Urban Infoscapes New Tools To Inform City Design A
Urban Infoscapes New Tools To Inform City Design A
Urban Infoscapes New Tools To Inform City Design A
net/publication/231783524
CITATIONS READS
8 455
2 authors, including:
Nick Baker
University of Cambridge
47 PUBLICATIONS 1,962 CITATIONS
SEE PROFILE
All content following this page was uploaded by Nick Baker on 29 June 2015.
This paper might be a pre-copy-editing or a post-print author-produced .pdf of an article accepted for publication. For the definitive
publisher-authenticated version, please refer directly to publishing house’s archive system.
environmental design
Urban texture has a profound influence on environmental quality.
The ability to scan maps into a computer and manipulate them
provides a new method of informing this aspect of urban design.
During the past decades, many computer tools have occupation coefficient. Other techniques might
been developed to assist in the environmental design allow a more complex exploration of urban form.
of individual buildings. Heat, light, sound and The first idea was to adapt algorithms developed in
especially energy consumption can be analyzed in the field of image processing to work with these
many different packages. This is not generally true black and white maps. Software was needed and the
for urban design, especially at the medium scale. investigation began by downloading from the
Although it is widely assumed that urban texture – Internet NIH Image for Macintosh (2003), a free
the pattern of streets, building heights, open spaces image-processing program. Many of the ideas in NIH
and so on – will determine environmental quality Image will be familiar from popular programs such
both in the buildings and outside, tools for as Adobe Photoshop; the difference is one of
investigating the connections are sparse. The need emphasis. Photoshop is concerned mainly with
for medium-scale understanding is confirmed by visual appearance, whereas NIH Image concentrates
Givoni (1989): on measurement and analysis. The basic data
‘The outdoor temperature, wind speed and solar radiation structure is the image, a 2-D array of pixels, each
to which an individual building is exposed is not the storing a single byte (8 bits), which is typically
regional “synoptic” climate, but the local microclimate as understood as an integer in the range 0–255. NIH
modified by the “structure” of the city, mainly of the allows any image to be modified and analyzed using
neighbourhood where the building is located.’ a number of pre-defined functions. In addition, it
This paper describes how novel image-processing can be customized with the introduction of extra
algorithms could be applied in urban areas to user-defined functions. These functions are defined
calculate a wide number of parameters. These through macros, that is scripts containing sequences
parameters allow the construction of what we could of operations in a Pascal-based programming
call ‘urban infoscapes’: a layered collection of language.1
information on cities, that can be successfully used
to inform urban design and planning.
This potential encouraged us to modify the can be devised to switch from any one of these
original black and white images to encode building alternative urban representations to another.
heights. This resulted in producing what
geographers call a Digital Elevation Model (DEM), an The processing of DEMs
image where each pixel has a value proportional to An urban DEM, as defined above, is a rectangular
its height, visually indicated by a grey level as in [2a] array of numbers which tells us the height of the
which represents a case study site in central London, urban surface (buildings and streets) at regularly
around Tottenham Court Road. Although an image is spaced intervals. As such, it can be considered a
a 2-D array of values, the DEM is equivalent to a full digital image and analyzed with digital image-
3-D description of the urban surface. All the processing techniques.
information that is contained in the axonometric A digital image is a collection of picture elements,
view of London shown in [2b] or that can be encoded or pixels, usually arranged in a rectangular array. A
in any 3-D surface model, such as a CAD model, is also pixel is the smallest graphic unit of the image and, as
contained in the DEM image. In fact, automated ways an atom, it is not further divisible. It is assigned one
2a
2b
3
Carlo Ratti and Nick Baker Urban infoscapes: new tools to inform city design and planning
environmental design arq . vol 7 . no 1 . 2003 65
or more numerical values, which define its applied to an input image and a digital result,
appearance (colour, brightness, etc). The position of such as a new image or a list of extracted data, is
each pixel in the image is given by its co-ordinates, output [4].
which specify horizontal and vertical location. They For an introduction to digital image processing see
can also be interpreted as column and row and Jain (1989) and Baxes (1994). A discursive overview of
follow the axis sign and conventions illustrated in [3] the digital treatment of images can be found in
– the pixel of co-ordinates (0,0) being in the upper left Mitchell (1992).
corner of the image.
