You are on page 1of 46

Urban texture and radiation exchange

Carlo Ratti , Nick Baker , Koen Steemers


a

SENSEable City Laboratory, Massachusetts Institute of Technology, 77 Massachusetts Avenue,

Cambridge, 02139 MA USA


b

The Martin Centre for Architectural and Urban Studies, Department of Architecture, University

of Cambridge, 6 Chaucer Road, Cambridge CB2 2EB, UK

Address for correspondence: Carlo Ratti, SENSEable City Laboratory, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, 02139 MA USA; e-mail: ratti@mit.edu

ABSTRACT: This paper explores how raster-based models of urban form (DEMs) and software algorithms derived from image processing can successfully be used to compute radiative parameters. It starts by reviewing an innovative and fast algorithm to cast shadows over extensive urban areas. This algorithm is then elaborated into a more complex function to calculate the urban sky view factors, a well-established parameter which is central to the comprehension of the urban microclimate. From the sky view factor the analysis extends to some estimates of illumination falling on the urban surface under different sky configurations, such as the uniform diffuse sky and the standard CIE overcast sky. The distribution and variation of illuminance values both outside and inside the buildings is calculated, allowing the comparison of the environmental characteristics of different urban textures. Results suggest that the analysis of DEMs could open the way to a new paradigm for urban analysis.

1. Introduction
Sunlighting as form-giver in architecture (Lam, 1986). Or, more poetically Sun rhythm form (Knowles, 1981). These are the titles of two books, amongst many publications on solar energy in architecture and planning, which reflect a similar concept: the importance of sunlight and daylight in shaping the built environment.

This very idea traversed the whole history of architecture and survived even the otherwise iconoclastic Modern Movement. For instance, the third Congrs International dArchitecture Moderne (CIAM), held in Brussels in 1930, addressed, among others, the problem of accessibility of daylight in modern buildings. An influential speech was delivered there by Walter Gropius, who tried to demonstrate analytically the relationships between building height, open space, sunlighting and orientation .
1

Le Corbusiers attitude towards sunlight and daylight was similarly central to his architecture, though possibly more poetic. In the city of Chandigarh, the capital of the Indian state of Punjab which he designed from 1948, the middle of the governmental Capitol complex is occupied by a Tower of Shadows. Often described as a cosmic sign alongside the monumental symbols of secular power, the tower is an airy structure entirely clad with brise-soleils, which cut the sun in each direction from the internal space. In Corbus words, this was to demonstrate that "that one can control the sun on the four cardinal points of an edifice and that one can play with it even in a torrid country and obtain lower temperatures (Chandigarh Planning & Architecture, 2001). This play was to be repeated on most of Chandigarhs faades through the sculptural use of concrete brise-soleils, supposedly contributing to the overall environmental well-being of the town (this fact was recently confirmed by monitoring carried out by Faruqui Ali, 1998).

The pressure on the use and control of sunlight and daylight in architecture will possibly be greater in the future than it is today, due to the increasing recognition of the resulting energy1

His findings were later reworked by Martin and March (1972), p. 71 ff.

saving implications. In the words of Littlefair (2000) solar energy in its various forms will be more and more important in the buildings of tomorrow. In fact, increasing importance is placed in public policy, on the utilisation of renewable energy through solar water panels or photovoltaics (PVs). The European White Paper on renewable sources of energy (EU, 1997), for instance, envisages 500,000 1KW building-integrated PV systems installed in Europe by 2010. Subsidised action is currently being undertaken by countries as diverse as the USA and India and is expected to have a considerable impact on the overall energy budget of cities.

In addition to energy considerations, sunlight and daylight have implications for the quality of urban life. Baker (2000) argues that rich and varied urban environments, with a close relationship with nature and natural light, generate a higher degree of satisfaction and comfort than artificial environments. Quantitative measurement to support this thesis is provided by Nikolopoulou (1998) and Nikolopoulou and Steemers (2000), who carried out a general study on urban comfort in outdoor spaces. A number of medical studies also exist, suggesting that the reduction of daylight in urban environments might have serious health consequences on the population, such as loss of visual acuity .
2

Despite the well-established importance of sunlight and daylight conditions in the urban environment, tools for measuring and simulating them over extensive portions of cities are inadequate. This was noted by Compagnon (1999), who referred in particular to three limits: the simplified approach in modelling the sky-vault, the neglect of inter-reflections amongst buildings and the difficulties of performing simulations on extensive urban areas, which can occupy several hectares.

The latter task namely the difficulty of modelling geometry on large portions of cities is partly related to computing difficulties. Traditional models work at the scale of the building but fail on larger areas because of too much vectorial complexity. This paper aims to test the possibility of using a very simple raster model: the so-called Digital Elevation Model (DEM), which is shown in Figure 1. The DEM is a compact way of storing urban 3D information using a
2

The extensive use of artificial lighting might have lessened the importance of the considerations below. However, in the early days of studies on daylight, research showed that reduced illumination causes a lack of visual acuity. In the preparation of The hygiene of the eye in schools (1886), Hermann Cohn (1838-1906) studied the visual acuity of 10,600 children in relation to lighting conditions. He noted the number, size and position of windows and the presence of urban obstructions. By standardising other factors to which children were exposed, so that findings

2D matrix of elevation values; each pixel represents building height and can be displayed in shades of grey as a digital image. The analysis of DEMs with image processing techniques has already proven to be an affective way of storing and handling urban 3-D information, and being very conducive to a number of urban analyses (Ratti and Richens, 2004). Could it contribute to the study of daylight and sunlight conditions in cities? This paper will address that question by describing a nmber of algorithms that have been written using the Matlab software (2004), a well known package for doing numerical computations with matrices and vectors. Matlab's extensive matrix capabilities are supplemented by different toolboxes, among which is the image processing toolbox, with elaborate graphics outputs.

A number of image processing techniques are developed below and applied to three case study sites in London, Toulouse and Berlin. Standard measures used to quantify the luminous flux will be examined in urban areas, as well as more general parameters that characterise the radiative exchange between surfaces (in this case the urban surface). Among these parameters, the sky view factor has a prominent role; it is a well-known urban climatology variable, whose easy calculation might prove extremely beneficial to urban planning.

Aside from its computational interest, this paper also presents some comparative results on different urban textures as a preliminary insight into the interrelationship between urban texture and radiation.

The study of the impact of urban geometry on the penetration and distribution of sunlight, daylight, etc. might go some way to clearing up the confusion that occurs in present scientific literature. In a recent publication aimed at informing solar design of cities, for instance, it was found: Compared to open country, built urban sites have a larger area of exposed surfaces per unit area of ground covered. Because of this larger area, potentially more solar radiation can be collected on a built urban site than on a flat open terrain, especially in winter . This assertion is clearly not true: the total radiation falling on a unit area of ground built or un-built is clearly the same. Furthermore, solar energy in built areas can only be collected with more difficulty than on flat open terrain, because of the effects of overshadowing and the changing
3

could be attributed generally to school illumination conditions only, he found that the closer and higher the urban obstruction, the higher the percentage of myopics among the pupils. 3 the authors name could be omitted, since this statement is used here in a symptomatic way.

distribution of radiation in space and time. This is caused by the complex patterns produced by urban geometry. We will try to quantify them below.

250

200

150

100

50

Figure 1. Case study site in central London in the Digital Elevation Model DEM format (left) and its axonometric view (right)

2. Context
Before embarking on the derivation of radiative parameters in cities, it is useful to review some basic terminology: Radiant flux is the power emitted, transmitted or received in the form of electromagnetic radiation [Unit: W]; Radiance (in a given direction, at a given point of a real or imaginary surface) is the radiant flux in a given direction per unit solid angle, per unit area perpendicular to that direction (Figure 2). Expressed as a formula:

L=

d -2 -1 [Unit: Wm sr ] dx cos d

where d is the radiant flux transmitted by an elementary beam passing through the given point and propagating in the solid angle d , containing the given direction; dx is the area of a section of that beam containing the given point; is the angle between the normal to that section and the direction of the beam; Irradiance (at a point of a surface) is the quotient of the radiant flux d incident on an element of the surface containing the point, by the area dA of that element [Unit: Wm ].
-2

Irradiation is the product of irradiance and time, i.e. surface density of the radiant energy received [Unit: Wsm ].
-2

The same amount of electromagnetic radiation, however, has substantially different effects on the human eye according to its wavelength. No effect at all is recorded outside the so-called visible band, while maximum efficacy is noted around a wavelength of 555 nanometres. This fundamental mechanism of human vision must be taken into account when studying lighting in architecture and is dealt with in building science with the introduction of photometric parameters. These correspond exactly to the radiative parameters defined above, except that the radiant flux is weighted by its spectral response to the human eye. In other terms, radiation is evaluated according to its action upon the so-called CIE (Commission Internationale de lEclairage, 2004) standard observer. The photometric quantities that correspond to the radiant flux, radiance, irradiance and irradiation are named respectively:
-

Luminous flux Luminance Illuminance Illumination (not very common, but used in this paper)

When radiant or luminous flux falls on a surface it is partly reflected, partly absorbed and partly transmitted (the latter if the medium is semitransparent). The surface is therefore characterised by three adimensional ratios, namely the reflectivity , absorptivity and transmittivity , defined as fractions of the total incident radiation. These ratios are usually a function of the wavelength and of the angle of incidence of the incoming radiation. However, the following equation holds:

+ + = 1
An alternative name for reflectivity, ratio of the reflected to incident radiation, is albedo, which is commonly used by the climatology and earth sciences community.

