Figure 6.1 Model of proposed museum, at Lake Tahoe, California. Landform, vegetation, water, and people (in the boat) make up this model. Modeled and rendered in PolyTrims software. Courtesy of John Danahy and the Centre for Landscape Research.
6.1.1 Structures, Vegetation, and Water on Landform
6.1.2 People and Other Animals
6.2.1 Accuracy, Precision, and Conversion
6.2.2 Digital Modeling Project Management
6.3.1 Camera and Viewpoint
6.3.1.1 Plan View
6.3.1.2 Plan Oblique
6.3.1.3 Eye-level Perspective
6.3.1.4 Animation Cameras and Viewpoints6.3.2 Media
6.3.2.1 On-Screen
6.3.2.2 Printed Output
6.3.2.3 35 mm Slides or Photographs
6.3.2.4 World Wide Web
6.3.2.5 Animations
6.3.2.6 Video
6.3.2.7 QTVR
6.3.2.8 VRML
6.3.2.9 Immersion
6.4.1 USGS, GIS Agencies
6.4.2 Surveyors
6.4.3 Digitizing
Figure 6.6 A scaleless imaginary landscape, created in Bryce software.
Figure 6.7 A useful file folder hierarchy for digital project management.
Figure 6.9 Plan view of University Commons landscape model.
Figure 6.10 Aerial oblique view of University Commons landscape model.
Figure 6.11 Eye-level perspective view of University Commons landscape model.
Figure 6.19 Digital Ortho Photo Quad sheet image (DOQ)
Figure 6.22 3D "point cloud" scans of a: terrain and b: trees. Courtesy of Cyrax Corp.
Previous chapters isolated and emphasized the individual elements of landscape - landform, vegetation, water, and atmosphere. But in real landscapes, and most modeled ones, these elements are combined and interact. Water in the landscape is inherently related to the landform, whether held in a basin, like a pond, or running in a channel formed by sloping banks. Trees and other plants in the natural landscape are located partly by preferences for sunlight, slope, aspect, soil type and moisture, and other plants. In a designed landscape, these elements are carefully composed to reinforce design ideas, visual effects, and choreographic sequences. Animals, including people, inhabit and add life to these landscapes.
... (Note: This Website contains abbreviated text. For
the complete text,
click here to order the Landscape Modeling
book.)
Getting landscape elements to sit properly on the landform - at the right elevation, without visible gaps or discontinuities where they touch, or grow out of the terrain - is an essential and often difficult task for landscape models. Some CAD modeling systems have snap settings that will let you position new elements so they touch the terrain object (or objects); in others you may have to invoke some batch operation to move the already placed elements to the surface. Oddly, and sadly, this is not the case in all systems, and not all landform modeling objects (tins, meshes, NURBs, e.g.) can be used for snap operations. Sometimes you may be able to use associated database technology to compute, or store, a spot elevation value for landscape objects and use that to set them on the terrain. For numerous and multiple objects, like the trees in a forest, some automated special-purpose system is usually the only feasible approach.
Figure 6.1 Model of proposed museum, at Lake Tahoe, California. Landform, vegetation, water, and people (in the boat) make up this model. Modeled and rendered in PolyTrims software. Courtesy of John Danahy and the Centre for Landscape Research.
Buildings offer a special case, since they touch over an area, typically rectangular, or polygonal, rather than at a point. On anything other than a flat site, this means the terrain will have a range of values around the perimeter of the building, at the foundation level. One solution taken by unimaginative builders is to simply clear a flat "pad" for every building, which forces some landform grading all around the building, cutting and filling to create the pad, and channeling to prevent water from flowing into it. A more sophisticated approach is to vary the levels of the building to fit more closely the levels of the terrain, and then use grading and paving to clarify the transitions, especially at building entrance points. If you have access to detailed building plans, they should show spot elevations or even contours all around the building. Rather than actually model the landform with a hole, or excavation, for the foundation of the building, you can just make the terrain continuous over the building footprint, and then set the building in on top of it; letting the rendering system hide the earth inside. (Of course, if you are building a detailed model of the interior spaces as well, as for a VRML model, this technique won't work, and you'll have to actually excavate the virtual terrain).
