Managing scenic resources has recently focused on computer simulations to illustrate changes to large-scale environments. Typical strategies merge ground photography with site data from a range of sources. Such approaches produce visualizations that combine graphic realism with quantitative accuracy and control. Merging data with images has been challenging, however, as issues of image rectification and ground/plan registration are involved. The research presented here addressed the accuracy issue of data visualization by employing geographic information system (GIS), global positioning (GPS) , and digital elevation model (DEM) technologies, teamed with an image processing component, to simulate color shifts in a forested landscape. At a site in southern Utah, a tree survey and a photographic inventory were developed using global positioning to coordinate both tree and camera/target positions. All tree data was input into a GIS, using workstation ArcInfo on a SUN platform. A DEM was constructed, using contour data and tree height attributes. This produced a surface that projected selected trees to their estimated heights. DEM perspectives, corresponding to each photograph, were generated using the GPS coordinates. These perspectives were then merged with processed photographs, that illustrated levels of a bark beetle event. Applying this image merging technique, different levels of beetle damage were positioned in the photographs, using the DEM's to locate zones for scenic simulation. A final image, showing beetle activity levels, was then generated using editing functions in Adobe Photoshop.
Techniques previously applied to vegetation inventories (remote sensing, aerial photo interpretation) offer the potential for documenting change relationships that could influence a region's scenic quality. Such techniques suggest analytical procedures for identifying the conditions or causes that produce visual change. Shifts in forest canopy reflectance provide clues that modifications have taken place. Interventions such as road construction or timber harvesting increase contrasts between the disrupted areas and the unchanged landscape. These contrasts are observed first as the soil is disturbed, and later accentuated as plants thriving in disturbance areas are introduced (Palmer et al., 1986; Hull and Bishop, 1988; Magill, 1990). The observed results are modified color relationships caused by a new distribution of sun light on the forest canopy and the surrounding ground plane. Applying a knowledge of reflectance characteristics based on physical composition, associations can be established as to the effects a particular change might create.
Recently, managers have utilized perceptual testing to gauge the impact that different change conditions produce (Buhyoff, Wellman and Daniel ,1982; Hull and Buhyoff, 1986; Bishop and Leahy, 1989; Orland et al., 1993). During testing, subjects are presented visual stimuli (e.g. color slides) and are asked to judge each scene on issues such as scenic quality. By regulating the scenic content, selected elements (trees, ground cover) can be isolated to determine the impact that any changes might have on overall perceived quality. Computer simulations have recently extended the options of perceptual testing. Utilizing a range of manipulative procedures, modifications can be controlled to test singular relationships such as shifts toward a particular color (Daniel et al., 1993). In this way, landscapes can be artificially modified to simulate change based on issues of disease, insect events, or ecological shifts. This facilitates the testing of planning proposals before tax dollars are expended.
Visualization efforts (Daniel et al., 1990; Ervin, 1992; Orland, 1994; Shang, 1994) have recently applied a data-driven approach utilizing site data to assist in graphic development. These efforts emphasize procedures that rectify plan view data with photographs from aerial or ground positions. Referencing tools have included geographic information systems (GIS), computer-aided design (CAD), digital elevation models (DEM) (Gimblett and Itami, 1988; Bishop and Hull, 1991), and more recently, global positioning (Zavala and Zavala, 1993). Vector DEM output has provided the greatest utility by allowing the base photographs to show through to assist in the final registration. Manual versus automated procedures have been mostly utilized, although rubber stretching techniques similar to GIS/satellite image rectifications (Dull et al. 1990; Zhou, 1989; Welsh et al., 1992), have been applied.
A texture mapping strategy (Bishop and Flaherty, 1990; Gimblett, 1990) applies portions of aerial images, which are draped over 3D surfaces set in perspective. High levels of scenic registration are achieved. Questions have been raised, however, as to the color integrity of the draped image because of the different orientations and reflectance differentials exhibited in plan versus perspective views (refer to; Jackson et al., 1990; Li and Strahler, 1992). A second approach develops a complex surface that articulates individual trees using graphic primitives (e.g. cones, domes) which are incorporated into the final surface (Kellomaki and Pukkala, 1989; Rich, Hughes, and Barnes, 1993). The resultant surface can then be mapped using some draping technique. Unfortunately, the utility of the final graphic presents some limitations, due to generalized color/shadow articulations, and averaged tree sizes.
