The most common types of scientific visualizations have, thus far, been two-dimensional plots, either simple cartesian plots or contour plots. Most analysis tools today give the option of plotting data in three dimensions. These tools typically rely on perspective plots – that is, they can create the appearance of, say, a topographic surface as it would appear without stereo vision by utilizing the ‘trick’ of perspective. Figure 1 shows bathymetric data (of a submarine canyon) from off the Washington coast as a standard (100m) contour plot (Fig. 1) and as a perspective plot of a three-dimensional surface (Fig. 2). Not only does the perspective plot give the observer a more

	Fig. 1. 2-D contour plot of Astoria Canyon showing 100 meter depth contours hiding spurious data errors.
	Fig. 2. 3-D representation of Astoria Canyon with data 'spikes' readily observable.

recognizable object, but the ‘spikes’ in the bathymetry data are immediately observable – no usable choice of contours in the contour plot could tell the user that these spikes exist, or whether there is any pattern to them. Perspective plots are definitely useful, but the next step in the evolution of scientific visualization is to add motion parallax and stereo rendering.

The ability to rotate and zoom in on an object allows the observer to utilize motion parallax to more accurately just space and distance. This is a crucial next step for gaining insight into physical systems. Looking at a detailed rendering of bathymetry, and velocity vectors may reveal topographic steering by features that were masked by more traditional contour plots. The relationship between a density isosurface and velocity may suggest processes that had not been obvious to the scientist without these tools. We feel that, as three-dimensional rendering becomes more and more standard output for analysis packages, the gain in insight to dynamics could be quite large. There is, in fact, objective evidence that this might be true: Hardy, et. al. (1996) found that radiologists, viewing three-dimensional flouroscopic skull images were faster and more accurate in their diagnoses when viewing a rotating, stereoscopic rendering of the image. The least effective representation in the radiology experiment was the monoscopic, static representation (a simple perspective plot). The three-dimensional nature of the earth sciences lends itself nicely to three-dimensional rending of data. We’ve found that using immersive 3-D stereo rendering is a relatively easy way to fully exploit our visual ability by using motion parallax and, stereopsis for depth cues to view oceanographic and atmospheric data. Scientists at NOAA’s Pacific Marine Environmental Lab use the large-format ImmersaDesk, as well as desktop rendering to view this data.

The ImmersaDesk is a large format (82.5" screen), projection-based virtual reality device which uses stereo glasses and head/hand tracking to offer a type of virtual reality that is semi-immersive. These devices provide the graphical illusion of being in a three-dimensional space by displaying visual output in stereo and in a three-dimensional perspective according to head position, and by allowing navigation through the virtual data. A Silicon Graphics Onyx2 drives the animations on the ImmersaDesk, using several software packages, including Cave5D, Iris Explorer, VTK, and vGeo. This format is well-suited for the conference environment, and there will be an ImmersaDesk poster session at this meeting. Five or six viewers can comfortably arrange themselves around the screen and view an animation as the scientist explains the phenomenon and fields questions. Day-to-day use of stereo rendering is usual done on the desktop.

We find that Vis5D is a great desktop tool for 3-D stereo rendering for several reasons. The data files are easily translated from netCDF, it’s fairly easy to learn data manipulation tool, and now it does full stereo rendering on the desktop. We use a PC-based X-term server called Exceed because it has a 3-D add-on that allows OpenGL windows to be drawn on the PC taking advantage of hardware acceleration using the PC video card (fig. 3).

Fig. 3. Cave5D rendering of model output of the Gulf of Alaska region showing salinity contour, currents, and drifters.

Several worlds representing data from different research groups at NOAA's Pacific Marine Environmental Labs will be shown, each taking advantage of different aspects of stereo rendering:

Model output from a nested biological/physical model from NOAA’s Fisheries-Oceanography Coordinated Investigations (FOCI) program (Herman and Stabeno, 1996) shows the western edge of the continental shelf of Alaska, and a vector field showing currents at 100 meters depth (Fig. 3, above). A contour of the 32.85 ‰ salinity field is color-coded by temperature showing warm water advecting up from the southern boundary, and lagrangian float tracks are seen showing paths diverging around an eddy in the north-west region.

Model output from a shallow-water wave model from NOAA’s Modeling and Forecasting Group (Titov and Synolakis, 1997) shows the 1993 Hokkaido-Nansei-Oki tsunami approaching the coastal town of Aonae. Using the Method of Splitting Tsunami (MOST) model a tsunami’s evolution can be simulated, including generation by an earthquake, transoceanic propagation, and inundation of dry land. Runup, current velocities, and overland flow can be predicted in an effort to develop tsunami forecasting and hazard mitigation tools. The 3-D rendering includes high-resolution topography overlaid with an aerial photograph, and the surface of the model color-coded by wave speed (fig. 4).

Fig. 4. Tsunami model output showing wave approaching peninsula.

Climate research by NOAA’s Coastal and Arctic Research division includes work on the Pacific Decadal Oscillation (PDO) – an index based on the North Pacific sea-surface temperature fluctuations on timescales of 10 to 40 years (Overland et al., 1999). Figure 5 shows the first EOF for the PDO during a major cool phase. This web-based 3-D animation was created using the Virtual Reality Modeling Language (VRML) and the Visualization Toolkit (VTK).

Fig. 5. PDO showing cool phase as a 3-D interactive VRML object.

Stereo 3-D rendering of oceanic and atmospheric data is the culmination of a natural evolution in scientific visualization from simple 2-D cartesian plots and early 3-D perspective drawings. Fully stereographic, interactive 3-D rendering takes advantage of the minds ability to grasp complex environments to its full extent. The 3-D nature of atmospheric and oceanic data lend themselves to viewing in full 3-D, and the desktop and ImmersaDesk techniques used today give the scientist a powerful tool in the quest to better understand the world around us.

This publication was supported by the Joint Institute for the Study of the Atmosphere and Ocean (JISAO) under NOAA Cooperative Agreement #NA67RJ0155, Contribution #798.

Hardy, J.E., Dodds, S.R., and A.D. Roberts, 1996:

Hermann, A. J. and P. J. Stabeno, 1996: An eddy

Overland, J. E., J. M. Adams, and N. A. Bond, 1999:

Servos, P., 2000: Distance estimation in the visual

Titov, T.V., C.E. Synolakis, 1997: Extreme