A simple technique for making a building, or other object, appear to sit perfectly in uneven terrain, is to model the building or other object with extra depth below ground level - in essence, make an extra deep basement on the building. Then when the building is set into the terrain (at a subterranean level), when rendered, the landform will appear to flow seamlessly all around the object. This same technique can be used with tree trunks, or telephone poles, or other embedded objects.
Figure 6.2 a. Trees and building are set below terrain level; b. When rendered, trees and building appear to be set perfectly in terrain.
... (Note: This Website contains abbreviated text. For
the complete text,
click here to order the Landscape Modeling
book.)
The four constituent elements of landscape - landform, vegetation, water and atmosphere - that were covered in the previous chapters are the essential elements to be sure, but landscapes are inhabited by animals too, who often play an important role in shaping the landscape. That is certainly true for people, and indeed often the reason we want to model, analyze, or represent landscapes is because of some human (or, in the case of science-fiction, perhaps non-human!) purpose. Modeling animals, including people, in the landscape is like other elements: in lieu of actual dynamic models of animal or human behavior, such as landscape ecologists and urban dynamics researchers might provide, the choice is between some form of polygonal solid models, or simple bit-mapped billboards. A number of sources exist for 3D models of human forms, which can be incorporated into your landscape model, at the cost of a relatively large number of polygons and or NURB surfaces; alternatively, photographic images can be artfully cut and pasted as needed, and texture-mapped in your scene. Adding even a few people can often add an element of interest and scale to your landscape models.
Figure 6.3 a) 3D solid models of people, modeled and
rendered in 3DStudioMax; b) "billboard" people in model (note rectangular shadow).
Modeled and rendered in Evans & Sutherland RapidSite. Courtesy of Evans & Sutherland.
Figure 6.4 Model of the campus of University of Toronto, with photographic texture-mapped buildings and central green space shown with texture-mapped trees and landform. Modeled and rendered in PolyTrims software. Courtesy of John Danahy, the Centre for Landscape Research.
Figure 6.5 Model of proposed residential development. Modeled and rendered in Evans & Sutherland RapidSite. Courtesy of Evans & Sutherland.
Some digital landscape modelers will have the freedom of painters, to invent landscapes without constraints, perhaps exploring scale-free, fantastic landforms and plants, subject only to aesthetic criteria. For these, digital tools are just another expressive medium. For others, such as landscape architects and other professionals, digital modeling tools are more than just expressive; they are controlled instruments of technical representation, and are most often used in the context of real-world, or design-world, constraints and limitations in which measurements count and some elements may be fixed. The broad continuum from freehand sketching and eyeball adjustment to mechanical drawing and precise alignment is familiar to designers, and most digital landscape modeling tools can be used in both modes.
Computers are notorious for their precision, but precision is not the same thing as accuracy. Precision simply refers to the detail in measurement - for numbers, the number of decimal places, for example, so that 3.14159262738 is a more precise value than 3.14. In landscape context, this may refer to the choice between 10 cm, 1 m, or 3 m contour lines, which give respectively less precise landform descriptions. 3 m contours are typically found on USGS topographic maps; good site surveyors can produce elevation measurements to within 10 cm (4 in.) with ease.
... (Note: This Website contains abbreviated text. For
the complete text,
click here to order the Landscape Modeling
book.)
Figure 6.6 A scaleless imaginary landscape, created
in Bryce software.
Modeling the landscape is just complicated enough, even with the best software and good data, that even the simplest rendering may take on qualities of a larger more complex project, requiring attention to process details and careful management. Even if the scope of the proposed modeling project is modest, management of files, folders, conversions, intermediate steps, alternative settings, and final products requires a systematic approach.
... (Note: This Website contains abbreviated text. For
the complete text,
click here to order the Landscape Modeling
book.)
Figure 6.7 A useful file folder hierarchy for digital project management.
The final goal of every digital landscape model is some sort of presentation, whether formal or casual, public or private, scripted or impromptu. Choosing the presentation format and media is as much of a design decision as anything else in the process. Many times the circumstances or conventions dictate the form, but often there are choices to be made. Recognizing the values and limitations of various output media, and different modes of presentation, is critical. A presentation is really a narrative, a story, and so it embodies a point of view, which is most often directly suggested by the choice of camera and literal points of view of, and from within, the landscape model.