It is now possible to review some methods which are found in the literature for assessing the distribution of solar energy in urban areas. The most used quantities are irradiance and illuminance either of these being chosen according to the focus of the analysis: solar energy or lighting. Each of these two terms is then separated into two contributions: the direct

component, coming directly from the suns disc, and the diffuse component, received from the whole sky hemisphere.

The direct component requires knowledge of shadowing in the city. It is zero on all surfaces in shadow, while it has a value which is a function of the angle of incidence on all lit surfaces. In fact, all lit surfaces at a certain time of the day receive from the sun the same amount of incident solar radiation (per unit area perpendicular to the radiation beam).

This simple problem of detecting areas in shadow might prove difficult in urban areas with complex geometry. A review of methods to quantify solar access can be found in Littlefair (1998). From a single faade, traditional graphic methods are generally used, based on sunpath diagrams. A wide variety of them exist in the literature; in addition to the indication of sun and shadows, in some cases they allow the estimation of the energy that can be collected at any point.

Despite their convenience, graphic diagrams can be used only on one faade at a time. Furthermore, they require the plotting of the urban geometry against the sun path, and this process can be time-consuming. Consequently, their use in architectural design is usually limited to the building scale, or to repetitive and simplified urban layouts. It cannot be easily extended to urban areas of complex geometry.

In the latter case, the use of computer software becomes imperative. To date, this is not a problem in terms of computational power: purpose-built programs for architects and urban designers are readily available. An example is Shadowpack (Peckham, 1985 ), developed at the CEC research centre in Ispra, Italy, which utilises its own CAD-type program to generate a layout and then allows the evaluation of the amount of solar radiation received by each surfaces. Other software developed specifically for shadow calculations is Townscope (Laboratory of Architectural Methodology, 2001), Sombrero (Niewienda and Heidt, 1996) and Shading (Yezioro and Shaviv, 1994). Furthermore, it should be noted that today most CAD packages incorporate their own tools for shadow casting.

The most serious limitation of all these computer methods is that they require full 3-D urban models in vectorial form, something that can be extremely costly; in this sense, DEMs have a 7

lot to offer. Another limitation is that they are not always very advanced in the assessment of the diffuse radiative component falling from the whole sky-vault. According to Littlefair (1998), this is their major weakness: some models do not include diffuse radiation at all; others include it but do not assume it is reduced by obstructions; one or two include it, but assume the sky is isotropic.

Progress in software development is of course happening at a fast rate: for instance, Compagnon and Raydan (2000) report on a successful coupling of a well-known package for computer rendering, namely Radiance (Ward Larson and Shakespeare, 1998), with a luminance model of the sky and a 3-D vectorial model of the city. They derive the total amount of energy falling on buildings, eventually plotted as a distribution of faades and roofs according to their illuminance or irradiance (their sky model is based on a method developed by Perez et al., 1993, which allows the computation of luminance values on the sky-vault from statistical observations). Although very satisfactory from a theoretical point of view, their approach still has limitations when dealing with extensive urban areas of complex geometry.

To conclude this review of literature let us turn to a completely different approach, coming from the geosciences. There, several techniques have been developed to model solar radiation incident on topographic surfaces from raster databases (such as DEMs). An introduction to these can be found in Burrough and McDonnell (1998). Originally, algorithms were developed to derive shading relief maps automatically maps that aim to ameliorate the 3-D comprehension of cartography by portraying shading. A fictitious sun position is assumed, from which the angle of incidence of light on slopes is derived.

In more recent years, simple shading algorithms have been extended to model solar radiation intercepted by complex topographic surfaces an exercise which has a wide range of applications in agriculture, forestry, hydrology, ecology and design. In doing this, shadow casting subroutines have been integrated to model the self-shading of topography. This is usually done by calculating on each pixel the horizon angle in a given direction, and by comparing it with the actual sun altitude. More details can be found in Dubayah (1992), Dubayah and Rich (1995), Kumar et al. (1997).

A similar approach will be followed below for urban areas, where techniques developed in the geosciences have not yet been used. However, due to the discontinuous nature of urban DEMs, it should be noted that a simple transposition of algorithms developed in the geosciences to urban areas is not possible.

Figure 2. Symbols used in the definition of radiance. Image from Sillion and Puech (1994).

3. How to calculate shadows on a DEM


Let us start with the simplest problem: shadow casting on a DEM. Even such a simple problem can become challenging in urban areas. Traditional vectorial model fails, developed to work at the scale of the building, fail because of excessive geometric complexity.

Shadow casting is the first macro that was written, at the beginning of this investigation into potential uses of image processing DEMs for urban analysis. The approach is to compute shadow volumes, that is, the upper surface of the volume of air that is in shadow. As explained in Ratti and Richens (2004), this can be done by repeatedly shifting the DEM and reducing its height.

This algorithm is very simple and impressively fast (it allows processing acres of city at a time, something unthinkable with traditional geometric models; Figure 3). It is used below as a basis for a number of more complex analyses.

1 0.9 50 0.8 0.7 100 0.6 0.5 150 0.4 0.3 200 0.2 0.1 250 50 100 150 200 250 0

Figure 3. Shadow casting on the London DEM, sun position: azimuth=30, altitude=30.

50

50

50

50

100

100

100

100

150

150

150

150

200

200

200

200

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

50

50

50

50

100

100

100

100

150

150

150

150

200

200

200

200

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

Figure 4. Shadow casting in central London on 31 December, hourly intervals.

4. Animating the sun: dynamic calculations and solar envelopes


The algorithm described above computes shadows for an arbitrary lighting angle; the next stage is to add a procedure to calculate shadows from the sun for any given latitude, time of year, 10

and time of day, using the usual astronomical formulae. This allows the possibility of performing simulations for a whole day or number of days. Results are shown for instance in Figure 4, which represents the patterns of sun and shadows in central London on 21 December.

The task of converting latitude, time of year, and time of day into altitude and azimuth of the sun has been carried out in this case by a subroutine. This subroutine however elaborate it might be is not novel, as it codes well-established astronomical data. Therefore it need not be described in detail here; references on the formulas used can be found in Muneer (1997).

The first operation made possible by this subroutine is the dynamic representation of shadowing. Different images can be produced and subsequently animated in Matlab, creating short movies which give the architect or planner an idea of the shadowing conditions of an urban site during an entire day.

From these images, simple quantitative parameters have been defined by Steemers and Ratti (1999a), by the analysis of all the hourly frames of a single day. Summing all black and white shadowing images, each made of 0s and 1s, it is possible to obtain grayscale images which have values in the range 0..n, where n is the number of sun positions considered for the day (equivalent to the number of hours of sun, if the sun positions are taken at hourly intervals). The resulting image portrays in an elementary way the number of hours of shadow for each pixel (this value is simply given by the resulting value of that pixel). Contours can be added to improve visualisation: a standard function in Matlab allows the detection of value changes on images such as between a pixel with x and x+1 hours of sun.

A mean shadow density can also be defined as the average number of hours of shadow on a certain region: its value is simply the mean of all pixel values in that region.

The above considerations lead us to a more articulated concept: that of the solar envelope. This was first introduced by Knowles (1981, 2000, 2004) as the geometric envelope that describes the volumetric limits of building that will not shadow surroundings at specified times. In other words, given a certain site in an urban context, the solar envelope defines the maximum built height that can be reached on that site without compromising the neighbouring buildings solar

11

accessibility. This is defined as the minimum number of hours of sun, during critical periods of the day and the year.

The imaginary boundaries specified by the solar envelope can act, according to Knowles, as an operational tool to inform urban design and planning. They open new aesthetic possibilities in architecture and urban design, by bringing rhythm and a closer sense of nature into our buildings and urban spaces, and could be used by public administration as an instrument of zoning to regulate urban development. In some countries bylaws already acknowledge this fact: for instance, in the Netherlands, every living room should receive a minimum of three hours of direct sunlight on its faade every day, from 21 March to 21 September (MVRDV, 1999).

The emphasis on solar rights considerations in the design of urban fabric has recently led to the extension of the solar envelope concept by Capeluto and Shaviv (2000). They make the distinction between solar rights envelope and solar collection envelope:

1) the solar rights envelope defines the maximum height of a building in order not to violate the solar rights of any of the neighbouring buildings during a given period of the year; it is basically equivalent to Knowles definition;

2) the solar collection envelope defines the lowest possible surface to locate windows and solar collectors so that they are not obstructed by neighbouring buildings, during a given period of the year. This is, in a certain sense, a symmetrical parameter to the solar right envelope: it describes the overshadowing of a neighbourhood on a given site.
4

These two envelopes can be interpreted as the upper and lower boundary of a solar volume , which represents the portion of space where new developments could be allowed without reducing the solar access of the neighbouring buildings and guaranteeing sufficient solar access.

