In August 2021, as the Caldor Fire burned more than 200,000 acres in northern California, satellites captured the dramatic changes to the landscape in real-time.
That satellite data fed into a new Google tool, called Dynamic World, which recognized that an area once covered by trees had been reduced to shrub and scrub. In the days after the fire, Dynamic World’s color-coded map of the region transformed from green, where trees had grown in large enough numbers to be seen from space, to yellow, indicating a transformation to low scrub, showing the devastating outcome of the natural disaster on the land itself.
In the past, researchers interested in how the planet’s landscape changes over time have had to rely on large, cumbersome datasets and infrequently updated maps. Now, they can turn to Dynamic World, which shows exactly what covering the land — from crops and wetlands to buildings and trees — in great detail. The tool is updated in near real-time (about once every two to five days, depending on location), which makes it possible to monitor ecosystems as they rapidly evolve because of floods, wildfires, deforestation and urban development.
To create the new tool, Google partnered with World Resources Institute (WRI), an international environmental protection nonprofit, to determine the most critical land cover types across the world: water, flooded vegetation, built-up areas, trees, crops, bare ground , grass, shrub / scrub and snow / ice. Then, they created an artificial intelligence (AI) model that can recognize and label those land cover types on high-resolution satellite images captured by the Copernicus Sentinel-2 every two to five days for the last five years.
The researchers described their novel approach to land cover mapmaking in a paper published last week in the journalNature Scientific Data.
“We’re changing so fast, and the impact is so fast, that satellites are now the way to go,” says Fred Stolle, deputy director of the World Resources Institute’s Forests Program, Fast Company‘s Adele Peters.
The AI model can determine what’s covering the land in every pixel, each of which represents about 1,110 square feet (about 1 / 40th of an acre.) Researchers have not had access to that level of detail before. Typically, land cover maps have broad labels: A city might be listed on a satellite map as “built-up,” even though parks, open spaces and trees scatter among the high-rises.
“With this knowledge, [decision-makers] can develop plans to protect, manage and restore land and monitor the effectiveness of those plans using alert systems to notify when unforeseen land changes are taking place, ”writes Tanya Birch, senior program manager for Google Earth Outreach, in a blog post announcing the new tool.
Anyone can use the interactive tool to explore changing landscapes, but its creators believe it will be most useful to researchers, governments and nonprofits working to “develop helpful solutions and minimize their effects on issues like climate change, food insecurity and biodiversity loss.” per Birch. Researchers could use the tool to monitor the effects of drought on small farms, for instance, or to understand which city neighborhoods could benefit from more green space, per Fast Company.
However people decide to use the tool, as Craig Hanson, a vice president at World Resources Institute, says in Google’s announcement, one thing is clear: As humans put more and more pressure on the planet, the world needs to find “smarter, efficient and more sustainable ways to use land. ” Google hopes its new tool will do exactly that.