The choice of viewpoint makes a great difference in the perception of a landscape. A traveler walking through deep woods, along a sinuous path, and finally reaching an open overlook forms multiple reinforcing impressions of the landscape. In digital rendering, you can control the same kinds of impressions by choice of camera angle, position, and other parameters. Three different viewpoints are in common use, and have very different effects: directly overhead (plan), aerial oblique, and eye-level perspective.
Figure 6.8 View inside the University Commons landscape model, looking towards the grove of Carpinus trees.
The first view, directly overhead, gives a plan view, or a "bird's eye" perspective. This is a view rarely enjoyed by humans, except in airplanes looking down, and so is the most detached and perhaps indifferent view. It is omniscient, seeing everything, but often objects are seen in unusual ways (trees almost as plan view symbols, buildings dominated by roofs, and so on). In this way, layout and spatial proximity relationships can be best visualized, but the visual aspects of the landscape portrayed may be lost. Ordinarily, a plan view is taken from a sufficiently high altitude that perspective effects of objects on the ground are minimal (although some haze, or blueing of more distant ground, can be seen in areas of high topographic relief). However, when the viewpoint is brought closer to the ground, the perspective can be dramatic, especially when looking straight down into tall vertical elements such as conifer trees, or telephone poles, etc. Direct overhead views have rather special illumination qualities as well, since the tops of all the landscape elements are always in full sun, and shadows tend often to be hidden underneath canopy.
Figure 6.9 Plan view of University Commons landscape model.
The plan oblique, is a type of projection drawing, that has a high angle of view where one vertical plane has more visual emphasis than the other. The advantage of the plan oblique is that it provides a dimensionally correct or true orthographic plan with the height of elements such as landform, vegetation, and built structures. Temporal or phenomenal conditions can be layered into the drawing through the csting of shadows, emphasizing the play of light amongst the plan elements. With one vertical plane receiving greater visual emphasis, some elements may be lost or hidden while others are more visible or foregrounded. The selection of the point of view, must be deliberate as to highlight a specific spatial relationships or temporal conditions. The oblique drawing or view is a constructed projection for the purposes of illustrating specific plan and sectional relationships, and is not intended to simulate human vision like the perspective projection.
Figure 6.10 Aerial oblique view of University Commons landscape model.
The most engaging viewpoint is the eye-level perspective, in which landscapes are seen as they would be from a typical (adult) eye-level view (of about 5'6" , or 1.7 m off the ground). In this view nearby elements obstruct further ones, and the view is limited to a cone of vision about 60 degrees wide, so that most of the landscape is not seen at any one time. This is the view most likely to resonate with viewers, inducing emotional feelings - of beauty, awe, mystery, suspicion, etc. This view is also the one most influenced by lighting, shadows, and atmospheric effects.
... (Note: This Website contains abbreviated text. For
the complete text,
click here to order the Landscape Modeling
book.)
Figure 6.11 Eye-level perspective view of University Commons landscape model.
The same basic views used for static visualization - plan, oblique, and eye-level - can be used alone or in combination, in dynamic animations as well. Many landscape scenes are large enough that an animated view (a walk-through or flyover) is the best way to illustrate all of the landscape. In addition, landscapes are often experienced in motion and these representations can be used to simulate that experience. In the case of a virtual environment, such as VRML, the viewer will control the camera, and so all you can do is define pre-set camera "viewpoints" which may be used, but you don't control the path through the model. In a pre-canned animation, where you control the view, the choice of camera parameters and path is crucial. You may choose to start with an aerial view, then zoom into eye level, or stay at eye level the whole time, or even use unconventional views, such as from underground, or sideways, in a cinematic way to emphasize perceptions of the landscape space. In animations, you control both the viewer's position, and also the camera's "look-at" direction and other parameters such as angle-of-view; controlling these is the key to making everything from simple documentary walkthroughs of a landscape model to impressionistic and even fantastic cinematic experiences.
Figure 6.12 Plan, Oblique, and Eye-level views with
their associated camera positions in relation to the model's geometry. a. Plan
b. Oblique c. Eyelevel
The final medium for the visualization of any landscape representation will also determine the modeling approaches and techniques. Choosing an output medium and format early on is essential to make sure that work is efficient and effective in preparing material for your final presentation.