The subroutine that governs shadowing on a DEM, together with the dynamic calculation of sun positions, allows the detection of solar envelopes. For instance, it will be shown here how this can be done in the case of the solar collection envelope. Let us consider a fictitious open site

12

in central London depicted in Figure 5 and let us ask the following question: which is the boundary surface where there are at least n hours of sun on 21 March?

The algorithm is particularly simple and elegant. It is based on running the shadowing volume program (see section 3 above) on a DEM, for all n hourly sun positions on 21 March. Each of these n shadowing volume images represents the height of the separation surface between light and shadow at a certain time. If on each pixel of the DEM, these n heights are sorted out by increasing values, n 3-D surfaces one on the top of the other are obtained. Each of these images can be read as a 3-D map (or DEM) which define the portion of space where there are 1, 2, 3,, n hours of shadows. In other words, each of these images represents in the DEM form the solar collection envelope which corresponds to the required condition of n hours of sun. Figure 6 represents some of these surfaces on 31 December on the urban site in central London. As expected, the height of the solar collection envelope increases with the increase in the number of hours of sun.

In its standard form the solar envelope criterion has a limit: it does not take into account the angle of incidence of sunlight. In other words it is just binary, 0 or 1, sun or shade. An hour of sun at midday and one at dawn count the same, although they are very different from an energy point of view. This limitation, however, could simply be overcome by introducing a weighting function based on the sun altitude.

The information embedded in images such as the one shown above have direct applications in architectural design. This aspect was investigated, amongst others, by Birks (2000), who has been using the above shadow casting algorithms to translate images in proposals of building form. He concluded that software of this kind could play an active role in generating design.

His analysis drew on a wider approach to the synthesis of architectural and urban form, based on the use of the datascapes technique. This has been theorised by the Dutch practice MVRDV (1999), as an observation technique which should let the limits and gravities inherent a certain design situation emerge. Constraints (such as shadowing conditions, noise level regulations, etc.) are laid onto space and used to inform the design process.

It should be noted, however, that when the solar collection condition is particularly restrictive, the solar collection envelope might be higher than the solar rights envelope. In this case the solar volume does not exist.

13

Although MVRDVs design proposals often look like far too naive responses to context constraints as when they take plans of noise contours produced from a motorway as faade generators for architectural shape their analysis proves that solar envelopes and similar techniques have potential for exploring primary design solutions.

20 40

60 80 100 120

140 20 40 60 80 100 120 140

Figure 5. Experimenting with solar envelopes: the image represents a fictitious open site in central London.

20 40

20 40

20 40

60 80 100 120

60 80 100 120

60 80 100 120

140 20 40 60 80 100 120 140

140 20 40 60 80 100 120 140

140 20 40 60 80 100 120 140

Figure 6. Experimenting with solar envelopes: the image represents the height of the surface where there are

1, 3, 6 hours of sun respectively (values for 31 December); on the far right the surface where there are 6 hours of sun is plotted in axonometric.

14

5. The sky view factor


All the algorithms presented above shadow casting and its derivatives are based on the following principle: fix an arbitrary sun position and from this position trace back rays to the city. The DEM pixels that are intercepted by these rays are illuminated by the sun (and therefore white). All other pixels are in shadow (black).

At the end of the previous section, however, another approach emerged, based on the following principle: fix a point on the DEM and count the number of previously defined sun positions that can be seen.

This method can have very interesting applications, such as determining the amount of sky that is visible from a given pixel. The mechanism is the following: spread a number of sun positions on the sky-vault, cast shadows each time on the DEM and count the number of whites and blacks obtained on each pixel. If the sun positions are distributed uniformly on the sky-vault, the count of whites on a given pixel divided by the total number of sun positions taken into consideration is directly the solid angle of view of the sky (with a proportionality factor of

2 ).
By a slight refining of this algorithm, a very interesting parameter, often used in urban climatology, can be derived: the sky view factor (also called configuration, shape or form factor). This parameter is similar to the solid angle of view of the sky, although two weighting coefficients are applied in order to weight different parts of the sky-vault in different ways. Expressed as a formula, the sky view factor FdAi sky reads:
cos i cos j

FdAi sky =

sky

R 2

dA j

The integration is performed onto all elemental surfaces dA j which compose the sky-vault; dAi is an elemental surface of the city, i and j are the angles between the vectors normal to
dAi and dA j and the line connecting them, whose length is R (Figure 8).

15

In fact, the sky view factor is simply the sum of cosine weighted elemental solid angles on a whole hemisphere. The reason for this weighting comes from heat transfer theory, where the view factor was first introduced to model radiative exchange between surfaces. Given two diffuse surfaces i and j , the view factor simply defines the fraction of the total radiation
5

leaving surface i which is intercepted by surface j . A more complete discussion of this aspect can be found in textbooks of heat transfer theory such as Incropera and DeWitt, 1990.

In practice, the sky view factor is quite difficult to derive. Analytical solutions and numerical tables are available only in a number of simple geometrical configurations. In the urban context, Glenn and Watson (1984) derived an analytical expression to determine it for standard canyon arrangements. In more general cases, approximate graphic methods can be used, based on the manual or computer-based analysis of zenithal fish-eye lens photographs (see for instance Steyn, 1980; Rich, 1989; Chalfoun, 1998, who used it in the assessment of outdoor thermal comfort; Brown et al., 2001) as well as numerical techniques. The latter, however, can be quite time-consuming.

Let us look at how the sky view factor can be calculated on a DEM. As anticipated, the algorithm is based on the repeated application of the shadow algorithm. We simply compute the shadows for a large number of light sources, distributed over the sky, and for each pixel count the number of times they are in light. So if we use 1000 samples, any pixel whose count is 1000 can see all the sky and has a sky view factor 1, while a count of 0 means that it cannot see the sky at all (Figure 7).

To get meaningful results, according to the above definition of view factors, it is necessary to distribute the sample points over the sky in the correct manner. If a uniform distribution is used, then what is measured is not the view factor but the solid-angle of sky visible from each point. In order to take into account the cosine correction, the density of samples must be higher at the zenith than towards the horizon. It is easy to see that the correct distribution to compute the view factor from the city to the sky can be obtained by spreading points evenly over a unit circle in the horizontal plane, and then projecting up to a unit hemisphere.

A diffuse (or lambertian) surface transmits or reflects light adhering to Lambert's cosine law. This law states that the reflected or transmitted luminous intensity in any direction from an element of a perfectly diffusing surface varies as the cosine of the angle between that direction and the normal vector of the surface. As a consequence, the luminance of that surface is the same regardless of the viewing angle.

16

The number of points being very large, it is possible to chose a uniform and random distribution of points on the circle. This will also avoid a well-known problem in image processing, namely the emergence of patterns of interference. The following question, however, should be answered: how can a uniform distribution of points on a hemisphere be achieved?

The Matlab function RAND produces random entries chosen from a uniform distribution on the interval [0,1]. An elemental area in polar co-ordinates is written:

dA = rdrd
If a uniform distribution is taken on r and , the resulting distribution of points on the circle will not be uniform. The density of points will be higher towards the centre and lower at the periphery, following an inverse proportionality to r (Figure 9).

In order to have a uniform distribution of points on the circle a function r = z must be used, where z follows a uniform distribution. A proof of this fact is given in the footnote , results
6

are shown in Figure 9.

From this uniform distribution of points on a disc of radius 1, the positions of the sun are derived by projecting on a hemisphere. The iterative application of the shadow-casting routine allows the calculation of the sky view factor. The higher the number of calculations, the more accurate the results. 100 iterations are usually enough to produce satisfactory sky view factors, such as those for London, Toulouse and Berlin shown in Figure 10.

The proof can be given as follows: Make a change of variable so that dA becomes independent from r . If a new function z = z ( r ) is chosen, this implies:

rdr = kdz

Where

is a constant of proportionality. The above equation can be integrated:

rdr = kdz
and gives:

z=k

r2 +c 2

Any function z will satisfy our condition, so we choose z = r 2 . This gives:

r= z

17

1 0.9 50 0.8 0.7 100 0.6 0.5 150 0.4 0.3 200 0.2 0.1 250 50 100 150 200 250 0 250 50 100 150 200 250 200 150 100 50

2.5

1.5

0.5

5 4.5 50 4 3.5 100 3 2.5 150 2 1.5 200 1 0.5 250 50 100 150 200 250 0 250 50 100 150 200 250 200 150 100 50

10 9 8 7 6 5 4 3 2 1 0

Figure 7. Viewing the sky from the city: the process is based on the iterative sum of shadow images. A similar

procedure, based on a multi-sun heliodon, had been used by Lionel March in the 1970s at the Martin Centre to add shadows on physical models. The figures above show results for the 1st, 3rd, 5th and 10th iteration; from the 100th iteration (sum of 100 shadow images) results are very accurate; the figures in Figure 10 were stopped at the 1000th iteration.

