Visualisation Techniques

Introduction

Use of Visualisation

The technology of Visualisation

Visualisation and Geographic Information Systems

Perceived realism and the validity of visualised images

Photo-realism, schematic images and validity

Realism

Schematic images

Validity

Example of an experiment in validity of simulations

Rendering

Present uses of technology

Degree of realism of visual displays

Relation of detail to distance - variable object representation

Geometric transformations

Earth curvature

Perspective projections

Specific details in rendering

Lighting models and illumination

Ray tracing

Colour and texture

Objects

Atmospheric effects

Variable object representation

Tree and forest simulation

Image creation techniques

Tree simulation

Block draping

2.5 and 3D tree patterns

The ramification matrix method - an example of how to grow a tree

Examples of projects using tree or forest simulation

The Sulphur Pass Project

SmartForest

References

Introduction

Visualisation provides additional insights to results which would otherwise be displayed as text or numbers (Loh et al, 1992). It is a form of communication which is universal, and which has the ability to form an abstraction of the real world into a graphical representation which is comprehensible to a wide range of people. Increasingly, computer Visualisation is used to communicate the implication of natural and management changes in biological systems in national parks and forests (Orland, 19??). However, it can be said that society is becoming dependent on information presented in three-dimensional visual format (Faust, 1995), virtual reality is no longer a word only applicable to computer games.

The emerging role of environmental managers is to mediate between the environment and its many users. That role has four important components (Orland, 1992b):

1. to identify and interpret the complex interactions of environmental systems;

2. to communicate their implications to environmental scientists, other managers, policy makers, decision makers, and the general public;

3. to enable the testing and evaluation of alternative scenarios by experts as well as non-experts; and

4. to implement the resource plans resulting from this wide range of inputs.

Visualisation techniques are a useful aid in the second and third components above.

Until recently, Visualisation has been an added feature, not an essential part, of the decision making process. Too often it has been regarded as decorative in function rather than substantive. Visualisations have been developed after the real work is completed, and often only to "sell" the resulting proposals. Presently, the use of GIS in Visualisation is limited to mere pictures of landscape. The aim of researchers and managers should be to ensure that the Visualisations are tied to underlying databases, and that the links between Visualisations and data are verifiable, reliable, and accurate (Imaging Systems Laboratory, 1995).

Use of Visualisation

For natural resource managers to plan for a more healthy environment, and to elicit public and political support for such plans, two needs have been identified by Orland (1994):

1. To predict the responses of public groups to changes in the environment, for some of which the visual impact may be the dominant indicator, and to plan to minimise any negative impacts;

2. Once a proposal is developed, to communicate the effects of proposed changes to other agencies and public review groups to facilitate decision-making.

It is possible to visualise the landscape impacts of changes in the building code or engineering standards as well as in natural resource management. Models may be used that predict changes in landscape preferences, economic behaviour, ecological succession, wave erosion or even rises in sea level. All of these and other models have implications for the assessment of regional visual landscapes and as as result they may be used to produce new landscapes which may be visualised using the modelling process described above (Mayall and Hall, 1994). Visual impact assessments (VIA) are beginning to depend on Visualisations and visibility analysis.

The technology of Visualisation

Traditional tools for visual communication of resource issues have included simple graphic devices such as maps, line charts, sketches, photographs, and renderings. The new tools include coloured computer maps, 3-d models, animations, and interactive virtual reality environments used to explore design ideas (Imaging Systems Laboratory, 1995).

Photomontage techniques use a combination of photographs, renderings and artistic license. Kennie and McLaren (1988) define photomontage as "a physical or image composite of photographs of the existing landscape with a registered computer generated image of the proposed design objects (s)".

Computer graphics is incapable of substantiating all the details recognisable in the immediate surroundings. The visual simulation program must therefore be considered as a support system to help a designer to outline the basic features of the environment (Pukkala and Kellomaki, 1988), it will never be able to represent the landscape as it is in reality. The closest images will be produced by reproductions of the phenomena based on the laws of physics (Kennie and McLaren, 1988).

The programs needed to manage and utilise natural resource data include: database management systems (DBMS); geographic information systems (GIS); simulation models; expert systems; report generation systems; and other relevant application programs (Loh et al, 1992). Such programs are continually being designed, combined and upgraded.