The roots of digital image processing are generally An image-processing algorithm for shadow casting
traced back to the early 1960s, when the US National Let us see how image-processing algorithms applied
Aeronautics and Space Administration (NASA) started to DEMs could help in getting useful information
developing its lunar science programme. Video about cities. Let’s take a simple problem: how to cast
images had to be collected and transmitted to earth shadows on a DEM. A program to address this
and this required the development of digital problem was written at the beginning of this study to
retrieving and processing techniques. Further compute shadows for any angle of illumination and
developments in image processing occurred in the is briefly reviewed below.
following years in the context of medical research Two basic image-processing operations are
and earth-orbiting satellites. Today, digital image required, which are both easily implemented in NIH
processing has a myriad of applications in such Image:
different fields as biology, defence, photography, • translation of the DEM in a given direction
publishing, medical diagnostic imaging, remote • reduction of building heights, based on the
sensing, space exploration, film special effects, subtraction of a constant value from the built
astronomy, etc. In Baxes’ (1994) words: parts of the image, while keeping the street level
‘this dynamic field truly touches us all, playing an constant.
important role in our world … Digital image processing The approach taken is to compute ‘shadow volumes’
techniques are used to do everything from reading your as a DEM, that is, the upper surface of the volume of
checks at the bank to automatically inspecting the fill-level air that is in shadow. We start by defining the three
of your pop bottle.’ components of the vector pointing towards the sun.
Image processing is exciting as it changes the form or Then we compute the components of an opposite
appearance of an image through computerized, vector, scaled so that the larger of the x and y
numerical techniques. Numerous and highly varied components is just 1 pixel, and the x component is
types of operations have been developed during the adjusted to the image calibration. If we translate the
past 30 years: while one of them may improve the DEM by the x and y components, and simultaneously
quality of an image, another may automatically reduce its height by subtracting the z component, we
extract information from it. Whatever the aim, the get part of the shadow volume. If we continue
same steps are followed: a digital technique is translating and lowering by multiples of this vector,
5a
5c
5b
Urban infoscapes: new tools to inform city design and planning Carlo Ratti and Nick Baker
66 arq . vol 7 . no 1 . 2003 environmental design
6a
6c
6b
6d
6e
6 Exploring the effects inside the buildings the double of 7 Computing computed at 0, 3, 6,
of urban texture on by assigning to each unobstructed radiative parameters 9, 12,…, 33m; results
building energy internal pixel the passive zones; values a Shadow casting were obtained by
consumption value of the nearest in [KWhm–2year–1] on the London case spreading 1000
a Passive (within 6m facade pixel; values e Value of the study on 31 fictitious ‘sun-
from the facade) and in degrees optimum glazing December, hourly positions’ on the
non passive zones c Maximum ratio, which intervals sky-vault
(more than 6m from obstruction angle on minimizes energy b Sky view factors d Experimenting
the facade) on the the London case consumption, on on the London case with solar envelopes:
London case study, study DEM looking in the London case study DEM the image
second floor the direction study, on the second computed at 0, 3, 6, represents the
b Orientation of azimuth=30∞. floor (height = 6m); 9, 12,…, 33m; results height of the surface
passive zones on the d Energy small courtyards were obtained by where there are 1, 3,
London case study, consumption on the and obstructed spreading 1000 6 hours of sun
second floor (slicing London case study areas can reach fictitious ‘sun- respectively (values
the DEM at height = on the second floor glazing ratios as high positions’ on the for 31 December); on
6m), obtained by (height = 6m); as 50%, while less sky-vault the far right the
filtering the image glazing ratio 50%; obstructed areas c Sky view factors surface where there
with the Sobel edge- non-passive zones rarely exceed 30 %; inside the buildings are 6 hours of sun is
detectors; facade consume values in on the London case plotted in
values are spread approximately percentage [%] study DEM, axonometric
Carlo Ratti and Nick Baker Urban infoscapes: new tools to inform city design and planning
environmental design arq . vol 7 . no 1 . 2003 67
7a
7b
7c
7d
Urban infoscapes: new tools to inform city design and planning Carlo Ratti and Nick Baker
68 arq . vol 7 . no 1 . 2003 environmental design
8a
8b
London
8c
8d
Carlo Ratti and Nick Baker Urban infoscapes: new tools to inform city design and planning
environmental design arq . vol 7 . no 1 . 2003 69
8e
8 Examining the distance from the portion of the Illustrations [6] to [9d] present how the analysis of
interaction between centre of the image London case study
city and wind (as can be DEM (framed on the DEMs with image-processing techniques allows:
a Directionality
analysis: Fourier
understood from
the mathematical
left image), which is
then scaled (centre)
• Exploration of the effects of urban texture on
transform for the definition of the and transformed in
building energy consumption. This includes the
London case study. Radon transform). black and white identification of all building areas that are within
The plot shows the Therefore, from (right). The latter
6m from a facade [passive areas, 6b and c]. An
sum of the squares of the position of image is used to
the real and these spots it is channel a Cellular established computer model to calculate energy
imaginary parts of possible to have Automata for noise- consumption in buildings, the LT Model (Baker
the Fourier an idea of the driven diffusion
transform using a directionality of (figure 6.39) and Steemers, 2000), has been coupled with the
colour-scale the urban texture e Modelling noise- analysis of DEMs, providing energy simulations
b Directionality c Directionality driven diffusion on a
analysis. Left: black analysis for the selected portion of over extensive urban areas [6d and e].