Figure 8. Symbols used in the sky view factor formulas, taken from Incropera and De Witt (1990).

18

Figure 9. Distribution of points on a disc. The image on the left shows a uniform distribution of points on

polar co-ordinates r and

; the image on the right shows a uniform distribution of points per unit area.

19

1 0.9 50 0.8 0.7 100 0.6 0.5 150 0.4 0.3 200 0.2 0.1 250 50 100 150 200 250

1 0.9 50 0.8 0.7 100 0.6 0.5 150 0.4 0.3 200 0.2 0.1 250 50 100 150 200 250

1 0.9 50 0.8 0.7 100 0.6 0.5 150 0.4 0.3 200 0.2 0.1 250 50 100 150 200 250

Figure 10. Sky view factors in central London, Toulouse and Berlin obtained after 100 shadowing

calculations.

20

6. The meaning of sky view factor: the urban heat island


The view factor from the city to the sky can be intuitively understood as a measure of the openness of the urban texture to the sky. Its interest, however, lies in its capacity to explain a number of climatological phenomena in cities, such as the urban heat island.

The urban heat island is a well-documented example of modification of the atmosphere by urbanisation, which is accompanied by an increase in air temperature. While the trend toward higher temperatures in cities was first noticed by Howard (1833) during the first half of the 19
th

century, the term urban heat island appeared in the scientific literature towards the end of the 1950s. It refers to the characteristic shape that isotherms assume in urban areas, an island of higher urban temperatures can be recognised in the sea of lower rural temperatures.

The features of the urban heat island vary both in space and time, as a result of meteorological, locational and urban characteristics. The maximum temperature differences are found with clear sky conditions and wind calm, while they diminish with cloudy and windy weather. The urban heat island peaks are generally in the core of the city.

A well known formula was introduced by Oke (1981) to relate the maximum heat island intensity between urban and rural sites and the sky view factor:

Tmax urban-rural = 15.27 13.88 sky


where sky is the view factor from the middle of the canyon floor to the sky and Tmax urban -rural the maximum air temperature difference measured between that canyon and a rural site . The formula was obtained by results from physical scale models representing rural and urban surfaces, to mimic the passive radiative cooling of these environments following sunset on a calm and cloudless night. It was also verified using experimental data from a number of cities.
7

It might be useful here to review the physics that is behind the dependency between urban heat island and sky view factor. Two processes should be mentioned:

21

1) The reduction of long-wave radiation from the street canyons. As has been seen (see section 5 above), the concept of sky view factor was originally introduced in heat transfer literature to model the radiative exchange between surfaces. The higher the sky view factor in a street, the higher the radiative exchange between that street and the sky. Because of its low temperature, the clear sky is a very important energy sink in the infra-red region of the electromagnetic spectrum. Therefore, a high view of the sky means elevated heat losses, especially during the night when there is no solar radiation.

In densely built areas the ground screening by flanking buildings reduces the sky view factor. An increasingly larger proportion of the cold sky is replaced by the relatively warm sides of buildings, reducing radiative losses. This phenomenon is in accordance with the occurrence of the urban heat island at night-time, when the long-wave heat loss is very important in the energy balance of cities.

2) The increase of short-wave energy absorption due to multiple reflections. The city surface is a mixture of vertical and horizontal elements that create urban canyons; this has consequences on the amount of solar radiation that is absorbed in urban areas. The physical mechanism can be described as follows: when solar radiation falls on a flat surface, it is partly absorbed and partly reflected to the whole hemisphere. When solar radiation falls in a urban canyon part of the reflected radiation will hit the canyon surface again. This results in a sequence of multiple reflections which increases the total amount of solar radiation absorbed (Figure 11).

The phenomenon of trapping solar radiation occurs on any crenellated or textured surface and has been discussed at length in the scientific literature. All approaches show that in general the albedo of a crenellated surface is lower than that of a flat plane composed of the same material. Numerical simulation was carried out by Aida and Gotoh (1982), who found that the urban reflectivity decreases as the urban irregularity increases. Their overall results were confirmed by physical model experiments carried out by Steemers et al. (1997). Reflectances on the London, Berlin and Toulouse case studies, whose DEM models are examined here, were measured. Findings can be summarised as follows: In broad terms the
7

The definition of an urban temperature is quite difficult. Okes formula (1981) is based on air temperature measures

22

study has shown that, for all paint reflectances, urban forms absorb more sunlight energy than flat planes Furthermore, it is apparent that the complexity, or occlusivity [it might be said view factor] of the urban texture affects the amount of light that is absorbed. Thus the reduction of light reflected from the modelled surface (compared with plane surface) is much greater for Toulouse than for London, which in turn is greater than for Berlin. (Steemers et al., 1997).

The combined effect of the two factors above the reduction of long-wave radiation from the street canyons and the increase of short-wave energy absorption due to multiple reflections is at the basis of the urban heat island. Both factors are well taken into account by the value of the sky view factor .
8

Taking the results for our three urban case studies (average view factor in the urban canyons, Figure 12), it can be seen that they rank in the order London, Toulouse and Berlin, with average values of the sky view factor at the street level of 0.529, 0.646 and 0.720 respectively. By applying the above formula by Oke (1981) it is possible to predict values of the maximum urban heat island, which are respectively 7.9, 6.3 and 5.3C a considerable difference, which is simply produced by urban morphology.
9

These results, however, pose a series of questions: How good is it to have a urban heat island in cities? Is there an optimum value for it, which maximises, say, energy savings or thermal comfort? And, in this case, would designers be able to implement urban design strategies in order to modify the sky view factor and therefore the intensity of the urban heat island ?

The answers to all these questions are likely to be climate dependent. Little doubt exists that in cold climates the urban heat island has a beneficial effect in energy saving. This is clearly stated by Oke (1981): Further work on the effects of geometry could also be of practical value. For example in mid- and high-latitude cities fostering the winter heat island effect can reduce

taken near the ground in the middle of urban canyons. 8 It is noted here that the sky view factor is a rather coarse parameter at the small scale, as it is orientationinsensitive and unable to predict precisely where solar absorption occurs. However, it has a fundamental role to play at the urban level, where the focus is on spatially averaged parameters. 9 By using average values of the sky view factor at street level we are slightly underestimating the mid-street sky view factor used in Okes formula. The approximation, however, is negligible. Furthermore the sky view factor values in Okes formula were derived from measured height-to-width ratios with a little geometrical simplification, which tends to compensate the approximation.

23

space heating requirements by 5-15 per cent (Summers, 1974) and in cities subject to excessive heat in summer minimising this effect could save on energy used for air conditioning and perhaps even reduce fatalities due to heat stress (Clarke, 1972). If the role of geometry is as important as suggested in this study it should be regarded as a fundamental input in urban design.

In another paper, the same author suggests that for mid-latitude cities with no overheating problems the minimum acceptable H/W ratio of the urban canyon would be approximately 0.4 (which for an infinite canyon corresponds to a sky view factor of 0.78). This value would maintain about one third of the heat island potential for a given city (Oke, 1988).

The interpretation of the effects of the sky view factor and the subsequent urban heat island in hot climates is more controversial. In fact, minimising the urban heat island would suggest sparse and scattered urban developments. However, this is the opposite of what is commonly found in vernacular architecture in hot arid regions (such as Arabic countries). There, a courtyard-based and compact urban matrix, which tends to minimise the sky view factor, is mostly adopted (Figure 13). Would this be a case of anticlimatic solution, i.e. of irrational response to climate (Rapoport, 1969)? Or could it be justified in urban climatology terms?
10

Let us consider in more detail the urban heat island phenomenon. This is often described using average temperatures: in the definition of Givoni (1998) on average the diurnal temperature, in a densely-built urban area, is warmer than the surrounding open (rural) country. The average however, does not take into account peaks. The urban heat island usually presents two of them: a maximum during the night and a minimum during the day, often described as an urban cool island.

In hot arid climates the unmodified night-time temperatures are usually low, and an increase in them would probably be accepted if it can concomitantly alleviate extreme temperature stress during the day. Furthermore, urban pedestrian comfort is not only based on air temperature,
11

In fact, rarely is an architectural response to climate completely rational or completely irrational. More often it is a question of balance, where some design variables are privileged in comparison to others. For instance, an urban form could be optimised in terms of thermal mass and not in terms of sky view factor. It will be shown below that the latter is not the case for the dense vernacular developments which are found in hot-arid regions. 11 It should also be noted that in some hot-arid countries, such as Morocco, people have developed the habit of sleeping on building roofs, to maximise radiative losses.

10

24

but also on radiative exchange: this benefits from low sky view factors, which mean an increase in direct shading and a reduction in reflected radiation.

Quantitative descriptions of these phenomena, which justify the adoption of compact urban structures in hot arid climates, are given by Pearlmutter et al. (1999), who simulated the energy exchange between a cylindrical body, representing a pedestrian, and the canyon environment. Pearlmutter (1998) analyses new Israeli towns, planned according to the garden city model imported from Europe, and concludes that they are climatically inappropriate, as they lack the thermal moderation effect of traditional compact developments.

