Google Earth is introducing a new initiative called Dynamic World today, which will produce maps linked with a unique deep learning AI model. At a resolution of 10 meters (32 feet), it can distinguish land cover by type (water, urban, forest, agricultural). Each pixel so covers around 10 meters of land. In comparison, earlier state-of-the-art equipment had a resolution of 100 meters (320 feet).
Dynamic World allows users to see how land cover varies on Earth from orbit, whether it's due to natural seasonal changes, climate-related storms and disasters, or long-term changes induced by human activities such as destroying wild areas for crops, livestock, or logging.This new effort will allow experts and academics to better understand how land cover evolves organically, as well as identify when certain unexpected changes appear to be occurring.
Users may check through the various datasets on Google's Dynamic World page and see what the marked maps look like. One map, for example, depicts how the volume of water and foliage in Botswana's Okavango Delta swells and recedes from the wet to the dry seasons.
The map model, which uses Sentinel-2 satellite images from the European Space Agency, may refresh its data stream every 2 to 5 days for worldwide land cover monitoring. Every day, the Sentinel-2 satellite sends out roughly 12 gigabytes of data. The data is then sent to Google's data centers and Google Earth Engine, a cloud platform designed to organize and convey Earth observations and environmental analyses. The Earth Engine is linked to tens of thousands of computers that analyse data and use computer models to draw insights before it is made available in the Earth Engine Data Catalog.
Google needs the aid of artificial intelligence to automatically categorize how the land depicted in all those satellite photographs is used. They trained its land cover labeling AI on 5 billion pixels labeled by human specialists as part of this effort (and some non-experts). They recognized pixels in Sentinel-2 photos and the land cover class they belonged to in the training set (water, tree, grass, flooded vegetation, built-up areas like cities, crops, bare ground, shrub, snow). They'd then show the model an image that wasn't in the training set and ask it to categorize the different types of land cover. Not only are there color variances on the maps to indicate the various terrain types, but there are also shading differences. This is due to the fact that pixels may also transmit likelihood. The model is more confident in its categorization accuracy if the color is brighter. When the geography changes from land to forest or land to water, this generates a textured impression.
“We are making it all available under a free and open license,” said Rebecca Moore, Google Earth's director, on a press call before to the announcement. “The datasets are free and open. The AI model is open source.”
Google and the World Resources Institute cooperated on Global Forest Watch around ten years ago, a project aimed at monitoring forest cover to safeguard these places while checking for changes caused by unlawful activity like logging or mining. They're now attempting to broaden their efforts beyond simply conserving and studying one sort of land cover.
The goal is to aid in the interpretation of the data that is already accessible.“We’ve heard from a number of governments, [and] researchers that they are committed to taking action, but they are lacking environmental monitoring information about what’s happening on the ground so they can create science-based data-informed policies, track the results of their actions, [and] communicate with stakeholders,” Moore said. “The irony isn’t that there isn’t a ton of data. But they’re thirsty for insights. They’re looking for actionable guidance to support the decisions they need to make. And dealing with the raw data in many cases is overwhelming.”
Dynamic World, according to Google, can bridge the data vacuum surrounding land use and land cover, as well as identify where basic ecosystems such as forests, water resources, agriculture, and urban development are situated. Moore believes that this sort of data can help guide decisions concerning the sustainable use of finite natural resources, food, and water. It may also assist with concerns such as how to manage catastrophe resilience, how to cope with sea-level rise, where to construct protected areas, where to build dams, and what compromises may be necessary, to mention a few.
.jpg)
