How is a flood extent mapped from satellite imagery?
Mapping a flood from space requires comparing water signals after the event with normal surface water, land cover and observation conditions before it. Optical sensors cannot see the surface beneath persistent cloud, while radar can partly overcome that limitation but dark radar pixels do not always indicate water. The result is a partial observation at a particular time, not a real-time declaration that a route is safe.
Preparing observations before and after the event
1. Why does acquisition time matter? A flood wave can travel within hours; a satellite may pass before or after the peak. Match image acquisition time with rainfall, stream gauge observations and local reports where available. A smaller water outline the next day does not rule out more extensive inundation overnight. Clouds conceal the surface in Landsat and other optical imagery. NASA Worldview notes that its near-real-time optical flood product cannot detect water beneath persistent cloud. Mark that region as unobserved rather than silently filling it with a guess and presenting it as measured dry land. 2. How is ordinary water excluded? Lakes, river channels, canals and seasonal wetlands can contain water before a flood. Calling all dark or blue pixels after the event newly flooded incorrectly includes permanent water. Use several clear observations from a comparable season to estimate the usual water extent, check land cover and map newly observed water as a difference. Reservoir operations, tides and irrigation can move the water edge for other reasons. Before-and-after layers must use compatible geometry, pixel size and classification rules. A seasonal floodplain may be wet normally in one month but dry in another. 3. What can optical bands reveal? Water often reflects differently from dry surroundings in visible, near infrared and shortwave infrared bands. Band combinations and water indices can help classify possible surface water. Shadow, dark pavement, muddy water, wet soil or thin cloud can confuse a simple threshold. A mountain shadow and deep water can both have low reflectance. Apply the sensor's quality flags for cloud and shadow, and keep those pixels out of the dry category. An index threshold chosen on one landscape may perform poorly on another, so inspect reference samples.
Radar, uncertainty and ground decisions
4. How does radar show water? Synthetic aperture radar sends microwave signals and measures their return, allowing imaging through cloud. Smooth open water often reflects energy away from the sensor and appears dark. Yet pavement, radar shadow and other smooth surfaces can also be dark; water under dense vegetation produces more complicated backscatter. Compare radar before and after the event, and consult permanent water and land-cover layers. USGS describes combining strengths of multiple satellite types to deal with clouds. Radar is valuable evidence, not a universal automatic water detector. 5. How is flooded area estimated? To sum the area of classified pixels, specify the projection, cell area, resolution and unobserved regions. Treating a square kilometer hidden by cloud as zero flood biases the total; report it separately as unknown. Building shadows and tree canopy can hide water in towns, and a coarse pixel cannot show whether a narrow street is passable. Sample independent ground observations, aerial photographs and gauge measurements to assess errors. A clean colored outline can suggest more spatial certainty than the data actually provide. 6. What are the limits of use? A satellite map can reveal broad flood patterns, effects on farmland and places needing priority inspection. It cannot alone determine dam safety, whether a bridge is open, or flood depth inside one house. Publish acquisition time, sensor, cloud share, permanent-water mask, radar or optical thresholds and validation samples. For personal safety, follow current warnings from responsible emergency authorities and local teams. Do not use an archived event outline as a present-day road closure or a prediction of the next flood.