A Viewshed Analysis of Potential Power Plant Sites






Ideal Conditions

Problematic Conditions







This project attempted to calculate, characterize and compare the 'viewsheds' of emissions stacks from two possible sites for an electricity generating facility. A viewshed can be defined as an analysis that "indicates not only what areas of a surface can be seen by one or more observers, but also, for any visible position, how many observers can see the position" (ArcView online help). Initially, the goal was to determine the local visibility of a hypothetical emissions stack approximately two hundred feet tall. This was not possible due to many factors, including technical difficulties with the software package (ArcView 3D Analyst), limitations from the available data and time constraints. An alternative plan was devised to gain an understanding of the potential viewsheds from the 'ground level' of the specific locations. This involved calculating the visibility of the sites respective positions on the surface itself. The two resulting viewsheds were quite different and perhaps some insight can be gained as to how visible the emissions stacks would be if constructed.


Controversy seems to become part of the usual pattern whenever industry is introduced into a rural section of this country. The Dartmouth, Massachusetts based company E.M.I. (Energy Management Incorporated) had a recent experience in Maine that was no exception to this routine. When a corporate subsidiary, Maine Power Associates, proposed building a 120 million dollar facility on one of two sites in western Maine, local reaction was mixed: it ranged from immediate acceptance, through indecision and on to complete rejection of the plan. The parcels under consideration were in the neighboring towns of Waterford (Site #1 was 94 acres in Oxford County) and Harrison (Site #2 was 63 acres in Cumberland County). In fact, the sites were only a little over a statute mile from each other. This was no coincidence, as any potential site would have certain, very specific, characteristics.

Location Map

The power plant would be designed to generate between 250 and 350 megawatts, utilizing natural gas as a fuel. Obviously, the best sites were those that fell along the major utility corridor running through western Maine, since this corridor had both a natural gas pipeline and transmission lines. The plant would simply siphon off a certain amount of natural gas, use it to generate electricity and then place that electricity back onto the local grid (the transmission lines) via a series of transformers. Other sitting factors were also taken into consideration. Despite the fact that the plant's design was very modern and efficient, it would still use between one hundred-fifty and two hundred thousand gallons of water daily. For this reason, it would have to be sited near an adequate water supply. Based on a few of these general parameters, these two properties appeared to meet the minimal requirements during the initial evaluation.

The local people who opposed building the facility based a vast majority of their opposition, not on other, more common environmental issues like pollution or water availability, but on the visibility of the emissions stack. (In order to comply with Maine D.E.P. regulations, the stack would have to be between two hundred and two hundred fifteen feet tall and seventeen feet in diameter. The final height would be calculated based on an average of the local meteorological conditions.) Since the local area is dependent on summer tourism, construction of the facility, especially the emissions tower, was perceived as a threat to the natural beauty of the area. Perhaps the acronym chosen by the group protesting construction of the plant says it best - P.O.R.CH stands for "Protect Our Rural CHaracter".


The idea for calculating and mapping out the viewsheds of the plant's locations came from a portion of a news article on how E.M.I. would handle the visual impacts of the emissions stack.

"I think the only visual impact, is it's hard to hide a stack 210- to 215-feet tall," [Vice President Craig D.] Olmsted said."
"We'll paint it any color you want," said [Manager of Environmental and Engineering Services Robert E.] Donahoe."
"As part of their process, the company will raise large balloons [to] the same height as the proposed stack and then take photographs of them from several different locations. This will better enable residents to get an idea of what the visual impact will be." - The Bridgton News, April 17, 1997

Following this line of thought produced the questions, "Just how visible would these structures be from the local area?" and "How does the viewshed of one site compare with the other?" The idea of photographing balloons from 'key' locations wouldn't appear to be a very satisfactory method, since the view from one's own picture window would likely constitute a 'key' location. Considering that the structure would probably be visible from hundreds, perhaps thousands, of households, and numerous scenic locations, there must be a better way to predict the extent of the stack's visibility. The answer may be provided by a three-dimensional surface generated within a Geographical Information System.


Ideal Conditions

Perhaps in the future most GIS projects will be completed without a hitch. Today there are potential problems with software as well as with data from unfamiliar sources. Actually, one can almost expect at least a couple of minor 'issues' to arise while completing any GIS activity. This particular plan was no exception to that rule. This 'ideal conditions' scenario gives a feel for the 'flow chart' approach, where one process blends smoothly into the next, and so on. This is how, in theory, the project should have gone. The next section outlines a few of the major obstacles that prevented the project from being completed as originally visualized.

The first step was of course, to gather the data. All the original project data was obtained from the Maine Office of GIS. This data set included six ARC/Info export coverages of contour lines, roads, utilities, ponds, streams and political boundaries for each quadrangle. These coverages were available on the 1:24000 scale for the two quadrangles of interest. These quadrangles were the Norway and Waterford USGS 'quad' sheets that basically covered the local study area.

