# Supervised tree-ensemble classification
*analysis . methodologies*

Assigns land-cover or change classes by applying a decision-tree ensemble such as random forests to per-pixel temporal and spectral metrics, trained on labelled examples. Used to produce wall-to-wall forest-cover and land-cover maps.

## Specifications
- **family**: Classification
- **entity type**: methodology
- **last verified date**: 2026-06-04
- **verified by**: agency-doc
- **claim status**: unclaimed
- **subtype**: analysis
- **attributes**: {"family":"Classification","kind":"analysis","summary":"Assigns land-cover or change classes by applying a decision-tree ensemble such as random forests to per-pixel temporal and spectral metrics, trained on labelled examples. Used to produce wall-to-wall forest-cover and land-cover maps."}

## Editorial
Supervised tree-ensemble classification answers the question: what land-cover class does each pixel belong to, based on patterns learned from labelled examples? The method trains a decision-tree ensemble on spectral and temporal features derived from satellite imagery, then applies the trained ensemble to produce wall-to-wall class probability maps.

The Hansen et al. 2013 Global Forest Change (GFC) product is the anchor implementation.[^hansen-2013-science] The algorithm computes per-pixel temporal spectral metrics from annual Landsat growing-season composites: statistical ranks, means, and linear regression coefficients of red, NIR, SWIR, and thermal bands, plus derived indices including NDVI, NBR, and NDWI. These metrics serve as features for a bagged decision-tree ensemble trained on geographically distributed labelled reference samples. Each tree in the ensemble votes on the class membership of a pixel (tree cover, forest loss, forest gain); the median vote probability across the ensemble is thresholded to assign the final class label. Annual updates apply the same pipeline to each successive year's Landsat stack, providing a record of global forest cover change from 2000 to present at 30m resolution.[^nasa-earthobservatory-gfc]

The GLAD-L alert system (Hansen et al. 2016) extends the same bagged decision-tree approach to near-real-time operation, applying the ensemble to the incoming Landsat stream with a 4-observation confirmation rule.[^hansen-2016-alerts] The GFC ensemble uses bagged decision trees, which aggregate bootstrap samples of the training data across a tree ensemble.[^hansen-2013-science]

The method degrades in persistently cloudy tropical regions, complex mixed-forest landscapes with ambiguous spectral signatures, and areas with rapid phenological variation that confounds the temporal metrics.

## Sources
- [hansen-2013-science] | High-Resolution Global Maps of 21st-Century Forest Cover Change | https://www.science.org/doi/10.1126/science.1244693 | tier=peer-reviewed | accessed=2026-06-04 | author=Hansen M.C. et al.
  (2013) Foundational paper: bagged decision trees on Landsat temporal metrics (ranks/means/regressions of red/NIR/SWIR); global 30m forest cover/loss/gain 2000-2012; Science 342(6160):850-853
- [hansen-2016-alerts] | Humid tropical forest disturbance alerts using Landsat data | https://iopscience.iop.org/article/10.1088/1748-9326/11/3/034008 | tier=peer-reviewed | accessed=2026-06-04 | author=Hansen M.C. et al.
  (2016) Alert extension of GFC: bagged DTs applied to near-real-time Landsat stream; red/NIR/SWIR temporal profile features; NDVI/NBR/NDWI indices; 4-observation confirmation rule
- [nasa-earthobservatory-gfc] | Mapping Forest Loss with Landsat | https://earthobservatory.nasa.gov/images/85824/mapping-forest-loss-with-landsat | tier=agency-doc | accessed=2026-06-04 | author=NASA Earth Observatory
  (2014) Accessible description of Hansen GFC decision tree methodology and Landsat temporal metrics approach

---
Source: https://eo-atlas.org/methodologies/supervised-tree-ensemble-classification
Maintainer: SpectraWorks B.V. (CC-BY 4.0)