A recent study of the city of Fez, Morocco, based on field measurements, confirms the same results. Rosenlund et al. (2000) monitored temperatures in two districts of the city, associated with different housing types: a traditional one, based on the compact assembly of buildings and a courtyard structure, and a more recent one, based on modern two- to three-storey houses arranged alongside wide streets. Temperature results are markedly different in the two sites. In the traditional district temperatures are higher during the night, but during the day a favourable cool island appears. Overall conditions are more stable than in the modernist development, with the tendency to smooth down maximum and minimum temperatures.

Interesting data have also been obtained by Perez-de-Lama and Cabeza (1998). They examined the environmental performance of patios in Seville using the sky view factor (which they call configuration factor). They focussed on maximum temperatures inside the patios during the day and concluded that in all cases but one maximum temperatures measured in the set of patios have been below the reference temperature considered. They also suggest that patio cooling performance in summer is proportional to the envelope to plan surface ratio. The smaller the opening to the sky in relation to the general envelope surface (i.e. the average sky view factor) the better the performance is.

These conclusions reassuringly show that vernacular urban structures found in hot arid regions are well adapted to the climatic context at least as far as air temperature and the urban heat island are concerned.

25

Figure 11. Schematic distribution of solar radiation in open flat country (top), built-up area with height-to-

width ratio of about 1 (middle) and high-density urban area with height-to-width ratio of about 4 (bottom); from Givoni (1998). London Average sky view factor at ground level Predicted increase in temperature [C] using the formula:
Tmax urban-rural = 15.27 13.88 sky

Toulouse 0.646

Berlin 0.720

0.529

7.9

6.3

5.3

Figure 12. Data for London, Toulouse and Berlin.

Figure 13. Courtyards in central Marrakesh, from Rudofsky (1964).

26

7. The meaning of sky view factor: urban illumination


Aside from its ability to explain the urban heat island, the sky view factor has a further meaning: it is proportional to the amount of light that would fall on a city under a uniform sky i.e. a sky that has the same radiance in all directions . The sky view factor can therefore be considered as a preliminary indicator of urban daylight conditions: when it assumes the values 0 and 1, it represents, respectively, nil and maximum illumination (where maximum illumination means an unobstructed view of the sky hemisphere).
12

Using this key, the grayscale views of London, Toulouse and Berlin shown in Figure 10 are easily interpreted. Small courtyards receive little light from the sky and are therefore dark. Roofs have an almost unobstructed view of the whole sky hemisphere, and are therefore very white. Street junctions are particularly interesting, because they make evident the linear additive nature of the radiation processes: their luminous intensity is approximately double of that of streets as two strips of the sky are visible at the same time.

In architectural and urban design, however, the greatest interest is not in illumination values in the streets, but on those on faades. These can be calculated on a DEM by cutting it at different heights (for instance each 3 metres, a value that approximates a standard floor height) and processing it slice by slice. Results represent sky view factors from an elemental horizontal area placed at a certain height on the faades.

The operation described above can be done with much the same computational time as for one single image: instead of first cutting the DEM to obtain multiple images and then running the sky view factor algorithm on each one, the slicing process is introduced at the end of the shadowing subroutine. This does not result in an increase in the total number of iterations.

View factors with height for the London, Toulouse and Berlin case studies are presented in Figure 14, Figure 15, Figure 16. It is interesting to see how available daylight increases with height. This is quantified in Figure 17, which show the variation of the average sky view factor in the streets with height. As expected, values approximate 1 at the top of the urban canopy.

27

A better evaluation of available daylight, however, is obtained by considering only faade values. These have been averaged and plotted with height (Figure 18 and Figure 19). As expected, they approximate the value 0.5 at the top of the urban canopy (an unobstructed faade will always see just half of the sky hemisphere). In Figure 18 a sudden reduction sometimes appears in the upper part of the city. This is due to the progressive cut-off of buildings with increasing slicing height; as a result of this the faades on which the average is calculated varies and so does the mean value.

Also, numerical results showing the average variation of the sky view factor on building faades can be linearly interpolated (Figure 20). The slope of the line is proportional to the variation of illumination with height: the steeper the slope, the smaller the difference in lighting levels between the top and the bottom of urban canyons. Results show that London, which has a more articulated texture (i.e. a complex mix of buildings heights) has the maximum slope (and therefore the most even vertical distribution of light). Toulouse, which is a more uniform city but still shows some vertical differences, follows and then Berlin, where most of the buildings are the same height. These results will be discussed in the next section.

Another parameter that would be of great use to architects is the distribution of illumination inside the buildings. This can be done using the sky view factor parameter, after some assumptions about the structure of the urban matrix: floor height and faade transmittance. In this case they have been assumed respectively 3 m and 100% (in reality faades will be partially windowed with glass of transmittance less than 100% and partially obstructed by opaque materials such as masonry; however, as far as they can be considered homogeneous and isotropic, a simple proportionality constant applies). A further assumption is that there are no partitions inside the buildings. With these hypotheses the city can be imagined as a series of slabs floating in the air at a distance of 3 metres.

It is worth noting here that under the condition of uniform sky, the view factor inside the buildings is the same as what architects call the sky component of daylight factor. This is defined as the ratio of that part of the illuminance at a given point on a given plane which is received directly from the sky, to the illuminance on a horizontal plane due to an unobstructed hemisphere of this sky (Baker et al., 1993).
12

This meaning of the sky view factor directly follows from its definition in terms of radiative exchange with the sky.

28

Results are presented in Figure 21, Figure 22, Figure 23. It is interesting to note not only the increase of illumination with height, but also its different penetration inside the buildings. The latter clearly is at a maximum on the top floors, which can see the largest area of sky. Important information can also be extracted from these data by applying a threshold (cf. Compagnon and Raydan (2000). If it is assumed that an area inside a building relies solely on daylight if it has a sky view factor (in the absence of a faade) greater than a given value h , all urban portions which are potentially naturally lit can be detected on the DEM.

Results shown by Figure 21, Figure 22, Figure 23 are of course very high, because of the hypothesis of dealing with buildings with completely trasmittive skin. In reality, a major reduction of illumination is caused by the glazing ratio of faades, the obstructions of framing and the transmittivity of glass: lighting levels would therefore be much lower. Also, a more accurate algorithm might take into account the directional properties of glass transmittivity: light falling from patches of sky high on the horizon has a low angle of incidence on vertical windows and is therefore mostly likely to be reflected than light falling with high angles of incidence (Figure 24 shows the variation of glass transmittivity with angle of incidence).

50

50

50

50

50

50

100

100

100

100

100

100

150

150

150

150

150

150

200

200

200

200

200

200

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

50

50

50

50

50

50

100

100

100

100

100

100

150

150

150

150

150

150

200

200

200

200

200

200

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

50

50

100

100

150

150

200

200

250 50 100 150 200 250

250 50 100 150 200 250

Figure 14. Sky view factors on the London DEM computed at 0,3,6,9,12,,39 m; results were obtained by

spreading 1000 fictitious sun-positions on the sky-vault.

29

50

50

50

50

50

50

100

100

100

100

100

100

150

150

150

150

150

150

200

200

200

200

200

200

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

50

50

50

50

50

100

100

100

100

100

150

150

150

150

150

200

200

200

200

200

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

Figure 15. Sky view factors on the Toulouse DEM computed at 0,3,6,9,12,,30 m; results were obtained by

spreading 1000 fictitious sun-positions on the sky-vault.

50

50

50

50

50

50

100

100

100

100

100

100

150

150

150

150

150

150

200

200

200

200

200

200

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

50

100

150

200

250 50 100 150 200 250

Figure 16. Sky view factors on the Toulouse DEM computed at 0,3,6,9,12,,18 m; results were obtained by

spreading 1000 fictitious sun-positions on the sky-vault.

40 35 30 25 20 15 10 5 0

40 35 30 25 20 15 10 5 0

40 35 30 25 20 15 10 5 0

0.2

0.4

0.6

0.8

0.2

0.4

0.6

0.8

0.2

0.4

0.6

0.8

Figure 17. Average sky view factor in the street canyons (from left to right: London, Toulouse and Berlin). As

expected, it increases with heights and approximates 1 at the top.

30

40 35 30 25 20 15 10 5 0

40 35 30 25 20 15 10 5 0

40 35 30 25 20 15 10 5 0

0.1

0.2

0.3

0.4

0.5

0.1

0.2

0.3

0.4

0.5

0.1

0.2

0.3

0.4

0.5

Figure 18. Average sky view factor on the building faades. As expected it increases with height and

approximates 0.5 at the top (where faades view half of the sky-vault); the slight decrease observed at the top and the various changes in slope are due to the fact that the average is taken on different urban portions, due to the progressive cut-off of buildings.

London

Toulouse

Berlin

London

Toulouse

Berlin

0m 3m 6m 9m 12 m 15 m 18 m

0.20 0.23 0.27 0.31 0.34 0.35 0.40

0.18 0.21 0.25 0.28 0.33 0.36 0.37

0.24 0.26 0.27 0.29 0.31 0.34 0.38

21 m 24 m 27 m 30 m 33 m 36 m 39 m

0.40 0.41 0.42 0.44 0.44 0.44 0.43

0.38 0.34 0.36 0.38 -

Figure 19. Value of the average sky view factor on the building faades for London, Toulouse and Berlin.