Visualisation and Geographic Information Systems

GIS-based Visualisation goes beyond the simple ability to discuss anticipated outcomes via traditional graphic tools. It offers the opportunity to visualise relationships across time and space, and to explore more comprehensive ranges of possibility (Orland, 1994). More flexible Visualisation methods would enable users to select their own viewpoints and be free of weather, seasonal and other restrictions. Currently the emphasis for providing that flexibility is on GIS. However, GIS-driven image creation does not currently provide a means of integrating detailed, small-scale Visualisation with large-scale regional views. The coarse grain of data sources such as digital elevation models and remote sensed imagery make GIS most appropriate for large-scale, synoptic, views of resource issues (Orland, 19??). It is as yet not so useful for small scale, detailed Visualisations.

Currently, Visualisation in GIS systems is generally limited to two-dimensional viewing either of individual GIS layers or of the results of GIS analyses. Pseudocolour representations of data variables are often used to identify different regions within a raster GIS whereas line styles, widths, colour and symbolisation are used for data representation for vector GIS. True colour combinations of multiple raster GIS layers can also be accomplished for viewing the spatial relationships between the layers (Faust, 1995).

Several GIS systems currently have the capability of creating three-dimensional perspective images by using elevation data for geographic areas overlaid with a GIS variables such as land cover or land use. In most cases this imagery is simply used for show, and there is little or no analytical work that can be accomplished with the perspective image. Measurements of size and shape are not valid in a perspective image without significant ancillary information being present along with the perspective image. In addition, these perspective image are generally well received as showing the relationships of the GIS data to the natural terrain; these factors have limited the usefulness of the perspective images to `show and tell' type applications (Faust, 1995).

Bishop and Hull (1991) have the following to say about GIS and Visualisation in the future:

"It is an attractive thought that, at some future time, we will have sufficient accumulated research to assess probable changes in visual resources entirely from a GIS without further recourse to psychophysics or video-imaging. This point is unlikely to occur, however, because even if the process of modelling from mapped /mappable information is shown to be valid and reliable, the landscape experience is dependent upon purposes and values and therefore varies from place-to-place and time-to-time. Recalibration of such models will therefore always be required."

Perceived realism and the validity of visualised images

A question to address in producing photo-realistic simulations is: how good is good enough? A good enough image is one that has a high degree of perceived realism, conveys maximum quality, contains enough data, yet is efficient in terms of equipment costs, storage and management (Perkins, 1992). Perceived realism may not necessarily vary directly with image quality. Image quality may be very high in technical terms, while perceived realism is not (Perkins, 1992), although image quality will affect perceived realism, so will the content of the image, the viewpoint of the image and the receptivity of the viewer. Daniel (1992) suggests that image quality is sufficient when additional inputs to improve image quality do not result in an increase in image validity, or indeed, in perceived realism.

Some basic understanding of the factors that influence the perception of image quality is therefore needed to increase the `fit' between computer-generated images and real world conditions (Perkins, 1992). Public perception studies have been conducted with images and they indicate that simulations are achieving a high degree of validity (Orland, 19??).

Photo-realism, schematic images and validity

Graphics in the realm of illustration have two characteristics: precision (or detail) and realism. Precision is necessary because small variations (in terrain, for example) can have large effects on design. What might be noise in the domain of inference becomes a high priority in golf course design (Buttenfield and Ganter, 1990).

Consideration of the validity of substituting computer-generated simulations for photographs only makes sense if it is accepted that photographs are themselves an adequate surrogate for direct experience of the landscape in question (Bishop and Leahy, 1989). This question is discussed in the review of landscape preference and perception.

Realism

The importance of realism is noted by Buttenfield and Ganter (1990):

"Realism provides the visual context within which constraints and externalities may be considered."

If the intent is to convey the potential impact of proposed management actions, with the goal of informing public review and approval processes, then more realism and detail may be demanded in the Visualisation. Techniques include photo-realistic simulations, and hand-crafted graphic renderings and models (Orland, 1994).

