The Global Land Analysis and Discovery (GLAD) laboratory at the University of Maryland, led by Matthew Hansen, built the first system to flag tropical forest loss in near-real time. Where annual forest-cover maps told you a forest had vanished only long after the fact, GLAD alerts process incoming satellite imagery continuously and raise a flag within days of trees being cut, fine enough for rangers and enforcement agencies to respond while intervention is still possible.
The original GLAD-L system analyzes Landsat imagery across the tropics at 30-meter resolution. A machine-learning classifier - bagged decision trees trained to recognize forest-loss signatures - scores each pixel, and an alert is generated when the model’s likelihood of canopy loss exceeds a 50% threshold against a baseline of forest defined as trees over five meters tall with more than 30% canopy closure. The approach became practical only after the 2008 opening of the Landsat archive and systematic dual-satellite coverage from 2013, which together gave the algorithm enough clean observations to work from. A later GLAD-S2 system uses 10-meter Sentinel-2 imagery to map primary-forest loss in the Amazon at even finer resolution.
The alerts are distributed publicly through Global Forest Watch, the World Resources Institute platform that pairs GLAD’s science with Google Earth Engine computing power and an interactive map. Studies of the system found measurable impact: subscriptions to GLAD alerts were associated with a meaningful drop in deforestation in monitored areas of central Africa.
Why business readers should care: GLAD is a textbook example of how the value of a model depends on a free, open data supply - the algorithm only became possible once Landsat imagery was opened to the public, turning a government archive into the backbone of global forest enforcement.