Digital Earth South Africa: turn an image archive into evidence

Digital Earth South Africa, or DESA, is SANSA’s platform for working with Earth-observation data. Its role is different from launching a new satellite: it helps users discover, access and analyze imagery collected by satellite missions. SANSA’s wider Earth-observation portfolio includes the long Landsat record and SPOT imagery. Landsat 8, carried inside the rocket shown in the launch photograph, belongs to that longer observing history; the launch scene is not a screenshot or product from DESA.

DESA offers several routes into the data. Its map interface supports browsing and comparison; its analysis platform provides a Python notebook environment; and its web services let other geographic software display data layers. These are different tools for different tasks. Seeing a layer on a map does not necessarily mean the user has downloaded the underlying numerical measurements needed for a particular calculation.

Analysis-ready data — open definition reduce repeated preparation work, but the label does not remove the need to understand the product. Check the sensor, collection date, processing level, quality flags and spatial grid. Clouds can hide the ground, seasons can change vegetation naturally, and different sensors can respond differently to the same surface. A visually striking difference between two pictures is a question to investigate, not yet a demonstrated long-term trend.

In an invented reservoir exercise, the same 100-square-kilometre study area is analyzed on two dates. Clouds leave only 70 square kilometres valid on both dates. Compare change within that common valid area, and report the missing 30 separately. Treating cloud-covered pixels as dry land would manufacture a loss of water. Filling those gaps with a model may be useful, but the modeled estimate must stay distinguishable from a direct observation.

Compare the same valid ground on both dates

A defensible map records its input identifiers, dates, area, masking rules and method. If a threshold separates water from land, explain how it was chosen and where it can fail, such as shadows or mixed shoreline pixels. Independent field observations can help evaluate a classification. The number of downloaded scenes is not the same thing as the number of reliable observations at every location.

DESA’s value is partly shared infrastructure: users can spend less effort assembling basic data access and more effort asking useful questions. Decisions about settlements, crops or water still require interpretation, local knowledge and an account of uncertainty. A final map should therefore tell the reader what was observed, how it was processed and what remains unknown. That makes the archive a source of evidence rather than just a gallery of attractive images.