Back to Bigger Picture

BIGGER PICTURE Disaster preparedness

Mapping the flood even when clouds hide the ground

After Cyclone Eloise, UNOSAT used radar images and automatic water detection to follow flooding across Mozambique over several days.

AI-generated illustration: Conceptual satellite radar revealing floodwater beneath clouds.
Satellite radar revealing floodwater beneath clouds.AI-generated illustration of the idea, not a photograph of the actual people, equipment or event.
Event
Mozambique flooding after Cyclone Eloise, 2021
Repeat observations
Sentinel-1 images from 25, 26 and 31 January
Delivered to
OCHA’s regional coordination team

How the idea works

  1. Observe with radar

    Sentinel-1 records the surface through clouds and darkness.

  2. Extract water extent

    FloodAI classifies prepared satellite imagery.

  3. Compare successive maps

    Analysts share changing flood coverage with coordinators.

The disaster keeps changing after the first map

Following heavy rains and Cyclone Eloise in January 2021, humanitarian coordinators requested satellite analysis for Mozambique. Sofala became a priority because much of the observed flooding was concentrated there. A single map could not describe the whole event: responders needed repeated views as water spread or receded, including an overview of the accumulated flooded area.

Sources: [1]

The cloud-penetrating part is the radar

Sentinel-1 supplies radar observations that work through cloud cover and without daylight. That physical capability is especially useful during storms; AI does not create it. UNOSAT’s processing pipeline downloads and prepares the images, then its FloodAI model extracts water extent. Historical flood datasets helped train the model to perform that classification across large image collections.

Sources: [2][3]

Three dates became a changing picture

For Sofala, UNOSAT produced AI-based analyses from images captured on 25, 26 and 31 January, plus an aggregate covering two weeks. Products were shared with OCHA’s regional coordination team and made available online. The documented gain was the ability to monitor more areas and repeat analysis over longer periods with a more automated workflow.

Sources: [1]

Water extent is one piece of the response

A satellite-derived water map does not directly tell responders the depth, building damage or needs of every household. It is also an observation of available image dates, not a forecast of tomorrow’s river level. Combined with local information, it gives coordinators a repeatable geographic picture for planning; the case report does not quantify lives or property saved.

Sources: [1][2][3]

The original sources

  1. UNITARUsing Artificial Intelligence algorithms in rapid mapping activation in Mozambique
  2. UNITAR / UNOSATUNOSAT introduces AI in its flood rapid mapping operations
  3. European Space Agency10 ways Sentinel-1 data lets us see our world

What to explore next

A phone on a table became part of a seismic networkA maths tutor that fits inside a WhatsApp conversation
Explore all 99 applications