and white image of London case study the London case • Exploration of a number of image-processing
London; centre: weighted using the study using Cellular
Radon transforms in Gaussian-like decay Automata. A
techniques to compute radiative parameters on a
polar co-ordinates; functions plotted pollution source is DEM. These include shadow casting [7a], sky view
right: both images above. Note that not positioned at the
factors inside and outside the buildings [7b and c],
superimposed. Note all the plots are to bottom of the
the precise fit of dark the same scale. From image, while and solar envelopes (as defined by Knowles, 2003,
spots in the street left to right: values pollutant particles and Capeluto and Shaviv, 2000) [7d].
network: they of sigma migrate upwards.
appear in each urban 0.1–0.2–0.4–0.8 Diffusion after 10, • Examination of the interaction between city
canyon at the point respectively 100, 1000 and 10,000 texture and wind. This can be done either via a
of minimum d Selection of a sub- iterations
number of geometric measures of urban porosity
and anisotropy [8a-c] or Cellular Automata
and take the maximum of this volume with that modelling where pollutant dispersal, for example,
previously calculated, we build up the whole shadow can be simulated by giving simple translation rules
volume [5a]. The process can be stopped when all to ‘digital smoke particles’ [8d and e].
levels are zero, or the translation has shifted the • Calculation of parameters to complement Space
volume right off the image. Syntax, a well-known method for studying urban
Results can be seen in [5b], which represents configuration which has been developed by
shadowing on the London case study. To reduce the Professor Bill Hillier at University College, London
shadow volume to an actual map of shadows on the (Hillier, 1996). A number of alternative indicators
roofs and ground of the city, the original DEM is of the accessibility of urban areas can be
subtracted from the shadow volume. Pixels with developed, based on their visibility [9a–c] and
negative or zero value are in light, positive values are accessibility [9d].
in shade, as in [5c].
This algorithm is very simple and impressively fast. Using these techniques in design
The running speed depends on the number of pixels Different analyses of urban DEMs were briefly
in the image, but not on the geometric complexity. reviewed in the above paragraph. Images shown
So it is feasible to process acres of city at a time, range from the computation of energy consumption
something unthinkable with traditional geometric in buildings to the assessment of sunlight and
models. daylight on facades, the potential for pollution
dispersion in urban canyons, and the analysis of the
A library of analysis functions accessibility of streets based on Space Syntax. While
Based on the same approach, a whole library of these techniques are perfectly justifiable on their
functions to analyze urban texture can be built. A own as a contribution to the characterization of
number of different results is shown here to make a urban textures, a more fundamental question is
proof of concept of the wide range of possibilities raised: could the analysis of urban DEMs help in the
that open up by using urban DEMs. For more detail design of cities? Could it be of any assistance to the
on the algorithms that have been written or on their planner and the architect?
relevance to the urban context, the reader is referred Some tentative and preliminary answers to these
to Ratti (2002). questions are given in this paragraph, which also
Urban infoscapes: new tools to inform city design and planning Carlo Ratti and Nick Baker
70 arq . vol 7 . no 1 . 2003 environmental design
9a
9b
9c
9 Calculating computed). From average inside the used to compute [192,113] in image central London
parameters to left to right, top to buildings all visible points co-ordinates, height case study:
complement the bottom: average, b Polar deformation from the centre of 187m. Right: deformation of the
Space Syntax urban maximum, on the London case the deformation shadowed (black) map so that radial
configuration maximum by study DEM from c London case and illuminated distance from the
method adding values in two point [220,110] in study: shadow (white) pixels, which centre point is
a Lines of sight opposite directions, image co-ordinates; volume in constitute the proportional to
analysis on the minimum, minimum deformation of 14, 35 grayscale (left) viewshed travelling time
London case study by adding values in and 70 pixels produced by a d Chronogeographic (concentric circles
DEM (32 different two opposite respectively; the virtual light bulb transformation of a are reached
directions directions, polar deformation is placed at the point portion of the simultaneously)
Carlo Ratti and Nick Baker Urban infoscapes: new tools to inform city design and planning
environmental design arq . vol 7 . no 1 . 2003 71
9d
review two urban projects in which DEMs have been fashion to the way Space Syntax software is
used. In general, different ways in which they can currently used in design exercises.
