MULTIVARIATE OPTIMIZATION OF NEIGHBORHOOD SCALE PROBLEMS FOR ECONOMIC,
ENVIRONMENTAL, AND SOCIAL SUSTAINABILITY
BY
GRANT NORMAN MOSEY
DISSERTATION
Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Architecture
in the Graduate College of the University of Illinois at Urbana-Champaign, 2020
Urbana, Illinois
Doctoral Committee:
Professor Brian Deal, Chair Professor Mohamed Boubekri Associate Professor Richard K. Strand Assistant Professor Yun Kyu Yi
ii
ABSTRACT
This thesis begins by arguing that the architectural profession has fallen out of balance.
I contend that the series of objectives which compete for the architect’s attention have been
gradually subsumed by economic concerns. As a means of empirically seeking to restore a
balance, a method is proposed for quantitatively determining the trade-offs between the social,
economic, and environmental sustainability of an architectural problem. The method is tested
on a neighborhood-scale mega-development. This scale of built environment intervention falls
between the building scale work of architects and the city scale of urban designers and
geographers.
The selected design intervention is in Chicago, Illinois with variables including the
number of buildings, their use, and the height of each type of building. Solutions are optimized
for Social, Economic, and/or Environmentally sustainability outcomes using a single-objective
genomic algorithm. A multi-objective genomic algorithm is then utilized to evaluate all three
sustainability objectives simultaneously. Solutions to the chosen problem are proposed and
contextualized within the “restoring balance” framework discussed above. The outcomes
appear to show that such a process can be efficacious in quantitativley balancing the
sustainability based trade-offs implicit to the design process.
iii
TABLE OF CONTENTS
LIST OF TABLES AND FIGURES .................................................................................................... v
CHAPTER 1: PROBLEM STATEMENT ........................................................................................... 1
1.1 INTRODUCTION .................................................................................................................... 1
1.2 THE PLANNER’S TRIANGLE ................................................................................................ 2
1.3 THE PRESENT METHOD ...................................................................................................... 4
1.4 ARCHITECTURE AND THE ECONOMIC NODE .................................................................. 4
1.5 ARCHITECTURE AND ENVIRONMENT ................................................................................ 9
1.6 ARCHITECTURE AND SOCIAL JUSTICE ........................................................................... 14
1.7 THE LACK OF PRESENT BALANCE BETWEEN NODES .................................................. 18
CHAPTER 2: LITERATURE REVIEW ............................................................................................ 20
2.1 OPTIMIZATION .................................................................................................................... 20
2.2 THE GENETIC ALGORITHM ............................................................................................... 20
2.3 ADDRESSING MULTIPLE OBJECTIVES AND THE PARETO SURFACE .......................... 22
2.4 OPTIMIZATION IN BUILDINGS AND ARCHITECTURE ...................................................... 23
2.5 OPPORTUNITIES FOR EXPANSION .................................................................................. 27
CHAPTER 3: METHODS ............................................................................................................... 30
3.1 EXPLORING A NEW METHOD ........................................................................................... 30
3.2 CASE DEFINITION .............................................................................................................. 30
3.3 VARIABLES ......................................................................................................................... 33
3.4 ECONOMIC OBJECTIVE .................................................................................................... 36
3.5 ENVIRONMENTAL OBJECTIVE .......................................................................................... 50
iv
3.6 SOCIAL OBJECTIVE ........................................................................................................... 59
3.7 CONSTRAINTS ................................................................................................................... 69
3.8 MODEL SCENARIOS .......................................................................................................... 70
CHAPTER 4: RESULTS ................................................................................................................ 72
4.1 MODEL PARAMETERS ....................................................................................................... 72
4.2 PRE-SCENARIO INFORMATION ........................................................................................ 73
4.3 SCENARIO A: ECONOMIC OPTIMUM ............................................................................... 78
4.4 SCENARIO B: ENVIRONMENTAL OPTIMUM .................................................................... 81
4.5 SCENARIO C: SOCIAL OPTIMUM ...................................................................................... 84
4.6 SCENARIO D: MULTI-OBJECTIVE OPTIMUMS ................................................................. 87
4.7 SCENARIO VISUALIZATIONS ............................................................................................ 98
CHAPTER 5: CONCLUSIONS .................................................................................................... 100
5.1 OUTCOME EVALUATION ................................................................................................. 100
5.2 LIMITATIONS ..................................................................................................................... 102
5.3 APPLICABILITY TO DESIGNERS ...................................................................................... 103
5.4 FURTHER RESEARCH ...................................................................................................... 104
BIBLIOGRAPHY .......................................................................................................................... 107
