Run GLOWfm yourself

Pick a flood and press Run to run GLOWfm, frozen as tested, in your browser on NISAR's radar. These are this season's floods, not the tested results.

Run

No flood chosen yet

Change GLOWfm's settings

For this run only; the published result uses the frozen settings.

Dry reference
Radar signal

How it works, and the data

The area excludes permanent lakes and rivers. US floods are graded against the river's gauges and lidar; elsewhere there's no ground truth, and the trust light decides.

The trust light is also on trial. Before NISAR acquired them, we sealed which of its next passes over US floods will be scored, and what the light predicts for each (the sealed test in the results).

Floods the job looked at but couldn't prepare (0), and why:

    Floods prepared but not offered to run (0), and why. Two rules, added on Nov 15, 2026, after the list grew to floods where runs scored near zero, and both settled before any run. A US flood is offered only if its gauge was at NOAA's flood stage when NISAR passed, read from USGS's 15-minute record: a gauge that tops it for an hour a day, or a pass after the river fell, leaves no flood to map. And GLOWfm reads flooding under trees, so a flood is offered only where at least half of the valley floor (land up to 6 m above the nearest water and within 4 km of the river) is forest or wetland.

      GLOWfm compares NISAR's pass during the flood with earlier passes on the same track, keeps forest and wetland close to the river, and lets the scene set its own cut-off. Its rules are in the results and in research/LIVE_PROTOCOL.md; since protocol v2 (research/LIVE_PROTOCOL_V2.md), floods outside the US use bare-earth heights from FABDEM, which tested better than treetop heights without lidar. Data: NISAR GCOV (NASA/ISRO, via ASF), NOAA river gauges, USGS river lines, gauges and 3DEP lidar, ESA WorldCover 2021, Copernicus GLO-30, FABDEM V1-2 (University of Bristol; Hawker et al. 2022; CC BY-NC-SA 4.0, height layers made from it are shared under the same license for non-commercial use; produced using Copernicus WorldDEM-30 © DLR e.V. 2010-2014 and © Airbus Defence and Space GmbH 2014-2018 provided under COPERNICUS by the European Union and ESA; all rights reserved), GDACS. Imagery © Esri.

      Roads under water are drivable roads from OpenStreetMap (© OpenStreetMap contributors, ODbL), with footpaths and bridges left out, counted on the ground GLOWfm flags. They're saved with each prepared flood (scripts/build_live_roads.py); a run of your own asks OpenStreetMap for its box. Mapped roads can be incomplete, and a raised road can stay dry where the ground beside it floods.