Compared to continental regions, oceans is generally data sparse. It has been proved that the data supplement over the oceanic region is very crucial for weather forecast in downstream continental area, especially when the data are supplied in the area where the verifying forecast is most sensitive to changes in the initial condition. Recently, several experiments, such as U.S. Air Force Reserves (USAFR) drowpwindsonde experiment, FASTEX and NORPEX offer opportunities to test the feasibility of this kind application by so-called "targeting" or "adaptive" observation technique. Based on those experiments, this presentation will introduce two useful targeting techniques: adjoint sensitivity and quasi-inverse linear method. The focus of this study is to examine the forecast sensitivity to initial analysis differences, forced by these observation by using both adjoint method and quasi-inverse linear method. The relationship between forecast error sensitivity and adaptive observations will be explained. Several suggestions will also be given for future adaptive strategies. Following these results, the real-time FASTEX results and NORPEX post-analysis results will be presented with the disscussion in several issues, such as the comparison of the different targeting techniques, the role of the data assimilation scheme in the adaptive observation, the impact of the changes of data assimilation aspects on the evaluation results etc.. Finally, the potential of integrating the adaptive observations or other data type into operational forecast system will be discussed.