LandTrendr temporal segmentation — PNW (Kennedy et al. 2010)
Detecting trends in forest disturbance and recovery using yearly Landsat time series: 1. LandTrendr — Temporal Segmentation Algorithms · 10.1016/j.rse.2010.07.008
Temporal segmentation of Landsat NBR time series for Pacific Northwest industrial forest. Detects clearcut harvests and regrowth using the LandTrendr algorithm (Kennedy et al. 2010) — running entirely in the browser.
Recipe source: demos/landtrendr-pnw/folia.yaml — view folia.yaml
- Input data
- Landsat time series
- Period
- 2000-2024
- Region
- Pacific Northwest, USA
- Bounding box
- -123.95, 45.82, -123.75, 45.95
| metric | observed | threshold | measured by |
|---|---|---|---|
| note | Reference pixel at -123.845, 45.889 (Oregon Coast Range, conifer- dominated industrial forest). Known history: ~17 years stable mature conifer, clearcut harvest in 2001-2002, subsequent regeneration. LandTrendr detects a major NBR drop (breakpoint) within +/- 1 year of the known 2001-2002 clearcut event, followed by a recovery segment. Verified visually against Landsat RGB filmstrip and LCMS annual change product (lcms_validation.py). | ≥ 1 | human · 2026-03-29 |
| visual match | Disturbance patterns across 5 PNW sample areas match LCMS v2024 annual change product (USFS GTAC). Oregon Coast Range: clearcut harvest detected (PASS). B&B Complex Fire 2003: wildfire detected (PASS). Blue Mountains: MPB insect damage partially detected (PARTIAL — our spectral-only approach misses slow decline). HJ Andrews: stable old-growth correctly shows no disturbance (PASS). Central Cascades: mixed disturbance detected (PASS). 4 of 5 sites fully agree; 1 partial. See RESULTS.md for per-site details. | ≥ 1 | human · 2026-03-29 |
| note | Fed GEE's own NBR time series into Rust LandTrendr and compared fitted trajectories at 10 sample points. On pixels with real disturbance signal, RMSE agreement within 0.001-0.009. On stable pixels, GEE returns RMSE=0 (simplest model) while our F-test model selection sometimes picks a more complex model — this is a known model selection threshold difference, not a fitting quality issue. See gee_comparison.py and RESULTS.md for full table. | ≥ 1 | human · 2026-03-29 |
- PaperDetecting trends in forest disturbance and recovery using yearly Landsat time series: 1. LandTrendr — Temporal Segmentation Algorithms (10.1016/j.rse.2010.07.008)
- Specdemos/landtrendr-pnw/folia.yaml
- DataLandsat time series
- Result—
- Validationunvalidated
The declarative recipe is the reproduction: paper linked to spec linked to data linked to result linked to validation. Full mechanical provenance ships with the Evidence Bundle.
No forks yet. Forking (change AOI / time / params and re-run) opens in the workbench — a community fork graph is planned (hub Phase 4).