Conceptually, I think I have brought up the notion of using, say, a quadtree to determine a specific "zoom" level to coalesce the network graph at. I think that the meter-projection variation of this is more ideal, given that it produces a more controlled mechanism (meter-distance) for coalescing the data.
Meanwhile, allowing for a second option, such as Google'sS2geometries, would allow for a more dynamic clustering operation which would account for high-intensity node clusters, while summarizing more sparsely populated areas with large geometry clusters.
Similarly, H3 may be appropriate for the same reasons.
Conceptually, I think I have brought up the notion of using, say, a quadtree to determine a specific "zoom" level to coalesce the network graph at. I think that the meter-projection variation of this is more ideal, given that it produces a more controlled mechanism (meter-distance) for coalescing the data.
Meanwhile, allowing for a second option, such as Google's S2 geometries, would allow for a more dynamic clustering operation which would account for high-intensity node clusters, while summarizing more sparsely populated areas with large geometry clusters.
Similarly, H3 may be appropriate for the same reasons.