GLAD alerts cut deforestation probability in Africa by 18%

AI system

System: GLAD (Global Land Analysis and Discovery) Landsat deforestation alerts on Global Forest Watch
Model: Not assessed
Release: Not assessed
Application developer: University of Maryland GLAD laboratory (Hansen et al.)
Role: Not assessed
Configuration: Not assessed
Inputs and outputs: Not assessed
Autonomy: Not assessed
Workflow: Weekly Landsat coverage and a conservative classifier. Free public platform with subscription-based alert delivery. Subscribers with enforcement or management capacity.
Human oversight: Not assessed
Relation to AI: The alerts are a machine-learning product (bagged decision trees). The evaluation attributes the 18 percent reduction to subscriptions to those alerts rather than to their mere availability. The evaluation's lead authors are independent economists, but co-authors are from WRI (operator) and the University of Maryland (producer).
AI contribution: The alerts are a machine-learning product (bagged decision trees). The evaluation attributes the 18 percent reduction to subscriptions to those alerts rather than to their mere availability. The evaluation's lead authors are independent economists, but co-authors are from WRI (operator) and the University of Maryland (producer).

Benefits

Outcome identifiers: Not assessed
Benefit: decreased the probability of deforestation in Africa by 18 percent
Category: Improved Environmental Sustainability
Status: Not assessed
Beneficiary groups: People. Organizations. Society.
Beneficiaries: Not assessed
Reach: Not assessed
Distribution and access: Not assessed
Duration: Not assessed
Expected?: Yes
Costs and harms: Not assessed

Measurement

Studies: Not assessed
Study design: Not assessed
Measurement: Not assessed
Setting and sample: Not assessed
Measures: decreased the probability of deforestation in Africa by 18 percent
Comparison: average 2011–2016 deforestation levels, with areas lacking alert subscribers as the counterfactual across 22 tropical countries
Results: Reference: average 2011–2016 levels. Result: probability of deforestation in Africa decreased by 18%; no effect on other continents.
Additional value from AI: Not assessed
Limitations: Not assessed
Denominators: Not assessed
Materials for replication: Not assessed

Money

Kind of value: value of avoided deforestation in Africa, priced at the social cost of carbon
Amount: Estimate: between US$149 million and US$696 million
Currency and prices: USD, stated in the source. Social-cost-of-carbon valuation; valuation and price year are not specified in the checked abstract.
Period and unit: Comparison with pre-2016 deforestation; the checked abstract does not specify the valuation horizon. Modeled value of avoided deforestation in the African study sample.
Realization: Modeled valuation.
Recipients and payers: Recipient: People benefiting from avoided climate damage; no direct cash recipient. Payer: Not applicable to a social-cost valuation.
Gross or net: Modeled climate value, not cash income, market revenue, or a fully costed net-benefit estimate.
Included costs: Not assessed
Comparison: Deforestation associated with subscriptions versus the study counterfactual. Alert availability alone had no significant effect.
Method: Not assessed
Return and payback: Not assessed

People and places

Participants: Developer: University of Maryland GLAD laboratory (Hansen et al.). Operator: World Resources Institute (Global Forest Watch) and alert subscribers.
Jurisdictions: Developer: United States. Operator: United States.
Ownership and control: Not assessed
Dependencies: Not assessed
Cross-border connections: Not assessed
Geopolitical implications: Not assessed

Follow-up

Conditions for replication: Alerts reach actors who can respond. Cloud cover does not suppress detections for long periods. Protected-area or concession managers retain enforcement authority.
Replication attempts: 1 follow-up entries. Benefit did not recur: 1.
Later findings: 13 January 2026. Global Forest Watch kept the Landsat-based GLAD alerts (GLAD-L) running and in 2022 folded them into an integrated disturbance alerts layer with GLAD-S2, RADD and, from January 2026, DIST-ALERT.
Availability: Not assessed
Institutional decision: Not assessed
Grounds: Not assessed
Action and result: Not assessed
Success criterion: Not assessed

Evidence

Source numbers: AIBD-0105-source-0001–0003
Citations: Numbered references below.
Language: Not assessed
Source relationships: Independent, Developer, Operator
Original evidence: Not assessed
Access and versions: Not assessed
Supported findings: Not assessed
Evidence links: Not assessed
Evidence strength: Not assessed
Missing evidence: Not assessed

Grades

Assessed benefits: People. Organizations. Society.
Depth: Not assessed
Reasons: People: Aggregate statistical effect across subscribed areas, no discrete attributable event. Organizations: Effect measured at the level of subscribed administrative areas and concessions. Society: Continent-level reduction estimate with no single identifiable beneficiary.
Observed or intended: Not assessed
Rubric: Not assessed
Assessor and date: Not assessed
Alternative assessments: Not assessed
Fill: Independently studied

Entry details

Kind

Not assessed

Related entries

Natural England's ML land-cover mapping from imagery misclassified granite and quarries as peat, misdirecting conservation; GLAD's Landsat classifier maps forest loss for the same users.

Field

Conservation and biodiversity. Forest monitoring and enforcement.

Purpose

Detect humid tropical forest disturbance in near real time from Landsat imagery to support land management and enforcement.

Dates

Date basis unassessed: 2021 (precision unassessed). Updated: 21 October 2026.

Stage

Not assessed

Version

SHA-256: 37e5d9f524fe…

References

Other identifiers: AI Incident Database: 1159.
Category mappings: Not assessed
Brief: Not assessed
Revisions: 5 revisions. Latest: 21 October 2026.

Sources

Peer-reviewed study Independent author

Moffette, F., Alix-Garcia, J., Shea, K., & Pickens, A. H. (2021). The impact of near-real-time deforestation alerts across the tropics. Nature Climate Change.
Peer-reviewed study Developer

Hansen, M. C., Krylov, A., Tyukavina, A., Potapov, P. V., et al. (2016). Humid tropical forest disturbance alerts using Landsat data. Environmental Research Letters 11(3).
Press release Deployer

Berger, A., Carter, S., Schofield, T., Pickens, A., Reiche, J. (2026). Explore GFW's Updated Integrated Disturbance Alerts. Global Forest Watch / Global Nature Watch blog, 13 January 2026.