Heights and locations of artificial structures in viewshed calculation: How close is close enough?
Introduction
The arrangement of the landscape significantly influences the visual experience of the environment. The sales price of a home or the aesthetic impact of a scenic lookout is greatly influenced by the view it offers, which is determined in large part by the landscape features directly visible from observer locations. While view quality is partially dependent on relatively unchanging landscape elements like mountains or valleys, views are also affected by more readily altered landscape features, particularly built structures such as buildings (Miller, 2001). The arrangement of these elements can significantly impact how humans experience their environment as even slight changes in these elements may drastically alter human perceptions of the visualscape. Changes in land use associated with urbanization often impact human perceptions of the landscape, frequently with negative economic and social consequences.
Advances in geospatial technologies have improved the ability of urban and resource planners to calculate views and thereby attenuate the negative impacts of new building construction. Nonetheless, it remains difficult to incorporate these built landscape features into view analysis due to the poor availability of high resolution spatial data for buildings, specifically their locations and heights. These features are rarely reflected in commonly available elevation maps, and adding the exact locations and heights of buildings to these existing data can be difficult as a result of the time, effort, and expense involved. Given these costs, it is important to ask: how necessary are these high-resolution data? In particular, how do viewsheds, or computational projections of views, change as we acquire higher accuracy information on building location and height?
Understanding the effect of data quality on viewsheds is important because views have aesthetic and economic impacts. View quality contributes significantly to visitors’ experiences in natural areas (Gourlay and Slee, 1998, Steinitz, 1990) and to property values (Benson et al., 1998, Bishop et al., 2004, Bourassa et al., 2004, McCleod, 1984, Oh, 2002, Paterson and Boyle, 2002, Pompe and Reinhart, 1995). Certain land use types are more desirable in views than others and have corresponding impacts on property values. For instance, views of urban green spaces (Jim and Chen, 2006, Luttik, 2000), water (Benson et al., 1998, Luttik, 2000), and forests (Tyrväinen and Miettinen, 2000) all have been found to significantly increase property values while views of buildings and industrial areas have been found to decrease values (Bishop et al., 2004). View structure has also been found to influence land prices such that properties with more diverse views have higher monetary values (Bastian et al., 2002). Given the relative ease with which views can be altered through human activities (such as forest clearing or building construction) and the influences such alterations may have on property values and thus on a community's tax revenue and the income accruing to land developers, assessing the visual impacts of changes in land cover has become increasingly important to natural resource and land use planners. The accuracy of such visual impact assessments is questionable, however, since the locations and heights of both current and future landscape elements used in viewshed computation are often uncertain.
Enthusiasm for visibility analyses has been bolstered by recent advances in computing power and software that make such analyses more accessible (Rana, 2003). Two basic approaches, isovists and viewsheds, are utilized in view assessments. These approaches differ in their methodologies and in the applications for which they are commonly used. While neither method provides a direct measurement of the actual behavioral aspects of a view in relationship to land markets, both are able to provide information on visibility and view structure that is useful in studies of scenic quality.
The isovist approach is often used in built environments, particularly in architectural studies. This approach calculates an isovist, or the subset of points in space that are visible from a particular vantage point (Benedikt, 1979) that can be generalized to a graph in two-dimensions or a polyhedron in three (Llobera, 2003). Isovists are usually derived from architectural or urban plans and are therefore often calculated without reference to information related to variability in the heights of elements. Isovists have been analyzed based on their numerical properties (e.g., area, perimeter, variance) (Benedikt, 1979) and using visibility graphs to calculate properties that relate to view quality, such as average, minimum, and maximum distance; perimeter compactness; and cluster ratio (DeFloriani et al., 1994, O'Sullivan and Turner, 2001, Turner et al., 2001).
The viewshed approach, used more commonly outside of urban and architectural fields, calculates intervisibility. A viewshed is a two-dimensional spatial layer that records the intervisibility between points in a tessellated landscape and accounts for the location, height, and angle of view of an observer located in three dimensions (for an overview of the algorithms used, see DeFloriani and Magillo, 2003). The viewshed procedure computes locations in an elevation model that are connected by a line-of-sight to a viewpoint location within a specified distance of an observation point. A digital elevation model (DEM) is used most often in this procedure, although triangular-irregular-networks (TINs) may also be used. The viewshed calculation is now a common function of many software packages, especially geographic information systems (GISs). Batty (2000) demonstrated that, in many respects, isovists and viewsheds are computationally similar when projected onto a surface defined by a regular tessellation (e.g., a raster grid or TIN). We focus on viewsheds in this paper.