A third technique attempts to minimize the color/texture anomalies realized during image draping by merging of ground photography with wire frame perspectives in corresponding view relationships (Shang, 1994). The perspectives (CAD or GIS) are typically merged with photographs using editing software such Adobe Photoshop. Because human interpretation is typically applied, realism and color integrity are maintained at the expense of spatial registration. Previously, the DEM surfaces have been void of trees, or have been constructed using a top-of-tree canopy approach. While such modeling might suffice in vista-scale applications, near-view modeling requires a more definitive process to correlate the imagery. This need for an improved registration that maintains high image fidelity was the motivation behind the research presented here. In the described work, a photographic inventory was developed, utilizing global positioning to articulate all camera/target positions. The coordinate data, plus information relating to the site's physical and biological characteristics, was then input into a GIS data base. Utilizing the camera/target coordinates, the photographs were merged with DEM perspectives to link the photographs to the GIS. This linkage provided a mechanism to transfer tree-related data to specific zones within the photographs. In this way, tree color/change relationships were transferred from GIS to photographic image, facilitating a means to geographically articulate color shifts in portions of the forest.
The second goal was to develop a mechanism to merge color/change data, obtained through digital image processing, with field-recorded measurements of the biologic and ecological characteristics of trees within a forest condition. This was achieved by using a relational data structure to reference all field information pertaining to the health and vigor of the measured trees. An integrated approach, using geographic information system (GIS) and global positioning (GPS) technologies, was utilized to coordinate the data, and to articulate all tree positions and associated attribute information. Extending this data in three-dimensional display, the tree attributes were projected in-perspective to mimic those camera relationships of the acquired ground photographs. This was accomplished using the Surface Modeling with TINs module in ArcInfo to produce a 3D surface that illustrated the surface terrain plus the trees extended to their estimated heights. The two image types; the scanned photographs and the DEM perspectives, were subsequently merged to display the extended trees in the DEM with the scanned photographs. Using this methodology, the tree attributes within the GIS were registered to the ground-oriented photographs.
A final goal was to utilize this referencing technique to position portions of scanned photographs, which were processed to simulate incremental change due to a bark beetle (Dendroctonus rufipennis) infestation, within secondary images that displayed no visible evidence of a beetle impact. The resultant compound graphic presented a forest scene with simulated color variation that represented a beetle event in different stages. These simulations would later be applied in perceptual testing to ascertain the reactions of people to the simulated levels of scenic change. To accomplish these goals, a research design was developed to facilitate the following: (1) the acquisition and processing of scenic color changes in a forest canopy, (2) the geographic registration of those processed color signatures, and (3) the creation of scenic visualizations, depicting the measured color relationships, to simulate a forest undergoing visible change. The extended research question of whether the simulations would produce perceptual responses that varied in relationship to the different levels of beetle activity will be addressed in a separate research report.
The second site; Sidney Valley, is located approximately ten miles to the north-east of Midway Face. Sidney Valley has similar environmental, biologic, and ecological characteristics as Midway Face. Like Midway Face, it is relatively narrow with mountainous slopes on both sides. The elevational range is between 9,750 feet (2971 meters) in portions of the lower meadow and 10,600 feet (3231 meters) at the top of the western ridges. Sidney Valley's axis, however, is approximately 45 degrees north- of-east as opposed to Midway's east-west orientation. The eastern and western mountainous slopes are populated by dense stands of conifers, with Engelmann spruce again being the dominant species. As opposed to Midway Face, Sidney Valley is relatively inaccessible. A dirt road to the south is the only vehicular access. Numerous trails do, however, meander in and around the valley, opening the landscape up to limited access. The mixture of Sidney Valley's tree composition, and the existence of a known biological change agent, made the site ideal for extracting color/change data.
A DEM was developed
from the GIS coverage, and perspectives relating the site photography were
generated. These perspectives, the developed GIS and DEM, and
photographic material from both sites would later be utilized in an image
processing/data visualization effort to replicate the color/change
characteristics that occurred at the study site.
After locating the control points, two linear transacts of fifteen points each were developed on the study area's east and west sides. A hand-held laser instrument was utilized to coordinate these points to the initial GPS data. The laser instrument measured the distance (in feet) and direction (degrees-minutes-seconds from north) from individual GPS positions. From these secondary control points, the 5000 trees were surveyed, using the same distance and direction procedures. All trees within the study area were surveyed. Data was recorded on-site in survey log books, and later compiled on a micro-computer, using Quattro-pro.
