H3/PostGIS: Hands-on example

Darafei Praliaskouski, member of the #PostGIS PSC, walks through an interesting hands-on case study of modelling #LoRa radio tower range in the mountains of Georgia using #H3 and PostGIS. The example covers data preparation, #DGGS indexing, and spatial analysis, and offers some food for thought.
Author

Ralph Straumann

Published

February 27, 2026

From Darafei Praliaskouski, member of the PostGIS PSC1 and geospatial consultant, comes a practical example of using H32 with PostGIS. The example was shown in a talk on PostGIS day. It demonstrates how to use H3 to index geospatial data in PostGIS, allowing for efficient querying and analysis.

The task involves modelling radio tower – more precisely: LoRa3 – range in the mountains of Georgia. Early in his talk, Darafei gives a light technical overview of H3. I liked the point that H3 is “optimized for the analyst’s mental health” and Darafei’s comment: “If you are getting annoyed by H3, you are doing it wrong” 😁.

Next, Darafei walks through the data and data sources for the case-study. The data includes all different types of geospatial data:

Next, Darafei shows the steps to prepare and DGGS5-index the data with H3 (this mainly involves ogr2ogr, raster2pgsql and SQL, as well as osmium for the OSM data) and, finally, how to use PostGIS to combine and query the indexed data. The final result is a proposed LoRa mesh optimized for population coverage, among other criteria.

Overall, Darafei’s case-study is a great demonstration of how H3 can be used to conduct, maybe simplify, geospatial analysis in PostGIS and Darafei gives some valuable, very applied H3-related advice, for example on resolutions.

Uncertainty and trade-offs

The case-study of course also shows how H3 (or any DGGS) simplifies6 the underlying data (especially, if the original data is of high resolution or high accuracy). This is the trade-off you will have to make when using H3 or any DGGS: You get a more efficient and simpler data structure, but you also lose some of the original data’s detail and accuracy.

There is certainly need for research in connection to this, for example:

  • How does indexing affect the accuracy of the data itself and of various geospatial algorithms (for example, terrain analysis as shown in the example)?
  • What amount of uncertainty is introduced?
  • What are sensible trade-offs in this area?

Not limiting itself to uncertainty, this paper by Hojati et al. seems to identify some of the key challenges related to DGGSs:

With renewed interest in DGGS brought on by a recently proposed standard, the demands of big data, and growing needs for monitoring environmental changes across a variety of scales, we seek to highlight current challenges that we see as central to moving the field(s) and technologies of DGGS forward. (…) Through articulation of these challenges, we hope to identify a clear research agenda, expand the DGGS research footprint, and provide some ideas for moving forward towards a scaleable Digital Earth vision.

Interesting stuff to think about.

Footnotes

  1. Project Steering Committee.↩︎

  2. H3 is a hexagonal discrete global grid system (DGGS) created by Uber.↩︎

  3. LoRa (Long Range) is a low-power, wide-area networking modulation technique used in IoT applications for transmitting small data packets over long distances.↩︎

  4. OpenStreetMap, a collaborative, open-licensed map of the world.↩︎

  5. Discrete Global Grid System.↩︎

  6. You may also say “distorts”.↩︎