Kyle Barron

New Lonboard release makes the first step towards rich 𝐫𝐚𝐬𝐭𝐞𝐫 data visualization in Python, starting with PMTiles!

- New RasterLayer, supporting PMTiles input via async-pmtiles
- New support for categorical data filtering on the GPU

Release post: https://developmentseed.org/lonboard/latest/blog/2026/02/12/releasing-lonboard-014
Full Changelog: https://developmentseed.org/lonboard/latest/CHANGELOG/#0140-2026-02-11

Building on Vincent Sarago's work, I updated and published 𝐚𝐬𝐲𝐧𝐜-𝐩𝐦𝐭𝐢𝐥𝐞𝐬, an asynchronous PMTiles reader for Python: https://github.com/developmentseed/async-pmtiles

This integrates with Obstore for efficiently reading remote files and Obspec for extensibility.

Published a new version of async-geotiff:

- Support for reading band-interleaved GeoTIFFs and Cloud-Optimized GeoTIFFs
- Initial support for decoding LERC, LERC_DEFLATE, LERC_ZSTD images
- Fix computation of image bounds and resolution for rotated images

https://github.com/developmentseed/async-geotiff/releases/tag/v0.2.0

Introducing 𝐀𝐬𝐲𝐧𝐜-𝐆𝐞𝐨𝐓𝐈𝐅𝐅, a new high-level library for reading GeoTIFFs and Cloud-Optimized GeoTIFFs. https://developmentseed.org/async-geotiff/latest/blog/2026/02/03/introducing-async-geotiff/

- High-level & easy-to-use
- Load from full-resolution or reduced-resolution overviews
- Fast, with a Rust core
- Uses a non-blocking thread pool for image decoding
- Integration with NumPy, PyProj, Affine, & Morecantile
- Remote data support with Obstore
- Lightweight with no GDAL dependency
- Full type hinting
- Broad decompression support

A @developmentseed@fosstodon.org project.

We've released a new version of async-tiff. Fast, asynchronous reading of TIFFs, GeoTIFFs and Cloud-Optimized GeoTIFFs in Rust and Python. https://developmentseed.org/async-tiff/latest/

- New Array class for zero-copy data sharing between Rust and numpy
- Easily separate CPU-bound and IO-bound work, with native support for decoding image data on a Rust thread pool
- Revamped block caching for reading metadata
- Expanding tests
- Revamped Rust API for fetching TIFF metadata and image data

https://github.com/developmentseed/async-tiff/releases/tag/py-v0.4.0

Visualize a mosaic of NAIP imagery loaded from Microsoft Planetary Computer's STAC Collection, all in your browser, *no server involved*.

Switch between true color, false color infrared, and NDVI renderings. https://developmentseed.org/deck.gl-raster/examples/naip-mosaic/

Each of these tiles loading are individual Cloud-Optimized GeoTIFFs, loaded directly from Azure. Only the COGs in viewport are loaded and rendered.

Built at @developmentseed@fosstodon.org with deck.gl-raster: https://github.com/developmentseed/deck.gl-raster

@cartocalypse@norden.social Yes, yes, strictly speaking you could call it a server, but for the purposes of this discussion it's at least a very reliable server that is very inexpensive and requires none of your own maintenance. The goal is to not need to always deploy your own additional server just to view raster imagery

@cartocalypse@norden.social Yes! It's loaded directly from S3 using geotiff.js using HTTP range requests.

New from Development Seed: deck.gl-raster. 𝐅𝐮𝐥𝐥𝐲 𝐜𝐥𝐢𝐞𝐧𝐭-𝐬𝐢𝐝𝐞 𝐂𝐎𝐆 𝐫𝐞𝐧𝐝𝐞𝐫𝐢𝐧𝐠. No server required.

This is a 1.3 𝐠𝐢𝐠𝐚𝐛𝐲𝐭𝐞 COG, streamed directly into the browser. Explore it yourself: https://developmentseed.org/deck.gl-raster/examples/land-cover/

- GPU-accelerated raster reprojection supports images from nearly all projections
- GPU-accelerated image processing to apply colormaps, filter out nodata values, convert color spaces, & more
- Efficient use of COG overviews to fetch the minimum data required

https://github.com/developmentseed/deck.gl-raster

Released Lonboard v0.13!

- New layer types: H3, S2, Geohash, A5 and fixed HeatmapLayer
- Render data layers beneath Maplibre text labels
- Map UI controls: scale, fullscreen, navigation
- Significant performance improvements: fixing earlier regressions

https://developmentseed.org/lonboard/latest/blog/2025/11/05/releasing-lonboard-013

I'm happy to announce I've been promoted to an Apache Arrow Committer, meaning I'm authorized to merge code patches into Arrow repositories!

I join this list of amazing people: https://arrow.apache.org/committers/

New blog post! https://kylebarron.dev/blog/fast-python-bindings

This dives into performance pitfalls of Python bindings, why projects like Lonboard are so fast, and why I care about Apache Arrow so much. The post is high level and you don't need to know C/C++/Rust to understand it!

New Lonboard release and new demo! Integrating Marimo and Apache DataFusion to visualize the NYC taxi dataset. https://developmentseed.org/lonboard/latest/examples/marimo/nyc_taxi_trips/

Published a new version of parquet-wasm: https://github.com/kylebarron/parquet-wasm/releases/tag/v0.7.0

- Update to latest version of Rust parquet crate
- Workaround to read Parquet data containing Arrow string view/binary view types
- Streaming writes
- Comparison between parquet-wasm and hyparquet, a pure-JavaScript Parquet reader

Introducing Obspec: A Python protocol for interfacing with object storage

Obspec is an alternative to fsspec. It defines a minimal Python interface to read, write, and modify data on object storage while acknowledging that object storage is *not* a filesystem.

Obspec formalizes the Obstore API. You can build around Obspec and be generic around backend implementation used at runtime. It also provides an avenue for caching to be built as a wrapper on top of Obspec.

https://developmentseed.org/obspec/latest/blog/2025/06/25/introducing-obspec-a-python-protocol-for-interfacing-with-object-storage/