contribute to urban design can be distinguished: • an even more ambitious level, the analysis of
At
• At a basic level, the analysis of DEMs adds to the DEMs could evolve into a tool to stimulate
knowledge of the designer. It helps in deciphering invention. While the techniques described in the
the built environment and provides relevant data first and second points above could be defined as
on, say, the shadowing conditions of a given site. It simply ‘reactive’, here a fully ‘proactive’ tool is
informs, therefore, the making of massing discussed – ie, a tool which would not only test
hypotheses and can orientate the design at a solutions, but suggest them. This would, of course,
preliminary stage. Analyses of this kind can and require some additions to the body of techniques
often are carried out on individual buildings. The that has been developed so far. They should be
use of DEMs allows them to be extended to the coupled, for instance, with the synthetic
urban scale (where current software fails due to generation of DEMs or with combinatorial
difficulties in modelling geometrical complexities) methods to produce designs. A similar approach
and to include new simulations, such as pollution was followed, among others, by the Hong Kong
dispersion analysis. architect Michael Chan (2000) whose work,
• Later in the design process, the analysis of DEMs exhibited at the Venice Biennale 2000, is based on
can help practitioners choose between different the automatic generation of building facades and
alternatives. Design is often considered not as a their iterative selection based on the optimization
one-off process, but as a circular one. In Simon’s of environmental criteria.
words (1969), it can be described ‘as involving first the This section will briefly focus on the second point,
generation of alternatives and then testing of these showing how the analysis of DEMs can be used to
alternatives against a whole array of requirements and compare different design options to inform
constraints. There need not be merely a single generate- architects and planners. The images in [10a], for
test circle, but there can be a whole nested series of such instance, show the results of an analysis carried out
circles.’ The analysis of DEMs could then be used in for the Richard Rogers Partnership to compare a
simulation mode; it would help designers in the number of different schemes for the Greenwich
recursive optimization process and in making Peninsula in terms of shadowing and accessibility to
choices between different alternatives. It would daylight. A sensible difference between the behaviour
receive as input different options, each of which of the different options was detected, allowing the
would then be ranked in terms of the selected designers to make informed choices in the
variable (say, energy consumption) – in a similar subsequent phases of design. Illustrations 10b
Urban infoscapes: new tools to inform city design and planning Carlo Ratti and Nick Baker
72 arq . vol 7 . no 1 . 2003 environmental design
10a
10b
10c
Carlo Ratti and Nick Baker Urban infoscapes: new tools to inform city design and planning
environmental design arq . vol 7 . no 1 . 2003 73
10d
and c address a more general planning problem, first because of the difficulties of handling complex
discussed by Leslie Martin and others at the geometry. These difficulties, as we have seen, can be
University of Cambridge during the late 1960s: what overcome by the use of the DEM – a versatile support
is the relative environmental performance of for storing, manipulating and analyzing the
courtyards versus pavilions? A number of different geometry of the city, allowing the fast computation
configurations have been represented with DEMs, of many parameters.
analyzed with image-processing techniques and later Some of these parameters are commonly used
compared in terms of accessibility to daylight and within the architectural community. Others are
sunlight at different times of the year. Finally, [10d] relatively new. In some cases, it has been the very
shows the simulated building energy consumption nature of DEMs that suggested analyses that had
value in a portion of central London with changing rarely been used in urban studies before. These
built densities. It shows that a simulated process of include, among others, the estimation of building
densification – which in recent years has been often energy consumption using the LT Model and the
suggested in the UK as an effective way to reduce computation of the sky view factor.
energy consumption in cities, mostly due to the This growing amount of data that can be obtained
reduction in transport costs – has its side effects: it from DEMs is naturally leading to the idea of an
results in reduced access to light and air and ‘urban infoscape’: a layered information system that
therefore in higher energy consumption in could successfully inform urban design and
buildings and a generally poorer urban planning. This seems, today more than ever, a
environmental quality. necessity: only a greater power of simulation can
satisfy the rapidly changing needs of contemporary
Conclusions design at the city scale. Today, as the leading architect
Most current software to assist design focuses either and urbanist Rem Koolhaas (2000) puts it;
on individual buildings or large portions of cities. A ‘no longer is the city visualised or composed as much as it
lack of tools occurs at the urban block scale, possibly is empirically computed’.
Urban infoscapes: new tools to inform city design and planning Carlo Ratti and Nick Baker
74 arq . vol 7 . no 1 . 2003 environmental design
Carlo Ratti and Nick Baker Urban infoscapes: new tools to inform city design and planning