AI draws weather "barometer"
Editor's note
The World Meteorological Organization (WMO) released data last week (May 17) showing that "there is a 66% chance that the global average annual temperature will temporarily increase by 1.5 degrees Celsius above pre-industrial levels in the next five years." At the same time that this warning was issued to the world, a report published in the Proceedings of the National Academy of Sciences on artificial intelligence (AI) predicting that "global warming will break through the critical threshold (1.5 degrees Celsius in 10 to 12 years)" has attracted attention again. The thesis that "the prediction results generated by artificial intelligence are consistent and more accurate than those obtained by traditional forecasting methods" has led to a new round of discussion on the relationship between artificial intelligence and traditional weather forecasting.
Digital Intelligence Interview
——Interview with Ma Zhuguo, researcher of the Institute of Atmospheric Physics, Chinese Academy of Sciences
■ China Economic Times reporter Li Hainan
Recently, the news that "the World Meteorological Organization (WMO) announced that the probability of a temporary increase of 1.5 degrees Celsius in the global average annual temperature from pre-industrial levels in the next five years has attracted attention and discussion about whether artificial intelligence will challenge traditional weather prediction methods, or even replace it.
Ma Zhuguo, a researcher at the Institute of Atmospheric Physics of the Chinese Academy of Sciences, has long focused on climate change research, and as an expert in climate models, he is more aware of the close relationship between computing power, algorithms and information data processing. From the perspective of observers, he expressed positive feedback to the reporter of China Economic Times on the current application of artificial intelligence technology in the field of meteorology and climate, including the positive changes brought by weather forecasting and future application prospects.
"Artificial intelligence brings technical methods to weather forecasting"
Ma Zhuguo believes that the WMO announcement is intended to serve as a warning, after all, the economic losses and personal safety caused by extreme weather and climate performance cannot be ignored. This is also a major reason why human society has continuously strengthened technical support to improve the accuracy of meteorological forecasting.
"Everyone is concerned about improving technical means to achieve higher accuracy of early warning, and then to minimize the losses caused by extreme weather and meteorological disasters." Ma Zhuguo said, "Artificial intelligence is bringing improvements in technology and methods to weather forecasting, thereby improving the efficiency and accuracy of forecasting and making more accurate warnings possible." ”
The application of artificial intelligence in the field of meteorology is essentially the use of big data, high computing power and other technologies and the deep combination of meteorological prediction to form an intelligent cross-domain, multi-scale and accurate meteorological system, to achieve a greatly improved meteorological forecasting speed, enhance the timeliness of forecasting, and mine laws and mathematical equations from data to complement each other.
The technical means used in weather forecasting have undergone rapid iteration and progress. Ma Zhuguo himself has personally experienced the iterative changes from the use of early microcomputers in climate research to today's artificial intelligence. He said that in the early 80s of the 20th century, a programmable pocket computer, commonly known as PC-1500, could integrate the experience information and professional data of meteorological analysts to form an index reference to forecast meteorological conditions such as rainfall, hail, and wind.
The technological progress brought about by more than 40 years of time has been particularly impressive. "Efficiently and accurately extracting useful and comprehensive information from large amounts of information requires increasingly intensive computing power. The large model algorithm and high computing power behind artificial intelligence are aimed at improving data information processing capabilities, which can quickly extract usable information. He said.
"Limitations are difficult to break through, artificial intelligence is not omnipotent"
The development of new technologies often makes it difficult to break through its own limitations. Ma Zhuguo believes that even artificial intelligence technology with big data and high computing power has only achieved the processing of "a large enough amount of existing information data". He said that using this "past data" to speculate about the future is based on a basic premise: the future development law and the past have a certain correlation and similarity, which can rely on the "past data" to form an algorithm model, and then realize the prediction of the future. "But predicting the future, there's a lot of unpredictability."
Combined with the recent ChatGPT, Ma Zhuguo gives a visual comparison to illustrate the limitations of artificial intelligence. Suppose a user enters "2+8=12" into ChatGPT, and it replies "No". The user then asks "That's right." It will reply "2+8=12 data does not exist in my database, maybe you are right".
"This invisibly reflects that the artificial intelligence represented by this obviously also has certain limitations." Ma Zhuguo believes that ChatG? PT represents the large language model, and the artificial intelligence technology derived from it, the essence of the core is not a creator, but a very skilled information processing and integrator.
"To some extent, meteorological and climate forecasting also has such problems." Ma Zhuguo said that taking the latest climate model for climate prediction as an example, once the accuracy of the data in a certain link of the model is insufficient, it will definitely cause errors in the results.