Computer image editing methods can be more realistic than hand rendering and less expensive than photo-retouching and they do much to convey a realistic visual experience of the planned new environment. The realism and transparency of the medium to a non-expert viewer is high, but the accuracy and validity less easy to defend (Orland and Daniel, 1995). In a study by Oh (1994) only image processing was successful in separating the visual attractiveness of one landscape from another in simulations; the other methods used were wire frame, surface modelling and combined surface modelling and scanned photographic images.

In the search for true realism, a compromise must be found between reality and costs (Zewe and Koglin, 1995).

Schematic images

Many studies have tended to focus on the realism or accuracy of simulation. But simulation cannot reproduce that reality completely. Rather, it selects critical aspects of that reality for the particular purpose at hand (Oh, 1994). In practice the simplification or abstraction of detail is directly related to the savings of effort, time and costs of simulation.

In many ways, schematic images, those which do not attempt photo-realism, are as useful as the more realistic images. They do not require the technician complications of photo-realistic visual simulation and do not have the theoretical problems associated with defining realism (Ervin, 1993). These views are approximations - more realistic than just plan views, but more schematic than photorealistic renderings or presentation drawings (Ervin, 1993).

In schematic renderings the images are rough; colours are shaded, the ground is broken up into half-acre coloured triangles, and all trees and buildings are alike and rather diagrammatic. There is no provision for subtleties of texture or curved surfaces. These representational conventions are no more limiting than any others commonly encountered in design, and they are no more difficult to understand (Ervin, 1993).

For some purposes, the Visualisation media may intentionally be highly abstract and intended to convey information most effectively to experts within the same discipline. In this category are coloured maps, where the need is primarily to represent single issues such as the extent or locations of an impact, or statistical displays of simulation models where the purpose of the Visualisation is to prompt further modification of resource models and hence to understand better the underlying physical system (Orland, 1994).

Validity

There are two questions which need to be answered when the validity of computer simulations is discussed (Daniel, 1992):

1) What is the validity of data Visualisation systems?

2) What level of data Visualisation is sufficient for environmental planning and management?

The primary concern for data Visualisation intended for decision support in environmental management is to achieve accurate and verifiable representations of existing and projected environmental conditions. The validity and sufficiency of a given data Visualisation system, therefore, depends in part on the purposes for which it is intended (Daniel, 1992). The Visualisations must also be accurate, because their power to convince is high, and any misrepresentations must be avoided.

The resolution and fidelity of environmental simulations seems limited only by the computer resources and peripheral devices allocated to the task. However, the validity of data Visualisation systems is not necessarily related to judgements of realism, believability, or other such qualities. Neither is there any necessary dependence upon resolution, colour fidelity or other technical criteria (Daniel, 1992).

The answers to the questions provided by Daniel (1992) are that the Visualisations are valid to the extent that responses to environmental representations correlate with appropriate responses made directly to the environments represented. Data Visualisations are sufficient to the extent that adding detail, higher resolution, colour fidelity, animation or other features does not improve the match between representation based and direct responses.

Example of an experiment in validity of simulations

Bishop and Leahy (1989) conducted an experiment to determine under what conditions computer simulations may be used reliably. They decided to select only those factors for which some objective measure could reasonably be applied in future simulations.

Prospect/depth, refuge, and complexity could conceivably be established from photographs of the type used by Shafer and Brush (1977) for estimating preference. Extent of background clearly relates to prospect, refuge, and perimeter lengths to complexity. Landform is largely a reflection of measurable relief. Water can be measured by area, flow and drop. Cultural modifications can be clearly outlined - although the effect of different types of modification is highly subjective (Bishop and Leahy, 1989).

In the scheme, colour is judged on the subjective basis of vividness (high score) or subtlety (low score) while vegetation ranges from the harmonious variation (high score) to little variation (low score) (Bishop and Leahy, 1989).

In looking at simulations, therefore, it is the degree to which the simulation can capture the subtlety of colour which may prove a decisive variable. Measures of both subtlety and veracity could be derived objectively by using image analysis colour filtration, and signal processing techniques (Bishop and Leahy, 1989).

