The electronic pages of this journal have, for the past few years, contained numerous examples of virtual reality maps:  from planning for extra residential units in downtown Ann Arbor to tracking rugged voyages of Lewis and Clark.  Many maps cover large areas of terrain; they are global in scale.   Virtual reality, however, often is best executed in the small, at a local scale (Beier (lectures in Engineering 477, The University of Michigan), Crispen).  The virtual maps tend to become large in file size quickly, causing the maps not to load properly.  One problem is that software that easily creates virtual maps may not also optimize file size.  That persistent problem of the virtual modeler can be partially addressed by importing Virtual Reality Modeling Language (vrml) files exported from Geographic Information System software (GIS software) into a software package that executes polygon reduction of the vrml code.  Beyond such reduction, however, there remains the geographer's dilemma of scale transformation and a need to map both globally and locally.  A classical way to execute such transformation is to arrange the spatial information in layers of a nested hierarchy and use a well-defined transformation to move from one level of that hierarchy to another.  In the case of virtual reality maps (VR-maps), one puts individual maps in separate layers, separating maps before they become too heavy to run smoothly, or adequately, on current computing equipment (the creator, of course, needs to decide the target audience and the computing environment in which its members are likely to function).
     Earlier work on creating a 3D Atlas of Ann Arbor has produced thousands of separate virtual maps of the downtown (inventory of previous work; click on the "Archive" button).  Some are maps that show the current stock of buildings.  Others are maps that suggest future buildout scenarios based on concepts provided by leading architectural and construction experts.  The images below show screen capture of separate virtual reality models.  The image on this offers a simple solution to link the different hyperlinked images, using kiosks that transform the user, via the internet, from one hierarchical level to another.

     This linked image shows a screen capture of an associated virtual reality model of downtown Ann Arbor with buildings represented as simple boxes extruded from digitized building footprints.  Over several years, such images, of both actual and projected cityscape, served as one set of basic materials for planning efforts for city officials involved in various aspects of municipal planning.  Visualizing alternative outcomes associated with different zoning types and land use patterns was helpful in understanding how decisions might be reached.  Maps and decisions often go hand in hand.
     These maps and box-models were useful in a general way:  for looking at patterns of vacant land, possible infill sites, general skyline resulting from proposed infill, relation of topography to the built environment, and as a general inventory of the existing downtown scene by building height, zoning type, and special designation (such as historic district or floodplain designation or State-owned property, as are University of MIchigan buildings).  Many buildings, however, are not boxes, and all buildings have texture.  In the municipal arena of professional planning, supplementary architectural drawings and models are therefore critical (until the day arrives that virtual models are easy and inexpensive to make at all levels of detail).  Thus, one wishes for virtual maps that show more detail:  not only for planning purposes but for others as well.  An important other use centers on emergency response:  of firefighters (using virtual training of firefighters in their downtime in the firehouse, for example), of police officers, of environmental management personnel, and of a whole host of first responders  (link to the 3D Laboratory, Duderstadt Center, The University of Michigan).
     The introduction of photographic textures on files causes rapid enlargement of file size.  The map in this linked image shows a screen capture of one view of a virtual reality file in which textures have been applied to a number of buildings in the downtown area.  The file is about at the limit of what a typical desktop modern machine running 512 MB of RAM can run in a reasonable (not optimal) time (anywhere from 1 to 5 minutes to load the model depending on various local conditions such as amount of free hard drive space, fragmentation of the hard drive, internet connection speed, age of machine, and so forth).  In this file, there are textures only on selected buildings.  There are no street or sidewalk textures and there is no street "furniture" nor are there any other related accessories such as traffic lights, lane markers, or cars.
     The extra detail that comes at high cost in terms of file size is quite attractive for many applications.  Another way to introduce extra detail is to put links on the virtual buildings.  Thus, in the last linked virtual reality model, one can click on selected buildings to look inside:   not at an image of the interior but to internet links to the interior via a firm's own webpage and so forth.  Try 219 S. Main Street: look at the bottom of Cosmo Player to see building addresses as the mouse is moved across virtual buildings with active links to html pages.  The corresponding html page that comes up when the building is clicked on in the virtual reality model is linked here.   The background color of the html page is significant.  All buildings in the same physical, four-sided, block have the same color background.  Buildings in adjacent blocks (with adjacency defined as being "across the street from") have a different color.***   When environmental managers choose to insert, in the html pages, detail about environmental hazards known to be in particular buildings (for private municipal use only), then this model provides even more information valued by these managers, and other first responders, in times of disaster.  One advantage to this approach using html files linked to virtual buildings is the small file size of the linked web page; another is that no special knowledge, beyond simple html, is required to maintain the model.
     Finally, one may wish to introduce even more detail with street textures, benches, trees, and so forth.  All of this detail quickly increases file size.  Over the course of two years, student groups worked for three months to produce more detailed files.  In the Fall of 2003, in Professor Klaus-Peter Beier's Engineering 477 course (The University of Michigan), with Faculty Advisors Arlinghaus, Naud, and Nystuen, students Kwon, Lazzaro, Oppenheim, and Rosenblum produced a detailed model of the intersection of Main and Liberty Streets that Beier edited substantially (look both at the screen capture and at the virtual reality model).  In the Fall of 2004, again in Engineering 477, students Nolan, Ramesh, and Shani,  produced a detailed model of Main and Huron Streets that Beier edited substantially (look both at the screen capture and at the virtual reality model).  Both of these models were edited by the author in terms of cleaning of photographic images (removing the foreground where it remained), adjusting building shape and height to be correct and to be compatible with materials from the City of Ann Arbor and The University of Michigan, and adjusting building location so that the detailed models would fit correctly and seamlessly with the City of Ann Arbor files and with The University of Michigan files.  These adjustments involved the use of ArcView GIS software from ESRI, Adobe PhotoShop, Windows notepad as a vrml editor, and 3D Studio Max from Discreet.  The adjustments were executed in the Usability Lab of the Duderstadt Center of The University of Michigan using computing equipment configured with a double flat-screen monitor for wide-angle viewing, and also on a home desktop with a large single flat screen monitor.
     While each model has its own uses, an obvious desire is to be able to move from one to another without returning to a page and links:  to move directly from one level of the hierarchy to another and stay within the virtual world.  One way to execute such movement is to insert markers, "kiosks,"  in files to link from one one level to another.  Each virtual kiosk has an internet link on it:  the virtual kiosk serves as a center of information based on the virtual matter of the internet much as its real-world kiosk counterpart serves as a center of information based on printed matter.  Thus, the virtual kiosks, when loaded with correct information regarding Uniform Resource Locators (URLs), serve as well-defined transformations from one hierarchical level of virtual map to another.  The table below shows how to use this single file.  The rows represent layers in a hierarchy:  moving across the rows is a lateral move within the same hierarchy or virtual image.  The columns represent movement from one virtual image to another; in this case, using kiosks loaded with appropriate URLs to execute that movement.
 