Climate models are the use of mathematical methods to simulate weather changes. "The processes of climate change are not well understood, because the study of certain climate phenomena has to be hypothetical, and the more accurate the model, the more observations are required." Ma Zhuguo believes that the application of artificial intelligence technology in the meteorological field, including the large computing power and the large improvement of information and data processing capabilities brought by algorithms, is worthy of recognition, but it is still inevitable to have limitations.
Just as "no matter how you read through the history books, it is difficult to plan for the future," Ma Zhuguo said, no matter how much meteorologists understand the laws of past climate development, it is difficult to predict weather completely accurately.
"AI and traditional mainstream forecasting methods reinforce each other"
In Ma Zhuguo's view, artificial intelligence into weather forecasting and atmospheric physics application scenarios, essentially brings or through computing power, algorithms to integrate big data and other information, to provide more new technical support and methods for model models, and then to improve the accuracy and efficiency of forecasting, "but can not expect or simply say that artificial intelligence to replace traditional mainstream prediction methods, at least not yet." ”
Taking the well-known satellite cloud image as an example, Ma Zhuguo explained that in weather forecasting, satellite monitoring is very effective, can intuitively see the trajectory and speed of cloud travel, once the data-based model algorithm replaces manual intuitive observation, the change brought about by the essence is also obvious.
The conventional method of weather forecasting today is to use weather forecasting models for quantitative forecasting. Ma Zhuguo introduced that the model is essentially a mathematical model, that is, using the law of fluid dynamics to establish a motion equation for atmospheric fluids, and then generate a forecast equation with time changes. For example, if the weather at this moment is known, the weather in the next period of time can be judged according to the equation of motion.
"But in the process of running the equation, there are many assumptions to be defined, and this process will have trade-offs in the data, which will bring a certain error, because only certain assumptions can be made under the conditions of fluid dynamics can speculation and calculations be carried out." Ma Zhuguo believes that once big data and artificial intelligence technology are cited and applied, a large amount of data information can be entered, and all the information that can be dataized, such as the relationship between satellite cloud movements, can be statistically entered, and artificial intelligence will generate and extract more accurate usable information.
At present, human research in the field of meteorology still has limitations, more new understanding depends on scientific research exploration and breakthrough constraints, Ma Zhuguo believes that artificial intelligence-related technology will definitely benefit scientific research and help achieve new scientific results. In the field of meteorological and climate prediction, artificial intelligence and traditional mainstream forecasting methods promote each other.
Digital Intelligence Forward
"AI + meteorology" may become a new blue ocean for business
■ Lin Chunxia, reporter of China Economic Times
As the core driving force of the new round of scientific and technological revolution and industrial transformation, artificial intelligence (AI) is not only a national strategy and a new growth engine, but also the core competitiveness of industry competition. In recent years, artificial intelligence has not only improved the accuracy of weather forecasting, but also diversified the application of meteorological data.
The International Meteorological Organization (WMO) recently released data showing that the probability of a temporary increase of 1.5 degrees Celsius in the global average annual temperature in the next five years from pre-industrial levels has reached 66%. This is consistent with previous artificial intelligence predictions in terms of time, but the direction of climate warming. For a time, the market began a new round of attention to "AI + weather".
Weather services embrace artificial intelligence
In recent years, the mainland meteorological department and related enterprises have actively embraced artificial intelligence and continuously explored the application fields and methods of artificial intelligence meteorology.
For example, the Shenzhen Meteorological Bureau and HUAWEI CLOUD have carried out in-depth cooperation to build an all-field in-depth cooperation model of "weather + cloud + AI + 5G", promoting breakthrough innovation in accurate weather forecasting for megacities and meteorological services for smart cities.
Ink weather, which has been deeply engaged in meteorological services for many years, is also constantly using AI technology to expand the blue ocean market in the meteorological field. As early as 2016, Ink Weather began to explore the B-side, exerting commercial value in many fields such as urban construction, transportation services, agricultural meteorology, disaster prevention, etc., and gradually opening up the 100-billion-level "meteorology +" blue ocean market.
Taking the transportation industry as an example, under complex meteorological conditions, the probability of traffic accidents on the expressway is very large, the road traffic product portfolio service launched by Moji uses refined data products such as kilometer grid forecast and early warning, minute-level short-term radar, real-time cloud map, typhoon path forecast, etc. combined with traffic visualization chart analysis, to send early warning prompts for dangerous road sections, greatly reducing the incidence of traffic accidents.