The only attribute to which an objective measure was eventually applied was complexity; the size of the run-length encoded file containing each digitised image was used. Regression analysis based on selected variables indicates the importance of maintaining both the greenness and subtle colour variations of the original slides, the inadvisability of basing simulations on scenes with significant background or major relief, and the advantage of having a recognisable visual focus even if this is itself simulated (Bishop and Leahy, 1989).

Rendering

Present uses of technology

It is possible to simulate landscapes as they might look as a result of environmental impacts such as building or forestry developments, or to test environmental models (Fisher et al, 1993). Terrain modelling is possible in a range of different guises from the simple vector (or wire-frame) surfaces displayed as rectangular or triangular matrices, to sophisticated colour perspectives generated from complex hidden-line and hill-shading algorithms (Moore, 1990).

Commercially available, generic toolkits, such as PV Wave, AVS and Explorer, can be used for the production of visual displays of different types of data. This generation of dedicated graphics software epitomise the Visualisation revolution in computer graphics (Fisher et al, 1993), however, they require powerful hardware, such as Sun or Silicon Graphics workstations. Less powerful PCs and Apple Macintoshes have an increasing potential for visual display and can be used as effectively in developing new Visualisation strategies.

Degree of realism of visual displays

Degree of realism is dependent on several factors including nature of the application, the objective of the Visualisation, the capabilities of the Visualisation, capabilities of available software/hardware package and the amount of detail required and/or available (Kennie and McLaren, 1988).

The majority of computer graphics techniques have been developed to address the Visualisation of objects that have their geometry defined using a mesh of planar surfaces such as triangles. This can lead to jagged edges on curves, this is often solved using anti-aliasing techniques (Kennie and McLaren, 1988). The anti-aliasing for montage methods should be different from other usual ones because the background scene includes pixels which change pixel by pixel, and image overlay operation is performed several times (Nakamae et al, 1986). The aliasing problem also occurs when foregrounds are superimposed onto the computer-generated images.

Special problems are posed by the realistic rendering of landscape because of the amount of detail required. Fractal methods have been proposed to allow database amplification, which is the generation of controlled random detail from a fairly sparse description. An alternative approach is to use texture mapping methods on a few simple primitives. The texturing is defined procedurally, which may be expanded without loss of high frequency detail, or shrunk without the occurrence of aliasing artifacts (Miller, 1986). Fractal subdivision methods are slow and generate defects due to what is known as the `creasing problem'. The texture map methods, on the other hand, display visible discontinuities in texture gradient where two surfaces intersect.

Relation of detail to distance - variable object representation

An accurate representation of each detail in the scenery is not possible and also not necessary (Gross, 1991). Using only satellite data, observer positions close to the ground or narrow fields of views result in large quadrilaterals in the foreground due to perspective foreshortening. Therefore, resolution should be high towards the foreground or local zone, but the resolution should decrease towards the background (Graf et al, 1994). Far away from the observer it is sufficient to present the scenery at a low level of detail using hierarchical data sets in order to limit the corresponding data. Photorealistic rendering is provided with great success by texture mapping of remote sensing data on the digital terrain model. This method is useful for great distances from the observer (Gross, 1991).

Geometric transformations

Earth curvature

Simulation methods must take into account the curvature of the earth and refraction of light; there is a formula using the true elevation of the view point from which reductions to all elevations can be calculated to allow for curvature and the refraction of light (Aylward and Turnbull, 1977).

Since all topographic data is based on an assumed flat reference plane, usually height above sea level, it is necessary to reduce all elevations from the specified view point to take into account the curvature of the earth and refraction of light, if the survey area is greater than 2 miles (3.2 km). Since the distance from the view point to all elevations can be calculated, a simple formula can be used to modify each elevation value (Aylward and Turnbull, 1977).

Perspective projections

3D terrain is normally mapped into 2D space by a perspective projection. For applications where the geometric fidelity of the rendered scene is of vital importance such as a photomontage for a Visual Impact Assessment (VIA), it is necessary to incorporate both earth curvature and atmospheric refraction corrections into the viewing model (Kennie and McLaren, 1988).

The geometric transformation described by Kennie and McLaren (1988) for landscape view generation is a perspective projection similar to that of a photographic system and of the human visual system. It uses perspective fore-shortening. No shadows are calculated since the model is illuminated and shaded already by the overlay image which is lighted by the sun.

