Post-fire recovery

WS4 can consume the public NTEMS post-disturbance recovery layer and attach a recovery status to blocks. The layer is a spectral recovery product, not a tree-survival or stocking survey.

Source

  • Canada Forest Post-Disturbance Recovery Rate, fire layer.

  • Natural Resources Canada / NTEMS, 30 m, EPSG:3978.

  • Wildfire disturbance 1986-2012; observation 1985-2017.

  • Classes: 1 faster than the ecozonal baseline, 2 slower, 3 not recovered by the end of 2017.

  • Open Government Licence - Canada.

  • Citation: White et al. 2022, Mapping, validating, and interpreting spatio-temporal trends in post-disturbance forest recovery, Remote Sensing of Environment 271, 112904.

Contract

ws4 recovery ingest samples the raster at block polygons and writes:

regen_status.csv

block_id, area_ha, burned_area_ha, valid_pixels, faster_fraction, slower_fraction, not_recovered_fraction, dominant_status, dominant_fraction.

recovery_metadata.json

Source version, disturbance/observation windows, resolution, CRS, licence, citation, class map, and notes.

A block with no class 1-3 pixels is outside_footprint. That is not a failure status and is never imputed as one. The dominant status is the largest class fraction; tied classes are resolved deterministically by the class order.

regen_status_summary.csv groups blocks, total area, and burned recovery area by dominant status.

Commands

ws4 recovery ingest \
  --raster recovery_tsa46_5km.tif \
  --blocks blocks.geojson \
  --block-id-field FEATURE_ID \
  --output regen_status.csv \
  --metadata recovery_metadata.json \
  --summary regen_status_summary.csv \
  --source-version "NTEMS post-disturbance recovery rate / White et al. 2022"

Pilot

examples/recovery_tsa23 applies the layer to the TSA 23 instance, which is the fire-oriented estate. The 30 m subset covers 102,128 blocks and 988,816 ha.

  • 1,555 blocks are in the fire-recovery footprint.

  • 1,497 blocks are not recovered: 21,948 ha, 5,727 ha burned.

  • 58 blocks are slower: 745 ha, 176 ha burned.

  • No block is dominated by faster recovery.

  • 100,573 blocks are outside_footprint, not failures.

  • The existing 2,000-row baseline schedule contains 29 not-recovered blocks (212 ha) and 2 slower blocks (8 ha); 1,969 scheduled blocks are outside the footprint.

Using the 44,388-block ft_warm schedule, 411 not-recovered blocks (6,048 ha) are scheduled. Deferring those treatments changes harvested volume from 138.50M to 137.12M m3 (-1.00%) and treated area by -3,579 (-1.13%). Deferring both not-recovered and slower blocks changes volume by -1.03%. The block area is larger than the treated-area change because some scheduled treatments are not operable.

On the full 20-year TSA 23 abatement_bc/sim_6400000000 schedule, 491 not-recovered blocks (7,502 ha) are scheduled. A 5-year deferral changes volume by -0.069%; a 10-year deferral by -0.093%. The aggregate effect is small, so the layer remains an informational overlay, not a hard constraint.

The earlier examples/recovery_tsa46 is retained as a 5 km mask-based smoke test.

Caveats

  • Spectral recovery is not survival or regeneration success.

  • The observation window ends in 2017; recovery may continue after it.

  • A block outside the footprint is not a failure.

  • A 30 m pixel is not a block-level validation plot.

  • The status is read-only; it does not enter the objective or constraints.