SIParCS 2022 - Philip Chmielowiec
Python Visualization, Analysis & Jupyter Notebook Development for Unstructured Grids
The current Scientific Python Ecosystem (SPE) has had decades of support and contribution for regular structured grids. However, the same is not the case for unstructured grids. This lack of maturity means that typical data science routines such as analysis and visualization often prove to be challenging.
Project Raijin is an NSF EarthCube funded project aimed at solving this issue. This summer, I worked on developing a tool to represent unstructured grids as a mesh of polygons for visualization. Due to the variable resolution and mixed-topology of grid cells, unstructured grids can’t be visualized with the same pipelines as standard structured grids. By representing our grid as a mesh of polygons through the use of highly-optimized geometry libraries (SpatialPandas, PyGEOS), we get a considerable speedup when constructing our mesh compared to other methods (Delaunay Triangulation). Combined with visualization libraries such as Datashader, it allows for the rendering of millions of polygons in less than a few seconds.
Mentors: Orhan Eroglu, Alea Kootz, Anissa Zacharias, Michaela Sizemore
Slides and poster