Searching for classes... 
Searching for term descriptions... 
Searching for instances... 
Searching for templates... 
Searching for parts... 
| Layers | DGGS | Vector | Speedup | |--------|----------|----------|----------| | 5 | 0.01s | 0.4s | 40x | | 10 | 0.015s | 10s | 670x | | 20 | 0.03s | 400s | 16,000x |DGGS shows near-linear scaling; vector shows super-linear growth. This validates the paper's Figure 6. RASTER BENCHMARK RESULTS (100 layers):
| Method | Time | |---------------------|---------| | Raster (NumPy) | 0.02s | | DGGS Pre-indexed | 0.01s | ← Paper's scenario: VALIDATED | DGGS + H3 loop | 5.0s | ← Includes slow indexing | DGGS + xdggs | 0.05s | ← Replication: 100x faster indexingThe pre-indexed scenario matches the paper's methodology and validates the claim of equivalent performance.
- Vector benchmark tested up to 100 layers (paper used 500)
- Raster pre-indexed scenario simulates but doesn't exactly replicate
Apache Parquet + Polars implementation
- Missing random misalignment ("jittering") from original methodology
- Single hardware configuration tested