In order to use satellite images and aerial photographs for perspective viewing, they must refer to the same geometric reference system as the digital elevation or terrain model (DEM/DTM). Although a vertical aerial photograph provides a map-like view of the earth's surface, it differs fundamentally in geometric terms from a map (Graf et al, 1994). There are many geometric distortions of satellite image data with severe effects: the satellite (platform) monitor and earth rotation, the imaging geometry of the sensor and the terrain variations in the scene.

Specific details in rendering

There are a number of details to take account of when rendering a landscape Visualisation. As well as the model and associated landscape features, a number of other parameters need to be defined (Kennie and McLaren, 1988).

1. viewing position and direction of view
2. lighting model to describe illumination conditions
3. `conditional modifiers' e.g. wet, snowy
4. `environmental modifiers' e.g. atmospheric conditions such a haze
5. sky and cloud model representing the prevailing conditions.

Some of the specific attributes of the Visualisations are discussed below; these include colour and texture, object, lighting, ray tracing and atmospheric effects.

Lighting models and illumination

The appearance of a surface is dependent on several factors: type of light; condition of atmosphere; surface colour; reflectance and texture; position and orientation of surface relative to the light source; other surfaces; and the viewer (Kennie and McLaren, 1988). Normal lighting models are simplified by assuming only a single parallel light source located at infinity (the sun). There are two types of light source - direct and ambient (reflected).

A number of models have been published for displaying natural scenes, terrains, flames, eruptions, glasses, and trees, such as the fractal model, the procedure model, the growth model, and the semi-transparent mapping model (Nakamae and Tadamura, 1995). Mirror images and transparent effects have been realised using refraction techniques.

Radiosity

Radiosity is defined as the simultaneous global solution for the intensity of light leaving each surface by constructing and solving a set of linear equations describing the transfer of diffuse light energy between all surfaces (Kennie and McLaren, 1988).

Shadows

Surfaces visible from both viewpoint and light source are not in shadow, those visible from viewpoint but not light source are in shadow (Kennie and McLaren, 1988).

Depth cueing

Depth cueing is used to increase the 3D interpretability to match the perceived computer generated image to human `natural' visual cue methods. Intensity depth cueing is concerned with putting the relative lighting in the sky (Kennie and McLaren, 1988).

Ray tracing

This approach (view dependent) involves tracing a ray from the viewpoint through a pixel and into the model where its interaction with objects in analysed. Each collision with an object produces 3 rays - diffusely reflected light, specularly reflected light and transmitted (refracted) light. Normally the last two continue to be traced (Kennie and McLaren, 1988).

Ray tracing is a process which considers each pixel in the image, a ray is defined as a line joining the viewpoint to the pixel. Ray tracing is a recursive algorithm, dealing with the rays as they branch through object interaction (Evans, 1993).

Colour and texture

Objects in computer graphics can be visualised using various attributes such as smoothed curved surfaces, 2D texture, half-transparency, specular reflection effects and bumped texture. There are also strong rendering techniques including z-buffer and scan-line algorithms and anti-aliasing algorithms (Nakamae and Tadamura, 1995).

For non-exact modelling of terrain surface, data amplification primitives such as fractals which automatically densify the model are used. Texture mapping shades a surface mathematically, bump mapping stores surface normal perturbations in the texture map, achieving roughness without explicitly modelling the geometry. The latter is good for terrain Visualisations. Stochastic fractal models, a class of irregular shapes that are defined according to the laws of probability, can accurately model natural terrain (Kennie and McLaren, 1988).

Objects

In traditional manual photo-montage, objects are painted over a background image. This technique has several drawbacks: the result depends much on the artist's skill; the position of an object can only be estimated; and the geometric accuracy therefore is limited (Graf et al, 1994). The montage system developed by Nakamae et al (1986) included some unique characteristics along with anti-aliasing processing enabling superior simulation of natural shading and shadows. Atmospheric moisture effects (fogginess) was added to simulate various weather conditions. The system allows the objects to be constructed, rendered and placed in the scenery without relying solely on an artist.