Before the 1993 activities, camera positions were staked out in three quadrants at Midway Face, and in one quadrant at Sidney Valley. Geo- referencing techniques, identical to those used during the tree survey, were employed. From these points, photographs were taken using three camera orientations: South 30� East, South, and South 30� West. A tripod mounted 35mm camera with a 50mm lens, a UV Haze filter, and Ecktachrome 100 slide film was employed for all photography. The photography at Sidney Valley was more simplistic, involving only three view points. An initial GPS point was provided by the Forest Service, and the two other points were positioned from that GPS point using the mentioned surveying methods. The Sidney Valley photography was taken using the same three camera bearings, and identical camera equipment and film as that used at Midway Face. All positional data was later converted to metric and input into the study's GIS. To verify its accuracy, each point at Midway Face was visually reviewed during GIS development using a geo- referenced aerial video mosaic of the site.
Pixel samples representing the three increments were extracted from the two images, using only the sunlit portions of individual trees (Graph 1).
The mean red, green, and blue (RGB) values were recorded per sample, and averages of the mean values were computed. Mean RGB differentials were then computed to produce multiplicative factors between (1) the healthy, non-visible beetle levels and the 1-2 year level of beetle damage, and (2) the healthy, non-visible beetle levels and the 3-5 year level of beetle damage. These multiplicative factors represented the statistical changes between the RGB levels of the unaffected and affected pixel samples. The multipliers were applied to the images (using ERDAS) to simulate the color relationships indicative of the three beetle increments. These multipliers were applied universally to all the trees within the images. During final data visualization, these uniformly processed images were merged with their respective DEM perspectives to locate only those desired tree zones within the scenes where the final visualizations would occur.
Data from the 1992 tree survey was converted from the initial Quattro-pro file format to a series of ascii text files. These text files were then imported into ArcInfo, and converted to GIS point coverage (Table 1).
Tree No. XUTM YUTM ZMTR LD SP DB HT __________________________________________________________________ 500 337575.52 4159387.70 3004.1 L ES 18 92 501 337576.74 4159385.60 3004.4 L AF 29 34 502 337574.91 4159391.60 3004.4 L ES 24 108 503 337576.44 4159394.70 3003.2 L ES 17 83 504 337577.35 4159395.00 3002.6 L ES 6 25 505 337634.05 4159394.40 2999.5 L ES 16 86 506 337638.62 4159396.50 2998.3 L ES 18 94 507 337638.92 4159393.80 3000.5 L ES 13 72 508 337641.36 4159393.50 3000.5 L ES 8 24 509 337641.36 4159392.00 2999.5 L ES 19 91 510 337642.88 4159394.10 3000.5 L ES 15 79 511 337641.97 4159395.60 2999.8 L ES 28 105 512 337641.67 4159395.90 3000.5 L ES 27 107 513 337643.49 4159395.30 3000.1 L ES 12 71 514 337644.41 4159399.60 2999.2 L ES 20 96 515 337648.37 4159391.40 3000.1 L AF 13 56 516 337649.59 4159389.50 3000.1 L AF 15 78 517 337652.03 4159390.70 3000.8 L AF 12 68 518 337648.68 4159388.00 3000.5 L ES 30 111 519 337658.12 4159369.40 3004.7 L ES 18 103 520 337652.94 4159384.00 3001.4 L AF 15 74 521 337655.08 4159381.90 3002.0 L AF 12 65 522 337655.08 4159380.40 3002.0 L AF 13 60 523 337656.30 4159376.70 3002.3 L ES 18 90 524 337655.38 4159373.40 3002.6 L ES 14 55 525 337657.82 4159388.00 3000.5 L ES 18 84XUTM= X Coordinate (meters), YUTM= Y Coordinate (meters),
*
Within ArcInfo, all tree locations (expressed in feet and positioned relative to the control points) were converted from relative distances to absolute UTM coordinates. Upon completion of this point data conversion, it was observed that several tree points were potentially in error. Additionally, there were duplications of points, and some were found void of necessary attribute data (e.g. species type, elevation, estimated height). This issue was magnified after a review of a preliminary TIN, which was generated to test the reliability of the tree survey data. Using two simple AML's, which were designed to detect potential positional and elevational errors (one reviewed each tree's northing and easting, and a second identified any excessive differentials in the height attribution of trees located next to each other), a series of corrections and reductions were performed on the tree coverage. These corrections changed the initial tree inventory from a list of 4923 trees to a revised total of 4473 individually attributed trees (Figure 2).
