If discoveries were often made thanks to Google Maps (archaeological sites, hidden treasures or unknown places), now We could multiply that by ten thanks to a new initiative.
Alphaearth Foundations, an artificial intelligence model (AI) that functions as a virtual satellite, integrates land observation data petabytes to generate a unified representation that revolutionizes cartography and global monitoring. To give us an idea of the amount of information he controls, 1 Petabyte is equivalent to half a billion text pages.
The system, developed by Google, characterizes precisely The entire earth’s surface and the coastal waters of the planet by integrating huge amounts of data of Earth’s observation in a unified digital representation, or “embedding”, which computer systems can easily process. For example, the system is not fooled by the light that is reflected in the ice and combines images with orographic and depth detection systems.
This allows the model to provide scientists with a more complete and coherent vision of the evolution of our planet, helping them to make more informed decisions about crucial issues such as food security, deforestation, urban expansion and water resources. Its qualities have been published in Arxiv.
To check your qualities Its use has been allowed for about 50 organizations that have been able to better classify non -cartography ecosystems, understand agricultural and environmental changes, and considerably increase precision and the speed of his cartographic work, among others.
Alphaearth Foundations operates in two different sections. First, it combines information from dozen different public sources: optical satellite images, radar, 3D laser cartography, climatic simulations and more. It combines all this information to analyze the land and coastal waters of the world in crisp squares of 10×10 meters, which allows you to track changes over time with extraordinary precision.
Secondly, it facilitates the use of this data. The main innovation of the system lies in its ability to create a very compact summary for each square. These summaries require 16 times less storage space that those produced by other AI systems and drastically reduce the cost of planetary scale analysis.
This advance allows scientists to do something so far impossible: to create detailed and consisting maps of our planet, with different characteristics, on demand. Whether they are monitoring the health of crops, tracking deforestation or observing new constructionsthey no longer have to depend on a single satellite that flies over. Now they have a new base for geospatial databases.
On average, Alphaearth Foundations had an error rate 24 % less than other similar models and that taking into account that the enters set (shared images) Google Earth Engine satellites is one of the largest of its kind, with more than 1.4 billion trags per year.
This collection is already being used by organizations around the world, as the United Nations Food and Agriculture Organization (FAO), Harvard Forest, the Earth’s Observations GroupMapbiomas, the State University of Oregon, the Space Computer Group and Stanford University, to create personalized maps that generate real world information.
For example, in Brazil, Mapbiomas is testing the data set for better understand agricultural and environmental changes throughout the country. This type of map reports conservation strategies and sustainable development initiatives in critical ecosystems such as the Amazon jungle.
“The satellite integration data set can transform the way our team works; now We have new options to create more precise and fast maps to produce, something that we would never have been able to do”, Concludes Tasso Azevedo, founder of Mapbiomas.