Will artificial intelligence make weather forecasts more accurate? my country's new achievement is published in a top international journal

Since ChatGPT came out and artificial intelligence (AI) has become a hot topic again, many people have a question in their minds: I hope AI can help me sweep the floor and wash dishes, so that I can be in the mood to write poetry and paint. But why are AIs now writing poetry and painting, while we humans are still sweeping the floor and washing dishes?

Image taken from social media

The situation that netizens complained about is related to the development path of artificial intelligence technology and the commercialization process of technology, but it is not "deliberate" by scientific researchers. At the same time, there are also countless artificial intelligence researchers who are working hard to make this technology help everyone achieve some more practical goals.

Recently, the Huawei Cloud team published a large AI model in the journal Nature, which attempted to solve a very practical problem that plagues all of humanity: weather forecasting .

Copyright images in the gallery. Reprinting and using them may lead to copyright disputes.

One of the biggest difficulties in weather forecasting is the chaotic nature of the evolution of the atmospheric systems that determine how the weather occurs.

Although current NWP methods have made significant progress, there are still some limitations. First, NWP models require a lot of computing resources to solve complex physical equations. This can be a "fatal" problem for large-scale global weather forecasts. Second, NWP models usually require detailed initial conditions , which are often provided by satellite and ground measurements, but the data may have uncertainties or errors. This will affect the accuracy of the forecast.

Therefore, although we are able to make fairly accurate short-term weather forecasts (such as forecasts for the next few hours or days), the accuracy of medium- and long-term forecasts (such as forecasts for the next few weeks or months) remains a challenge. Today, the emergence of AI big models may allow us to move one step closer to more accurate medium- and long-term forecasts.

In the July 5, 2023 issue of Nature magazine, the Huawei Cloud team reported a work on medium-term weather forecasting based on the Huawei Cloud Pangu Weather (Pangu-Weather) large model. This work solved for the first time the global problem that AI weather forecasts are less accurate than traditional numerical forecasts , and the forecast speed is 10,000 times faster, achieving "second-level" global weather forecasts .

The core of Huawei Cloud Pangu Meteorological Model is a 3D Earth-Specific Transformer that can capture complex patterns in weather data. It uses about 40 years of global weather data for training. At the same time, the team adopted a hierarchical time-domain aggregation strategy to reduce the cumulative error in medium-term forecasts. As a result, the Pangu Meteorological Model surpasses the world's best NWP system in terms of accuracy and speed in some cases .

The Huawei Cloud R&D team found that there were two main reasons for the lack of accuracy of previous AI weather forecast models: first, the original AI weather forecast models were all based on 2D neural networks and could not handle uneven 3D weather data well; second, the AI ​​method lacked mathematical and physical mechanism constraints, so iterative errors would continue to accumulate during the iteration process.

A key breakthrough of Huawei Cloud Pangu Weather Model is the understanding of weather patterns on Earth. By integrating altitude information into new dimensions, the system can understand weather patterns in three dimensions , thereby predicting the weather more accurately. In addition, the hierarchical temporal aggregation strategy is also an important technical breakthrough, which greatly reduces the number of iterations required for medium-term weather forecasts, thereby reducing the cumulative error.

Copyright images in the gallery. Reprinting and using them may lead to copyright disputes.

The industry has given high praise to Huawei's research. Relevant experts generally believe that the emergence of the Pangu meteorological model represents a major breakthrough in the field of artificial intelligence in weather forecasting. Institutions such as the China Meteorological Center and the European Centre for Medium-Range Weather Forecasts (ECMWF) have confirmed the superiority of the Pangu model's predictions in actual measurements. The results include temperature, humidity, wind speed, sea level pressure, etc., which can be directly applied to multiple meteorological research scenarios.

However, although the Pangu model has opened up a new forecasting path in a sense, it still relies on NWP for training , and the results do not crush NWP. Therefore, from an objective point of view, peer experts have also pointed out some shortcomings and improvement directions of the Pangu meteorological model, which still need further research and verification by the research team.

In May 2023, Typhoon Mawar attracted worldwide attention as the strongest tropical cyclone so far this year. The National Meteorological Center used artificial intelligence to quickly enhance recognition technology and achieve trend forecasts 12 hours in advance . China Meteorological News reported that Huawei Cloud Pangu Model performed well in predicting Mawar's trajectory, predicting its turning trajectory in the waters east of Taiwan Island 5 days in advance. At the 19th World Meteorological Congress, the European Center for Medium-Term Forecasts also pointed out that Huawei Cloud Pangu Meteorological Model has undeniable capabilities in accuracy, and the purely data-driven AI weather forecast model has demonstrated forecasting capabilities comparable to numerical models.

The Huawei Cloud R&D team also proposed an adaptive learning strategy that allows the model to make real-time adjustments when making predictions based on new data. The successful implementation of this technology will further improve the prediction accuracy of the Pangu Meteorological Model , making it more valuable in practical applications.

Accurate weather forecasts have significant social and economic value in the fields of agriculture, aviation, energy, disaster warning, etc. For example, in agricultural applications, accurate precipitation forecasts will help farmers arrange farming activities reasonably and improve agricultural production efficiency. In the field of air transportation, accurate wind speed forecasts will help airlines arrange routes reasonably and reduce operating costs.

Over the past few decades, meteorological scientists have been working hard to improve the accuracy of weather forecasts, but there are still many challenges in this field. In the future, artificial intelligence technology may become the key to meeting these challenges.

Notes

[1] https://www.nature.com/articles/s41586-023-06185-3

[2] The abstract of the paper states that it is 39 years. The news report on Huawei’s official website states that it is 43 years.

[3] https://www.cma.gov.cn/en2014/news/News/202306/t20230607_5560758.html

[4] https://www.ecmwf.int/en/about/media-centre/science-blog/2023/rise-machine-learning-weather-forecasting

Planning and production

Author: Wenfei Science Reporter

Review丨Yu Yang, Head of Tencent Xuanwu Lab

Editor: Cui Yinghao

The cover image and the images in this article are from the copyright library

Reprinting may lead to copyright disputes