Nakamae and Tadamura (1995) looked at creating photorealistic images based on optics, taking into account inter-reflection between illuminated objects, skylight with spectral effects and atmospheric scattering and absorption. They also looked at complex objects such as fur and trees. In order to reduce the cost for creating such complex phenomena, research on effective polygonal surface techniques, interactive modelling tools, and powerful graphics engines for rendering has increased (Nakamae and Tadamura, 1995).

Atmospheric effects

Due to scattering effects by water molecules, dust and pollution, objects appear to lose their colour and intensity with increasing distance. The reduction rate of colour and tone depends on the various factors such as seasons, weather conditions and time. The colour and tone become grey as the distance becomes large. The hazing affect created due to the atmospheric moisture content leads objects to undergo exponential decay of contrast with respect to distance from the viewpoint (Kennie and McLaren, 1988; Nakamae et al, 1986; Graf et al, 1994).

The human eye relies on colour fading for correct depth perception. Therefore in order to enhance realism and to maintain the estimation of distance, atmospheric effects must be modelled (Graf et al, 1994). The fog effect increases the sense of perspective in the montages. The shading and shadows of the computer generated images help a montage match to the background scene (Nakamae et al, 1986).

Variable object representation

Any image that is attempting to simulate realism in a spatially deep scene should include a mechanism to vary the perceived level of detail of an object based on its distance from the viewer (Kennie and McLaren, 1988).

Tree and forest simulation

It is important that any landscape Visualisation system allows realistic tree and forest simulation. The symbols used can be very effective in improving the visual realism of a rendering. Often, logged areas which are visible before trees are drawn will become invisible once tree symbols are added; the occluding effect of trees is especially noticeable in gently sloping landscapes (Smart et al, 1991).

At a basic level, trees have been reproduced as simple geometric forms, circles or spheres with thin cylinders as trunks, or as pattern of branches rotated through 360 degrees to give wire-line form skeletons. When images are drawn on colour raster devices, more impressive effects are possible: shapes can be flood-filled in subtle green hues, textured patterns can be applied and give a naturalistic, organic quality to the vegetation (Moore, 1990).

Image creation techniques

To create an image with many trees for environmental assessment the following conditions should be satisfied (Nakamae and Tadamura, 1995): easy construction of a tree database in view of the variety of species of trees; easy display of a cluster of trees in view of the necessity for making a variety of images; creating still images from arbitrary viewpoints as well as animation.

The branching patterns of higher plants are evident everywhere and are relatively easy to formalise, providing excellent examples for study (Aono and Kunii, 1984). There are two basic categories of tree branching patterns - dichotomous and monopodial, most trees have the latter. A branch divides in two at the growth point but one follows the direction of the main axis and the other goes in a different direction to form a lateral branch.

Some examples of tree and forest simulation are discussed below, including block draping (a method often used by the Forestry Commission) and 2.5D rotating trees. The ramification matrix method, a method of growing trees in the computer rather than using simulated or scanned fully grown trees, is also discussed.

Tree simulation

In the method of tree simulation described by Nakamae and Tadamura (1995), two textures digitised from two photographs taken from the right side and from above the tree were mapped onto a set of transparent planes. For shading and shadowing, the shape of a tree is approximated by a transparent polyhedron surrounding it. Shadows cast onto trees, as well as tree shadows cast onto objects, look natural whether the trees is lush or has sparse leaves. Shadowing is available for the following four cases: shadows cast by trees onto their own trunks; shadows cast by trees onto objects; shadows cast by objects onto trees; and shadows cast by trees onto other trees (Nakamae and Tadamura, 1995).

Block draping

Block draping is a technique which uses a straight forward digital terrain model, and then in regions of forest stands, the elevation values are increased by the height of the trees (Evans, 1993). Stylised trees use wire frame symbolised trees created using the simple graphics entities of lines and filled circles and triangles. Coniferous and deciduous trees are easily distinguishable due to the different symbol trees. Using colour on the terrain model and within the filled symbols provides a greater degree of realism.

2.5 and 3D tree patterns

If trees are needed for close range drawings, 2D tree patterns are not sufficient; using 3D tree patterns, however, requires much computational time and memory storage. If the location of the viewpoint is fixed, some parts of the data will not be visible. It is not necessary, therefore, to use up the memory storage with useless information (Sasada, 1987).

