GLAD alerts vs RADD alerts for supply shed screening

If you screen supply sheds for new clearing, you've already run into the alphabet soup: GLAD-L, GLAD-S2, RADD, sometimes all three stacked in the same Global Forest Watch layer. They're not interchangeable, and the gap between them matters most in exactly the regions where supply shed risk is highest: the Congo Basin, Sumatra and Kalimantan in wet season, parts of the Amazon during the cloudier months.

What GLAD alerts detect

GLAD alerts come out of the University of Maryland's GLAD lab. The original version (GLAD-L) runs on Landsat, 30-meter pixels, weekly updates across the pantropics. A newer version, GLAD-S2, uses Sentinel-2 optical imagery and gets down to 10-meter pixels in the areas it covers, which catches smaller clearings the Landsat version misses.

Both are optical. That's the catch. A cloud-free pass has to happen before the algorithm can flag anything. In a region running six or eight weeks of near-continuous cloud cover, which is most of the tropics for stretches of the year, GLAD just queues up and waits. The clearing happened on day one. The alert might land on day forty.

What RADD alerts detect

RADD comes out of Wageningen University and runs on Sentinel-1, which is radar, not optical. Radar doesn't care about cloud cover. It bounces a signal off the canopy and reads the change in backscatter whether the sky is socked in or clear. Coverage started in the Amazon and Congo Basin and has expanded since.

This is the part that matters for screening: RADD can flag a clearing during a cloud-locked stretch that would otherwise leave GLAD blind. The tradeoff is that radar backscatter is noisier than a clean optical image. RADD alerts run through a confirmation step, unconfirmed then confirmed, precisely because the first pass throws some false positives that a second pass filters out.

The practical difference for supply shed screening

For a trader doing supply shed work, the question isn't which system is right. It's which one flags the clearing first, and that depends on the weather over that specific patch of forest that week.

In a dry, clear stretch, GLAD-S2 at 10 meters can catch a smaller clearing than RADD will reliably flag. In a wet, cloud-heavy stretch, RADD is often the only feed producing anything, because GLAD is sitting on a backlog of cloud-obscured scenes. Teams that rely on one feed alone end up with a blind spot shaped exactly like their local weather pattern, because the two systems exist side by side precisely to cover for each other's gaps.

The teams that get burned are usually the ones checking a single alert type on a fixed schedule, say once a month, and treating a quiet month as a clean month. In a cloudy basin, a quiet GLAD month typically just means the satellite never got a clean look at the ground, not that nothing was cut.

Using both without doubling the workload

Running both feeds properly means pulling GLAD-L, GLAD-S2 and RADD for the same draw area, deduplicating the overlap, and routing whatever's left to someone who can act on it. Done by hand across a dozen supply sheds, that's a standing weekly task for whoever owns it, and it's the kind of task that slips the first week that person is out sick or pulled onto a different fire.

This is the gap Deforestation Alerts is built to close: you draw the supply shed once, and a weekly alert list comes back flagged within days of whichever feed, optical or radar, got the first usable look that week, instead of you checking three dashboards on three different schedules. If your team is still stitching GLAD and RADD together by hand every Monday, it might be worth seeing what that list looks like for your own supply shed.