Applying this technique, a series of ground/canopy DEM's were constructed from the GIS data to simulate a spreading pattern of beetle activity through the study area's spruce inventory . This spreading pattern, as projected by Forest Service personnel, started from a central cluster of Picea engelmannii, which was designated the hypothetical starting point of a beetle event. Approximately 100 Picea engelmannii, grouped in the center of the study area, were selected. This selection was achieved utilizing the species type and estimated age attributes to allocate each tree for inclusion (Figures 3, 4 & 5).
Next, a second DEM surface was generated, using the height attributions from both the initial central cluster of Picea engelmannii and an additional group of Picea engelmannii peripheral to this core. This simulated the radial spread of the beetle activity across the study site. (Figure 6).
In later stages, other enlarged spruce clusters would continue the radial pattern in all directions from the starting point. As each subsequent DEM was created, perspective views similar to those of the site photography were generated. The observer point, view orientation, and focal length of each perspective was based on the positional data acquired during the photographic survey.
The mergers, using Adobe Photoshop, were created by super-imposing each DEM perspective over their photographic equivalent. The GPS coordinate data was utilized in this effort, as well as using visually identifiable site elements to support the process. DEM/photographic composites were generated using three processed versions of each image (one for each simulated beetle activity level) and two versions of the DEM (one showing a small central core of trees, and a second showing an expanded central core of trees).
Three types of DEM/photographic composites were then created. The first merged the smaller central core of trees in the DEM perspective with the image that had been processed to simulate a 1-to-2 year level. A second composite merged the expanded central core of trees with the image that had been processed to simulate a 1-to-2 year level. The third composite merged the smaller central core of trees with the photograph that had been processed to simulate a 3-to-5 year level.
The three composite image types were then used to develop the final scenic visualizations. Four sets of final visualizations were created. These corresponded to four camera/target relationships at Midway Face. The visualization strategy was to maintain the integrity of the base images while modifying only those tree portions that were spatially defined by the DEM perspectives as being areas of simulated change. Generally, the process can be summarized as follows: (1) construct an irregular boundary around the desired tree clusters for both the 1-2 year and 3-5 year beetle levels/DEM composites, (2) copy only the irregular boundary to a clipboard, (3) open the images with the different simulated levels of beetle activity, (4) paste the saved irregular boundary in its appropriate location, (5) using this boundary, isolate and clip the desired portions of infested trees, save these pixels to the clipboard, (6) insert the clipped trees into the appropriate scenes, save as new images, and (7) position the 3-5 year infestation levels on top of the previously positioned pixels displaying 1-2 year beetle levels. Resultant images were made into 35mm slides using an Agfa Forte film recorder, at a resolution of 4000 horizonal by 2666 vertical screen pixels. Figure 8 illustrate an original scene and a simulated image of damage.
The utility of the described approach presented several benefits which equated into improved data registration and transfer procedures. First, because scenic change has spatial implications, the described methods improved the positional accuracy of the graphic output. These improvements can be credited to the combinations of global positioning and surveying procedures utilized during field data acquisition. Further, the GIS data structure provided a convenient mechanism to catalog and later query the distributions and patterns of the inventoried trees. A second benefit was realized when the GIS was applied during DEM development. Extending the tree coordinates to their measured heights produced a viable canopy model for illustrating the mass and form of clustered trees. Previous modeling efforts (Rich, Hughes, and Barnes, 1993) might have provided similar results, however, the simplicity of this approach justified its application. A third benefit became apparent when the tree data was transferred from the initial plan view registration to a ground-based projection. Because the hypothesized beetle spread patterns were controlled through the UTM coordinate geometry, a 2D/3D transfer was facilitated by using the DEM. By applying the positional relationship between GIS and DEM data, the trees' physical/biological characteristics were three- dimensionally related with the processed color signatures representing increment beetle impact. Issues of data integration and perspective display were, therefore, resolved.
The camera/target coordinate data further assisted in maintaining the positional relationships between the DEM perspectives and those of the site photographs. This insured common view orientations, which assisted in the ultimate merging of the two image types. Image merging was also facilitated by distinct or recognizable terrain characteristics, such as the spaces within the forest canopy, or the irregular tree edge as it met the open meadow. These common identifiers had the advantage of being both visually apparent in the photographs, and geographically defined within the GIS. Image matching was, therefore, assisted by using the combined referencing of the visual as well as computer-generated information. A final advantage of GPS referencing was in the actual location, on-site, of photographic positions. Because of the site's snow loads and severe winter conditions, markers to delineate reference points were typically shifted or lost. This potentially placed limitations on the acquisition of photography during the multi-year research effort. Knowing the coordinates of each position provided a means to locate the previous year's camera and reference positions that otherwise might have been lost.