While urban and resource planners have constructed isovists and viewsheds for decades, research that analyzes the effect of input data quality on the realism of viewshed outputs has a shorter history. The increased availability and use of GIS has lead to a greater number of metrics with which to analyze viewsheds, including areal extent, depth, topographic relief, landscape diversity, and amount of edge (Bishop et al., 2000, Germino et al., 2001, Llobera, 1999) and attempts have been made to quantify changes in the values of these metrics with landscape change (Miller, 2001). These metrics have made viewsheds much more amenable to quantification and analysis.
Input data quality influences the realism of viewsheds created by computational means. Bishop et al. (2000) identify two key challenges in this area. First, most viewshed analysis depends heavily on 2.5D data from DEMs, or two-dimensional data with height attributes not rising to the sophistication of a fully three-dimensional data model. Viewshed calculations that rely on these data may not adequately represent the effects of vertical landscape features on views. Second, viewshed studies often completely omit vertical elements (e.g., built structures and vegetation) from viewshed calculation, relying instead on elevation and topography alone to identify views. In such cases, the viewsheds generated lack important elements that may obscure views, and instead identify views that are solely a function of landscape elevation. These two issues are particularly problematic in practice because view change analyses are often executed to identify the impacts of new structural elements on views. The use of models that do not accurately represent or that omit key components of the built and natural environment may make identifying these impacts accurately impossible.
Thus, there is a need to better incorporate built structures into the analysis of viewsheds in three dimensions (Bishop, 2003) and to do this in a manner that does not significantly increase calculation time and effort. Several studies have endeavored to include information related to structures and, relatedly, vegetation in viewshed calculation (see Lake et al., 2000, Germino et al., 2001). In a case study that exemplifies the challenges of incorporating building information into a viewshed analysis, Lake et al. (2000) used two different approaches to incorporate building heights into viewshed computation. These authors used a GIS layer containing building outlines (footprints) for the entire city of Glasgow, United Kingdom and employed two methods for assigning building heights to them. In the first, they determined building heights as a function of their number of stories after physically observing the number of stories in each building over the course of approximately 10 days. In the second approach, the authors assigned a uniform height to all buildings. For both approaches, building heights were assigned to the locations corresponding to actual building footprints and then summed with base elevations from a DEM to produce two elevation surfaces containing building heights assigned using each method. The authors compared viewsheds calculated using the two different methods and found that the areas of viewsheds calculated using uniform building height were somewhat smaller than those of viewsheds calculated using field observed heights. This result suggests that, in order to accurately calculate viewsheds in urban areas, it may be necessary to accurately identify building footprints and heights.
Identifying structure locations and heights is more easily achieved in some locations than in others. For example, in the Minneapolis-Saint Paul metropolitan area, some counties have access to spatially referenced structural data while others do not. Rural communities in Minnesota generally do not have these datasets. A similar situation exists in other areas of the United States, for instance in Massachusetts where such data are available for much of the metropolitan Boston area, but have spotty availability for jurisdictions in the rest of the state. The availability of such data may largely be a function of the GIS capabilities, economic status, and degree of urbanization of a jurisdiction. Wealthier, more urbanized communities may be better able to support a GIS program than poorer, more rural communities. More urbanized communities may also better recognize the usefulness of such data. Where these structural data are not available, it may be necessary to invest significant time and effort into digitizing building footprints and determining building heights if the results of viewshed analyses based on them are to be accurate. This may be prohibitive, particularly in poorer jurisdictions with limited GIS capabilities.
What is the best way to calculate viewsheds for the many locales that do not have readily available data for either building heights or building footprints? Digitizing building footprints from aerial photographs, plat maps, or engineering plans is resource intensive. An exhaustive ground survey of every building is similarly time consuming and expensive. Other options are also costly, including the use of LIDAR (light detecting and ranging) data or stereographic aerial photos. However, it may be possible to generalize building locations and heights based on existing land use or zoning data with a reasonable amount of effort. In such situations, building location information is not likely to be 100% accurate. However, where building location information is not readily available and intensive methods like those described above are cost-prohibitive, reductions in the accuracy of calculated viewsheds may be minimal or at least acceptable in the interest of time and cost-savings.
In this study, we examine the extent of differences between viewsheds based on building heights and locations generalized from land use data versus those created with higher quality footprint and height data. We address the following questions:
- (1)
Do viewsheds calculated using generalized building locations and footprints differ significantly from those calculated using exact locations and footprints?
- (2)
Do viewsheds calculated using building heights based upon dwelling style or the number of stories in a building differ significantly from those calculated by assigning all buildings a uniform height?
By answering these questions, we hope to both advance understanding of the role of error in viewshed analysis and help guide future viewshed studies in urban and urbanizing areas in choosing among alternative means for incorporating structural information into viewshed calculation without drastically increasing data acquisition and processing time.