v
LIST OF TABLES AND FIGURES
Figure 1: Planners' Triangle, example from Campbell and Fainsten (2003) ................................. 3
Figure 2: Trade-off's Out of Balance. Adapted from the Planners' Triangle ............................... 18
Figure 3: Pareto Front (Left) and Surface (Right) Image adapted from the University of Sheffield
https://www.sheffield.ac.uk/acse/staff/peter_fleming/intromo ..................................................... 23
Figure 4: Levels of Digital Explorations in Architecture: Adapted from Carl Steinitz (2012) ....... 31
Figure 5: Six Proposed Mega-Developments in Chicago. Adapted from Google Earth and
Photos by Respective Developers ............................................................................................... 32
Figure 6: Cost Premium per Height using Multiplier Method ...................................................... 39
Figure 7: Per-Unit-Floor Area Costs by Building Type and Height .............................................. 39
Figure 8: Relationship between Unit Height and Rental Price per unit Floor Area for Three
Buildings ....................................................................................................................................... 42
Figure 9: Relationship between Unit Height and Sale Price per unit floor area for Three
Buildings ....................................................................................................................................... 44
Figure 10: Floor Count vs. Floor Plate Efficiency with data from Barton and Watts (2013) and
Sev and Ozgen (2009) .................................................................................................................. 48
Figure 11: Greenhouse Gas Emissions per Unit Floor Area for Subject Building Types ........... 53
Figure 12: Embodied Energy Premium For Height, Calculated based on Values from Forabachi
et al (2013) .................................................................................................................................... 55
Figure 13: Net Floor Area per Occupant, not accounting for vacancy loss ................................ 57
Figure 14: Use Fraction multiplies applied to fuel consumption of each building type .............. 58
Figure 15: Employees for a 1,000,000 NSF Building of Each Use Type ..................................... 66
Figure 16: Per Unit Floor Area Profit per Building by Type and Height ...................................... 74
Figure 17: Return On Investment by Building Type and Height .................................................. 75
vi
Figure 18: Total Greenhouse Gas Emissions by Building Type and Height ............................... 77
Figure 19: Per Unit Floor Area Greenhouse Gas Emissions by Building Type and Height ........ 77
Figure 20: Best and Average Solutions by Generation for Single-Objective Economic Model .. 79
Figure 21: Best Solution by Generation for Single-Objective Economic Model ......................... 79
Figure 22: Optimum Solution for Single-Objective Economic Model ......................................... 80
Figure 23: Best and Average Solutions by Generation for Single-Objective Environmental Model
....................................................................................................................................................... 82
Figure 24: Best Solution by Generation for Single-Objective Environmental Model .................. 82
Figure 25: Optimum Solution for Single-Objective Environmental Model .................................. 83
Figure 26: Best and Average Solutions for Single-Objective Social Model ................................ 85
Figure 27: Best Solution for Single-Objective Social Model ........................................................ 85
Figure 28: Optimum Solution for Single-Objective Social Model ................................................ 86
Figure 29: Histograms showing Distributions of First Rank Solutions by Objective Cost Function
....................................................................................................................................................... 88
Figure 30: Non-Dominated Solution Set by Economic and Environmental Performance .......... 90
Figure 31: Non-Dominated Solution Set by Economic and Social Performance ....................... 91
Figure 32: Non-Dominated Solution Set by Environmental and Social Performance ................ 92
Figure 33: Normalized Economic, Environmental, and Social Performance of Non-Dominated
Solutions ....................................................................................................................................... 93
Figure 34: Solution A Building Types and Heights ...................................................................... 96
Figure 35: Solution B Building Types and Heights ...................................................................... 96
Figure 36: Solution C Building Types and Heights ...................................................................... 97
Figure 37: View of Scenario A from south along Wells/Wentworth Connector ........................... 98
Figure 38: View of Scenario A from northwest along the Chicago River .................................... 99
Figure 39: Optimization Solutions on the Planner's Triangle ..................................................... 101
vii
Table 1: Summary of Design Variables ........................................................................................ 36
Table 2: Summary of Buildings and Regression Date for Residential Rental Model .................. 43
Table 3: Summary of Building and Regression Data for Residential For-Sale Model ................ 45
Table 4: Single-Objective Genetic Algorithm Settings ................................................................. 71
Table 5: Multi-Objective Genetic Algorithm Settings ................................................................... 71
Table 6: Population Information for First Rank Solutions ............................................................ 94
1
CHAPTER 1: PROBLEM STATEMENT
1.1 INTRODUCTION
The great structural engineer Fazlur Kahn, father of some of Chicago’s most noteworthy
landmarks, spent his career grappling with the notion of trade-offs. As demand for tall
buildings exceeded what could be easily produced with a standard steel frame, Kahn realized
tall buildings would have what he referred to as a “premium for height.” (Mufti and Bakht 2011).