Most digital modeling work is done interactively in front of a computer screen, normally limited to 72 - 100 dpi, and less than 2000 x 2000 pixels, in full (24-bit) color. Having two screens, side by side, can be a very great convenience, but requires special hardware setup. Work that is to be presented over the Internet, or the World Wide Web, is usually confined to screen size or smaller (800 x 600 is a common compromise), for transmission efficiency, and often smaller, low-resolution "thumbnail" images are made as a preview, so that the viewer can choose among several images without downloading each in its entirety.
Figure 6.13 The four most common screen display sizes: 320 x 240 , 640 x 480 (VGA), 800 x 600 (SVGA), 1024 x 768 (XGA) Not shown: SXGA (1280 x 1024)
For printed output, either letter-size (8.5" x 11") or tabloid size (11" x 17" or 12" x 18") is most common for color laser printing, at 600 - 2000 dpi. Larger sizes, up to 36" wide by 20' long, are produced on ink-jet plotters, typically at around 600 dpi. Printers and plotters always use CMYK, rather than RGB, printing techniques, and some color changes can occur when converting between them, so careful calibration as well as trial and error are often required for color matching. File formats for all of these vary, but Postscript format is a common standard. Postscript can encode linework as well as raster images, and can include text in specified fonts which can be scaled without loss of detail or introduction of jagged edges. Pure photographic images, including screen-snapshots, are recorded as raster files, and may be in TIFF, BMP, or JPEG format (or a wide range of others, but these are the most common). Low resolution photographs may look coarse, grainy, and jagged when enlarged to very large sizes, and so are best printed in the smaller formats.
If 35 mm slides are to be made, then the 3:2 aspect ratio (or smaller) should be respected. Slide-writing hardware and software can generally take a combination of file formats, like those listed above, and can handle a range of resolutions up to 4000 dpi or higher. Slide recorders typically employ RGB, rather than CMYK format. The newer APS (or IX240) system allows for three different standard sizes of exposures:
Figure 6.14 The four common photographic film formats: 35mm and the newer APS film sizes. Many digital slide writers have removable camera backs, and so can create output for many different formats.
The World Wide Web (WWW) - or more generally, the Internet - provides opportunities and constraints all its own for presentations. The distributed nature of the Web means that it is often the best choice for sharing models and other information with multiple parties, especially if they are geographically remote, but also even if they are in adjoining rooms. The set of graphical format standards that make the Web work (HTML, XML, and many others, constantly evolving) are of necessity well documented. Many commercial and open-source computer programs are available for creating and manipulating Web content (individual pages and complete sites). The format requirements are generally the same as those of on-screen presentations, but in many cases an additional effort is required to minimize graphics complexity in the interest of transmission time (time for the remote viewer to download the images). Many compression schemes, and special image formats for the web, both proprietary and open-source, are in constant evolution. This is an area of technology changing rapidly, so be sure to consult an up-to-date reference (on-line on the Web is often best!) about current Web publishing standards and techniques.
Animations require enormous file sizes, since anywhere from 15 - 30 frames are required per second. Consequently, they are usually produced at small image resolutions (like 320 x 240 pixels); if done at high resolution like 800 x 600, the resultant files can be enormous (hundreds or thousands of megabytes in size), making them hard to transfer, or play without very large disks (tens of gigabytes), very fast processors, and often special image compression and encoding hardware. The standard way to save and present an animation is in some form of video, analog or digital. The "QuickTime" format and QuickTime movie player are common standards for animation on the web and on Website ROM.
Figure 6.15 A standard application for playing animations
from Website -ROM or over the web: Apple's Quicktime Player.
Video format at the time of this writing is evolving from an analog low-resolution format to a predominantly digital, much higher-resolution format. If destined for analog television, for example on a videocassette for a VCR, the normal standard is NTSC/VHS, which has a limited resolution of around 512 x 380, so creating 640 x 480 is usually more than adequate (and still creates enormous files). Special video-editing software is required to match frame rates and format for NTSC video exactly. The additional technological requirements of producing television "broadcast-quality" imagery is beyond the scope of this book, and a suitable reference should be consulted.
Digital video is much more flexible, and can support multiple, different resolutions. The QuickTime format is commonly used to encode digital video for delivery over the web, or on DVD, Website -ROM, etc.
