Abstract: AbstractThe atmospheric science has been advanced from a subjective “art” to a mathema tical and physical science during the past century, with an accelerated pace in the recent 50 years. Major advances in atmospheric measurements, numerical weather prediction and climate research were reviewed and prospects of the atmospheric sciences in the 21st century were speculated. Undoubtedly, the rapid progress in atmospheric sciences in both the past and the future is highly determined by the technical developments in atmospheric measurements and computing power as well as information technology. As a result, the model grid resolution has continued to increase with time; a growing amount of in situ and remote sensing (e.g., satellite) data, after retrieval and assimilation, has been incorporated into the model initial conditions and the global meteorological analysis; model physical processes and parameterizations have become more realistic, thereby leading to the persistent improvements in the global and regional meteorological predictions. Similarly, the statistic analysis of tremendous observed data and the conduct of many coupled climate-model simulations have considerably advanced our understanding of the dynamical mechanisms by which climate change and its impact occur, particularly for those associated with the global warming, the effects of human activity, and the development of the ENSO-related events. Recently, it appears today that unified regional and global models tend to be developed; model prediction periods for our daily weather and larger-scale flows will be extended; and more efforts will likely be devoted to ensemble forecasts with sufficient numbers of members to represent uncertainty in both initial conditions and models. Climate research will focus more on minimizing uncertainties in the parameterization of various physical and chemical processes, and in the modelgrid resolution, in order to reasonably predict the regional to global climate changes at the monthly to intraseasonal and decadal time scales. It is speculated that in the near future measuring and forecasting various (meteorological or biological) variables in the earth-space system will be gradually digitized and automated; the location and timing of severe weather events may be predicted 3-5 days in advance; and some breakthroughs in weather modification could be made in the 21st century. Meanwhile, we may see the coordinated efforts from all the countries that are devoted to the establishment of a comprehensive earth-space-observing system, and the development of a unified, sophisticated earth-space-disaster mitigation modeling system. Then, the observed and modeled data will be used to monitor closely the natural variability of the earth-space system at all spatial scales to ensure the persistent advances in human society.