Abstract: The adaptive observation refers to those effective supplement observations in a sensitive area/duration to the regular observing network, which, through data assimilation, optimally improves forecasts of high-impact weather events of importance to society in a specified region. This new strategy is aimed at improving the skill of the numerical weather prediction has been paid great attention in the world in the past decade. A number of field observation experiments have been carried out successfully. Many research works on the adaptive observation strategies have been conducted to identify sensitive areas and quantitatively evaluate the impact of targeted observation assimilation on numerical weather predictions. In this paper, after a brief introduction of the objectives of targeted observations, the progress on the various methods to identify sensitive areas and the impact of targeting observations on the skill of numerical weather prediction is reviewed, and some key scientific issues for the use of targeted observations in the future to maximize the positive impact on forecasts are also discussed.