The new standards for HDTV include much higher resolution, and a 16:9 image proportion, better suited to landscape images. Video formats, compression techniques, and related technologies are rapidly evolving, and continually improving in resolution and quality.
One useful format for landscapes is the panoramic view, or animation (including the QuickTime Virtual Reality or QTVR format). This provides 360¡ viewing around a fixed viewpoint. These can be constructed by stitching together a series of photographs, taken in a 360¡ circle, using special-purpose software, and some modeling systems can directly export QTVR format. QTVR also allows for embedded "hyper-links" so that portions of one scene can be linked to another QTVR scene, much like rooms connected by doors.
...(Note: This Website contains abbreviated text. For
the complete text,
click here to order the Landscape Modeling
book.)
Figure 6.16 a. The QTVR format encodes a 360¡ image
in a single file; b. A QTVR viewer, such as Apple's QuickTime player, allows
the viewer to scroll completely around, and to zoom in and out to a limited
extent.
Other Virtual Reality (VR) display systems use a special head-mounted display to create a stereo view by projecting two synchronized images directly in front of each eye of the viewer, and use motion tracking hardware and software to change the viewpoint of the scene as the viewer's head moves, from side to side or up and down. This can give the illusion of being inside a virtual landscape. These systems must create real-time imagery, at up to about 30 frames per second, and so usually have very simple, and highly stylized contents. Very expensive, high-tech, military systems, such as flight training simulators, use similar technology but with much more detailed images.
For the most "realistic" presentation of landscape visualizations, some form of fully immersive projection or display is the best, albeit most expensive and complicated. In these environments, an image is displayed on a screen or series of screens completely encircling the viewer, 360¡ around, so that wherever the eye is looking, the peripheral vision, outside of our normal 60¡ cone of vision but very important to our perceptual sense of context and continuity, is engaged. This requires between 6 and 24 individual projectors, each projecting an overlapping, carefully synchronized image. Another variation is the so-called "CAVE" visualization environment, in which a cubic volume of space has images projected, usually from the rear, on at least four surfaces, and up to all six surfaces including floor and ceiling. Both of these systems require special expensive, hardware, software, and presentation spaces, as well as special model formats.
Figure 6.17 A 3-screen immersive panoramic display environment, requiring 3 projectors. Rear-projection avoids shadows of viewers on the screen.
Figure 6.18 3-screen immersive display environment at the Unversity of British Columbia, Landscape Visualization Laboratory. Courtesy of Stephen Sheppard.
When creating a purely imaginative landscape model, with no context or constraints, you may be able to start with a blank screen and create the landscape de novo, using the tools and techniques described in previous chapters. But when you are modeling a landscape in the real world, you are ordinarily dependent upon data from a number of sources to set the context and fill in the details. These can include contours and planting plans from landscape architects, site photographs and satellite images. Acquiring and managing the necessary data for any project is an essential first step in any digital modeling undertaking.
The United States Geological Survey (USGS) is the federal agency responsible for mapping the national landscape, and has a huge inventory of cartographic resources including the 1:25,000 topographic map series, color and black and white aerial photographs, thematic maps such as land use and transportation, and others. Since the early 1990s the USGS has been at the forefront of developing standards and means for distributing digital maps. Their web site offers a range of options for acquiring digital data, some of which can be downloaded for free, some of which can be ordered for a small cost. For basic contextual digital maps, especially relatively large areas, the USGS maps are an obvious first start. Many of their products are derived from the 1:25,000 mapping series and so are not particularly reliable at the detailed site scale, for which higher resolution and more precise data are necessary.
Figure 6.19 Digital Ortho Photo Quad sheet image (DOQ)
... (Note: This Website contains abbreviated text. For
the complete text,
click here to order the Landscape Modeling
book.)
Figure 6.20 LIDAR image, showing trees and vegetation as well as buildings and infrastructure. Courtesy of the Houston Advanced research Center (HARC).
For more detailed surveys of specific areas, a commercial contractor is usually required. Surveyors, engineers, aerial photographers, and remote sensing specialists all offer services to produce detailed digital surveys, often to within one centimeter precision if required. Increasingly, these are accomplished with the use of Global Positioning System (GPS) survey equipment, which depend upon a network of high-altitude satellites to achieve high-precision, three-dimensional readings of location. Coupled with laser-based survey equipment, and new techniques for "3D image capture," devices which actually measure three-dimensional coordinates of large objects such as buildings, or sites, digital models in a variety of formats can be produced.
