Google Earth Engine is taking its closest look yet at how landscapes are changing

Google Earth Engine is taking its closest look yet at how landscapes are changing thumbnail

Today Google Earth launched Dynamic World, a project that creates maps paired with an innovative deep learning AI model. It’s able to classify land cover by type (water, urban, forest, crops) at a resolution of 10 meters, or 32 feet. That means each pixel covers about 10 meters of land. For comparison, previous state-of-the-art technology had a 100 meter resolution (320 feet).

Dynamic Earth allows people to see from space the many ways that land cover changes over time. This includes natural seasonal changes, climate-exacerbated storms, and disasters, as well as long-term changes caused by human activity, such clearing wild habitats for crops, cattle, and logging. This project allows researchers and experts to observe how land cover changes naturally and flag any unexpected changes.

Users can visit Google’s Dynamic Web website and browse through the different datasets to see the marked maps. One map, for example, shows how water and greenery in Botswana’s Okavango Delta changes from the rainy to the dry seasons.

The map model draws satellite imagery from European Space Agency’s Sentinel-2 and can update its data stream every 2-5 days for global land cover monitoring. In fact, around 12 terabytes of data comes from the Sentinel-2 satellite every day. It then goes into Google’s data centres as well as Google Earth Engine, which is a cloud platform designed to organize and relay Earth observations and environmental analysis. The Earth Engine is connected with tens of thousands computers that process the data and derive insights using computer models.

To automatically label the land in satellite images, Google needed artificial intelligence. This project involved the development of an artificial intelligence that can label land cover. It was trained using 5 billion pixels labeled both by experts and non-experts. They identified pixels in Sentinel-2 images, and determined what type of land cover they were (water/tree, grass, flooded vegetation or built-up areas such as cities, crops, bare soil, shrub, snow). They would then present the model with an image that was not in the training data and ask it to classify these land cover types. There are shading differences as well as color differences that help distinguish different land types from the maps. Because pixels also convey probability. The more confident the model is with its classification accuracy, the brighter the color. This creates a textural effect as the topography changes from forest to land or water to land.

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

A detailed description of their dataset has been published in the journal Nature Scientific Data.

“We are making it all available under a free and open license,” Rebecca Moore, director of Google Earth, said in a press call ahead of the announcement. “The datasets can be downloaded for free.

About 10 years ago, Google and the World Resources Institute collaborated on Global Forest Watch, a project aimed at monitoring forest cover to protect these areas while looking for changes from illegal activities such as logging or mining. They are now trying to expand their efforts beyond observing and protecting one type of land cover.

The idea is to help make sense out of all the data available. Moore stated that while we have heard from many governments and [and] scientists that they are determined to take action, they lack environmental monitoring information so they can create science-based, data-informed policies, track their results, [and] communicate effectively with stakeholders. The irony isn’t that there aren’t tons of data. They are hungry for insights. They are looking for actionable guidance to help them make the right decisions.

It can be overwhelming to deal with raw data in many cases.

Google believes that Dynamic World can play a role in filling the data gap about land use and cover. It can also describe where important ecosystems such as forests, water resources, and urban development are located. Moore stated that this type of information can help guide decisions about sustainable management for scarce natural resources, food, water, and other resources. It can help answer questions about disaster resilience, sea-level rise, creating protected areas, putting in dams, and what tradeoffs may be required, just to name a few.

Read More