【Wen / Observers News Zhang Jingjuan】 In the tide of the era where artificial intelligence is reconfiguring the weather forecasting system, meteorological data has risen to become a core national strategic resource.
Bloomberg reported on the 22nd that China is vigorously competing for the status of a meteorological superpower, promoting the development of its own meteorological data sets in an effort to break dependence on Europe.
According to the report, ERA5 (the fifth generation of atmospheric reanalysis data set from the European Centre for Medium-Range Weather Forecasts) is a benchmark in the field of climate data. It covers multiple meteorological elements such as rainfall, temperature, and wind, with a time span of over 80 years, and is continuously updated. Up to now, this data set has been the core support for the AI revolution in weather forecasting. Several major Chinese companies' AI weather models use this data set for training work. However, this contradicts China's strategy of data security and technological self-reliance.
The strategic value of meteorological data has long gone beyond weather forecasting itself. ERA5 reconstructs complete historical climate records by integrating global observation data, which is crucial for analyzing climate trends and improving forecast accuracy. Currently, many governments around the world use ERA5 data for flood, wildfire and other disaster risk management, while insurance companies rely on it to build disaster models. The EU estimates that the economic value created by this data set can reach hundreds of millions of dollars annually.
However, Professor Andreas Prein from ETH Zurich, a meteorology and climate modeling expert, said that weather forecasting is related to national security. If only relying on external data sources, the country will be in a passive and vulnerable position.
To avoid potential risks and gain control over the meteorological field, the National Data Bureau stated in a statement in September that the China Meteorological Administration has launched a project to develop a global atmospheric reanalysis system, one of the core goals being "to break China's reliance on Western reanalysis products in meteorological operations."
In the same month, the China Meteorological Administration opened download access for the new version of the global atmospheric reanalysis product CMA-RA V1.5 data set to the world for the first time. The department stated that some AI weather models in the country have already used this data set for training.
Monitoring Chart, National Meteorological Information Center
This domestically developed data set is gradually showing the potential for replacement, and through three technical system innovations, it is driving China's reanalysis from "following" and "running side by side" to "partial leadership".
First, it achieved an upgrade in assimilation technology. Introducing four-dimensional ensemble variational hybrid assimilation technology, overcoming multiple key technologies, the application volume of satellite data in the first 20 years increased by 13%; constructing a flow-dependent Be matrix to enhance assimilation efficiency, with product quality better than CRA-40 and JRA-55.
Second, it achieved autonomous control of domestic observation data. Integrating unique domestic data, independently developing radiosonde bias correction technology, assimilating global 116 satellites with 215 types of data, among which 37 satellites with 45 types are domestic, accounting for as high as 18%.
Third, it has international leading resolution and timeliness. The model resolution is 13 kilometers (post-processed to 10 kilometers), with hourly real-time updates, better than ERA5 (25 kilometers, delayed 5 days update).
Professor Hui Su from the Hong Kong University of Science and Technology is applying it to her startup company Stellerus, for training regional AI weather models and evaluating numerical weather prediction models. She frankly said that one of the advantages of this data set is that its global grid division accuracy is higher than ERA5. This high spatiotemporal resolution characteristic can provide massive data support for model training.
The release of the domestic data set has also attracted international industry attention. David Whitehead, head of the meteorological risk management department at Vaisala Oyj, said, "If the international market can obtain more Chinese meteorological data, a large number of companies will invest in designing and brokering weather derivative contracts." This Finnish listed company, which specializes in providing financial hedging meteorological data, has recently begun to study the potential application scenarios of CMA-RA V1.5.
According to Rémi Gandoin, product development manager at Danish engineering consulting company C2Wind, ERA5 has data biases and defects. Integrating multiple data sets will help researchers studying climate change and extreme weather to work, and also provide reference for wind farm developers' wind farm engineering design decisions.
"In the future, having multiple rather than a single meteorological data set will bring practical benefits to all parties," Gandoin said.
This article is exclusive to Observers News. Unauthorized reproduction is prohibited.
Original: toutiao.com/article/7586904404851278378/
Statement: The article represents the personal views of the author.