Suggestions
Chris Holmes
Planet Fellow, Open And Vice President Product And Strategy at Planet
Highlights
Have you been working with GeoParquet and embeddings? If so we'd love to hear from you - we're looking to gather all the practices of how people are storing their data, and hopefully work together to figure out a set 'best practices' in a mini-specification.
The first step is to join our new github repo at https://lnkd.in/dMX8r9qm. So far we've got examples from Earth Genome's Earth Index and from Clay's foundational model - thanks Ben Strong for helping start the repo, and Bruno Sanchez-Andrade Nuño for some of the first contributions.
This came out of Jason Gilman's post asking about combining GeoParquet plus plus embedding search in https://lnkd.in/dAFpVHjj
In a couple weeks we'll survey the contributions, so if you'd like to take part in the conversation just make a PR or even just an issue explaining how you're using GeoParquet with embeddings.
My talk for FOSS4G North America just went up! It presents the idea of a 'Cloud-Native Spatial Data Infrastructure' - a radical simplification of the core idea of making geospatial data accessible & interoperable, using Cloud-Native Geospatial Foundation standards.
Check the talk out at https://lnkd.in/gGgVxkpq
The quick overview is that the driver for me for this concept is The driver for me is to enable 'Queryable Earth', presented by Will Marshall 5 years ago in https://lnkd.in/gRJpetz8 which I dove deeper on in my 2018 Foss4g-NA talk at https://lnkd.in/gBcKzUsh
The Spatial Data Infrastructure concept has been around for years, but it's clearly not lived up to its promise when only very limited spatial data is queryable from ChatGPT or other LLM's. I believe that's because the architecture astronauts drove the SDI effort & made things unreasonably complicated: the way for a GIS person at a municipality to share 3 vector layers is to stand up a database plus WMS, WFS and CSW, with dynamic mapping from the DB to a GML schema.
The rapidly maturing Cloud-Native Geospatial Foundation standards of #COG, #STAC, #Zarr, #COPC provide an alternate route. I went more into #PMTiles and #GeoParquet as the two specs to complete the suite, and touched on concepts I explored in https://lnkd.in/dQTgM_mx I believe this approach should be cheaper, scale better and become easier to use than the existing SDI approach. Scale is handled by cloud object stores, the most robust part of their entire infrastructure. Publishing data is just uploading a file, and the main cost is storage.
It's an exciting time in geospatial, as I believe we as a community are close to realizing the potential of the SDI vision. The key to me is that we focus on global datasets, like Overture Maps Foundation, and build the software and standards to support data distribution and access.
Thanks to Eddie Pickle and the FOSS4G North America organizing committee for inviting me and putting together a really great panel and conference. I look forward to seeing everyone at the next one, and to collaborating on building the cloud-native spatial data infrastructure!
