar and look forward to the future development of the GeoViews project!