Dashboard / Problems / landtrendr-change

LandTrendr Forest Disturbance Detection (PNW, 1985-2023)

change-detection difficult USGS Landsat Collection 2

Ground Truth

22.0
percent
25.0%
Tolerance
1
Runs
1
Passed

Calibrated from GEE execution on 2026-03-29 (actual: 22.03%). Will be validated against QGIS and folia outputs. Pipeline uses fitted-value range (max - min) from LandTrendr array with threshold > 300 NBR*1000 units. Captures disturbance + recovery over 1985-2023. High tolerance due to sensitivity to LandTrendr parameters, cloud masking, and disturbance threshold definition.

Run Results

Workflow Model Backend Status Answer Error Cost Latency
exec gold gee PASS 22.028571621854176 0.1% $0.004350 78.3s

Requirements

Data Sources
  • Landsat Collection 2 (TM/ETM+/OLI) annual composites
Operations
landtrendr temporal_reduce raster_threshold_mask

LLM Prompt

The prompt given to LLMs in single-shot workflow benchmarks. 

Problem: Detect forest disturbance in the Pacific Northwest using
LandTrendr temporal segmentation on Landsat imagery from 1985-2023.

Method:
- Build annual NBR composites from Landsat TM/ETM+/OLI
- Run LandTrendr with default parameters (maxSegments=6)
- Identify disturbance as fitted-value range > 300 in NBR*1000 space
- Report percentage of area showing significant change

Study area: -122.5, 45.0, -122.0, 45.5 (western Oregon).
Expected answer: approximately 22% of area with significant change.