Another potential area of refinement is the actual registration technique employed. While the applied techniques transferred data from a 2D to a 3D display, the process does not represent a direct GIS/photographic rectification. Additionally, some visual assistance during the rectification process is still warranted. This short coming does not necessarily limit the utility of the process, however, because both photograph and DEM perspective are, in reality, two-dimensional representations of three-dimensional surfaces. The point only emphasizes the need for improved automation in the rectification process. A final area for refinement relates to the computer hardware and software employed during the research. During the data development and processing, three hardware configurations (MsDos, Macintosh, Unix) were used in conjunction with eleven application softwares. Any transfers of technology to governmental agencies will require a more simplified approach with common hardware/software arrangements.
This research provided support for utilizing the described methods in spruce forest conditions, using surveyed trees as the basis for the image/data integration. While it must be stated that few researchers have the luxury of a study site with 5000 individually surveyed trees, new technologies are being tested to automate the process. A pattern recognition approach, (Orland, 1994) shows promise for identifying trees from aerial photographs. Future data collection using combinations of remote sensing and pattern recognition could further articulate issues of health, insect/disease activity and nutrient deficiencies. Image resolution has been a limiting factor; however, as data capture extends into other, improved areas, the individual tree as a discrete data source will become a viable alternative. Global positioning, teamed with a GIS, suggests benefits that extend beyond the ability to create scenic image composites. The described data registration techniques can relate research material to existing resource inventories. This flexibility to incorporate new data could expand existing research efforts beyond their initial scope, and potentially provide extended benefits to agencies funding the efforts.
Bishop, I.D. and Flaherty, E. 1990. Using Video Imagery as Texture Maps for Model Driven Visual Simulation. In Proceedings, Resource Technology 90. Washington, D.C.: American Society for Photogrammetry and Remote Sensing. pg. 58-67.
Bishop, I.D. and Hull, R.B. 1991. Integrating Technologies for Visual Resource Management. Journal of Environmental Management, (1991)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.
Brown, T.C. and Daniel, T.C. 1986. Predicting Scenic Beauty of Timber Stands. Forest Science, Vol.32(2), 471-487.
Buhyoff, G.J., Wellman, J.D., and Daniel, T.C., 1982. Predicting scenic quality for mountain pine beetle and western spruce budworm damaged forest vistas. Forest Science, 28, 827-838.
Daniel, T.C., (1990). Measuring the quality of the natural environment. American Psychologist, 45, 633-637.
Daniel T.C. and Boster, R.S. 1976. Measuring Landscape Aesthetics: The Scenic Beauty Estimation Method, USDA Forest Service Research Paper RM-167. Fort Collins, CO: Rocky Mountain Forest and Range Experiment Station.
Daniel, T.C., Orland, B., Lynch, A., Hetherington, J., and LaFontaine, J. 1990. Integration of GIS and Video Imaging Technology for Data-Driven Visual Simulations. In Proceedings, American Society for Photogrammetry and Remote Sensing. Tucson, AZ: The University of Arizona. pg. 126- 138.
Daniel, T.C., Orland, B., Hetherington, J., and Paschke, J.L. 1993. Public Perception and Attitudes Regarding Spruce Bark Beetle Damage to Forest Resources on the Chugach National Forest. Fort Collins, CO: USDA Forest Service. Forest Pest Management Region 10.
Daniel, T.C. and Vining, J., 1983. Methodological issues in the assessment of landscape quality, in I. Altman & J.F. Wohlwill (eds.) Behavior and the Natural Environment. New York: Plenum.
Dull, C., Rubel, D. Spears, B. and Uhler, R. 1990. Integration of Remote Sensing and GIS Databases to Monitor Forest Conditions in the Southern Region. In Proceedings, The Third Forest Service Remote Sensing Application Conference, Tucson, AZ: The University of Arizona. pg. 22-32.
Ervin, S.M. 1992. Integrating Visual and Environmental Analysis in Site Planning and Design. GIS World, July, 26-30.
Gimblett, H.R. 1990. Visualizations: Linking Dynamic Spatial Models with Computer Graphic Algorithms for Simulating the Effects of Resource Planning and Management Decisions. URISA: Journal of the Urban and Regional Information Systems Association, (6) 26-34.
