In this exercise, I needed to perform a visual impact analysis of an actual proposed wind farm in Sweden. The specific objectives were:
Identify locational features in the study area that might be important to people.
Examine whether consented but not yet built wind turbines are visible from these locations.
To achieve this, I chose to conduct a viewshed analysis of the Lursäng wind farm in Sweden. Specifically, the construction of three 200m tall Siemens Gamesa (SG-155) with 155m rotor diameter with an installed capacity of 6,6MW were to be erected in Tanum municipality in Västra Götaland (see image to the left). They would stand 106,51m, 103,69m, and 95,83m above sea level, each producing an estimated 16,9GWh of electricity each year, or a total of 50,7GWh.
The point of the viewshed analysis was to assess the visual impact of the turbines on the surrounding environment and to see how it may impact features and points of interest that are important to people. For my simplified analysis, I chose to look only into the visibility of the turbines from the viewpoint of buildings, scenic viewpoints, attractions, camping and caravan sites, and memorials in the area. See the screenshot to the right.
I chose these POIs based on activities that primarily have to do with spending time outdoors, such as camping or traveling with a caravan in the countryside while visiting viewpoints to enjoy a beautiful landscape, and other various attractions in the area. I also decided to include memorials as they are sacred places that arguably ought to be minimally disturbed. Although this may not be a complete list of things that people care about in this context, it was a good starting point for this exercise.
The effective area of potential scenic and aesthetic disruption is taken to be a maximum of 16 km radially from each turbine.
For this analysis, I needed Lidar data, points of interest (POIs), bodies of water, and building footprints. The table to the left summarizes the workflow using this data.
Step 4 indicates how I created both the Digital Elevation Model (DEM) and Digital Surface Model (DSM) which were both essential for the whole analysis. The DEM would represent the bare ground where no trees or buildings are taken into account in the viewshed analysis, only the landscape.
Conversely, in addition to the landscape itself, the DSM would account for anything that extrudes from the landscape such as trees and buildings, and therefore be a more accurate and realistic representation of the area. The figures below show this, where the DEM (left) is smoother, while the DSM (right) is coarser.
↑The DEM clipped to the
16km visibility range limit.
↑The DSM clipped to the
16km visibility range limit.
Moving on to step 5, I ran the main analytical tool, the ‘Geodesic Viewshed’ tool in ArcGIS Pro on both the DEM and DSM. I chose ‘geodesic’ instead of ‘planar’ in order to stay as close to reality as possible; accounting for Earth’s curvature. The results from the DEM-based viewshed can be seen to the right.
The result provided the locations indicating how many turbines can be seen from any given raster cell. However, it did not provide information showing which raster cells see none of the turbines.
I needed NoData/Null information to calculate statistics later on. Therefore, I ran the ‘IsNull’ tool, creating a new raster with two values, 1 and 0, i.e. “True” and “False”, thus filling in the empty spaces. Finally, since I wanted to do a simple analysis of determining which POIs and building footprints are in the line of sight of the turbines, as well as how many turbines each one sees, I needed to use ‘Select by Location’ to check for where they intersect as indicated in step 8.
↑Building footprints (red) partially intersecting with the vectorized viewshed raster cells (blue: 3 turbines seen, purple: 2 turbines seen, and pink: 1 turbine seen).
However, it is not possible to check where a vector-based feature class intersects with a raster. In order to do that, I needed to convert the output viewshed rasters to polygons, as indicated in step 7. The figure to the left shows how this works. Specifically, logging the number of intersections would enable me to generate general statistics about how many turbines are seen.
It has to be emphasized, however, that these are simply just the footprints of buildings, and therefore lack their extrusion in the z-dimension. This means that a building might be seeing more or less turbines, depending on its true shape and obstructions in its vicinity. The final results can be seen below.
↑Results of the DEM-based viewshed analysis.
↑Results of the DSM-based viewshed analysis.
Looking at the maps above, we can see that no matter if using a DEM or DSM, the majority of the total area either sees 3 turbines or none. Although, in the case of the DSM, is that really so? Probably not, at least not for observers on the ground. Since trees and other structures are factored in, the majority of the areas where all 3 turbines can be seen, are only visible from the top of tree canopies (see figure to the right), which is not a typical observation point, unless you are a bird. However, the difference between what is seen on the ground (DEM) and on top of the canopies (DSM) can be visualized with a simple 'Minus' operation in ArcGIS Pro (which, for some reason, I had skipped when working on this but realized later on).
Nonetheless, after compiling the aforementioned intersections in a table, I could chart the proportions of POIs and building footprints that intersect with a given visibility area, i.e. 0 turbines, 1 turbine, 2 turbines, or all 3 turbines.
Generally, both charts indicate that there is not that much difference between using the DEM and DSM, especially for the visibility of 0, 1, and/or 2 turbines - which is good. The exception is that with no obstructions playing a part in the DEM, the more POIs and building footprints will see all 3 turbines.