SAR flood mapping — seeing water through clouds · Learn

In one lineSAR is a radar: it sends out its own microwave pulses and listens for the echo. Calm water bounces that echo away from the satellite, so flooded land shows up dark — through cloud, day or night.

Why radar sees what cameras can't

An optical satellite is a camera — it needs sunlight and a clear sky. Synthetic Aperture Radar makes its own illumination at microwave wavelengths that pass straight through cloud, and works at night. That alone makes it the right tool for a storm.

The clever part is what the echo (backscatter) tells you:

Calm water is smooth, so it reflects the radar like a mirror — away from the satellite. Little comes back → it looks dark.
Land, vegetation, buildings are rough, scattering the pulse in all directions, including back to the sensor → they look bright.

So a flood is simply new dark patches where there used to be bright land — and OPERA DSWx packages that into a ready-made water map.

Play with it

Raise the flood and roll in the storm clouds. The optical camera goes blind; the radar keeps mapping the water.

Flood level Sky storm clouds

Optical camera

SAR (radar)

flooded area:

Do it yourself

editable · runs in your browser

import numpy as np
THRESH_DB = -18.0                                # VV backscatter below this is roughly smooth water
pre_vv  = np.array([-8, -10, -7, -9, -6])        # before: mostly bright land
post_vv = np.array([-8, -22, -25, -9, -20])      # after: some pixels went dark
water_after = post_vv < THRESH_DB
water_before = pre_vv < THRESH_DB
new_flood = water_after & ~water_before
print("newly flooded pixels:", int(new_flood.sum()), "of", new_flood.size)
print("water = LOW backscatter; new_flood = dark now but not before.")
print("note: roads and smooth sand are also dark (look-alikes); OPERA DSWx avoids hand-thresholding.")

The honest caveats

Rough water fools it. Wind-whipped or fast floodwater scatters radar back and can look like land — calm-water assumption.
Cities and forests hide flooding. Buildings and canopy scatter the pulse, so water beneath them is hard to see — combine with optical (Sentinel-2 SWIR) where skies clear.
Look-alikes. Smooth dry surfaces (tarmac, sand, radar shadow behind hills) can also be dark — use a before/after difference, not a single image.
Revisit. Each satellite passes every few days; timing may miss a flash flood's peak.