Due to the shift from gravity-controlled design to laterally controlled designs, the taller a
building was to get, the more robust its structure would have to be per unit floor area.
Kahn’s career focused on the development of schemes to solve this problem. While the
clearest steel structures were cheap and economical, they were often limited to twenty stories
in height. While inclusion of shear walls allowed for greater height, it did not totally obviate
height restrictions and often left buildings with areas of undesirable opacity, anathema to an
era when clear, trabeated solutions were in vogue. Framed-tube buildings could be built
much taller, but only if their exterior columns were spaced densely together, limiting access to
views and natural light. Tube-in-tube buildings could be constructed higher still, but in addition
to the problems of the framed-tube, they also required an interior core directly at the center of
the floor plate. “Braced-tubes” added strength, but often left individual apartments or offices
staring at a large diagonal brace rather than a pristine view. “Bundled tube” buildings could be
building the tallest of all, but their plan organization was broken into several smaller “spans,”
with interior columns frequently interfering with interior program.
Designers will recognize in the above example a fundamental (if anecdotal) truth: that
design is governed by a process of trade-offs whereby improving one metric of a design often
involves limiting another aspect. This condition is not particular to buildings. In a 1991 article
about engineering design, authors Otto and Antonsson note:
2
“Design trade-off strategies are always present in the design process… ...[A] designer
may wish to slightly reduce some of the weaker goals in a design if large gains can be
made in the other goals, which would more than compensate for the slight loss. For
example, in the design of a sports car, the designer might reduce the safety margin of
some variables (like stress) to gain performance in other variables (like horsepower),
even though the stress may already be quite high.” (Otto and Antonsson, 1991).
1.2 THE PLANNER’S TRIANGLE
Considering specifically the built environment, one useful framework for understanding
some of the most important trade-offs governing the building sciences is offered by the
Planner’s Triangle (Campbell 1996) which suggests a balance between nodes of social
sustainability, environmental sustainability, and economic sustainability. While others had
touched on this multi-variate approach to solution evaluation, most notably Elkington with his
notion of the “triple bottom line,” (Elkington 1997), Campbell’s framework is unique it that it
identifies the three nodes of the triangle as being perpetual conflict with one another.
The first conflict identified, the “Property Conflict,” describes situations in which social
equity is placed at odds with economic growth. Campbell notes that this conflict is particularly
complex because “each side not only resists the other, but also needs the other for its
survival.” (Ibid.) One may call to memory recent debates regarding the merits of gentrification
as a possible example of this conflict in architecture. The second axis of the triangle, the
“Resource Conflict,” explains situations in which economic growth clashes with environmental
sustainability. Campbell characterizes this conflict through the lens of resources, describing
“tension between their economic utility in industrial society and their ecological utility in the
natural environment.” However, observers in architecture may be more likely to recognize this
conflict in the constant pressure to sacrifice long-term environmental performance for short-
3
term economic incentive. Finally, Campbell describes the third axis of the triangle in which
social outcomes and environmental outcomes are at odds, which he calls the “Development
Conflict.” While Campbell points to “miners, lumberjacks, and mill workers” as examples of
people who “see a grim link between environmental preservation and poverty,” this conflict is
perhaps the most difficult to understand in an architectural context. Still, little imagination is
required to envision a situation in which forest preserve must be left pristine or be torn asunder
for the construction of affordable housing.