The blue ocean of "AI + climate services" is in the ascendant, and its prospects are promising. According to the "China Meteorological Industry Development Report", the scale of China's meteorological service industry can reach 300 billion yuan in 2025, and the follow-up growth potential is strong. Of course, its potential is based on accurate forecasts.
Huo Zhiguo, a researcher at the Chinese Academy of Meteorological Sciences, said in an interview with the China Economic Times that with the help of digital technology and artificial intelligence and other high-tech means, weather forecasting has a higher accuracy and refined level than in the past.
High accuracy of weather forecasting plays a great role in empowering economic and social development. For example, when sowing seeds in agriculture, combined with weather forecasts, some targeted measures can be taken, and rain can reduce irrigation. If it does not rain for a long time, some remedial measures such as irrigation can be taken. For example, disaster forecasting, such as typhoon forecasting, allows fishermen to return to Hong Kong in time to avoid the wind, which can greatly reduce losses.
Explore the cutting-edge technology position of "Meteorology+" service
In 2022, the Outline for High-quality Development of Meteorology (2022-2035) clearly proposed to "strengthen the deep integration and application of artificial intelligence, big data, quantum computing and meteorology". In the AI era, smart weather has become one of the foundations of China's new digital economy infrastructure.
Jiang Qiping, director of the Information Research Center of the Chinese Academy of Social Sciences, said in an interview with the China Economic Times that artificial intelligence will play an increasing role in services in the future, including industrial Internet, including applications in all walks of life, including meteorological applications. Meteorological applications are complex large systems, and improving efficiency for complex systems is exactly what artificial intelligence is good at.
In the future, how will artificial intelligence be deeply integrated and applied with meteorology, bringing new changes and explorations to the meteorological field?
On May 18, at the seminar on the development of artificial intelligence meteorological application held by the China Meteorological Administration, experts and scholars from universities, research institutes, enterprises, etc., with the theme of "artificial intelligence empowers 'meteorology+'", in-depth discussion of new ideas and measures for artificial intelligence meteorological application.
Zhang Jun, Secretary of the Party Committee of Beijing Institute of Technology and Academician of the Chinese Academy of Engineering, said at the meeting that artificial intelligence and meteorological work are methodologically connected, and artificial intelligence has huge room for high-quality development of meteorology. Through the deep combination of artificial intelligence technology and meteorology, a global intelligent cross-domain, multi-scale and accurate meteorological system is established, and artificial intelligence can play a role in perception, transmission, computing, service and other fields.
Chen Yunji, deputy director and researcher of the Institute of Computing Technology of the Chinese Academy of Sciences, believes that based on intelligent meteorological science research, its focus is to improve the prediction ability of seasonal prediction and long-distance spatial connection modeling across multiple time scales, so as to achieve accurate forecasting and control of meteorological systems.
Xie Lingxi, a senior researcher at Huawei Cloud Computing Company, proposed that artificial intelligence has entered the field of weather forecasting, bringing many new ideas and paths. For example, it can greatly improve the speed of weather forecasting and enhance the timeliness of forecasting. Mining laws and mathematical equations from data complement each other. Rapidly developing new methods can help break the monopoly of old forecasting technology.
Digital Intelligence Lecture Hall
The focus and direction of artificial intelligence to help mankind cope with climate change
■ Zhou Hongchun
Climate change is a common challenge facing all humankind. Human beings are passengers on the "spaceship" of the earth, a community with a shared destiny, and must work together to address the challenges of climate change.
"AI predicts that the Earth's temperature will warm above the 1.5C mark between 2033 and 2035" is still lingering, and the World Meteorological Organization (WMO) has a 66% chance that "the global average annual temperature will rise by 1.5°C above pre-industrial levels in the next five years; There is a 98% probability that at least one year between 2023 and 2027 will be the hottest year on record."
At present, the use of artificial intelligence to help humanity to cope with climate change is bringing us great opportunities: it can be reflected in the implementation of carbon reduction action plans, climate change adaptation and public understanding.
Artificial intelligence, with the iterative advancement of big data models, is increasingly recognized as the next generation of general technology, becoming a breakthrough technology in the technological revolution and industrial revolution. AI is having a growing impact by analyzing, modelling and forecasting, as well as optimizing production processes, improving supply chain efficiency and productivity.
In addressing climate change, artificial intelligence has broad application space and will play an increasingly important role.
