Explore how the world uses the earth with Dynamic World

Today, Google Earth is launching a project called Dynamic World, a new effort to create maps that are paired with a new deep learning AI model. It can be classified according to the type of land cover (water, urban, forest, crops) at a resolution of 10 meters or 32 feet. This means that each pixel takes up about 10 meters of land. For comparison, the state-of-the-art technology had a resolution of 100 meters (320 feet).

The dynamic world is a way for people to observe from space Earth cover in a variety of ways, such as natural seasonal changes, storms and catastrophes that have exacerbated climate change, or long-term changes caused by human activities, such as cleaning. wild habitats for crops, livestock or exploitation. Experts and researchers can use this new project to understand how land cover changes naturally, and point out when some unexpected changes are happening.

Users can go to Google’s Dynamic World website to browse different datasets and see what the marked maps look like. For example, a map shows how the volume of water and green in the Okavango Delta in Botswana blooms and recedes from the rainy season to the dry season.

The map model, which captures Sentinel-2 satellite imagery from the European Space Agency, can update its data flow every 2 to 5 days for global land cover monitoring. In fact, about 12 terabytes come from the Sentinel-2 satellite every day. From there, it accesses Google’s data centers and Google Earth Engine, a cloud platform built to organize and deliver Earth observations and environmental analysis. Earth Engine is connected to tens of thousands of computers that process information and gain knowledge of computer models before being available in the Earth Engine Data Catalog.

In order to be able to automatically tag how the land represented in all these satellite images is used, Google needed the help of artificial intelligence. The ground cover labeling AI developed in this project was trained in 5,000 billion pixels tagged by human experts (and some non-experts). In the training data, they identified pixels in Sentinel-2 images and built areas such as what kind of land cover they were (water, tree, grass, overgrown vegetation, cities, crops, bare land, shrubs, snow). The model was then presented with an image that was not in the training set, and asked to classify the types of land cover. There are not only color differences but also differences in shadows to distinguish different types of land on the maps. That’s because pixels also transmit probability. The brighter the color, the more confident the pattern is in the accuracy of the classification. This creates a texturing effect when the topography moves from land to forest or from land to water.

[Related: Google Street View just unveiled its new camera—and it looks like an owl]

A detailed description of their data set has been published in the journal Natural Science Data.

“We’re making everything available with a free license,” Google Earth Director Rebecca Moore said in a press release before the announcement. “The data sets are free and open source. The AI ​​model is open source. ”

About 10 years ago, Google and the World Resources Institute partnered with Global Forest Watch, a project to control forest cover to protect these areas while seeking changes in illegal activities, such as timber or mining. Now, they are trying to expand their efforts beyond just protecting and observing a single type of land cover.

The idea is to help make sense of the data available. “We have heard from some governments, [and] that researchers are committed to taking action, but lacking information to monitor the environment about what is happening on earth, to create policies based on data based on science, to monitor the results of their actions, [and] “The irony is that there isn’t a lot of data. But they’re thirsty. They’re looking for viable guidelines to help them make the decisions they need to make.

Google believes that the role of Dynamic World is to fill the gap in land use and land cover data, and to describe the location of basic ecosystems such as forests, water resources, agriculture and urban development. This type of information, Moore said, can be useful in guiding decisions about the sustainable management of natural resources, food and water. It can also help you manage disaster resilience, deal with rising sea levels, create protected areas, put up dams and make commitments, to name a few.

Leave a Comment