Figure 1: Planners' Triangle, example from Campbell and Fainsten (2003)
If one accepts, then, that trade-offs are intrinsic to the design process, and that
Campbells model, shown in Figure 1, is a useful formwork of characterizing these trade-offs,
one is quickly leads to further questions regarding the present condition of these trade-offs,
how these competing variables might be quantified, and how architectural outcomes can be
improved through this process. This thesis will attempt to quantify select trade-offs for
problems at a specific design scale. First, it will examine the present manner in which these
competing objectives are being managed (or failing to be managed, as the case may be).
4
Second, it will propose a method by which these trade-offs may be better understood and their
outcomes optimized. Third, it will identify an appropriate scale and select a problem to which
this new method may be most efficaciously applied and attempt to implement the method for
these problems. Finally, it will evaluate the results of this limited implementation, drawing
insights as to how the method may be more broadly applied.
1.3 THE PRESENT METHOD
Even in the absence of an empirically valid method of quantifying the trade-offs in the
design process, construction in the building environment has (unsurprisingly) continued to
advance. The planners’ triangle will likely have a ring of truth to practicing architects, who are
in the business of deftly navigating the give-and-take of the design process. However, in the
absence of a clear way to quantify this give-and-take, there is compelling evidence that the
economic/environmental/social sustainability model is not presently at equilibrium.
Architects, being a heterogenous group, would likely value each node of the triangle
differently if given the choice. However, by examining the built environment collectively and
seeking to understand recent trends in the profession, it may be possible to draw some broad-
strokes conclusions about the present state of the trade-offs in design. What follows is a brief
assessment of how architecture, has accounted for each point of the triangle, culminating in an
argument that too much emphasis has been placed on economic prosperity without sufficient
consideration of the trade-offs implicit in this decision.
1.4 ARCHITECTURE AND THE ECONOMIC NODE
Seeking to quantify the economic fortunes of building as a discipline is a difficult task even
without trying to parse the economic roll of the architect from the noise. It is possible, to some
degree, to measure the fortunes of architects themselves as a group (although even this
exercise is often reliant on self-reported data) and individually. However, there is little evidence
5
that the economic prosperity of architects is correlated with the importance each puts on the
economic trade-offs of design. A more reasonable task may be to look at the prosperity of the
built environment itself, as well as what architects are doing as a result of the present course of
this prosperity and draw conclusions for this point.
It is possible to quantify the profitability of certain types of buildings to some degree. The MSCI
World Real Estate Index, for example, offers a “free float-adjusted market capitalization index
that consists of large and mid-cap equity across 23 Developed Market countries.” (MSCI 2018).
More specifically, the MSCI World Real Estate Management and Development Index zooms
more narrowly, focusing from the real estate sector at large to companies “engaged in real
estate development and sales, real estate operations and management, or in real estate
services.” In both cases, the general prosperity of the sector in the evaluated countries
appears to be on the rise. The former cumulative performance index, which uses February
2003 as its basis (i.e. a value of 100), now reflects (as of February 2018) a value of 422.12, near
all-time highs. The latter cumulative performance index, which uses April 2006 as its basis (e.g.
a value of 100), now reflects a value of 170.95. One-year returns on the former index are
positive at 3.95%, but appear to lag slightly behind the growth experiences since 1994, which is
annualized at 6.71%. One-year returns in the latter index are much more strongly positive, with
a rate of 20.01% vs. an annualized value of 4.63% since April of 2006. The picture here is
somewhat mixed, yet the upward trend, subtle in some places and more obscure in others, is
undeniable.
Another data point is provided by evaluating capitalization rates, essentially the ratio
between the value of commercial realestate and it’s net operating income. Low capitalization
rates indicate buys are willing to pay more for a given property than high capitalization rates,
given static cash flows. Cushman & Wakefield’s mid-2017 survey of capitalization rates
6
indicate that they are relatively flat, with mild growth. More noteworthy in the report is that
relatively constant capitalization rates are being experienced despite addition of square
footage. The report notes that, since 2015, the “supply pipeline has markedly increased.” New
construction, only 2.0% of inventory in 2015, has risen sharply to 2.9% of inventory in 2017
(Cushman & Wakefield, 2017). This suggests that while new construction is abundant,
demand and prices for real estate remain high in the private sector.
