It will allow faster and more efficient work with Earth data.
IBM and NASA have joined forces to create what can be called a kind of analogue of ChatGPT in the world of science, although the analogy is inaccurate, since it is not a chatbot.
More specifically, we are talking about the launch of a first-of-its-kind open source geospatial AI core model for Earth observation data.
NASA and IBM launched AI for scientists
According to NASA’s own estimates, all missions of the Earth Exploration Agency will generate about 250,000 TB of data next year alone. And often scientists get to some very important data only years later. The Geospatial AI Foundation Model, which will be based on Harmonized Landsat Sentinel-2 (HLS) data, will allow scientists to process huge amounts of information faster and more efficiently. The model has a wide range of applications, such as tracking changes in agricultural land use, monitoring natural disasters, deforestation, yield forecasting, studying greenhouse gas processes, and so on. The model is already available in the Hugging Face database, an open source machine learning model repository.
The model has already been used for flood mapping and fire impact assessment. In addition, using time series data, the teams demonstrated the benefits of using the model to map land cover and crop types across different geographic regions.
