Its a well researched area. My understanding is for most use cases and data like this R trees outperform as bounding box comparisons are fast to run and the bounding boxes tend to be well organised to chunk data efficiently. H3 is a looser area and you may find lots of your points are clustered in a few grids so you end up doing more expensive detailed intersection calculations. Of course it all depends a little on your data, use case and to some extent the parameters chosen for the spatial index. But I think safe to say now based on industry experience that r trees do a very good job 99.9% of the time.
You can of course also use h3 in postgis directly as well as r trees. Its helps significantly for heatmap creation and sometimes for neighbourhood searches.
Comments
Its a well researched area. My understanding is for most use cases and data like this R trees outperform as bounding box comparisons are fast to run and the bounding boxes tend to be well organised to chunk data efficiently. H3 is a looser area and you may find lots of your points are clustered in a few grids so you end up doing more expensive detailed intersection calculations. Of course it all depends a little on your data, use case and to some extent the parameters chosen for the spatial index. But I think safe to say now based on industry experience that r trees do a very good job 99.9% of the time.
You can of course also use h3 in postgis directly as well as r trees. Its helps significantly for heatmap creation and sometimes for neighbourhood searches.