A third data point may be the AIA’s Architectural Billing Index, seen as a bellwether of
performance for the AEC sector. The January 2018 report shows Billings presently stand at
54.7 (indexed against a value of 50.0 for January of 2017). Inquiries and Design Contracts are
also up slightly from the year before. While these metrics evaluate performance on a year over
year basis, it is worth noting their continued rise, considering the economic strength
experienced in 2017 (AIA, 2018). A broader picture view can be provided by evaluating the
billing index on a three-year moving average. This method shows that even by 2014, early in
the present building boom, the ABI has rebounded from values near 35 during the 2008
Recession to values near 50 (AIA, 2014). Generally, demand for architect’s services, correlated
closely with building demand, appear to be high and rising.
A final data point may be offered by the cost of buildings incurred by consumers. It
would stand to reason that if consumers are paying more for buildings, constructing said
buildings is likely more profitable. Evidence for strength in consumer spending can be found in
the S&P Case Shiller U.S. National Home Price Index. This index reports a value 197.01 for
February 2018, up from a post-recession low of 133.99 (FRED, 2018). The Consumer Price
Index’s Housing Component also shows strong upward movement, with an April 2018 value of
123.11, up from pre-recession lows which frequency sank below 100. At the consumer level,
demand for buildings appears high, with the price consumers pay for shelter on the rise.
7
While none of these metrics offer perfect proof that building is becoming more profitable for
developers and other stakeholders, there does seem to be compelling evidence that the
Building Industry has become much more prosperous in recent years and, in some cases,
months. This jibes intuitively with popular discourse on the economy and the increasing
number of tower cranes visible each day in America’s urban cores.
Whether this prosperity has any connection, causal or otherwise, to architect’s valuing
of economic principles over those of environmental and social good is not totally clear.
However, amongst the critical literature in architecture, there is certainly qualitative support for
the notion of architects having become more concerned with economics than with other
metrics of success. One incisive criticism of the profession on these grounds is offered by Paul
Knox:
“A starkly neoliberal political economy, in which progressive notions of the public
interest and civil society have been eclipsed by the bottom line in corporate and public-
private investment, means that design solutions have to be commercially attractive.
Compromises have to be made, projects have to be hustled, ideas have to be sold, and
unpalatable truths have to be spun into palatable propaganda. Planning and urban
design, deflected from issues involving regulation or social expenditure, were
increasingly pushed toward contributing to artful fragments of upscale suburbia as a
means of sustaining professional identity and credibility” (Knox, 2010)
Knox, not one to mince words on this particular topic, goes on to argue that while
modernist designers strived for progress, postmodern architecture has instead sought
“consumption, hedonism, and the generation of profit, with little regard for the social
consequences.”
8
Another such indictment, of sorts, can be found in Klingmann, who argues in her 2014
book, Brandscapes, that the design of public space has shifted from the more egalitarian aims
of the earlier 20th century to a more privatized experienced, controlled by large-scale corporate
interests:
“As the symbolic definition and appropriate of spaces is increasingly reliant on
attributes that communicate “what space may be used and by whom,” the question
arises as to what extent social diversity is wanted or encouraged. The accompanying
dangers of exclusion, and of the privatization of city live, have been noted by many
critics…” (Klingmann, 2014)
Certainly, these types of theoretical criticism would appear to indicate that at least some
amongst the profession believe that the trade-off between profitability and the other goals of
architecture (among them social and environmental sustainability) have not been properly
accounted for.
Evidence that the names and likeness of architects are being co-opted for economic
purposes is certainly not difficult to come-by, if anecdotal in nature. Consider the residential
high-rise located at 8 Spruce Street in New York. Originally known by its address, the building
was re-branded as “New York by Gehry,” in reference to Frank Gehry, the projects architect.
The Projects website, which refers to Gehry as “one of the most celebrated architects
practicing in the world today,” also notes that “master architect Frank Gehry has reimagined
the classic Manhattan high-rise,” and that “Gehry’s distinctive aesthetics is also reflected in the
residences and three floors of amenity spaces, all designed with custom furnishings and
installations.” While the so-called High Modernists, forged in the egalitarian fires of
9
Weissenhof, would eventually design building for IBM and Pan-American Airlines, one finds it
difficult to see Mies van der Rohe conceiving of the idea of referring to 860-880 Lakeshore Drive
as “Chicago by Mies.”
Clearly, architects in today’s world are incentivized to focus on profitability. This is not
necessarily problematic, in and of itself. A profitable building is not necessarily a good or bad
building, or a sustainable or un-sustainable building. If architects continue to focus on the
Social and Environmental poles of design, and consider these elements in balance, profitability
becomes one of many competing goals vying for the attention of the architect. However, as
the following pages will show, it appears as though far less progress has been made on social
and environmental goals as on economic goals.