A total of twelve coverages were imported into ArcView as themes. Once here, similar themes (streams from both quads, for example) would be joined into one theme by using the sample 'merge themes' script in Avenue. (Avenue is ArcView's native scripting language.) So the original twelve themes were combined to make six new themes. After completing this process, the contour line theme was made active and the 'generate TIN' command was chosen from the surface menu. With the ArcView 3D Analyst extension, one is able to generate a TIN, in the active view, from either elevation/contour line or point themes. TIN stands for Triangulated Irregular Network and is simply a vector-based way of depicting topology as a three dimensional surface using Delaunay Triangulation. (Delaunay triangles are the result of partitioning space into an array of irregularly spaced points and optimally constructing non-overlapping triangles from the points.)

Once the TIN surface was generated, it was assigned a hypsometric (elevation based) shading scheme, with a subsequent color band for each hundred feet of elevation change. Then the other themes were placed over the base surface. The idea was to on-screen digitize the approximate point locations for both the power plants and the actual parcels of land that the sites would sit upon. This was done using a base map supplied in an E.M.I. press release. With the roads and utilities themes active, this process was straight forward, as a majority of the property lines followed those features. Both site locations were saved as individual 'point' shapefiles, while the two parcels were combined onto one 'polygon' shapefile.

This is what the TIN with all the available theme information looked like: Landscape

This was a simpler version of the TIN with 'hillshading' applied to the surface: Hillshade

Initially, all the themes were brought from the active view into the 3D Scene. However, this proved to cause very slow 'redraw times', so the roads, streams, and political boundaries were removed, leaving the TIN, ponds, utilities, site locations and site parcels. Once the themes for the 3D Scene were chosen, all other themes were referenced to the TIN under the 3D-properties selection. To give the surface a sense of realism, the ponds were given an elevation of one foot above the surface while the stack points were projected two hundred feet above the TIN. Additionally, the transmission lines were given a relative height of forty feet and the parcels were assigned an elevation of twenty feet above the surface.

Finally the data was in the appropriate relationship to allow a viewshed to be calculated. This was done by simultaneously activating the surface and the site #1 shapefile. Then under the surface menu choosing the option 'calculate viewshed' and entering parameters for the output visibility grid, completed the operation. The result was a grid theme with visibility attributes assigned to every cell. The assigned values are the number of observation points from which a particular location can be seen. The extent of the grid was the same as the active surface and produced 228 columns and 313 rows (for one quadrangle). The resulting cell size was approximately 46 meters wide by 68 meters tall. The visibility theme was elevated thirty feet above the TIN surface and given a transparency of 50%. The transparency function allows the viewer to see through a layer and keep some sense of orientation or reference to a particular base map. (In this case the base map was the TIN with the ponds theme active.) It was while navigating through these 3D scenes, that it became apparent that the maps were not really the viewsheds for the top of the stacks, but were actually the viewsheds for the point on the surface at 'ground level' (more on this in the next section).

Problematic Conditions

One of the first problems that the project ran into was with the contour line data set. In order to calculate viewsheds for the entire area (both quads) the contour data would have to be combined or merge into a single continuous theme as mentioned above. ArcView refused to merge these themes and perhaps part of the reason why was because the contour line interval was different between the two quads! The Waterford quad had been digitized or generated at the twenty-foot interval while the Norway quad had data for every ten feet.

If you have Adobe Acrobat Reader you can view the contour line data.

Warning: the file is large (~ 6 Meg), but it is possible to zoom in and see details.

Interested? Contours

This fact was discovered after the two coverages were appended and 'tic-matched' in ARC/Info. Needless to say it was quite a mess! So, this project utilized two separate themes placed side by side, instead of a single theme. This really limited the amount of analysis that could be done, since the area covered by the Norway quad would not 'recognize' a point over the Waterford Quad and vice versa. This is the reason why the resulting viewsheds only cover one-half of the study area.

A second major problem developed while attempting to calculate the viewsheds for the points representing the tops of the stacks. As was previously mentioned, the viewsheds that were supposedly calculated from a point two hundred feet above the surface were apparently calculated from the point on the surface directly below the point of interest. So, the 'elevated viewsheds' were actually produced from a ground level perspective of the surface. At least that was the way things appeared to be as this was only a guess, but the viewsheds just 'looked' to small.

To verify this observation two-dimensional viewsheds were generated using the same shapefiles, only in this case the position was known to be on the surface. (It is impossible to elevate themes in the normal view, they can only be assigned relative positions like the layers in a cake.) In this case, the visibility grid was assigned parameters that produced 427 rows and 311 columns, and a cell size of 34 by 50 meters. This slightly finer resolution was also possible with the 3D scenes; they would have taken much, much longer to generate.