... (Note: This Website contains abbreviated text. For
the complete text,
click here to order the Landscape Modeling
book.)
Sometimes, when sufficiently accurate paper maps or plans are available, the fastest and most direct method of getting digital data is to digitize or scan the paper maps. Large-format digitizing tablets can be used to trace lines - contours, roads and streams, parcels, etc. - directly into a 2D CAD system. In this process, human judgment is often required to set the resolution of digitizing (for example, when tracing a curved line into several short straight line segments, how many segments, and how long?) and to extract important features. It is important to remember when digitizing from paper maps to set the coordinate systems and units according to the standards required for the project, otherwise the result may be unusable with other information, such as a USGS site map. Especially with contour lines, it is important to set the required attributes, including elevation, during the digitizing process, or else the 3D information may be hard to extract.
... (Note: This Website contains abbreviated text. For
the complete text,
click here to order the Landscape Modeling
book.)
Figure 6.21 A field unit for performing automatic 3D digitizing surveys. This laser device captures "point clouds" in the field with up to 1" accuracy. Courtesy of Cyrax Corp.
Figure 6.22 3D "point cloud" scans of a: terrain and
b: trees. Courtesy of Cyrax Corp.
Making landscape models, like engaging in any other representational activity, is not always a value-free activity. When you are just expressing yourself, or making art, you have no particular obligation except to yourself in making your choices. When you are making a model that you or others will use as a basis for decision making, however, the responsibilities are greater.
In the design process, designers make models all the time, to test their ideas, to explore and evaluate alternatives, to form the basis for a next round of models and explorations. In this role, digital models and other representations have an ambivalent status: they need not be realistic, but they must be informative and appropriate. Choosing an angle of view, for example, is an implicit act of choosing between possible experiences of a landscape, for example between the omniscient but almost-never-experienced aerial perspective, or the ubiquitous eye-level view. Making these choices is a part of any design process. A power of digital models is that from a single model, multiple views can be rendered, at will (assuming that no view-dependent simplifications have been made, of course, such as rendering a grove of trees as an image, rather than 3D objects, in which case the view from within the grove would not be possible).
... (Note: This Website contains abbreviated text. For
the complete text,
click here to order the Landscape Modeling
book.)
Figure 6.23 Four slightly different images showing
visual simulation of cell-phone tower installation a. initial proposal; b. tower
painted brown; c. building at base painted brown; d. screened with evergreens.
These kinds of visualizations depend on a high degree of accuracy in geometry
and color, in order to help effectively evaluate alternatives. Courtesy of Curt
Westergard.
The techniques discussed in this text have been, and are, under constant development. In general there have been experiments and prototypes in research labs and academic settings that have developed one feature or another, sometimes in the specific context of landscape visualization as part of a landscape architecture or planning department, or sometimes geography or computer science, which have then been taken over and turned into features of commercial off-the-shelf software. Military and aerospace visualization systems for flight simulators and "virtual battlefields" provided some of the early development of computational techniques; more recently the motion picture special-effects industry has also played a substantial role in the development of techniques for creating digital environments for feature films, as has the entertainment/games industry in the development of backgrounds for video games. The annual SIGGRAPH meetings of the ACM (Association for Computing Machinery, Special Interest Group in Computer Graphics) have provided both an impetus and a showcase for new techniques in the field, and the proceedings of those meetings from the 1980s and 1990s are rich in detailed techniques for many aspects of landscape visualization.
Though there is a certain maturity to some of the techniques for modeling and rendering, there are still many developments to come and substantial challenges remaining. In general, the tension between realistic "acts-like" models and more superficial "looks-like" models remains a driving force. Many natural systems such as light, gravity, and wind which shape the natural environment are deceptively complex in their real behavior, especially in their dynamnics and in their interactions. Although techniques for simulating each of them have been developed in the context of computer modeling systems, these are all still approximations at best, and tend to be limited in their practical application, often requiring substantial computer resources such as memory, processing power, and time. Compromises for the sake of expediency or even tractability are found throughout all techniques for computer modeling and rendering, and landscape modeling because of the size and complexity of the subject matter is a prime example.
