H3 to World Cover

OVERVIEW

An experiment in building a pipeline that downloads ESA WorldCover 10m land-cover tiles and maps their pixel data onto H3 hexagon cells, producing a per-cell breakdown of land-cover composition (built-up, trees, water, cropland, etc.) as percentages. The pipeline is designed to run unattended on a single machine against an external SSD - fetching tiles one by one, computing partial results, and aggregating everything into a PostgreSQL table - while keeping raw tiles around so cell composition can be recomputed at other H3 resolutions later.

FEATURES

• →Fetch ESA WorldCover 10m tiles from the official S3 grid (2,651 tiles covering the globe)
• →Map WorldCover land-cover pixels onto H3 hexagon cells at a configurable resolution (starting at 7)
• →Compute per-cell land-cover composition across all 11 ESA WorldCover classes as basis-point percentages
• →Resumable, tile-by-tile pipeline that survives interruption via a persisted state file
• →Produces a PostgreSQL table mapping H3 index to land-cover composition, ready for querying
• →End-to-end run against an external SSD, holding raw tiles, partial results, aggregates, and the database in one place
• →Single-cell CLI tool for quickly inspecting one H3 cell's composition against a downloaded tile
• →Retains raw downloaded tiles so cell composition can be regenerated at a different H3 resolution later

TECHNOLOGY