These 2D viewsheds were then compared to planimetric views of the three-dimensional viewsheds, and they were found to be almost identical, proving that the 3D viewsheds were not calculated from elevated positions. The two viewsheds patterns are quite comparable even with the slightly different resolutions. Quite obviously, the three dimensional viewsheds were generated for a point on the surface and not for an elevated position above the surface. Delving into the online help files for both ArcView and ARC/Info provided no answers as to why the software was not recognizing the elevated Z position of the point shapefiles. Elevations were even assigned to the site's attribute tables, but to no avail. Essentially, time ran out before this issue could be adequately resolved to produce the desired results. As of this writing, no answers to this problem were actually found. It seems as though it should be possible to calculate viewsheds for elevated positions, so further research is needed in this area.


Once the 3D scenes were positioned correctly by manipulating them with the navigation tool, JPEG snapshots were taken and imported into a layout template. These files were small enough to be exported in the ArcView EPS (encapsulated postscript) format. The layouts derived from 2D views of the TIN had to be exported as bitmaps because the EPS format produced file sizes approaching 60 megabytes! (Actually, Illustrator would not parse the ArcView EPS files, so all the layouts were finally exported as bitmaps.) The 1.8 Meg bitmaps were transformed into 16 megabyte Adobe Illustrator EPS files, as an intermediate step, which improved the resolution of the final maps. As a final step the encapsulated postscript files were transformed into the CompuServe GIF format using Adobe PhotoShop, for viewing on the Web.

The following links lead to the various viewshed maps that were produced from this project. The quality of the maps is not particularly high; because they were exported as ArcView bitmaps (.bmp format), which has a maximum resolution of 144 dots per inch. Even though they were transformed, the effects of the bitmap format are still obvious. If time had allowed, the layouts (especially the text) would have been reworked while the files were in Illustrator.

Two-Dimensional Viewshed of Plant Site #1: 2DView1

Two-Dimensional Viewshed of Plant Site #2: 2DView2

Planimetric Snapshot - Pseudo 3D Viewshed of Plant Site #1: Map3D1

Planimetric Snapshot - Pseudo 3D Viewshed of Plant Site #2: Map3D2

Perspective Snapshot - Pseudo 3D Viewshed of Plant Site #1: 3DPer1

Perspective Snapshot - Pseudo 3D Viewshed of Plant Site #2: 3DPer2

A close up of the elevated site locations: Stack Tops

As you can see from viewing these files, the maps are reasonably straight forward, that is, the areas that are either red or have a reddish tinge are areas that were not visible from the site locations, while the blue or bluish areas were visible. While viewing the last map, hopefully one can tell that the points were, in fact, well above the polygons that represented the site parcels.

From the maps, it should be obvious that the viewshed for site #2 is larger in the Waterford Quadrangle, than the viewshed calculated for site #1. Any further interpretation of the results would be misleading, since the Norway Quadrangle could not be analyzed. A through viewshed analysis would also include a much larger area with a coarser level of elevation data, perhaps every 50 or 100 feet. Anyone unfamiliar with the area would not know that there are substantial mountains to the northwest of these sites, and there would certainly be large areas included in the viewsheds of these two emissions stacks. In order to accurately assess the visual impacts of these structures, a wider scope should be applied to this local viewshed analysis.


Even though the project did not go quite as planned, the results do show that viewshed analysis is possible. If one uses a little imagination, then it should be clear that a point a few hundred feet above the points that were actually analyzed would have a much larger viewshed than those depicted on the accompanying maps. The 'relative ease' with which ArcView 3D Analyst generated the surface on a desktop PC, should make this type of study much more common in the future. The quality of the resulting TIN surface was actually much higher than expected.

When compared to a photographic series showing a balloon flying at an approximate elevation, the results have a much more 'qualitative' feel. Since visibility is shown simultaneously for many locations, this sort of study may actually be more damaging to the case of those companies siting industrial facilities. On the other hand, it does give industry the ability to take into account yet another aspect of the siting process. As geographical information systems gain in popularity for this particular application, it could be utilized to improve relations between industry and communities, by removing that unpleasant 'Surprise!' factor.


Various press release materials and fact sheets from E.M.I. / Maine Power Associates

The Bridgton News (April 17, 1997 edition)

The Lewiston Sun-Journal (Date not included on photocopy)

Clarke, K. C. (1997) Getting Started with Geographic Information Systems. Upper Saddle River, NJ: Prentice Hall

The Maine Office of GIS

ARC/Info online help

ArcView online help



The Maine Office of GIS




Submitted by Steve Harmon - 8 May, 1998

Thanxs to C.P. for his help/computer on this project.