"Can numerical models help weather forecast?" On the afternoon of March 25, Zhang Pingwen, an academician of the Chinese Academy of Sciences, member of the Standing Committee of the Party Committee and vice president of Peking University, visited the School of Mathematics and gave a speech on the application of artificial intelligence to current weather forecast.
Zhang Pingwen started with the limits of the model. Numerical model is one of the most important means of weather forecast, which is composed of the atmospheric dynamic system and computable modeling modules. Taking the temperature forecast of Winter Olympic Games area as an example, he explained the limitations of the numerical model in actual weather forecast. "But we must notice that during the course of improving weather forecast, people proposed quite a few new mathematical concepts such as theChaos theory" Zhang said during the speech.
As Zhang put it, in actual forecast, model forecast data should be combined with the actual observation data from radar, satellites, weather stations, aviation source and other channels, and the forecast results should be determined through the weather consultation panel participated by experts. “Our goal is to use algorithms to merge information and data, and replace or even surpass weather consultation panel through machine learning post-processing.” Zhang Pingwen introduced model post-processing as an indispensable sector in weather forecast. To serve the 2022 Beijing Winter Olympic Games, he studied and shared a large number of research cases on the forecast of temperature, relative humidity, wind speed and wind direction of the stations in Winter Olympic Games sites, showing significant advantages in temperature forecast.
Finally, he told the differences between machine learning post-processing and statistical post-processing, as well as the trade-off between weak mechanism and accuracy.
In the Q & A session, Zhang Pingwen answered some specific questions about the prediction of climate by artificial intelligence, the optimization of empirical parameters by machine learning, and the combination of community weather forecast and urban model. In view of the advantages of traditional computing mathematics and suggestions on data resource acquisition, Zhang Pingwen shared his opinions on scholarship and making research plans for young scholars. He proposed that choosing a topic was very important for any study, and it often depended on personal goals and values. “Simplicity and beauty are the shared values of mathematics and physics, and to solve practical problems by way of interdisciplinary research is also quite fun".
“What kind of person you want to be determines what kind of research you will do”, Zhang’s words reflected his love for the beauty of algorithm and mathematics.
By the School of Mathematics
Editor: Yang Fan & Eva Yin