20 18 16 14 12 10 8 6 4 2 0 0.2000 0.2200 0.2400 0.2600 0.2800 0.3000 0.3200 0.3400 London Touluse Berlin

Figure 20. Linearly interpolated variation of sky view factors with height in London, Toulouse and Berlin

(showing just the first 20 metres near the ground).

31

50

50

50

50

50

50

100

100

100

100

100

100

150

150

150

150

150

150

200

200

200

200

200

200

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

50

50

50

50

50

50

100

100

100

100

100

100

150

150

150

150

150

150

200

200

200

200

200

200

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

50

50

100

100

150

150

200

200

250 50 100 150 200 250

250 50 100 150 200 250

Figure 21. Sky view factors inside the buildings on the London DEM, computed at 0,3,6,9,12,,39 m; results

were obtained by spreading 1000 fictitious sun-positions on the sky-vault.

50

50

50

50

50

50

100

100

100

100

100

100

150

150

150

150

150

150

200

200

200

200

200

200

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

50

50

50

50

50

100

100

100

100

100

150

150

150

150

150

200

200

200

200

200

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

Figure 22. Sky view factors inside the buildings on the Toulouse DEM computed at 0,3,6,9,12,,30 m;

results were obtained by spreading 1000 fictitious sun-positions on the sky-vault.

32

50

50

50

50

50

50

100

100

100

100

100

100

150

150

150

150

150

150

200

200

200

200

200

200

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

250 50 100 150 200 250

50

100

150

200

250 50 100 150 200 250

Figure 23. Sky view factors on the Berlin DEM computed at 0,3,6,9,12,,18 m; results were obtained by

spreading 1000 fictitious sun-positions on the sky-vault.

8. A brief mathematical diversion: optimisation of urban daylight


This section contains some elementary mathematical developments, inspired by previous results. It was shown in section 7 above that London behaves better than Toulouse, which in turns behaves better than Berlin in terms of daylight and sky view factor distribution. Although the difference is not striking, it prompts the following question:

Is there an urban texture that optimises the distribution of daylight?

This question could be addressed in a numerical way by taking different DEMs of both real and imaginary urban cities, by analysing them and by comparing the results. However, under a number of simplifying assumptions, the problem can be treated mathematically.

The concept of optimum daylight distribution should be discussed first. Clearly, urban texture does not affect the total amount of radiation falling on a city, but only the way this radiation is distributed amongst faades, streets or roofs (the image of Marrakesh, Figure 13, is an example where radiation is mostly intercepted at roof level). Furthermore, the notion of optimum is likely to be climate dependent: introverted urban texture with reduced light admittance might prove effective in hot and sunny climates, while open textures with large faade areas might be preferred in cold climates.

33

Here a simple and rather general definition of optimum urban texture is proposed: that where available daylight is most evenly distributed, as in a sort of democratic access to light principle. We will test against this law a simplified urban component, widely used in urban climatology: the infinite and symmetrical urban canyon. Under these conditions the analysis becomes simply 2-D, and the previous question can be more accurately rephrased:

What section of an urban canyon results in the most uniform distribution of sky view factors on faades?

An interesting geometrical property is noted first: all points on a semicircle, such as that shown in Figure 25, see the segment AB with a constant angle (if AB is the diameter, then

= / 2 ). If this law is extended to a semicircular urban canyon of infinite length, i.e. half an
infinite cylinder, it is possible to conclude that all points of the canyon would see the sky under the same solid angle or, in other words, would see the same amount of sky.

Going back to the sky view factor, which can be thought of as a kind of sky-view solid angle with a cosine weighting imposed by radiative transfer principles (see section 5), by analogy, it is possible to infer that the canyon profile that results in a uniform sky view factor on the faades should have a shape similar to that of Figure 25: a smooth and concave symmetrical curve.

Analysing the problem in more detail, mathematical tables exist that give analytical expressions for calculating sky view factor between surfaces. Using the expression reproduced in Figure 26, it is possible to impose the condition of uniform view factor from an elemental horizontal surface to the sky. This leads to the following equation (in a Cartesian co-ordinate system):
13

The equation can be obtained in the following way. Figure 26 gives the view factor between an elemental surface dA1 and a rectangular surface A2 parallel to it at a distance h :

13

FdA1 A2 =

a b b a 1 tan 1 tan 1 + 2 2 2 2 2 2 2 a 2 + h 2 a +h b +h b +h

Due to the additive property of view factors (Figure 26) it is possible to write:

FdA1 Atot = FdA1 A2 + FdA1 A3 + FdA1 A4 + FdA1 A5


In an infinite canyon it can be assumed that b . Taking the limit for each term gives:

34

1 x 2 (1 x) + y
2 2

1+ x 2 (1 + x) 2 + y 2

= const

Plotting this equation gives the curve shown in Figure 27 (for const = 0.5 ), somewhat similar to a parabola. Any elemental horizontal surface (which could represent, for instance, a working desk) placed on this curve will have a constant view factor of the sky .
14

The profile of Figure 27 is certainly difficult to realise in architectural practice, as it would require tilted and glazed building faades. Despite Frank O. Gehrys efforts, most of todays buildings still consist of vertical and horizontal surfaces. It can be pleasantly noted, however, that two environmentally conscious projects dating back to the late 60s namely the Brunswick Centre in London and the archetypal courtyard forms proposed by Leslie Martin and others, Figure 28 and Figure 29 tried to take advantage of sloping faades. In the Brunswick centre, in particular, the upper part of the glazed verandas is tilted and the resulting canyon profile somewhat approximates to the curve of Figure 29.

The above analysis, however, would not be valid if a zig-zag faade is considered, made of horizontal and vertical surfaces, however close it might be to the profile of Figure 27. The sky view factor would then not be constant anymore, but higher in the upper part of the canyon than in the lower one. The beneficial increase in view of the sky provided by the tilt of the faade would be lost by the view cutting effect of the horizontal and vertical surfaces.

The interest of the above analysis, however, might lie elsewhere. If extended, it might confirm something that was already suggested by the comparison of sky view factor distributions in
a 1 2 a 2 + h 2 2

lim FdA1 A2 =

Adding together all 4 terms and adopting a Cartesian co-ordinate system, where the y axis is vertical and the top of the canyon has co-ordinates A(1,0) and B(1,0) , gives:

1 x 2 (1 x) + y
2
14

1+ x 2 (1 + x) 2 + y 2

= const

If instead of the sky view factor the CIE overcast sky model were used (see section 9), the canyon profile would be changed further (it would be deeper, as the CIE model weights patches of the sky high on the horizon more than the sky view factor).

35

London, Toulouse and Berlin: that a mixed distribution of building heights, as in a parabola, is beneficial to a more uniform accessibility to daylight on faades. This would mean that, for a given built density, cities with irregular skylines might be better than homogeneous ones. Formal proof of this fact, however, could only be given by a complex statistical analysis, which lies beyond the scope of this work.

Figure 24. The variation of the trasmittance of ordinary clear window glass with angle of incidence, according

to different authors. Almost 100% of light is reflected when the angle of incidence is very low, compared with just 8% when it is 90. Image from Hopkinson et al. (1966).

P1

P2
Figure 25. It is easy to prove with elementary geometry that

is constant in all triangles inscribed within a

circle; in particular, when AB is the diameter, = / 2 .

36

Figure 26. Calculating the view factor between two surfaces. Left: the view factor between a rectangular

surface A2 and an elemental surface dA1 parallel to it at the distance h can be written as
FdA1 A2 = . a b b a 1 tan 1 tan 1 + 2 2 2 2 2 2 2 a 2 + h 2 a +h b +h b +h

Right: the view factor from dA1 to Atot is additive and can be written as the sum of the partial view factors
FdA1 Atot = FdA1 A2 + FdA1 A3 + FdA1 A4 + FdA1 A5 . From Boffa and Gregorio (1976).

0.5

x 0 0.2 0.4 0.6 0.8 y 1 1.2 1.4 1.6

0.5

Figure 27. Plot which represents the section of an infinite long urban canyon where all points have the same

view of the sky (i.e. the sky view factor, and therefore illumination under a uniform sky, is constant). The function in its implicit mathematical formulation is given in the footnotes and is plotted here for a value of
const = 0.5 . The plot was produced using the mathematical computer program Maple (see, for example,

Nicolaides and Walkington, 1996).

37

Figure 28. The court form, or pavilion antiform, according to Martin (1972).

Figure 29. The Brunswick Centre, London, an icon of British architecture of the 1970s. Designed by Patrick

Hodgkinson, it is characterised by its slanted glass faade. Images from Progressive Architecture (1973).

