DS-EO: turn a color image into a map

DS-EO is a Singapore optical Earth-observation spacecraft jointly owned by DSTA and ST Engineering. It launched with NeuSAR and SCOOB-I on June 30, 2022. The three spacecraft shared transport while serving different purposes. DS-EO’s multispectral — open definition payload was intended to support land classification — open definition and humanitarian or disaster-response applications, connecting an engineering project with the need for timely information about places on Earth.

A multispectral image records several wavelength bands. An analyst can compare those measurements to help distinguish surfaces such as vegetation, built areas and water. A full-color picture is a useful view of the scene, but the map’s labels are an interpretation added through analysis. Light, clouds, haze, mixed surfaces and the instrument’s sampling can affect that interpretation. The mission’s application goals do not make every label automatically correct.

Consider a fictional exercise with two hundred clear image cells. An algorithm labels sixty as water. The mapped water fraction is thirty percent of those evaluated cells. That calculation does not establish that thirty percent of the entire district is flooded: some cells may contain permanent water, some may be misclassified, and cloud-covered or unobserved areas are absent from the denominator.

A useful change map needs an appropriate earlier observation and a consistent definition of the classes being compared. If a river already occupied forty cells, calling all sixty water cells new flooding would overstate the change. Even the difference of twenty cells needs spatial comparison and validation, since the same count can describe different locations. Numbers become meaningful when readers know what was counted and which places were actually observed.

Define the denominator

ISRO photographed visitors watching the PSLV-C53 launch from the viewing gallery at Sriharikota. Their phones, seating and flags show the public setting of a launch event that carried three Singapore satellites. The photograph does not show an imaging result. It is a reminder that the visible moment of ascent is followed by quieter work: checking the spacecraft, calibrating measurements and preparing products that people can use.

For a disaster-response map, analysts should retain acquisition time, cloud masks, classification methods and uncertainty information. A later user needs to know whether an apparent absence of water is a measured result or simply an unobserved area. DS-EO illustrates the difference between obtaining a picture and supplying decision-relevant information. The spacecraft contributes measurements; the complete map also depends on ground processing, comparison evidence and clear communication of what remains unknown.