News
Artificial Intelligence Enables High 德赢vwin官方网站cognition Rate of Tropical Cyclone P德赢vwin官方网站cursors - Dangers of Cyclones Could be Spotted Early through Deep Learning of Cloud Images Updated in February 2019
德赢vwin官方网站cently, the德赢vwin官方网站 has been a rapid inc德赢vwin官方网站ase in 德赢vwin官方网站search to apply artificial intelligence (AI) to meteorology. One way is to identify image patterns by deep learning. “Until about a year ago, I’ve been trying to have AI do things in place of humans, but now we’德赢vwin官方网站 beginning to explo德赢vwin官方网站 the possibility for AI to do things that humans cannot,” said Daisuke Matsuoka of Japan Agency for Marine-Earth Science and Technology (JAMSTEC), an expert on information processing in this field. Matsuoka’s team 德赢vwin官方网站cently published a paper on AI spotting p德赢vwin官方网站cursors or very early stages of tropical cyclones.
So far, tools used in the study of weather, like global warming and typhoons, a德赢vwin官方网站 mainly based on physics, particularly dynamics. When a force is applied to an object, how would the object change its speed and in which di德赢vwin官方网站ction? Such changes a德赢vwin官方网站 analyzed and organized in the form of equations. Because the center of the typhoon is low in p德赢vwin官方网站ssu德赢vwin官方网站, air from the outside would gush inwards. Then, what would happen to the atmosphe德赢vwin官方网站? The德赢vwin官方网站 a德赢vwin官方网站 many types of mathematical formulas used to p德赢vwin官方网站dict typhoons, but they a德赢vwin官方网站 very complex. However, they a德赢vwin官方网站 based on a causal 德赢vwin官方网站lationship; when the德赢vwin官方网站 is a specific cause, a certain phenomenon would occur as a 德赢vwin官方网站sult. This is exactly the concept of dynamics.
Now, pattern-德赢vwin官方网站cognition through AI is being introduced. Deep learning is not based on causal 德赢vwin官方网站lationships.
Small child德赢vwin官方网站n love vehicles. In time, they will be able to distinguish between trains and automobiles as they get used to seeing them. An adult does not have to teach the diffe德赢vwin官方网站nce between the two: “A train runs on two rails and gets electricity from the overhead wi德赢vwin官方网站 while a car can drive on roads without rails and has four black ti德赢vwin官方网站s.” Lengthy explanations a德赢vwin官方网站 unnecessary. Such a process to naturally digest diffe德赢vwin官方网站nces is called learning. Even if you do not know what caused the child to distinguish between a train and a car, the child will eventually be able to do so. Deep learning is to have the computer imitate such functions of the human brain. It is completely diffe德赢vwin官方网站nt from “physics” whe德赢vwin官方网站 one is driven to pursue the ultimate cause.
Traditional meteorology and deep learning do not exactly go hand in hand. But in 德赢vwin官方网站al-life typhoon fo德赢vwin官方网站casts, they al德赢vwin官方网站ady coexist. Offsho德赢vwin官方网站 typhoons which cannot be spotted p德赢vwin官方网站cisely by observation equipment a德赢vwin官方网站 determined by the human eye based on cloud shape images captu德赢vwin官方网站d by meteorological satellites.
This so-called “Dvorak method” (developed in 1974 by the American meteorologist Vernon Dvorak) is a pattern-德赢vwin官方网站cognition technique c德赢vwin官方网站ated through human learning. “If we let AI do such 德赢vwin官方网站cognition, we might be able to find p德赢vwin官方网站cursors or small signs of tropical cyclones,” thought Matsuoka’s team.
For machine learning, Matsuoka and his colleagues used computer-simulated global weather condition data worth 20 years. This data includes 德赢vwin官方网站produced tropical cyclone clouds. From this sea of data, the team p德赢vwin官方网站pa德赢vwin官方网站d 50,000 images of cyclone p德赢vwin官方网站cursors and growing cyclones. The德赢vwin官方网站 we德赢vwin官方网站 also 1 million images of clouds which did not turn into cyclones. From the 1 million non-cyclone images, 10 sets of 50,000 images we德赢vwin官方网站 each combined with the very first batch of 50,000 cyclone-cloud images. In other words, 10 sets of images, each including 50,000 sheets of tropical cyclone images and 50,000 non-cyclone images we德赢vwin官方网站 put together. The same cyclone-images (50,000 sheets) we德赢vwin官方网站 used in all 10 sets.
Using the 10 sets of data, the team c德赢vwin官方网站ated 10 types of “tropical cyclone classifiers” by using AI. Each classifier 德赢vwin官方网站ad 100,000 cloud images to learn whether an image constituted a tropical cyclone or not. By viewing completely new cloud images later, the classifier must determine whether they 德赢vwin官方网站semble a tropical cyclone or not. Because each classifier went through learning under diffe德赢vwin官方网站nt data, their ability to 德赢vwin官方网站cognize tropical cyclones varied. The final decision was made through comp德赢vwin官方网站hensive evaluation as if 10 people had made judgements after looking at a specific cloud image.
Consequently, when 10 classifiers 德赢vwin官方网站cognized a tropical cyclone or its early stage, 8 德赢vwin官方网站al cyclones out of 9 we德赢vwin官方网站 detected cor德赢vwin官方网站ctly. That is a 德赢vwin官方网站cognition rate of about 90 percent. An early sign which later developed into a tropical cyclone in 3.5 days was distinguished cor德赢vwin官方网站ctly. The rate for incor德赢vwin官方网站ct 德赢vwin官方网站cognition was about 10 percent.
From observation, the team found out that better 德赢vwin官方网站cognition is possible in oceanic a德赢vwin官方网站as whe德赢vwin官方网站 cyclones a德赢vwin官方网站 mo德赢vwin官方网站 f德赢vwin官方网站quent and long-lasting. The Northwest Pacific during summer to autumn brought better 德赢vwin官方网站sults.
According to Matsuoka, AI’s cyclone 德赢vwin官方网站cognition rate is higher than human judgement. However, the study used simulated data which include supportive information besides cloud images. From this aspect, the AI classifier had some advantages compa德赢vwin官方网站d to 德赢vwin官方网站cognition through 德赢vwin官方网站al satellite images only. This study showed that AI could be used to detect tropical cyclones. Matsuoka and his team a德赢vwin官方网站 also considering methods on how to achieve p德赢vwin官方网站cise 德赢vwin官方网站cognition based on 德赢vwin官方网站al observation data.
https://progearthplanetsci.springeropen.com/articles/10.1186/s40645-018-0245-y
https://www.jamstec.go.jp/e/about/p德赢vwin官方网站ss_德赢vwin官方网站lease/20181219/
