The exponentially increasing amount of Earth Observation (EO) data requires novel approaches for data mining and exploration. Visual analytic systems have made valuable contribution in understanding the structure of data by providing humans with visual perception of data. However, these systems have limitations in dealing with large-scale high-dimensional data. For instance, the limitation in dimension of the display screen prevents visualizing high-dimensional data points. In this paper, we propose a virtual reality based visual analytic system, so called Immersive Information Mining, to enable knowledge discovery from the EO archive. In this system, Dimension Reduction (DR) techniques are applied to high-dimensional data to map into a lower-dimensional space to be visualized in an immersive 3D virtual environment. In such a system, users are able to navigate within the data volume to get visual perception. Moreover, they can manipulate the data and provide feedback for other processing steps to improve the performance of data mining system.