Artificial intelligence can improve the accuracy of weather forecasts. Studies have shown that 87% of AI experts confirm that AI will become an effective tool to combat climate change. Specifically, artificial intelligence can play an important role in the following aspects: first, predict areas with a high probability of floods, droughts, fires and other risks; The second is to predict severe climate (meteorological events) and issue early warnings to reduce labor costs; The third is to better allocate water resources for different uses of land within the administrative jurisdiction; The fourth is to choose infrastructure projects such as dams and fire protection projects with better utility investment.
The Boston Consulting Group's report, How Artificial Intelligence Can Be a Power Tool to Combat Climate Change, shows that using AI can help an organization reduce greenhouse gas emissions by 5 to 10 percent, and if extended globally, it will reduce greenhouse gas emissions by 2.6 billion to 5.3 billion tons of carbon dioxide equivalent.
Relevant studies show that artificial intelligence can help humans cope with climate change, including improving energy efficiency, reducing emissions in important areas, public participation, energy conservation in data centers, and developing and implementing adaptation plans for climate change.
For improving energy efficiency.
In the next 3 to 5 years, artificial intelligence will improve energy efficiency in related fields by 15%. Machine learning can support everything from automated maintenance, to leak monitoring, to process optimization, facility management, to power generation and distribution efficiency. Some AI tools can predict wind direction up to 36 hours in advance, thereby optimizing wind farm operation and reducing wind curtailment. Artificial intelligence can select and optimize the development and utilization of renewable energy, link various types of renewable energy generation nodes, bases and other links, and adjust and balance supply and demand.
Industrial manufacturing, transportation, construction, consumer goods, utilities and other fields can all use AI to achieve carbon reduction effects.
This is done by monitoring carbon emissions first. Use AI-powered data analytics to track the carbon footprint of your own operations, suppliers, users, and other parts of the value chain, and supplement missing data to improve monitoring accuracy. Through detailed insights into all aspects of the value chain, AI can improve the efficiency of production, transportation and other aspects of the company, reduce carbon emissions and reduce costs. AI can optimize traffic paths, traffic signals, continuously reduce operational emissions, and contribute to mitigating climate change.
In agriculture and carbon sequestration, voluntary emission reduction projects such as forestry carbon sequestration are an integral part of the roadmap for carbon peaking and carbon neutrality. AI research and calculations can also greatly improve crop yields, efficiency, and sustainability cycles by analyzing and modeling conditions such as atmospheric temperature, soil, bird migration, and conditions such as planting, irrigation, pesticide and fertilizer use and harvesting.
Carbon footprint and climate change advocacy. Climate change has had a significant impact on the global ecological environment, social and economic systems, and "achieving carbon peak and carbon neutrality is a broad and profound economic and social systemic change". In addressing climate change, AI can help people build platforms to track the carbon footprint of individuals and even enterprises, and design targeted countermeasures to reduce people's carbon emission reduction in daily life such as clothing, food, housing and transportation, so as to achieve the improvement of people's living standards and well-being with lower energy consumption and carbon emission reduction.
AI can help develop climate change adaptation plans. For example, the countries most vulnerable to climate change are using artificial intelligence systematically to initiate their adaptation actions; Some countries are using AI to initiate accurate mapping of crop distributions and predict the impact of climate change on crop harvests, among other things.
The carbon footprint of the development of compute-intensive technologies such as machine learning cannot be ignored. Therefore, in addition to using renewable energy wherever possible, general-purpose neural networks, artificial intelligence or machine learning models should be designed to be universal, and best practices and tools for measuring carbon efficiency should be used to reduce carbon emissions. Start small; Design AI solutions using the lowest cost concept, and iterate, integrate, and refine the solution in a timely manner. Strengthen capacity building, develop enabling technology platforms, and implement new governance models to maximize the benefits of AI to help reduce carbon emissions.
(The author is a researcher at the Development Research Center of the State Council)
The copyright of this official account belongs to China Economic Times. Reproduction or quotation of the content of this article must be licensed and indicated to be reproduced from the China Economic Times.
Hotspot links
●How long will international oil prices fall?
●Hot spot丨What does the "zero breakthrough" mean for the forced delisting of A-share convertible bonds
●What "hurdles" do new energy vehicles need to pass to the countryside
●Han DM is renewed, and the B-class sedan market ushered in "great changes"
Chief Producer丨Wang Hui Wang inherited
Producer丨Li Piguang, Wang Yu, Liu Weimin
Editor-in-chief丨Mao Jinghui Editor丨Gu Yun
【China Economic Times--China Economic News Network http://www.cet.com.cn 】