Skip to main content
Advertisement
Live broadcast
Main slide
Beginning of the article
Озвучить текст
Select important
On
Off

Russian experts have taught artificial intelligence to detect incipient storms and hurricanes on satellite images within half an hour after their formation. The development will allow forecasters to warn about dangerous weather events faster, and forecasts will become more accurate. According to the scientists, processing the data manually would take months, whereas the AI handles it in minutes. Climatologists note that unusual snowfalls this winter can lead to anomalies in spring and summer, and with global warming, extreme weather will become more frequent.

AI for hurricane prediction

Scientists from MIPT and the A.M. Obukhov Institute of Atmospheric Physics of the Russian Academy of Sciences and colleagues have created an AI—based system for automatic detection of dangerous weather phenomena on satellite images - mesoscale convective systems (ISS). They often cause heavy rains, hail, and hurricanes. The development is able to identify the ISS as early as 15-30 minutes after the start of its formation, thanks to this, storm warnings can be made earlier and more accurately than now.

Первые снимки Земли

The first images of the Earth taken by Meteosat satellites from December 9, 1977 to August 4, 2015

Photo: ESA

— Currently, the classical weather forecast is used to predict anomalies in a few days, but it is not accurate enough. With the help of satellite images, we can detect a dangerous weather phenomenon in a few hours. First, identify a mesoscale convective system, and then use the model and additional data to make calculations based on how similar events developed in the past. This will provide more accurate predictability of the position of dangerous storms and squalls and allow for earlier warnings," said Alexander Chernokulsky, the author of the study and Deputy director of the Obukhov Institute of Atmospheric Physics of the Russian Academy of Sciences.

According to the scientist, the hurricane of May 29, 2017 in Moscow can serve as an example of the consequences of the ISS development. Then 18 people died, about 170 were injured, and the speed of the squally wind reached 28 m / s. The operation of urban public transport was disrupted, more than 300 settlements were left without electricity. However, in addition to the created system for accurate forecasts, high-quality satellite images are also needed, which are currently in short supply, the specialist noted.

Ураган

A tree broken by a hurricane, in the courtyard of an apartment building in Moscow, 2017

Photo: RIA Novosti/Yana Burmistrova

Mesoscale convective systems are a complex of phenomena that occur against the background of strong convection in the atmosphere. One of their manifestations is clusters of thunderstorms with intense downpours, large hail, and hurricane—force winds. According to the developers, thousands of people in Russia suffer from the ISS every year, and the economic damage amounts to billions of rubles.

It is extremely difficult to detect or predict the appearance of such systems in the first few minutes. Weather models capable of describing the ISS require significant computing resources, and the formation process depends on many factors. In addition, visual processing of large amounts of satellite data takes a lot of time.

AI for effective analysis

The new technology makes it possible to detect the ISS using satellite data in the first 15-30 minutes after formation. When implemented into operational monitoring systems, it will be able to radically reduce the response time of special services and reduce damage.

— The main strength of our development is efficiency. Where a meteorologist would have to manually analyze mountains of satellite images for months, artificial intelligence can handle it in a matter of tens of minutes. Fast processing of large amounts of data allows timely warning of services, reducing damage and creating detailed maps of the appearance of mesoscale convective systems. It also helps to study their changes in the context of climate change and increases the accuracy of modern climate models, which are currently being developed by more than a hundred research teams around the world," said Mikhail Krinitsky, head of the Machine Learning Laboratory in Geosciences at MIPT.

In the future, the system will allow residents of cities to be notified precisely about the approach of squalls, hail or showers associated with the ISS in a couple of hours so that they can prepare. It will help farmers protect crops from large hail, and transport companies to change cargo transportation routes in time, the developers believe.

Now scientists plan to expand the database, improve algorithms, adapt the system to data from Russian meteorological satellites, and eventually create an automatic monitoring and tracking system for the ISS for the entire territory of Russia. This is especially important in a changing climate and the associated increasing frequency of extreme events.

Vladimir Semenov, head of the Laboratory of Climatology at the Institute of Geography of the Russian Academy of Sciences, told Izvestia that an abnormally large amount of snow in Moscow could cause unexpected weather events in the coming spring and summer.

— Snow anomalies affect soil moisture anomalies, and these, in turn, are related to atmospheric circulation. However, summer is still three months away, and a lot can change during that time. Snow, for example, can quickly evaporate. Everything will depend on the weather conditions in April and May," the scientist said.

Due to global climate change over the past 40 years, summers have become warmer by one and a half degrees. Heat waves will appear more often. The number of extreme weather events will also increase, and precipitation will become stronger, the climatologist noted.

Employees of the Typhoon Scientific and Production Association and Perm State National Research University also participated in the development of the system.

Переведено сервисом «Яндекс Переводчик»

Live broadcast