Abstract: The experiment deploying dropsonde was, for the first time in mainland China, launched on 7 August 2009 over the South China Sea. Eleven dropsondes are released when the aircraft flied across the relatively thin cloud zone between Typhoon Morakot and the tropical depression Goni. Based on the dropsonde data taken from this experiment, the reliability of dropsonde data is tested, then a preliminary analysis is done on the characteristics of atmosphere such as the wind field and humidity field over the South China Sea. The feature of wind profile in the southwestern part of Morakot is revealed as a deep layer of southwestern airflow from low to middle level of the troposphere, whereas the feature of the humidity field exhibits both relatively low humidity at the lower and upper level of the troposphere, and relatively moist at the middle level. To investigate the impact of the dropsonde data upon the typhoon's track forecast, two experiments are designed. One experiment with the initial condition obtained with the assimilation of dropsonde data (the assimilation experiment) and the other without it (the control experiment), are run for 72 hours by using the GRAPES model, respectively. The results are shown as follows. (1) With the assimilation of dropsonde data, southwesterly weakens below 800 hPa, then strengthens from 800 hPa to 100 hPa, with the humidity weakening below 500 hPa at initial time. (2) The difference of wind field between with and without the assimilation of data at the initial time propagates to downstream regions with the increase of integration time, affecting the environment of typhoon and modifying the typhoon steering flow. The forecast track error of typhoon is reduced by 18% in 48 hours with the assimilation of dropsonde data. It is considered that the assimilation of dropsonde data released in the upstream regions of Typhoon Morakot improves the typhoon forecast track in this case.