Traces of Some VRML Experiments

We lost a number of the VRML files off of this server in 1999, but the remaining links should now work. -Matt (Sept '06)

The following are samples of VRML 2.0 models of the hydrography of Honolulu built from USGS DEM and DLG data and of contours rendered in VRML, and intended to suggest how VRML might be used to display and explore these data.

Several cartographic design issues are involved in building VRML representations from USGS digital cartographic data. The effects of generalization and vertical exaggeration can be seen in the VRML 2.0 models of the 1:24,000 Honolulu Quadrangle's hydrography.

Contours

This VRML 2.0 model of Oahu's 1000 foot contours gives a sense of the utility of VRML to support virtual examination of (skeletal) 3D geographic features. The points were originally digitized from USGS 1:24,000 quadrangles by Tom Giambelluca more than twenty years ago. (Odd how data can stick around and resurface.)

VRML 2.0 Oahu countours

Hydrography

These surviving samples show DLG hydrology for the Honolulu Quadrangle at only one level of vertical exageration (7X) and four levels of generalization (selection and simplification).

All lines, All points, 7X vert.

All lines. End points. 7X vert.

Stream lines, All points. 7X vert.

Stream lines. End Points. 7X vert. ,

Comments

Generalization impacts the amount of detail in a visualization as well as the volume of data which will be moved across the net and manipulated in your browser. Common cartographic data sets overwhelm the capabilities of many PC browsers. In the models, generalization is modified by selection, of either all DLG hydrology lines only the ones that are stream features, and by simplification, here keeping either all points along each reach or only the ends of each reach. Notice the impact of these generalization choices on loading, rendering and manipulating these models.

Vertical exaggeration refers to using different scales for horizontal and vertical components of map locations to exaggerate the vertical component. It has generally seemed that with printed maps, the vertical component has to be exaggerated for people to notice and "believe" the portrayed relief. It is not clear how much vertical exaggeration is needed with emersive VRML worlds. The contour models use 7X vertical exaggeration. Notice the effects of differing vertical exaggeration in the hydrographic models. The models without exaggeration seem "flat" but it is not clear whether and how much exaggeration is appropriate. Perhaps people have unreasonable expectations about local relief and would benefit from being shown "un-exaggerated" landscapes.

Clearly, reformatting data collected from 1:24,000 scale source maps for VRML browser presentation at unknown scales is fraught with problems. Astute observers will note that some streams in these models seem to run up and over rises in the terrain. These are artifacts of errors in the DEM data and of the interpolation method used to attach DEM elevations to DLG planimetry. Visiting Oahu will assure you that our streams really do drain to the ocean.

The impact of level of generalization is felt even more when we consider trying to visualize all of Oahu. The highly generalized (using straight lines between topological nodes in the stream network) hydrography of Oahu takes almost 2 MB of VRML 2.0 and can be manipulated on our lab machines, albeit slowly. The "full detail model" . taken from the USGS 1:24,000 Hydrographic DLGs takes about 8MB *WARNING* . We can download it quickly enough, but browsers (ca 1999) really churned and could not realistically be used to display or manipulate it. How are things now? Who still does VRML?

Please, let me know of any problems with these models and particular browsers. Thanks. -matt