Deforestation Alert Systems: GLAD, RADD and Integrated Alerts

Updated 2026-11-02 · 4 min read · By the akar.earth specialist team

Deforestation alerts GLAD RADD explained: sensors, resolution, latency, confidence levels and how to use integrated alerts for REDD+ and EUDR monitoring.

GLAD and RADD are free, near-real-time systems that flag tree cover disturbance: GLAD uses optical Landsat (30 m) and Sentinel-2 (10 m) data, while RADD uses Sentinel-1 radar (10 m) that sees through cloud. Integrated alerts combine them for faster, more confident detection, which makes them a strong first filter for REDD+ monitoring and EUDR supply-chain checks, as long as every important alert is verified.

The main systems

Newer products, such as land-surface disturbance alerts built on harmonised Landsat-Sentinel data, are also appearing. Check Global Forest Watch for current layers and coverage, as these change.

How alerts work

Each system builds a model of normal conditions for every pixel, then flags observations that depart strongly from that expectation. A single anomaly gives a low-confidence alert. Repeated anomalies, or agreement between systems, raise confidence. Alerts are usually published within days to a few weeks of the observation, depending on cloud and processing.

Radar matters most in Indonesia. Optical systems can miss weeks of clearing in cloudy Sumatra or Papua; RADD keeps observing. Optical systems, in turn, catch some changes radar misses, which is why integration helps.

Using alerts in practice

1. Load your boundaries. Project areas, supplier plots, concessions. Accurate polygons reduce false matches.
2. Subscribe or pull data automatically for those areas through Global Forest Watch or the underlying datasets.
3. Filter. Set minimum alert size, confidence and persistence rules suited to your use. Small single-pixel alerts in agroforestry are often noise or pruning.
4. Triage. Overlay with land-use maps, concessions, peat and protected areas. Prioritise alerts inside or next to project areas and supplier plots.
5. Verify. Check high-resolution imagery, then send field staff or ask the farmer or community for photos with location. Record cause: smallholder clearing, plantation expansion, fire, logging, mining, natural (landslide, storm).
6. Act and log. Feed confirmed events into REDD+ activity data, leakage monitoring, supplier engagement or EUDR risk assessment.

Uses by project type

• REDD+ and IFM: activity data for the monitoring report, early warning for patrols, leakage belt monitoring. See drivers of deforestation in Indonesia.
• ARR: detect disturbance or fire in planted areas and control plots; trigger reversal reporting where needed.
• EUDR: screen supplier plots for clearing after 31 December 2020, alongside baseline forest maps. See EUDR satellite verification.

Limits to understand

• Disturbance is not deforestation. Alerts flag canopy loss, including harvest in plantations, rubber replanting, or storm damage. Legal and definitional judgement is still needed.
• Agroforestry confusion. Shade-tree pruning in coffee and cocoa gardens can trigger alerts.
• Small clearings below the minimum mapping unit are missed.
• False positives on steep terrain and in wetlands, especially for radar.
• Degradation under intact canopy is largely invisible.

Setting up a verification routine

Alerts are only useful if someone acts on them. A simple weekly routine works for most projects and sourcing teams:

1. Pull new alerts for all monitored polygons.
2. Discard alerts below your size or confidence rules, but log them.
3. Check the rest against recent high-resolution imagery.
4. Assign field checks for confirmed or uncertain cases, with a target response time.
5. Record outcome, cause, area and photos in one register.
6. Report monthly summaries to management and, for EUDR, to the due diligence file.

Common mistakes

• Treating every alert as a breach, or ignoring all of them as noise.
• Monitoring only the project area and missing leakage outside it.
• Not recording verification outcomes, so the same false positives recur.
• Using alerts without a baseline forest map for the cut-off date.

What good evidence looks like

For a VVB or EUDR auditor, keep: alert data extracts with dates and versions; filtering rules; triage and verification records with imagery, photos and field notes; the cause and area of each confirmed event; and actions taken. A summary table per monitoring period linking alerts to reported activity data shows the system works.

Indonesia specifics

Alerts are well suited to Indonesia’s mix of plantation expansion, smallholder clearing and fire. During El Niño dry seasons, fire-related alerts surge in peat areas of Riau, South Sumatra and Central Kalimantan; combine alerts with fire hotspot data. See peatland REDD+.

We set up automated alert monitoring with field verification through our dMRV implementation service. The dMRV hub explains how alerts sit within a full monitoring system, and Sentinel-2 for forest monitoring covers the optical side.