Skip to content

Future COG Optimization #1

@rellimylime

Description

@rellimylime

Plan: COG optimization experiments (tiling, compression, overviews)

Motivation

COGs are currently generated using terra::writeRaster(filetype = "COG") with mostly default GDAL settings which may not be optimal for tile extraction or data storage.

Plan: run controlled optimization experiments.


Scope

This issue proposes a structured optimization phase to evaluate:

  • Runtime
  • Output file size
  • Practical usability

Experimental design

1. Select test rasters

Choose 1-2 representative rasters:

  • Domain score level (values 0-100)
  • One other (values 0-50)

2. Parameter grid

Test combinations of:

Compression:

  • LZW
  • DEFLATE
  • NONE

Blocksize:

  • 256
  • 512
  • 1024

Overview resampling:

  • AVERAGE
  • NEAREST

Threads fixed (e.g. NUM_THREADS=50).


3. Measurement

For each run, record:

  • input filename
  • parameters
  • wall-clock runtime
  • output file size
  • GDAL/terra version

Write results to a CSV for comparison.


4. Evaluation

Select defaults based on:

  • Reasonable runtime for large rasters
  • Acceptable file size
  • Simplicity and maintainability

Deliverables

  • Benchmark results CSV
  • Short written summary
  • Updated COG generation defaults

Notes

This optimization should occur after correctness is established and before final publication.

@FlukeAndFeather

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions