More than 700 imaging satellites orbit the earth, but only governments and companies with wealth and expertise can access the data they produce. Now, researchers said in a recent paper that they have invented a machine learning system using low-cost, easy-to-use technology that could bring satellite analytical power to researchers and governments worldwide. “To plan infrastructure like roads and bridges or to target food aid, we need to know where people live and what their needs are,” Jonathan Proctor, a co-author of the paper, told Lifewire in an email interview. “Satellite imagery and machine learning can help measure socio-economic conditions in places where other measurements are insufficient.”
Eyes in the Sky
The growing fleet of imaging satellites beams about 80 terabytes every day back to Earth, according to the research paper. But often, imaging satellites are built to capture information on narrow topics such as supplies of freshwater. The data doesn’t arrive as neat, orderly images, like snapshots. Instead, it’s raw data, a mass of binary information, and researchers who access the data have to know what they’re seeking. Storing so many terabytes of data is costly. Distilling the data layers embedded in the images requires additional computing power and human experts to decipher it. To solve these problems, researchers at UC Berkeley developed MOSAIKS, short for Multi-Task Observation using Satellite Imagery & Kitchen Sinks. It can analyze hundreds of variables drawn from satellite data—from soil and water conditions to housing, health, and poverty—globally. The research paper shows how MOSAIKS could replicate with minimal investment costly reports prepared by the US Census Bureau. “Combining machine learning and remote sensing has the potential to help us monitor ecological change, plan future infrastructure developments, and respond to natural disasters in real-time,” Esther Rolf, a co-author on the paper, told Lifewire in an email interview.
Help From Above
“Several developing nations are fusing emerging technology (AI, automation, cloud, etc.) with satellite data to help accelerate national infrastructure projects,” he added. “Satellite data can include temperature measurements that support global warming studies,” Iain Goodridge, senior director of marketing at Spire Global, a company that uses satellites to provide data and analytics, told Lifewire in an email interview. Soil moisture readings can aid early warnings for droughts and wildfires, even in remote areas. The same weather data that helps predict rain in the afternoon may also help identify communities at risk of infectious disease, Goodridge said. “That is because environmental conditions can impact transmission,” he added. “To account for these factors, epidemiologists sometimes include weather data—such as temperature, humidity, and ultraviolet index—in models that forecast how diseases spread.” Satellite data can also help analyze weather patterns and the risk of natural disasters for a region when planning infrastructure from residential homes to power grids. The recent invention of MOSAIKS could bring the benefits of satellite data to more people. “Overall, the uptake of remotely sensed predictions of economic outcomes to inform public decision-making is in its infancy,” Proctor said. “The increasing abundance of satellite imagery and machine learning algorithms, however, is likely to initiate a growth spurt in the coming years.”
