American Researchers Develops AI-based system to Predict Storms, Cyclones

American Researchers Develops AI-based system to Predict Storms, Cyclones
The Siliconreview
08 July, 2019

A team of researchers from America have developed an AI-based algorithm to identify cloud formations which can indicate cyclones, hurricanes, and storms. The new model can assist weather scientist to spot threatening storms more accurately in less time.

During the development of the system, the team made an ML-based framework – that spots possible unnoticed rotational movementsin the clouds taken from the satellite images.

The researchers studied over 50,000 US weather satellite pictures for examining the behaviour of the cloud; especially spotted and labelled the shape and motion of clouds that resemble the shape of a comma.

According to the researchers, the identified cloud patterns are strongly linked to cyclone formations which can cause austere weather phenomena such as thunderstorms, hail, high winds and blizzards.

The team used ML and computer vision based methods in order to train computers to automatically spot and detect comma-shaped clouds in satellite images. After completing the process, the computers were capable enough to help experts by pointing out the correct data in real-time and showing the right method to spot the onset of severe weather.

The results of the study highlighted that the new method is capable of identifying a comma-shaped cloud accurately upto 99 percent. Apart from that, it was also able to forecast 64 percent of weather phenomenon which is threatening; eventually leaving behind all other existing severe weather detection techniques.

The research work was published in the journal IEEE Transactions on Geoscience and Remote Sensing.