2.5D tree patterns are a logical alternative to using 3D patterns. A 2D tree pattern that rotates around a vertical line passing through the centre of the trunk can be used, this kind of rotating 2D pattern in called a 2.5D pattern. In a program that produces perspectives, the tree pattern automatically rotates with the viewpoint's rotation so that the front view always shows. If, however, the viewpoint rotates above the tree, a 3D representation will be required to see the top of the tree instead of just a line (Sasada, 1987).

The ramification matrix method - an example of how to grow a tree

This method provides a simple, easily understood process of tree growth. A tree structure is developed by hand to portray the style of the natural tree required. This is then represented in a computer by the binary tree data structure. From the binary tree, which can only split into two at each branching point, the contained nodes or branching points, are assigned a branching biorder, derived from the node's position within the structure (Evans, 1993). This follows the monopodial branching mentioned by Aono and Kunii (1984). This branching biorder is then used to develop the ramification matrix. This stochastic matrix is a lower triangular matrix consisting of the probability of a branch forming in a particular way (Evans, 1993).

New tree structures are formed by computation, using the ramification matrix, each being similar to the original hand designed tree structure. However, each tree produced is slightly different due to the random choice of branching biorder from the matrix. Development of the tree from the computer's binary structure is a straight forward process. The lengths and widths of the branches are controlled by the order weighting for each node in the binary tree. Linear and quadratic functions are used for the length calculations and polynomial or exponential functions are utilised for the width calculations. The branching angles are also governed by the order values of the nodes. The width, lengths and branching angles will all vary according to the shape and size of the tree being simulated (Evans, 1993).

As a final step leaves are added to the structure, the colour and shape again depending on the type of tree to be shown on the image. A more natural tree structure has been developed, varying the ramification matrix, length, width and branching angle functions, which allows both coniferous and broad-leaved trees to be portrayed (Evans, 1993).

Examples of projects using tree or forest simulation

The Sulphur Pass Project

The Sulphur Pass project showed that landscape Visualisation by DTM produces realistic and accurate images. Of the variety of enhancements used to increase realism, the use of tree symbols seemed to be the most effective. Not only did they make the landscape more lifelike, but they also improved the accuracy of the model. Another way to have improved accuracy would have been to allow trees in different polygons to have different average heights to correspond to their ages (Smart et al, 1991).

SmartForest

SmartForest uses three dimensional modelling based on a simple stem list to generate Visualisations that can be rotated and "walked" in real time. Visualisations can be created entirely from outside sources such as GIS-based models or can be developed interactively using built in biological models for tree growth, pest spread, and various silvicultural processes (Imaging Systems Laboratory, 1995). Forest prescriptions can be applied and the results modelled using the incorporated growth models. Although the resulting images are an accurate representation of the gross spatial characteristics of the forest, they are not realistic, neither in the sense that each tree symbol relates to a real tree, nor in the sense that the image faithfully shows the colour and texture of a photographic image. However, as discussed in the section on photo-realism and schematic images, it is not always necessary to have truly realistic images to understand what the simulation means.

References

Aono, M. and Kunii, T.L. (1984) Botanical Tree Image Generation. IEEE Computer Graphics and Applications, 4, 10-34.

Aylward, G. and Turnbull, M. (1977) Visual analysis: a computer-aided approach to determine visibility. Computer-Aided Design, 9, 103-108.

Bishop, I.D. and Hull, R.B. (1991) Integrating technologies for visual resource management. Journal of Environmental Management, 32, 295-312.

Bishop, I.D. and Leahy, P.N.A. (1989) Assessing the Visual Impact of Development Proposals: The Validity of Computer Simulations. Landscape Journal, 8 92-100.

Buttenfield, B.P. and Ganter, J.H. (1990) Visualisation and GIS: What should we see? What night we miss? In Proceedings of the 4th International Symposium on Spatial Data Handling, Vol 1, 307-316.

Daniel, T.C. (1992) Data Visualisation for decision support in environmental management. Landscape and Urban Planning, 21, 261-263.

