Rivers meander. Lakes grow and shrink. Engineers build dams and farmers flood fields.
In the 30-year animation* above, Bangladesh's Brahmaputra Riversweeps unpredictably across the landscape. Much of the world's water is in constant motion.
Scientists with the European Commission’s Joint Research Center in Ispra, Italy, working with Google engineers, have used millions of satellite images to illustrate how rivers, lakes and other bodies of water have changed over three decades.
The project, which is freely available, will allow researchers to improve climate models and find evidence of the effects of climate change around the world. But even the untrained eye can see the results of drought in the Western United States, for example, or of warming temperatures in the high country north of the Himalayas.
In the Western United States, drought and growing populations have led to a drawdown in the region’s biggest reservoir.
Lakes in other parts of the world are growing. Warming temperatures are melting East Asia’s glaciers, and large lakes have formed from the runoff.
The project shows, in more detail than ever before, how nature and people, including human-induced climate change, alter the pathways of water covering about a million square miles of Earth's land surface.
Even a simple activity, like digging river sand, can alter the landscape. Agriculture has an enormous influence, as farmers move water to cope with changing conditions.
Abandoned pits, remnants of sand mining along a river in Brazil, have filled with groundwater to create new lakes.
When fields grow too salty for crops, farmers flood them to create fish ponds. Adjacent protected areas that contain the world's largest mangrove forest remain relatively unchanged.
The project’s scope, as described in a paper published online Wednesday in the journal Nature, included the entire archive of three Landsat satellites from 1984 to 2015 — about three million high-resolution images, or nearly two quadrillion bytes (1.8 petabytes) of data. The images were processed in the cloud by 10,000 computers using Google’s Earth Engine platform, classifying each 100-foot by 100-foot pixel as water or land.