Follow the path suggested in the table above to become familiar with navigation through the various layers of the model.  Use viewpoints or drive freely within layers.  One entry point to the virtual world offers access to a wide range of virtual files.  An important aspect of such transformation is thus that one can have many files together in a single file no heavier to view than the single heaviest layer of the hierarchy.  It would not have been possible, using merging techniques or some such equivalent strategy, to create the entire set as a single file.  When kiosks (or some conceptually equivalent mechanism for jumping between hierarchical levels) are employed, it is easy to imagine emergency managers training first responders to fly through a small file of the entire city with chunky buildings in order to learn the general city layout.  When each street corner or other strategic location in this general model is marked with a kiosk then with one click of the kiosk a detailed model pops up, while additional clicks on individual buildings in this detailed view reveal the interiors of buildings as html files.  This sort of arrangement has been created for the City of Ann Arbor (and given to them).  Creation of the model has taken several years and involved the use of a wide variety of resources, both human and digital.  Maintenance, however, can be achieved using a simple text editor to alter the underlying code in either the html files or in the vrml files (part of the code for one building is attached).  Thus, highly computer-literate City employees, armed with digital cameras and good computing resources can maintain the model and insert changes and additions to suit a variety of needs and interests:  the process is in place.  The arrangement of spatial information in a hierarchy is an efficient arrangement not only in the classical worlds of pen and paper but also in the contemporary world of digital models as well as in the practical implementation of such hierarchical arrangements in the real world.

Prof. Klaus-Peter Beier, Ph.D., Director of the 3D Laboratory of the Duderstadt Center of The University of Michigan has offered extraordinary insight and support on various phases of this project.  The author is indebted to his graciousness in sharing computing as well as intellectual resources.  She is also extremely appreciative of the assistance and wise advice from staff members in the 3D Laboratory of the Duderstadt Center, Scott Hamm, Steffan Heise, Brett Lyons, Eric Maslowski, and Lars Schumann, and also to Graduate Student Instructors in Engineering 477, Thana Chirapiwat, Jamie Cope, and Bonnie Congxi Bao.
 

The name "Kioskland" derives from a paper by that name by the author (1976, Department of Geography, The University of Michigan, unpublished).

**  Matthew Naud, M. S. and M. P. P.,  is Environmental Coordinator for the City of Ann Arbor and a principal member of the Emergency Management team.
John D. Nystuen, Ph.D. is Professor Emeritus of Geography and Urban Planning, The University of Michigan.

In the Fall of 2003, Taejung Kwon was a Ph.D. Candidate in the Taubman College of Architecture and Urban Planning, The University of Michigan, Adrien Lazzaro was a Master's degree Candidate in the School of Information, The University of Michigan, and Paul Oppenheim and Aaron Rosenbaum were undergraduates (Seniors) majoring in computer science.

In the Fall of 2004, Nikolai Nolan was an undergraduate majoring in General Studies, Rasika Ramesh was a beginning graduate student in the School of Information, and Itzhak Shani was an undergraduate (Senior) majoring in Engineering.

***The Four Color Theorem is used in color assignment; thus, in a grid street pattern, two colors are both necessary and sufficient to distinguish all blocks (animated Four Color map link).  No more than four colors are ever required (viewing the block geometry as part of a plane or a sphere).  Thus, background color of the linked html page also gives subtle information as to location.