9. A more precise estimate of illuminance in urban areas


It has been seen above that the sky view factor is a precise estimate of illuminance in urban areas if the sky has a uniform luminance. This condition, however, is not very realistic. Even under optimal overcast conditions, the sky is brighter at the zenith than near the horizon. More generally, the sky luminance changes from point to point and is a complex function of the azimuth and elevation of the sky patch considered. Expressed as a formula:

Lsky = L( , )
38

The function Lsky , representing sky luminance at a certain time, can be given either with an analytical expression or with a table of measured values, from which intermediate values are calculated by interpolation. In general, one of the following is used:

1) The CIE overcast sky is the simplest sky luminance model, which is bound to represent ideal overcast conditions. The zenith is 3 times brighter than the horizon. The luminance distribution has cylindrical symmetry (i.e. it is solely dependent on the elevation ) and is given by the relation below:

L( ) =

LZenith (1 + 2 sin ) 3

where LZenith is the luminance of the zenith.

2) The CIE clear sky, a model like the one above but representing ideal clear sky conditions; its full formulation can be found in Commission Internationale de lEclairage (1973).

Both models above represent ideal overcast or clear sky conditions. More generally, however, the sky luminance will take intermediate values and present more complex patterns. This can be accounted for with the formulations below:

3) Tables with measured sky luminance distributions at given locations. Although these tables are the most accurate and realistic sky representations, they are available only for a handful of geographical locations. This has prompted researchers to produce empirical models that can produce sky luminance from easily-measured quantities, such as direct and global irradiance at a given location, as reviewed below.

4) A number of sky luminance distribution models existing in the scientific literature. For a comparison of their respective performances see Ineichen et al. (1994). The model most referred to in the scientific literature is the one by Perez et al. (1993). It is based on an experimental data set of more than 16000 full-sky scans from Berkeley, California, and provides an all-weather luminance distribution based on measured diffuse and global radiation at certain locations (these quantities are commonly available for a large number of locations across the globe). This model has been incorporated into a number of software 39

packages such as Radiance, by the Fraunhofer Institute for Solar Energy Systems in Freiburg (Germany). Assuming that the function Lsky = L( , ) is available from one of the formulations above, how can it be used to calculate illuminance values on a DEM ?

The algorithm is similar to the one developed to calculate sky view factors, and is based on casting shadows from a number of points spread over the sky-vault. However, to calculate the sky view factor (see section 5 above), these points were distributed uniformly on a circle and then projected onto a hemisphere, in order to obtain the appropriate cosine weighted function. Now, in order to calculate the luminous energy falling on the DEM from a sky of given luminance, it is necessary to distribute points uniformly on the sky-vault and then weight the shadow images using the function Lsky = L( , ) .

A random distribution of points on a hemisphere with a uniform density can be obtained as explained below.

An elemental area on a sphere of unit radius is (in spherical co-ordinates): dA = cos( )dd where is the azimuth and the elevation. If a uniform distribution is taken on and , the resulting distribution of points on the circle will not be uniform. The density of points will be higher towards the poles and lower at the equator, following an inverse proportionality to cos( ) (Figure 30). In order to have a uniform distribution of points on the circle a function = arcsin( z ) must be used. A proof of this fact is given the footnotes ; results are shown in Figure 30.
15

The proof can be given as follows: Make a change of variable so that dA becomes independent from . If a new function z = z ( ) is chosen, this implies: cos d = kdz Where

15

is a proportionality constant. The above equation can be integrated:

40

With this approach it is possible to easily obtain illuminance values on the DEM for a given sky luminance function. For instance, in Figure 31 the CIE overcast sky luminance distribution has been used to compute illuminance values in London. The images have been normalised, i.e. scaled so that the urban surfaces which have an unobstructed view of the sky and receive a maximum illuminance take the value 1. They have been represented with a grey colourmap and resemble at first sight the view factor images. However, the difference is telling (Figure 31). In streets and highly obstructed urban areas, illumination values from the CIE sky are higher than sky view factors. This is indeed what would be expected, as the CIE overcast sky gives relatively more importance to sky patches high on the horizon. Had another function Lsky = L( , ) been chosen, different patterns of illuminance in the city would have appeared. The difference with the sky view factor would have been more evident by using the Perez model, which, unlike the uniform and CIE skies, does not have cylindrical symmetry.

Figure 30. Distribution of points on a sphere. The image on the left presents a uniform distribution of points

on and . In the image on the right the distribution on follows the function = arcsin(z ) , where z is a uniform distribution.

cos d = kdz
and gives: z = k sin + c1 Any value of k and c1 will satisfy our conditions, so we choose z = sin . This gives: = arcsin( z )

41

1 0.9 50 0.8 0.7 100 0.6 0.5 150 0.4 0.3 200 0.2 0.1 250 50 100 150 200 250 250 50 100 150 200 250 200 150 100 50

0.1 0.09 0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0

Figure 31. Illumination distribution on the London case study under the CIE overcast sky (left). The image

has been normalised, so that pixels with an unobstructed view of the sky have the value 1. Although at a first sight this image resembles to the sky view factor image (illumination under a uniform sky, see Figure
10), a more accurate examination shows diversity: the image on the right represents the difference between

uniform and CIE sky. This is approximately nil on roofs, which have an almost unobstructed view of the skyvault, and maximum for streets and highly obstructed courtyards, due to the fact that the CIE sky is brighter near the zenith than low on the horizon.

10.

Further developments: integrated energy values and methods to

estimate light inter-reflections


An algorithm to calculate illuminance values on a DEM for a given sky luminance function
Lsky = L( , ) was described above. Exactly the same algorithm, where just the weighting

function has been altered, can be used to calculate a wider set of parameters. The use of radiance Rsky = R( , ) , instead of a luminance sky distribution function, allows the calculation of received irradiance on the urban surface a variable which is of great interest in estimating solar gains in buildings. Furthermore, instead of an instantaneous function Lsky = L( , ) (or Rsky = R( , ) ), a cumulative one can be selected to represent sky luminance (or radiance) distributions integrated over a given period of time: for instance, a week, a month or a year. This leads to the construction of what Compagnon and Raydan (2000) call statistical photometric (or radiometric, if dealing with radiance) sky models (Figure 32). These sky models simply show average luminance (or

42

radiance) values computed over a number of hours. Their use allows the calculation of total illumination (or irradiation) received on the urban surface during a certain period of time. As explained above, the resulting images can be thresholded by detecting all pixels higher than a given value. This produces the immediate identification, for instance, of all roof areas where annual irradiation is higher than 1000 KWhm and therefore PV applications are economically viable (Compagnon and Raydan, 2000).
2

Another extension of the analysis techniques developed so far allows the modelling of light inter-reflections on the DEM. In fact, all the algorithms above deal with illuminance (or irradiance) falling directly from the sky. This is the main contribution to the total, especially in cities at northern latitudes, where the reflected component is limited due to a relatively low reflectivity (dark buildings and streets). More accurate estimates, however, should take into account inter-reflections of radiation.

This can be done on a DEM. However, the procedure is complicated, time-consuming and poses problems of accuracy, as the DEM is not well-suited to the representation of vertical surfaces. This makes DEM analysis less competitive in accounting for light inter-reflections than other kinds of modelling tools, such as Radiance and its derivatives (Compagnon and Raydan, 2000; Mardaljevic and Rylatt, 2000) and will thus not be examined here.

Simplified alternatives could also be envisaged, analogous to methods that are of common use in the calculation of daylight factors inside buildings. These methods are based on the exact determination of the so-called direct component (radiation falling directly from the sky) and a rough estimate of the reflected component (amount of radiation coming from internal reflections) via the introduction of an average surface albedo. Transposing a similar approach to urban areas would not allow a precise knowledge of illumination at a given point, but might instead suggest an average correction to be applied to a whole portion of a city in order to take into account the effects of inter-reflections.

43

Figure 32. Statistical sky model for Fribourg, Switzerland, for June. The image represents mean radiance

values computed over 445 daylit hours; the image was kindly provided by Raphael Compagnon, Ecole dIngnieurs et darchitectes de Fribourg (Switzerland); a similar diagram can be found in Compagnon and Raydan (2000).

11.

Conclusions

This paper opened with a question: could the analysis of DEMs with image processing techniques contribute to the study of daylight and sunlight conditions in cities? The answer is in the affirmative: the DEM has proven an extremely versatile support for the calculation of a number of parameters from simple shadowing to more elaborate view factors and illuminance distributions under different skies.

Various algorithms have been tested on different case study images, showing a significant variation of the parameters considered with urban morphology. This is particularly evident in the case of the sky view factor, which accounts for differences in the magnitude of the urban heat island in different cities; it varies greatly between London, Toulouse and Berlin and therefore suggests a significant difference in urban temperatures.

Nevertheless, the link between morphological parameters and environmental variables is complex. Many processes are involved, such as the absorption of radiant and luminous energy from the sun and the emission of long wave radiation. The different behaviours of urban texture cannot be described with a single morphological parameter. Furthermore, varying and conflicting requirements are faced: the time of the day and the year, the patterns of occupancy of the building, the overall external environmental conditions, etc. redefine the optimum each 44

time. For instance, in a cold climate maximising the accessibility to daylight (which is regulated to a certain extent by a high view factor) conflicts with urban warmth from the urban heat island (which requires a low view factor).

