国家气象中心

Artificial intelligence (AI) is rapidly transforming Earth science, offering unprecedented capabilities to tackle the most pressing challenges in the field. This work explores significant advances and emerging challenges across the AI for atmosphere-ocean sciences, while outlining critical ways forward. We review deep-learning methods and their application in weather and climate forecasting, which outperforms dynamical models in accuracy and computational efficiency. The role of AI in detecting complex phenomena, enhancing data assimilation and reconstruction, bias correction and downscaling coarse model outputs is also examined. However, the 'black-box' nature of complex AI models necessitates a focus on explainable AI to build trust and extract mechanistic insight. The most promising path forward is identified as the development of hybrid physics-AI modeling, which integrates the data-driven power of AI with the foundational constraints of physical laws to ensure generalizability and causal consistency. A new framework for AI-based model intercomparison is essential for rigorous benchmark performance. Finally, we contextualize these technical developments by discussing the usefulness and applicability of AI to society, including the improvement of multi-hazard early-warning systems and green energy production. We conclude by envisioning the future of AI agents for Earth science-autonomous, goal-oriented systems capable of designing and running experiments, generating and testing hypotheses, and learning dynamics from multisource data. This synthesis underscores that AI is not merely a tool, but a paradigm shift, which will significantly improve how we understand and adapt to a changing climate.

The development of traditional numerical weather prediction (NWP) relies on continuous advances in observation technology, data assimilation methods, numerical and parameterization algorithms, and the steady growth of computational resources, resulting in lengthy development cycles and relatively slow improvements in forecast skill. In recent years, machine learning (ML)-based weather forecasting models have advanced rapidly, and in some aspects, outperform traditional physical models, particularly in forecasting large-scale circulation. However, these ML-based models suffer from notable deficiencies, such as over-smoothing in forecasts and inadequate capability for predicting extreme weather events. In this study, an online correction system based on the spectral nudging (SN) method is developed. In this system, the China Meteorological Administration Global Forecast System (CMA-GFS) is used as the underpinning physical model, and a correction term is integrated into the governing equations, such that during numerical integration, the large-scale circulation is constrained to evolve toward the forecasts produced by the ML model FuXi. In this proof-of-concept study, both of the CMA-GFS and FuXi are initialized with ERA5 data. The performance of the hybrid system on large-scale circulation prediction is comparable to that of the FuXi model, with a substantial extension of forecast leading time and a marked improvement in the stability of forecast skill. Verification against high-impact weather events, including heavy rainfall and tropical cyclones, demonstrates that the hybrid system integrates the strengths of the FuXi model in forecasting circulation patterns, precipitation distribution and tropical cyclone tracks, while preserving the advantages of the CMA-GFS in representing precipitation intensity, tropical cyclone intensity and fine-scale details. This study realizes the independent implementation of the SN method based on a different combination of physical and ML models and further verifies its effectiveness in precipitation and western North Pacific tropical cyclone forecasts, providing additional evidence for potential operational application.

After landfall, tropical cyclone (TC) remnants may maintain or even rejuvenate and incur catastrophic disasters. What leads to the revival of TC remnants over land remains elusive. In this study, the revival mechanism of Typhoon Doksuri (2023) remnants is extensively explored. Doksuri brought severe damage to the Chinese mainland after its landfall. The remnants vortex of Doksuri sustained an inland trajectory for 3 days and underwent a total maintenance of 60 h, with a revival of 18 h. Based on multi-source observations and ERA5 reanalysis data, by calculation of moist potential vorticity and analysis of slantwise vorticity development (SVD), this study unveils that while maintaining a significant warm-core structure over the course of maintenance and revival, the Doksuri remnants transported sufficient moisture in the mid–lower troposphere, which intensified the north–south temperature and humidity gradients, causing tilting of the isentropic surfaces remarkably. According to the SVD theory, the tilting gave rise to vorticity development and forced upward air motion on the northern side of the remnant vortex. Moreover, numerical sensitivity experiments based on the WRF model reveal that the topography of Taihang Mountains and the diabatic heating associated with surface and convective latent heat fluxes also played important roles in the revival of the Doksuri remnants. The dynamic and thermodynamic mechanisms derived by this study will help improve understanding and prediction of the disasters induced by TC remnants.

Weather forecasting is the core of meteorological operations.Following the establishment of the People's Republic of China,systematic development of weather forecasting services and operational capabilities began.The evolution of weather forecasting over the past seven decades were reviewed,summarizing the operational characteristics of three key phases:Empirical synoptic forecasting,modern development driven by numerical weather prediction(NWP),and specialized fine-scale forecasting based on NWP advancements.Furthermore,this paper examines technological progress in specific forecasting areas,including numerical models,meteorological elements,and severe weather predictions.This study highlights that the modernization of weather forecasting has benefited from advancements in meteorological science and technology,particularly the widespread application of numerical prediction techniques and the integration of diverse observational methods.By reviewing this developmental journey,the present paper emphasizes the importance of adopting advanced technologies,enhancing the synergy between research and operational practices,and advancing the modernization of meteorological services.

The similarities and differences in environmental conditions between daytime and nighttime torrential precipitation processes in North China have not been fully clarified.Based on precipitation data collected at 981 surface meteorological stations and ERA5 reanalysis data,the spatial and temporal distribution characteristics and environmental conditions of daytime and nighttime types of torrential precipitation processes in North China during the period from May to September of 2013-2023 are comprehensively analyzed.The objective classification method of the obliquely rotated T-mode principal component analysis is used to classify the circulation situations of daytime and nighttime types of torrential precipitation processes,and the characteristics of environmental physical variables of their corresponding circulation situations are then compared and analyzed.Direct comparison reveals that the environmental conditions of the nighttime heavy precipitation process in North China are different from that of the daytime heavy precipitation process,which provides an important basis for deepening our understanding of the formation mechanism of nighttime heavy rainfall in North China.The results show that the nighttime type torrential precipitation processes in North China develop more often after midnight,and have more occurrences over more concentrated regions,while the daytime processes and the first half-night precipitation of the nighttime processes have stronger convection and they mainly occur in July and August.Moisture of nighttime type is richer than that of daytime type,while CAPE of daytime type is higher than that of nighttime type.The distributions of both 850 hPa and 500 hPa temperature difference and 850 hPa vertical velocity are similar between the two types.Low-level wind speed and 0-1 km vertical wind shear are significantly higher in the nighttime type than in the daytime type.Low troughs and vortices at the edge of the subtropical high are the main synoptic systems influencing torrential precipitation processes in North China.The distribution characteristics of physical variable of the environments in different types of circulation situations are somewhat different.Moisture of deep trough circulation of daytime type and cold vortex circulation of nighttime type are the worst.0-6 km vertical wind shear(SHR6)and 0-3 km vertical wind shear(SHR3)are generally not strong.SHR6 of daytime torrential precipitation processes is slightly stronger than that of nighttime type,and SHR3 of nighttime torrential precipitation processes is slightly stronger than that of daytime type.The above results indicate that the nighttime heavy rainfall over North China is closely related to the East Asian summer monsoon,which is characterized by abundant water vapor,high θse value,appropriate CAPE value and strong wind speed in the lower atmosphere.The low-level wind field and SHR3 distribution indicate that one of the dominant factors of nighttime heavy rainfall over North China is the diurnal variations of low level jet or strong wind speed.