Ervin, S.M. (1993) Landscape Visualisation with Emaps. IEEE Computer Graphics and Applications, 13, 28-33.

Evans, J.A. (1993) Simulation of realistic landscapes. Mapping Awareness and GIS in Europe, 7, 36-40.

Faust, N.L. (1995) The virtual reality of GIS. Environment and Planning B: Planning and Design, 22, 257-268.

Fisher, P., Dykes, J. and Wood, J. (1993) Map design and Visualisation. The Cartographic Journal, 30, 136-142.

Graf, K.Ch., Suter, M., Hagger, J., Meier, E., Meuret, P. and Nuesch, D. (1994) Perspective terrain Visualisation - a fusion of remote sensing, GIS and computer graphics. Computers and Graphics, 18, 795-802.

Gross, M. (1991) The analysis of visibility - environmental interactions between computer graphics, physics, and physiology. Computers and Graphics, 15, 407-415.

Imaging Systems Laboratory (1995) SmartForest: An Interactive Forest Data Modelling and Visualisation Tool. Department of Landscape Architecture, University of Illinois at Urbana-Champaign.

Kennie, T.J.M. and McLaren, R.A. (1988) Modelling for digital terrain and landscape Visualisation. Photogrammetric Record, 12, 711 - 745.

Loh, D.K., Holtfrerich, D.R., Choo, Y.K. and Power, J.M. (1992) Techniques for incorporating Visualisation in environmental assessment: an object-oriented perspective. Landscape and Urban Planning, 21, 305-307.

Mayall, K. and Hall, G.B. (1994) Information Systems and 3-D Modeling in Landscape Visualisation. In Urban and Regional Information Systems Association Annual Conference Proceedings, Vol 1, 796-804.

Miller, G.S.P. (1986) The definition and rendering of terrain maps. ACM Computer Graphics, 20, 39-48.

Moore, R. (1990) Landscapes on Pluto: improving computer-aided Visualisation. The Cartographic Journal, 27, 132-136.

Nakamae, E. and Tadamura, K. (1995) Photorealism in computer graphics. Computers and Graphics, 19, 119-130.

Nakamae, E., Harada, K., Ishizaki, T. and Nishita, T. (1986) A montage method: the overlaying of the computer generated images onto a background photograph. ACM Computer Graphics, 20, 207-214.

Oh, K. (1994) A perceptual evaluation of computer-based landscape simulations. Landscape and Urban Planning, 28, 201-216.

Orland, B. (1992a) Data Visualisation Techniques in Environmental Management. Landscape and Urban Planning, 21, 237-244.

Orland, B. (1992b) Evaluating regional changes on the basis of local expectations: a Visualisation dilemma. Landscape and Urban Planning, 21, 257-259.

Orland, B. (1994) Visualisation techniques for incorporation in forest planning geographic information systems. Landscape and Urban Planning, 30, 83-97.

Orland, B. (19??) SmartForest: A 3-D Interactive Forest Visualisation and Analysis System. Unknown conference proceedings.

Orland, B. and Daniel, T.C. (1995) Impact of Proposed Water Withdrawls on the Perceived Scenic Beauty of Desert Springs and Wetlands: Image Generation. Imaging Systems Laboratory, Department of Landscape Architecture, University of Illinois at Urbana-Champaign.

Perkins, N.H. (1992) Three questions on the use of photo-realistic simulations as real world surrogates. Landscape and Urban Planning, 21, 265-267.

Pukkala, T. and Kellomaki, S. (1988) Simulation as a tool in designing forest landscape. Landscape and Urban Planning, 16, 253-260.

Sasada, T.T. (1987) Drawing natural scenery by computer graphics. Computer-aided Design, 19, 212-218.

Shafer, E.L. and Brush, R.O. (1977) How to measure preferences for photographs of natural landscapes. Landscape Planning, 4, 237-256.

Smart, J., Mason, M. and Corrie, G. (1991) Assessing the Visual Impact of Development Plans. In GIS Applications in Natural Resources (eds Heit, M. and Shortreid, A.), GIS World Inc., 295-303.

Zewe, R. and Koglin, H.-J. (1995) A method for the visual assessment of overhead lines. Computers and Graphics, 19, 97-108.