University of Arizona researchers have devised a way to measure wind using machine learning and algorithms – an innovation that could help predict extreme weather events and save lives in some of Earth’s most populous areas.
The novel approach pulls water vapor information from National Oceanic and Atmospheric Administration (NOAA) satellites. Scientists have long used NOAA raw data for modeling but now can process it with algorithms not available a decade ago.
This approach, led by Xubin Zeng, professor in the College of Science, overcomes limitations that have compromised models for predicting hurricanes, tracking airborne pollutants and more.
Deducing wind data from cloud movements, for example, offers very limited information for certain layers of atmosphere. Similarly, vast areas of the planet, including oceans, have few or no surface stations for pulling measurements from balloon-borne sensors.
Wind is a major player on the atmospheric stage, moving not just water vapor but also aerosols like dust and sea salt. It also drives the formation of clouds and precipitation, including extreme weather.
The innovation provides critical data for improving a range of climate and weather models and has given scientists the first comprehensive picture of winds across the tropics and midlatitudes – areas that are home to roughly 90% of Earth’s human population.
The team is now developing a satellite concept optimized for wind research based on this breakthrough and preceding research from the past five years.