Charting the Future of Earth Observation 2024

Charting the Future of Earth Observation: Technology Innovation for Climate Intelligence17DE Africa WMS for over 700,000 waterbodies across Africa FIGURE 7 Source: Digital Earth Africa. (2024). DE Africa Waterbodies: A Waterbodies Monitoring Service. https://docs.digitalearthafrica.org/fr/latest/data_ specs/Waterbodies_specs.html.the open data cube technology to organize large amounts of satellite imagery into multidimensional arrays (i.e. sorting data by time, location or spectral bands), making it easier to extract and analyse useful insights. DE Africa announced the release of the Waterbodies Monitoring Service (WMS) in 2024. WMS is a tool that delivers a dynamic time series of wet, dry and unobserved surface area as a proportion of the historical water body extent. Overall, it does this for over 700,000 unique water bodies across the African continent on a weekly basis. It is projected that the amount of urban land at high risk from frequent flooding could rise by 270% in North Africa and by as much as 800% across southern African zones; mid-latitude areas could see a dramatic increase of nearly 2,600% by 2030.26 Open-source AI platforms are also able to quickly process new data and feed it into early warning systems.27Clay, an open-source and non-profit project, is democratizing access to complex Earth analysis by developing large-scale foundation models on satellite-borne data, starting with Sentinel-1 and Sentinel-2 imagery. The outputs of these models are called vector embeddings, which effectively compress the relevant spatial and temporal signals of EO data cubes to be accessed through simple point-and-click web interfaces. While there are several open EO foundational models, Clay’s approach to open-source mission-specific product development, coupled with its non-profit status, sets it apart. It assembles donated compute resources to train the largest open models, sets meaningful benchmarks to assess model quality and publishes gold-standard, easily accessible training data. Open-source AI platforms are also able to quickly process new data and feed it into early warning systems.