Section snippets
Methods
We calculated viewsheds using different methods for determining building heights (actual versus uniform heights) and locations (actual footprints versus generalized locations). We calculated these viewsheds for a sample of 220 residential properties in the City of North Oaks in Minnesota, United States. Each viewshed was based on natural topography and view obstruction by 1850 built structures, chiefly residential and commercial buildings. To support the subsequent analysis of differences among
Elevation surface analysis
We generated four elevation surfaces for the study area. These elevation surfaces were quite similar to each other in terms of their elevation ranges (Table 2). Surfaces 1A and 1B, which both have with realistic heights [1], but actual [A] and generalized footprints [B] respectively, had identical maximum elevations. The difference in maximum elevations between the highest surfaces (1A and 2A, uniform heights [2] and actual footprints [A]) and lowest (Surface 1B) was 4.18 m. The greatest
Discussion
In their innovative work, Lake et al. (2000) speculated that the locations and shapes of buildings had a much greater impact on the accuracy of calculated viewsheds than building heights because a building of any height is likely to block a person's view. Their results indicate that while building locations significantly affect views, precise building heights are only marginally important. As a result, these authors suggest that such simplification may be acceptable in viewshed calculation.
We
Conclusions
View quality can contribute both positively and negatively to property values and to the experiences of visitors to natural areas. View quality is often reduced, however, as a result of new development. As reduced view quality can cause economic losses for communities and developers, planners, policy makers, and developers are interested in assessing the impacts of proposed land use changes on view quality. When the impact of new structures on views can be predicted, development can proceed in
Acknowledgements
This work is supported in part by the National Aeronautics and Space Administration New Investigator Program in Earth-Sun System Science (NNX06AE85G) and the University of Minnesota's College of Liberal Arts and the McKnight Land-Grant Professorship Program.
References (42)
- et al.
Environmental amenities and agricultural land values: a hedonic model using geographic information systems data
Ecol. Econ.
(2002) - et al.
Identifying and predicting visual preference in the southern Appalachian forest recreation vistas
Landscape Urban Plann.
(1994) - et al.
Impacts of urban environmental elements on residential housing prices in Guangzhou (China)
Landscape Urban Plann.
(2006) The value of trees, water and open space as reflected by house prices in The Netherlands
Landscape Urban Plann.
(2000)Human responses to vegetation and landscape
Landscape Urban Plann.
(1986)Exploring isovist fields: space and shape in architectural and urban morphology
Environ. Plann. B
(2000)To take hold of space: isovists and isovist fields
Environ. Plann. B
(1979)- et al.
Pricing residential amenities: the value of a view
J. Real Estate Finan.
(1998) - Beyer, H.L., 2004. Hawth's analysis tools for ArcGIS. Last accessed September 20, 2006 at:...
Assessment of visual qualities, impacts, and behaviors, in the landscape, by using measures of visibility
Environ. Plann. B
(2003)
Estimation of the influence of view components on high-rise apartment pricing using a public survey and GIS modeling
Environ. Plann. B
Using depth variables as predictors of visual quality
Environ. Plann. B
What's in a view?
Environ. Plann. A
The impact of parks on property values: a review of the empirical evidence
J. Leisure Res.
Algorithms for visibility computation on terrains: a survey
Environ. Plann. B
Line of sight communication on terrain models
Int. J. Geogr. Inform. Syst.
The influence of wetland type and wetland proximity on residential property values
J. Agric. Resour. Econ.
The value of an ocean view: an example of hedonic property amenity valuation
Aust. Geogr. Stud.
Estimating visual properties of Rocky Mountain landscapes using GIS
Landscape Urban Plann.
Public preference for landscape features: a case study of two Scottish environmentally sensitive areas
J. Rural Stud.
Potential impacts of changing agricultural activities on scenic beauty—a prototypical technique for automated rapid assessment
Landscape Ecol.
Cited by (60)
Reaching beyond GIS for comprehensive 3D visibility analysis
2024, Landscape and Urban PlanningThe Antenna Coverage Location Problem in the context of cattle tracking in the Austrian Alps
2023, International Journal of Applied Earth Observation and GeoinformationRapid viewshed analyses: A case study with visibilities limited by trees and buildings
2023, Applied GeographyInvestigating the time evolution of a rural landscape: How historical maps may provide environmental information when processed using a GIS
2019, Ecological EngineeringCitation Excerpt :The main data are the Digital Surface Model (DSM) distributed by Geographic Database of the Basilicata Region (under Italian Open Data License 2.0) with a pixel resolution of 5 m and 1 m of accuracy of height and a set of points on the ground. Vegetation heights, buildings and other vertical structures are already included in the DSM elevation values, in order to partially mitigate the problem of the alteration of the conditions of visibility from these elements (Sander and Manson, 2007). These points have been extrapolated from the scenic roads that cross the study area, selected by the Italian Touring Club Atlas of Roads (Touring Club Italiano, 2006).
Classification of cityscape areas according to landmarks visibility analysis
2019, Environmental Impact Assessment Review
- 1
Tel.: +1 612 625 4577; fax: +1 612 624 1044.