Introduction of GeoVista at PyVista tutorial in SciPy2023

[tkoyama010]

@bjlittle

I will be working on the 7. PyVista in Action section of the pyvista tutorial in SciPy2023. I plan to spend a lot of time explaining GeoVista in this section.

If there are any GeoVista features or benefits you would like to see covered in the tutorial, please send us your comments. Suggested examples are also welcome.

Edit: The example here is particularly touching.

smc-timeseries.mov

[bjlittle]

Hey @tkoyama010,

This is amazing, thanks so much!

Sadly, I can't join you at the pyvista tutorial as I don't arrive at SciPy until late on Monday evening 😞. But I'd love to help support you as best as I can.

How can I easily do that? What do you need/want?

I've already banked quite a few examples, see geovista.examples, which showcase various initial use-cases. All the examples have associated data to make them executable, which geovista will download automatically. Also see the CLI i.e., geovista --help where you can easily download any/all geovista assets (geovista download --all) or run any/all of the examples (geovista examples --all --verbose).

Note that, geovista uses pooch to manage downloads of versioned assets.

Also see, geovista.samples (which uses geovista.pantry) which underpins the geovista.examples, and has a stock of meshes ready for you to visualise out-the-box and play with.

Obviously, one major feature is projection support, which allows you to convert your geo-located mesh in cartesian space to a target CRS (Coordinate Reference System). At the moment, only basic cylindrical and pseudo-cylindrical projections are supported, but as geovista matures a fuller treatment of those CRS will be available, and also other classes of projections, such as azimuthal and conic et al.

At the moment, I'm currently working on supporting coastline projections, so that should be available this week in geovista 0.3.0, along with gridline (graticule) support and kd-tree searching support.

We've also hinted at how the power of pyvista and vtk can be leveraged to extract mesh regions without using shapely i.e., see geovista.geodesic, in particular geovista.geodesic.BBox, geovista.geodesic.panel, geovista.geodesic.wedge and geovista.geodesic.line.

Essentially, further primitives will be provided in the near future that will allow users to easily extract regions from their data using caps, bands, shapefiles or other geometries e.g., extract the State of Texas, or the UK or whatever random polygon the user cares about.

In general, I'd say the following about geovista:

• It's early days for geovista, but I think the potential is kinda clear for people to see; for me the potential is huge
• I'm keen that geovista doesn't reinvent the wheel and simply "wrap" pyvista functionality. I want geovista to be a clean and simple gateway into the world of pyvista, but with cartographic conveniences
• I've put quite a bit of effort into geovista.bridge, which allows users to convert their uniform, rectilinear, and curvilinear grids and unstructured meshes to pyvista. So that's an attractive point to perhaps labour to your audience. We also recently added point-cloud support, see geovista.bridge.Transform.from_points
• I'm super excited about pyvista-xarray and the potential of geovista underpinning some of it's capability, but also I'm very keen to see the xarray community rally behind that project, simply because @banesullivan is going to do amazing things! I mean, a paradigm shift game changer, and that's pretty exciting... well, it is to me at least 😉

Let me know what you think... but I'm sure I can give you some additional material to support your introductory talk on geovista 👍

[tkoyama010]

@bjlittle
I have a question about the tutorial you created, about LFRic, is this article the most reliable information?
My specialty is the use of finite element methods in earthquake engineering, so I am impressed by the attention to the use of finite element and finite difference methods in the field of meteorology as well.

[tkoyama010]

I found the same link in README. I don't need to ask if I read the document more. Very sorry.

[tkoyama010]

Since there are still 3 minutes left in the tutorial time, I will explain how the Transform.from_2d method allows users to define their own longitude, latitude, and data with NumPy's masked_array to GeoVista.
Participants will be able to experience the possibility of connecting their numpy and xarray geographic data to GeoVista (and to PyVista)!

[bjlittle]

@bjlittle
I have a question about the tutorial you created, about LFRic, is this article the most reliable information?
My specialty is the use of finite element methods in earthquake engineering, so I am impressed by the attention to the use of finite element and finite difference methods in the field of meteorology as well.

@tkoyama010 You've done your research! I'm impressed!

I guess that's a reasonable public facing article to reference. It's a bit old now, but still highly relevant 👍

[bjlittle]

Since there are still 3 minutes left in the tutorial time, I will explain how the Transform.from_2d method allows users to define their own longitude, latitude, and data with NumPy's masked_array to GeoVista.
Participants will be able to experience the possibility of connecting their numpy and xarray geographic data to GeoVista (and to PyVista)!

@tkoyama010 Brilliant! That's a great idea. It sounds like the tutorial is shaping up to be really fun. I hope it all goes well.

There are plenty of examples (and associated data) that come with geovista if people don't have data but want to play around. Take a peek in geovista.examples for executable examples, where geovista will automatically download and cache the required data.

The geovista.examples get their data from the geovista.pantry, but there are also "ready-to-go" meshes built from the geovista.pantry data in geovista.samples.

Hopefully, you'll have enough data there to keep people busy 🥳

BTW I've just release geovista 0.3.0, so please update the dependencies of the tutorial 😀

[bjlittle]

Focusing on pyvista-xarray is a very good point of view! I will add an explanation about it in the tutorial. Can you tell us more about how the GeoVista project plans to utilize pyvista-xarray in the future?

I think the relationship is really the other way around. geovista won't be utilizing pyvista-xarray, it will be pyvista-xarray that will be utilizing geovista. Perhaps have a chat with @banesullivan to get a feel for his vision and roadmap of where he wants to take pyvista-xarray. But the potential is huge, given the size of the Earth Sciences xarray community.

[bjlittle]

@tkoyama010 Are you okay with me closing this discussion now?