Gimblett, H.R. and Itami, R.M. 1988. Linking GIS and Video Technology to Simulate Environmental Change. In Proceedings, GIS/LIS '88. Falls Church, VA: American Society for Photogrammetry and Remote Sensing, pg. 208-219.
Hull, R.B. and Bishop, I.D. 1988. Scenic Impacts of Electricity Transmission Towers: the Influence of Landscape Type and Observation Distance. Journal of Environmental Management, Vol.27(1988), 99-108.
Hull, R.B. and McCarthy, M.M., 1988. Change in the landscape. Landscape and Urban Planning, 15, 265-278.
Jackson, R.D. et al. 1990. Bidirectional Measurements of Surface Reflectance for View Angle Correction of Oblique Imagery. Remote Sensing Environment, Vol.32(1990), 189-202.
Kellomaki, S. and Pukkala, T. 1989. Forest Landscape: A Method of Amenity Evaluation Based on Computer Simulation. Landscape and Urban Planning, Vol.18(1989), 117-125.
Li, X. and Strahler, A.H. 1992. Geometric-Optical Bidirectional Reflectance Modeling of the Discrete Crown Vegetation Canopy: Effect of Crown Shape and Mutual Shadowing. IEEE Transactions on Geoscience and Remote Sensing, Vol.30(2), 276-291.
Magill, A.W., 1990. Assessing public concern for landscape quality: a potential model to identify visual thresholds. Research Paper PSW-203. Berkeley, California: Pacific SW Forest and Range Experiment Station, Forest Service, UDSA.
Munson, A.S. and DeBlander, V., 1992. A Biological Evaluation of Spruce Beetle Activity on Midway Face-Dixie National Forest. Ogden, Utah. USDA Forest Service Report No. 3420, Forest Pest Management Group, Region Four.
Orland, B., 1994. Visualization techniques for incorporation in forest planning geographic information systems. Landscape and Urban Planning, Vol.30(1994), 83-97.
Orland, B., Daniel, T.C., Paschke, J.L. and Hetherington, J., 1993. Final Report: Visualization of Forest Management Issues on the Dixie National Forest. Ogden, Utah: USDA Forest Service, Forest Pest Management Region Four.
Orland, B., Onstad, D. Obermark, J. and LaFontaine, J. 1992. Visualization of plant growth and pest models. In Proceedings ,ASPRS/ASCM/RT 92, Vol.5, 246-251.
Palmer, J.F., Gobster, P.H. and Kokx, T., 1986. Long term visual effects of alternative clearcutting intensities and patterns. USDA Forest Service; North Central Experiment Station, Chicago,IL.
Rich, P.M., Hughes, G.S. and Barnes, F.J. 1993. Using GIS to reconstruct canopy architecture and model ecological processes in pinyon-juniper woodlands. In Proceedings of the Thirteenth Annual Esri User Conference. Redlands, CA. pg. 435-445.
Schmid, J.M. and Frey, R.H. 1977. Spruce Beetle in the Rockies. General Technical Report RM-49, USDA Forest Service; Rocky Mountain Forest and Range Experiment Station. Fort Collins, CO.
Sell, J.L. and Zube, E.H., 1986. Perception of and response to environmental change. Journal of Architectural and Planning Research, 3, 33-54.
Shang, H. 1994. An integrated application of gps, gis, cad and photorealistic simulation in assessing visual impact of silvicultural systems. In Proceedings from Resource Technology '94, Melborne, Australia. pg. 126-139.
USDA Forest Service 1993. Environmental Assessment for Midway Face Viewshed Management Project. Ogden, Utah: USDA Forest Service, Intermountain Regional Office.
Vining, J. and Orland, B., (1989). The video advantage: a comparison of two environmental representation techniques. Journal of Environmental Engineering, 29, 275-283.
Welch, R., Remillard, M. and Alberts, J. 1992. Integration of GPS, remote sensing, and GIS techniques for coastal resource management. Photogrammetric Engineering and Remote Sensing, Vol.58(11), 1571-1578.
Zavala, I. and Zavala, M.A. 1993. Global Positioning System as a Tool for Ecosystem Studies at the Landscape Level: An Application in the Spanish Mediterranean. Landscape and Urban Planning, Vol.24(1993): 95-104.
Zhou, Q. 1989. A method for integrating remote sensing and geographic information systems. Photogrammetric Engineering and Remote Sensing, Vol.55(5), 591-596.
Zube, E.H., Sell, J. L. and Taylor, J.G.,1982. Landscape perception: research, application and theory. Landscape Planning, 9, 1-33.