Native Canopy And Tree Product Contract

Purpose

This page is the short source-of-truth contract for FEMIC’s native CHM, canopy structure, VRI uncertainty comparison, and optional individual-tree products.

Quick Contract

  • Canopy QA schema: femic.lidar.canopy_qa.v1.

  • Individual-tree schema: femic.lidar.individual_tree.v1.

  • CHM: DSM - DEM on exactly aligned native grids.

  • CHM negative heights: clamped by default and counted.

  • CHM cover thresholds: 2, 5, 10, and 20 m.

  • VRI height metric: CHM p95.

  • VRI cover metric: CHM cover at >=2 m.

  • No reprojection or resampling.

  • No new individual-tree segmentation algorithm.

  • No field-accuracy claim.

CHM Contract

DSM and DEM must share:

  • width and height;

  • CRS;

  • transform and native resolution;

  • projected metre units;

  • vertical datum.

Misaligned or geographic inputs are rejected. The CHM preserves the DSM grid and writes float32 values with a recorded nodata value. Negative DSM-DEM pixels are clamped to zero by default, counted, and warned about.

Canopy Structure Contract

The canopy structure report records:

  • valid and nodata pixels;

  • valid area hectares;

  • min, mean, median, p90, p95, p99, and max height;

  • percent cover at the configured thresholds.

The report is derived from the CHM, not a field inventory.

VRI Comparison Contract

VRI polygons must already match the CHM CRS. FEMIC clips each polygon to the CHM, requires a minimum valid-pixel count, and records:

  • polygon ID, VRI height, CHM p95, CHM mean, and height delta;

  • VRI cover, CHM cover at >=2 m, and cover delta;

  • valid-pixel count;

  • height bias, MAE, and RMSE;

  • cover bias and MAE.

Polygons outside the CHM or with too few valid pixels are skipped and counted. The comparison is uncertainty evidence, not proof of field accuracy.

Individual Tree Contract

An individual-tree product is optional. It must be point-based, have a CRS, and carry:

  • required tree_id and height_m;

  • optional crown_radius_m and species.

The summary records feature count, unique tree count, duplicate IDs, missing heights, height/crown distributions, species counts, geometry types, and warnings. Block summaries require tree and block CRS to match; trees outside all blocks are excluded and reported.

FEMIC does not implement tree segmentation. A source without a suitable public tree product is not given a synthetic one.

No-Resampling Rule

The canopy and tree workflows reject:

  • DSM/DEM grid or CRS mismatch;

  • geographic or non-metre CHM inputs;

  • VRI polygons in a different CRS;

  • tree/block CRS mismatch.

Consumer-side reprojection, if needed, is an explicit separate decision.

See Also