For this reason it is impossible to establish a general equation valid in all cases of urban planning and design. Ad-hoc simulations are required for decision-making as in the case of the solar envelope. The importance of the DEM is that it constitutes an accessible medium to support this process. Its increasing availability at low cost due to the rapid development of surveying techniques such as Synthetic Aperture Radar or Laser Altimetry is bound to make it a key tool for urban analysis in the coming years.

12.

References

Baker N., 2000, We are all outdoor animals, Proceedings of the International Conference on Passive and Low Energy Architecture PLEA 2000 (Cambridge, UK, July 2000). Baker N., Fanchiotti A., Steemers K. (editors), 1993, Daylighting in Architecture - A European Reference Book, (James & James, London). Brown M., Grimmond S., Ratti C., 2001, Comparison of Methodologies for Computing Sky View Factor in Urban Environments, Proceedings of the 2001 International Symposium on Environmental Hydraulics (Tempe, AZ, December 2001). Burrough P A, McDonnell R A, 1998, Principles of Geographical Information Systems (Oxford University Press, Oxford). Chalfoun, N.V., 1998, Outdoor thermal comfort assessment using MRT and fish-eye lens photography, Proceedings of the International Conference on Passive and Low Energy Architecture PLEA 1998 (Lisbon, Portugal, June 1998). Clarke J. F., 1972, Some effects of the urban structure on heat mortality, Environmental research 5, pp 93-104. Cohn H., 1886, The hygiene of the eye in schools, (Simpkin, Marshall & Co, London). Commission Internationale de lEclairage, 2001, Retrieved September 30, 2001 on the World Wide Web: http://www.cieuk.org.uk/index.htm (British section) Compagnon R., 1999, "Evaluation du gisement solaire en milieu urbain", Proceedings of CISBAT'99, Conference Internationale Energie Solaire et Batiment, (EPFL, Lausanne, Switzerland, September 1999), pp 137-142 Compagnon R., Raydan D., 2000, Irradiance and illuminance distributions in urban areas, Proceedings of the International Conference on Passive and Low Energy Architecture PLEA 2000 (Cambridge, UK, July 2000), pp 436-441. Dubayah R., 1992, Estimating net solar radiation using Landsat thematic mapper and digital elevation data, Water resources research, 28:9, pp 2469-2484. Dubayah R., Rich P., 1995, Topographic solar radiation models for GIS, International Journal of Geographical Information Systems, 9:4, pp 405-419. EU, 1997, Energy for the Future: White Paper for a Community Strategy and Action Plan, Retrieved September 30, 2001 on the World Wide Web: http://europa.eu.int/comm/energy/library/599fi_en.pdf. Faruqui Ali Z., 1998, Le Corbusiers Chandigarh from an environmental point of view, Proceedings of the International Conference on Passive and Low Energy Architecture PLEA 1998 (Lisbon, Portugal, June 1998). Givoni B., 1998, Climate considerations in building and urban design (Van Nostrand Reinhold, New York). Glenn T J, Watson I D, 1984, The Determination of View-Factors in Urban Canyons, Journal of Climate and Applied Meteorology, 23, pp 329-335. Hopkinson R.G., Petherbridge P., Longmore J., Daylighting (Heinemann, London). Howard L., 1833, Climate of London deduced from meteorological observations (Harvey & Darton, London) Incropera F.P., DeWitt D.P., 1990, Introduction to heat transfer (John Wiley & Sons, New York) Ineichen P., Molineaux B., Perez R., 1994, Sky luminance data validation: comparison of seven models with four data banks, Solar Energy, 52:4, pp 337-346. Knowles, R.L., 1981, Sun rhythm form, (The MIT Press, Cambridge, MA). Knowles, R.L., 2000, The solar envelope Its meaning for urban growth and form, Proceedings of the International Conference on Passive and Low Energy Architecture PLEA 2000 (Cambridge, UK, July 2000). Knowles, R.L., 2004, The solar envelope: its meaning for energy and buildings, Energy and Buildings, 35:1, pp. 15-25 Kumar L., Skidmore A., Knowles E., 1997, Modelling topographic variation in solar radiation in a GIS environment, International Journal of Geographical Information Systems, 11:5, pp 475-497.

45

Laboratory of Architectural Methodology, 2001, Retrieved September 30, 2001 on the World Wide Web: http://www.lema.ulg.ac.be/tools/townscope/xx Lam W, 1986, Sunlighting as formgiver in architecture (Van Nostrand Reihnold Company, New York). Littlefair P., 1998, Passive solar urban design: ensuring the penetration of solar energy into the city, Renewable and Sustainable energy reviews, 2, pp 303-326. Littlefair P., 2000, Overshadowing in the solar city, Energy World The magazine of the Institute of Energy, February 2000, 276, pp 11-13. Martin L., 1972, The grid as generator, in: Martin L. and March L., Urban space and structures (Cambridge University Press, Cambridge). Matlab, 2004, Retrieved September 30, 2001 on the World Wide Web: http://www.mathworks.com xx Muneer T., 1997, Solar Radiation and Daylight Models for the Energy Efficient Design of Buildings (Architectural Press, Oxford). MVRDV, 1999, FARMAX Excursions on density, (010 Publishers, Rotterdam). Nicolaides R., Walkington N., 1996, Maple: a comprehensive introduction (Cambridge University Press, Cambridge). Niewienda A., Heidt F. D., 1996, Sombrero: a PC-tool to calculate shadows on arbitrarily oriented surfaces, Solar Energy, 58:4-6, pp 253-263. Nikolopoulou M., Steemers K., Thermal comfort and psychological adaptation as a guide for designing urban spaces, Proceedings of the International Conference on Passive and Low Energy Architecture PLEA 2000 (Cambridge, UK, July 2000). Nikolopoulou M.H., 1998, Thermal comfort in outdoor urban spaces, unpublished PhD Dissertation at the University of Cambridge Department of Architecture, October 1988. Oke T. R., 1981, Canyon geometry and the nocturnal urban heat island: comparison of scale model and field observation, Journal of Climatology, 1, pp 237-254. Pearlmutter D., Bitan A., Berliner P., 1999, Microclimatic analysis of compact urban canyons in an arid zone, Atmospheric Environment, 33, pp 4143-4150. Peckham R., 1985, Shading evaluations with general three-dimensional models, Computer-Aided Design, 17:7, pp 305-310. Perez R., Seals R., Michalsky J., 1993, All-weather model for sky luminance distribution preliminary configuration and validation, Solar Energy, 50:3, pp 235-245. Perez-de-Lama J., Cabeza J.M., 1998, A holistic approach to the Mediterranean patio Extending the new method of configuration factors to semi open spaces, Proceedings of the International Conference on Passive and Low Energy Architecture PLEA 1998 (Lisbon, Portugal, June 1998). Progressive architecture, 1973, Brunswick Centre, London: forward through the past, May issue, 54, pp 100-105. Rapoport A., 1969, House Form and Culture (Prentice-Hall, Englewood Cliffs, NJ). Ratti C., Richens P., 2004, Raster analysis of urban form, Environment and Planning B Planning and Design, 31:2, pp. 297-309 Rich P.M., 1989, A manual for analysis of hemispherical canopy photography, Los Alamos National Laboratory Report LA-11733-M, Los Alamos, NM 87545, December 1989, http://www.hemisoft.com/doc/canopy89. Rosenlund H., Johansson E., Grundstrm K., El-Kortbi M., Mraissi M., 2000, Urban micro-climate in the city of Fez, Morocco, Proceedings of the International Conference on Passive and Low Energy Architecture PLEA 2000 (Cambridge, UK, July 2000). Rudofsky B., 1964, Architecture without architects: a short introduction to non-pedigreed architecture (Academy Editions, London). Sillion F.X., Puech C., 1994, Radiosity and global illumination (Morgan Kaufmann Publishers, San Francisco). Steemers K., Baker N., Crowther D., Nikolopoulou M., Dubiel J., Ratti C., 1997, Theoretical research Chapter 3, Contribution to the EU project ZED (Towards Zero Emission Energy Consumption), The European Commission, Directorate General XII: Science, Research and Development, Contract RENA-CT94-0016, Brussels. Steemers K., Ratti C, 1999a, Informing Urban Bioclimatic Design, Proceedings of the Architecture and Engineering Conference, European Association for Architectural Education (EAAE), (University of Plymouth, Plymouth, UK, February 1999). Steyn D.G., 1980, The Calculation of View Factors from Fish-Eye Lens Photographs, Atmosphere-Ocean, 18, pp 254-258. Summers P. W., 1974, Use and abuse of the urban mixing depth concept, Paper presented to the Eight Annual Congress of the Canadian Meteorological Society (Toronto). Ward Larson G., Shakespeare R., 1998, Rendering with radiance: the art and science of lighting visualisation (Morgan Kaufmann, San Francisco). Yezioro A., Shaviv E., 1994, Shading: a design tool for analizing mutual shading between buildings, Solar Energy, 52:1, pp 27-37.

46

You might also like