Geospatial visualization using hardware accelerated real-time volume rendering

We present a visualization framework using direct volume rendering techniques that achieves real-time performance and high image quality. The visualization program runs on a desktop as well as in an immersive environment. The application is named HurricaneVis, and it uses OpenGL, GLSL and VTK. For immersive visualization VRJuggler is added. To achieve real-time rendering rates for 4D scalar data we use the programmability of the GPU and in particular store the transfer functions as well as the 3D volume of scalar data on the GPU in texture memory. The initial use was visualization of scalar data from numerical weather model simulations of tropical cyclones, namely Hurricanes Isabelle and Lili. We are expanding that to include visualization of other types of data sets. We conducted a user study to compare the implemented volume rendering technique with state-of-the-art isosurface rendering. The subjects were students in the Dynamic Meteorology II and Physical Meteorology classes in the Department of Geosciences at Mississippi State University. The results establish that both volume rendering and isosurface visualizations are effective in examining data from computer simulations of hurricanes. Because of the higher image quality and the higher frame rates, direct volume rendering using ray-casting or view-aligned texture slicing was preferred.