As of this writing there are still some notable shortcomings in commercial software, and a number of developments which are still to come that will make a big difference to digital landscape modelers.
"Snapping" to a terrain model is one: the simple desire to place objects such as trees, buildings, and people in a 3D model and have their elevation interactively set with respect to an underlying surface is an obvious one, still not met in most software systems. Tracing paths and complex spline curves on an underlying terrain model is another example unsupported by most software (although the VRML specification for navigation of a virtual model allows for "gravity" to constrain the viewer's path to the terrain surface). This is just a special example of the more general need to have constraint-based modeling systems in which attributes of objects are constrained by specified relationships to others. Other examples include intersection-checking to prevent objects from physically intersecting (such as the branches of a tree next to a building), or dependencies such as groundcover next to a curb; if the curb is moved, the groundcover should be resized appropriately. Such developments in CAD software are underway, both in research and in the commercial arena, and will make modeling landscapes, as well as other complex systems, more effective.
Similarly, good techniques for dealing with the problems of levels-of-detail (LOD) are imperative and still mostly missing. The forest, the trees and the leaves, the fields and the blades of grass, the crowds and the individual people are all essential constituents of the landscape. Present techniques require too much detailed hand-configuration and are too limited in their scope. More automatic, default-driven, generic techniques for managing LOD will make for far more usable landscape models.
As a last example, the whole field of modeling dynamics is still in its infancy. Particle systems are useful but still limited, and procedural techniques and scripting languages are awkward at best, especially when powerful. More robust dynamic modeling will depend partly on the development of the constraint-based systems described above, and on the convergence of the best features of numerous different special-purpose dynamic modeling languages and techniques, many of which are still non-spatial. The field of landscape ecology for example has given great insights into processes such as animal movement, soil erosion and vegetative succession, but good visual models of these processes have yet to be developed outside of special-purpose research labs.
Gravity-constrained, level-of-detail savvy, ecologically aware dynamic landscape models are still some way off, but the trajectory toward them is clear. And it is important to remember that we will never have "perfect" models, for there is no such thing. Rather, landscape modelers will tinker and tailor models to suit various purposes, for some of which physical-based realism is important, but for many of which it is not. Making representations will always be an art guided by science. The science has many pathways still unexplored, and so has the art of landscape modeling.
Managing digital projects requires considerable attention to detail, from file naming conventions to data conversion parameters. Landscape modeling further requires attention to the sequence of elements, usually starting with landform, then water, then buildings, infrastructure, vegetation, and often animals, including people; followed by decisions about atmospheric effects, lighting, and rendering and output options. Making these decisions is sometimes a matter of personal artistic expression, but often is bound up in a professional design process, with communication required between team members, and serious responsibilities to colleagues, clients, or the public. In these cases, ethical concerns about making representations must be integrated with the many technical considerations. The next developments in computer graphics and dynamic modeling will add to the technical repertoire, but only practice will perfect the art of landscape modeling.
(Note: This Website contains abbreviated text. For the
complete text,
click here to order the Landscape Modeling
book.)
Brinkman, Ron. The Art and Science of Digital Compositing. San Diego, CA: Academic Press, 1999.
Ching, Frank. Architectural Graphics. New York, New York: Van Nostrand Reinhold Company, 1975
Ebert, David S., Kenton Musgrave, Darwyn Peachey, Ken Perlin, Steven Worley. Texturing and Modeling, 2nd edition. San Diego, CA: Academic Press, 1998.
Flake, Gary W. The Computational Beauty of Nature: Computer Explorations of Fractals, Chaos, Complex Systems and Adaptation. Cambridge, MA: MIT Press, 1999.
Fleming, Bill. 3D Photorealism Toolkit. New York, NY: John Wiley and Sons, 1998.
Fleming, Bill. Advanced 3D Photorealism Techniques. New York, NY: John Wiley and Sons, 1999.
Sheppard, Stephen. Visual Simulation: A User's Guide for Architects, Engineers and Planners. New York, NY: Van Nostrand Reinhold, 1989.
Figure 6.24 Image of a synthetic landscape. Modeled and rendered in Animatek World Builder. Courtesy of Igor Borovikov.