Thursday, April 25, 2024
1...
min
Artificial intelligence is now capable of predicting atmospheric and weather conditions with almost the same level of reliability as conventional forecasts, while also reducing calculation times and the environmental impact.
Knowing how the weather will behave with the greatest possible accuracy plays an essential role in minimizing the social, material, and environmental impacts of a changing climate. The emergence of certain technological tools, such as artificial intelligence, helps with the development and progress of forecast modeling, as experts have confirmed. "Thanks to AI, we are better prepared to tackle climate challenges," summarizes José Miguel Viñas, physicist and meteorologist at Meteored.
We are not talking about something that could happen in the future; it is already here. Without going any further, there is a team focused on this technology at the European Centre for Medium-Range Weather Forecasts, a global benchmark entity. According to the expert, the current method is based on sophisticated mathematical models that simulate the behavior of the atmosphere and are analyzed by a supercomputer.
However, "AI uses a different methodology that performs these calculations in significantly less time and produces results, in the case of weather forecasts for the next few days, that have a similar level of reliability as the traditional method," points out the specialist. How did we go from taking hours to complete these operations to performing them in just two minutes? Although it seems like magic, everything has an explanation. "The key is that AI works through machine learning, which means that it collects all the existing information about previous forecasts generated by traditional models and learns as new predictions are created," Viñas points out.
Tools like GraphCast, powered by Google, have started a journey based on technology and AI to discover almost unpredictable phenomena in advance so we can prepare for them. "This will have a very broad reach, especially because it uses Google’s search engines, which have transferred their abilities to the tool. It is capable of generating a forecast in just a few seconds with the same level of reliability as classic methods," explains the meteorologist.
The company itself published a study that showed how 90% of the predictions generated by GraphCast surpassed those of the European Centre for Medium-Range Weather Forecasts, the most reputable system in the world. In addition, this AI is a thousand times cheaper to use than the conventional method in terms of energy consumption. Time is another one of its great benefits. GraphCast can forecast the weather conditions up to ten days in advance in less than one minute.
Before GraphCast, Huawei had developed Pangu-Weather, another AI system that improves times and results. In this case, it allowed users to generate precise forecasts 10,000 times faster than the conventional method. What’s more, the creators of Huawei stated that the tool even works with extreme weather forecasts.
The developer Nvidia has also created an AI system for use in this field: FourCastnet. This application tracks tropical storms with greater accuracy than usual procedures. According to the latest data, FourCastnet is able to generate a 21-day weather forecast in just one-tenth of the time and requiring a thousand times less energy than the classic method.
However, beyond this aspect, these advances in production systems make it possible to minimize energy consumption and increase the sustainability of these processes, or create more detailed, reliable weather maps in shorter times. This could help increase citizens’ climate comfort and improve the plans of local government entities.
¿Te ha parecido interesante?