1.5 ARCHITECTURE AND ENVIRONMENT
In contrast to the world of economics, where the built environment appears to be
performing well, the complicated relationship between architects and environmentalism
appears to be offering less optimal outcomes. While a focus on economic sustainability is
clearly present, environmental sustainability appears to be insufficiently accounted for in the
trade-offs of design.
To the credit of some, select designers can be seen among the vanguard of the modern
environmental movement. Sustainability in design professions can be seen at least as early as
1969 with the work of Ian McHarg. McHarg, and his opus, Design with Nature, take their place
among the early turning points of the environmental movement, falling between Silent Spring
(Carson, 1962) and the 1973 Arab Oil Embargo. Within the realm of architecture specifically,
early adopters like Hassan Fathy and Laurie Baker were among the first to recognize and revive
the sustainability of indigenous forms, often as a matter of necessity. By the mid-1990’s, the
notion of architectural sustainability was receiving widespread publication in the works of Ken
10
Yeang (Yeang, 1995). By the end of the century, the Congress for New Urbanism has elevated
some ideas regarding sustainability to the urban scale (Duany et al, 1999), although these
came mixed with the odd, retrograde language of post-modernity.
To a viewer watching the contemporaneous literature as the sustainability movement
evolved, it may seem as though architects and designers were among early adopters and
remain some of the most ardent defenders of the environmental performance of the built
environment. For some designers, this may very well be true. It is certainly not a difficult task
to find a building advertising it’s architecture as sustainable. However, it is additionally clear
that while sustainability is a driving force for some, others, perhaps even a majority of others,
remain uncommitted to its principles. Further, it is evident that some projects advertised as
ecologically sensitive are, at best, of uncertain ecological merit.
While sustainability may be common in so-called “high design,” a drive down any
suburban street will showcase numerous buildings that appear to be designed without
considering ecology (and likely some that appear to have not been designed at all). The split
between the sort of practice that can indulge in the luxury of environmental performance and
the sort of practice which dictates how most of the world looks is an industry-wide
phenomenon. Klingmann notes the division in practices:
To this day, two types of architectural practices continue to split architectural production
– commercial practices – which pride themselves on being consumer friendly and which
for the most part offer conservative solutions that are based on tried and true formulas –
and critical practices, which offer innovate design but attract few customers. This
enforced schizophrenia between business and cultural ambitions, commercial
objectives and ideological pursuits, is hardly a desirable condition to maintain.
(Klingmann, 2014).
11
Berardi, writing about the profession at large, offers a similar observation:
The process of specialization was pushed to the extremes, the common ground of
intellectual exchange-which was the public sphere of the social movements-was erased.
Everybody is busy working in conditions of isolation and competition; engineers and
poets belong to two distant dimensions that today never meet.”
Knox connects this disconnect to the foundations of architectural education:
“Architecture schools, in particular, tend to have a fraught relationship with the
profession, resenting the restrictions on curricula imposed by professional accrediting
agencies and eschewing the issues that grads will face in the transdisciplinary and
commercially oriented practices of medium- and large-sized firms (that represent about
80 percent or more of the profession) in favour of idealized notions practice (sic) in
small, boutique firms committed to ‘good’ design.” (Knox, 2010)
Even those firms which consider themselves among the environmentalist vanguard can
often deliver projects which are marketed or justified as “sustainable,” but in reality, offer no
particular improvement on the performance of buildings which do not adapt this mantle. Even
the United States Green Building Council, arbiter of the much-publicized LEED Program, must
acknowledge that buildings awarded a plaque for sustainability are not necessarily
performative. While a survey of 121 LEED Buildings found that the energy use intensity for
those buildings was 24% below that the of the total building stock (69 kBtu/sf vs. 91 kBtu/sf),
the basis for comparison was the total commercial building stock, representing not just new
12
construction buildings, but also leaky and inefficient buildings built decades ago. (Turner and
Frankel, 2008). Fully one quarter of buildings certified by LEED failed to achieve an Energy
Star rating of 50 (making them worse performatively than a median building of comparable
use). (Ibid.) Further, while LEED Buildings have been found to use roughly 28% less energy
than the code baseline, these buildings are split about equally between buildings which used
more than anticipated and buildings which use less than anticipated. At least 10% of LEED
buildings sampled, included six Gold/Platinum rated buildings, actually performed worse than
the energy baseline. (Ibid.) Another survey of 100 LEED buildings yielded similar results:
while LEED buildings used, on-average, 18-39% less energy per floor area, 28-35% of LEED
buildings actually used more energy than the comparative standard. (Newsham et al 2009)
In the case of LEED, architects or other design professions have deliberately chosen to
meet the onerous reporting requirements to receive laudation for their building performance.
