Predicting the weather is an unenviable task – meteorologists often feel the brunt of the public’s disdain when a forecast just isn’t quite right – snow when it should be sunny, or dry when they’re hunkered down for a thunderstorm.
The weatherman may need to watch out, because a new study suggests that he may become obsolete, set to be replaced with forecasting by machine learning and artificial intelligence.
Study first author Lei Han, a professor at Ocean University of China, stresses the importance of his team’s work.
“Accurate weather forecasting is critical in protecting socioeconomic conditions in many areas such as transportation, agriculture and water resources management,” said Professor Han.
Traditionally, meteorologists have used a narrow band of data, known as numerical weather prediction (NWP), that relies on various measurements of humidity, air pressure and observations. Without being informed by an understanding of atmospheric dynamics and physics, these models are prone to under or over-forecasting.
The researchers have created a new forecasting method based on inspiration from medical imaging software called U-net. This new software, CU-net (or correction U-Net), is getting progressively better at predicting weather patterns.
Rather than using purely localized data to predict the weather of a single locale, CU-net takes data from adjacent blocks and layers them on top of each other. The artificial intelligence then gets to work predicting what weather patterns may occur at the location, refining the variables of temperature, humidity, wind-speed, and wind direction to come up with what the researchers believe is a more accurate forecast.
“The confidence gained from this research can directly lead to innovative diagnostics for weather and climate variability, as we have never been able to imagine before the era of AI,” explained study co-author Haonan Chen from the University of Colorado.
These advances in technology allow meteorologists to make large-scale predictions about weather patterns. According to Dr. Chen, artificial intelligence will be a game changer in Earth system research.
The study is published in the journal Advances in Atmospheric Sciences.