Hilbert indexing on parquet compressed files

This is very interesting. I'm wondering if you could expand this to index compressed files, like parquet, which is inherently columnar storage, e.g. for a use case with 100Gb geospatial data daily (uncompressed) over multiple years. I could see perhaps having chuncks of parquet files that are hilbert indexed and perhaps even subindexed within each file chunk. Thoughts?