Chapter 3 – Data and Software Design
3.1 Solar Resource Data Development
3.1.2 Topographic Model Development
Solar Analyst requires a digital surface model (DSM) that effectively describes features at the scale level being investigated. A DSM is a raster model where each cell contains a discrete height value of a continuous surface. A DSM is unique from other topographic models as it is inclusive of all human-made and natural surface features such as buildings and vegetation. At small scales, a topographic model based on contours or a triangulated irregular network (TIN) can effectively represent how a surface’s topography affects solar flux over a large area as these features are insignificant. At large scales, however, minor features have a larger impact in affecting surface solar radiation. Developing a topographic model that is suitable for investigating rooftop solar potential requires an accurate representation of building rooftops and surrounding features that impact global solar flux. For rooftops, this includes the position, aspect, and slope of roof faces as they play a role in total incident radiation and self-shading (Dean et al. 2009).
While there are numerous methods for generating a DSM, three approaches described in literature on solar resource modeling are investigated. By briefly discussing the methodology behind each, an understanding of their benefits and drawbacks when used in the context of solar resource modeling can be developed. This aids in determining the ideal methodology for generating a DSM based on the constraints of a specific case, depending on available data. The approaches described in this section include the use of vector-based three-dimensional (3D) geometry, aerial light detection and ranging (LiDAR) scanned surfaces, and the generation of planimetric buildings by extruding footprints based on an average height value.
3D Building Models
3D building models have been commonly used for urban planning practices such as investigating land use development, viewshed analysis, air quality, and other climate issues.
The transition from physical models to computer-aided-design (CAD) based models and the
integration with GIS have further improved analytical capabilities (Hu et al, 2003). In the context of urban solar radiation modeling, 3D building models are an attractive source of topographic data due to their highly precise representation of roof structure. This also provides an opportunity for the modelled data to be cleaned up, removing spurious features and noise (Jakubiec and Reinhart, 2012).
The development of 3D building data requires real-world measurements and various levels of manual and automated processing. Hu et al. (2003) investigate a number of data sources and methodologies for generating these data, including the digitization of ground or aerial imagery, stereoscopy, and ground or aerial LiDAR scanning. These sources have the potential to perform well in producing accurate 3D building data, but have high financial and data processing costs associated with them. The exclusion of features that can obstruct sunlight such as trees is also a concern, requiring further data sources to compensate if those features want to be modelled. Furthermore, as Solar Analyst is designed to use a raster surface model, the precision of continuous vector data would be lost when the models are converted.
There has been little use of 3D building models with Solar Analyst. In a review of solar resource modeling approaches, Jakubiec and Reinhart (2012) investigated 11 solar potential mapping websites. Of these, five used Solar Analyst to model solar resource data. None of these used 3D models to generate the DSM. In this review, a 3D model dataset of over 17 000 buildings in the city of Cambridge, Massachusetts was used both with the RADIANCE / DAYSIM raytracing engine and Solar Analyst. The authors identify that 3D models are best suited to vector-based raytracing modeling approaches rather than Solar Analyst as the latter approach does not account for surface and wall reflectivity and cannot model vertical surfaces.
LiDAR
Pulse-based airborne LiDAR is a remote sensing technology used to generate digital elevation models (DEMs) by measuring the three-dimensional position of discrete points on a surface. These points are measured using a high frequency laser pulse mounted on an aerial platform. A rotating mirror is used to reflect the pulses across a swath along the flight path.
The distance from the platform to the surface is measured by the time delay of the laser. This is converted into a point with x, y, and z coordinates using on-board global positioning system (GPS) and inertial measurement unit (IMU) data to solve the point’s position trigonometrically (Jensen, 2007). A raster surface model can then be derived from these points by identifying the highest point in each cell of a grid and applying its elevation value to that cell. Various interpolation techniques can be applied to fill any holes that may exist in the data.
Since modern airbone lidar systems can measure points at a very high rate, resulting point clouds are very high density, often at less than one metre average spacing (Baltsavias, 1999).
This results in the ability to resolve not only rooftops but other features such as trees without requiring additional sources of data. Furthermore, the LiDAR scanner can return multiple elevations as the beam is not discrete to a one-dimensional point in space but rather extends over a period of time and a two-dimensional area on the surface. Commonly known as different ‘returns’, a discrete pulse may produce a number of responses as the pulse reflects off different surfaces. This is most commonly caused by vegetation such as trees where a number of returns are recorded through the canopy structure. The first return (FR) is the first response recorded and generally represents the top of the tree. Intermediate returns may be recorded for reflections in the tree canopy and the last return (LR) is the final response and typically corresponds with the surface. By excluding all returns but the last, a point cloud and resulting DSM can be generated that represents ‘leaf-off’ conditions where penetrable tree canopies are ignored (Figure 3.6).
Figure 3.6: FR and LR based DSMs showing leaf-off (right) and leaf-on conditions (left)
There are a number of drawbacks to using aerial LiDAR data as a source for generating a DSM. Steep terrain and dense buildings often block the beam’s path, resulting in ‘shadow’
areas where topographic data was not recorded. In areas with many tall buildings with a lot of glass or metallic surfaces, reflection of the beam can create false return values, creating a large amount of noise in the data. LiDAR point clouds are also very large in file size and can require a considerable amount of time and effort to process and develop a DSM from (Dean et al., 2009).
Examples of projects that use aerial LiDAR data for Solar Analyst include the New York City Solar Map website and the Solar Salt Lake Project (New York, 2012; Salt Lake City, 2012). Both of these are Web-GIS applications that let users query rooftops and calculate estimated photovoltaic potential.
Extruded Buildings
In contrast to the previous two methods, the vertical extrusion of building footprints provides a much simpler alternative to generating a DSM. Building footprints are generally more readily available and less expensive than aerial LiDAR data or imagery, and require only a basic digital terrain model (DTM) and building heights. The DTM represents the basic topography of an area, excluding any surface features, and are typically generated from contour and spot height data or the production of a TIN. In order to maintain the horizontality of each building when combined with the DTM, either they are added to a modified TIN where footprint-touching triangles are flattened, or the grid cell values are normalised for each footprint area.
While this approach is computationally simple and requires data that are typically available at a lower cost than alternative methods, there are a number of drawbacks. First, it makes the assumption that all buildings have flat rooftops. This can impact solar radiation modelling as the sun’s apparent position changes through a year and has different northern or southern maximum extents depending on latitude. Second, it makes the assumption that all buildings share a uniform shape from the base to the top. This means that some buildings that decrease in area toward the top may have an over-representation of roof area.
An example project that uses this data source for the Solar Analyst model is the City of Boston Solar Map (City of Boston, 2012). Much like the Salt Lake City and New York City applications, Solar Boston lets users select rooftops or draw custom areas to query PV feasibility.