Yet, even in these buildings, performance is far from guaranteed. Meanwhile, the have-nots of
the architectural practice world, miles from LEED certification, rely mostly on Energy Codes to
regulate the performance of their buildings. Fortunately, the International Energy Conservation
Code and ASHRAE Standard 90.1 continue to impart greater and greater requirements on
building assemblies, operations, and performance. While the forefront of architecture may
make broad overtures to the aesthetics of sustainability, architects have largely ceded the
regulatory apparatus to other design professionals. There are a frustrating number of
architects standing ready to build the worst building allowable by code.
All of this, of course, expresses only concern for operational energy. A true analysis of
Building Lifecycle Energy consumption remains elusive. Sartori and Hestnes (2007), review
several articles on the calculation of embodied energy and it’s relation to operational energy,
finding that various sources quote embodied energy as making up anywhere between 5 and
32% of total Lifecycle Energy. Buildings with lower operational energy were found to generally
13
have higher embodied energy, although additional embodied energy was almost always
worthwhile if it resulted in substantial reduction of operational energy. Drew et al (2015)
expound on this idea even further, arguing that to be truly complete, transportation would
energy associated with building modalities would have to be evaluated contemporaneously
with the buildings themselves.
While LEED and some other architect-friend tools do make nods to lifecycle energy
costs (e.g. LEED credits for Sustainable Sites and Local Materials,) the prescriptive method
used in codes relied upon by many architects often specify only material properties (e.g. wall
thermal resistance) or, at best, energy use intensity per unit floor area. These metrics, the most
common in use, miss the place of buildings in in its larger context from an energy perspective.
Perhaps the most compelling evidence that environmental factors have become
subservient to economic factors is offered by the very world around us. US Commercial
Buildings built between 1990 and 2012 use an average of 80.9 kBtu/sf of floor area. While this
is a marginal improvement from the 84.5 kBtu/sf used by buildings built between 1960 and
1989, it actually represents worse performance than buildings built in 1959 or before, which
had average EUI’s of only 68.6 kBtu/sf (EIA, 2012). While this method fails to account for
improvement in thermal comfort standards or continuous ventilation requirements, it remains
damning that brand new buildings, for all their standards, actually use more energy than older
ones.
Residential Buildings fare little better. The median and average size of new single-
family homes, which took a brief respite from its meteoric rise during the Great Recessions, has
now resumed its sharp upward trajectory (Census, 2010). 23.1 million housing units, 19.54% of
all those in the United States, are presently over 3,000 square feet in gross floor area. These
changes are occurring at a time in which average household size is shrinking. In 2015, the
average home used 30.3 million BTU per household member per year. This certainly
14
represents an improvement from 1993, when this measure was first calculated, when the
average home used 40 million BTU per household member. What’s more, much of this gain
has been recent, with values of 34.9 million BTU per household member being published as
recently as 2009. However, these gains have been largely offset by the growth in population.
In 1993, all US homes used 10.01 quadrillion BTU’s. By 2009, that number had grown to
10.183 quadrillion BTU. It wasn’t until 2015 that a substantial decline, to 9.114 quadrillion
BTU’s was observed (EIA, 2015). While residential buildings were becoming more
performative, the energy savings of these performance enhancements were, to some degree,
offset by a demand for more units and units of greater size.
The evidence for the growing profitability of buildings certainly seems more convincing
that the evidence for growing environmental performance, where gains have been modest. To
what degree the architect bears responsibility for the valuation of economy over environment is
unclear. A compelling argument can be made that architects, always serving at the pleasure of
their clients, are creating the most performative buildings allowed within the framework of a
growing demand for economic return.
Of course, perhaps a more convincing argument can be made that architect’s ultimate
responsibility is to the health, safety, and wellbeing of the general public, and not exclusively to
their clients. Operating under this assumption, it can be easy to see why architects can be
seen as at best complicit and at worst willing participants in placing building economics ahead
of environmental performance.
1.6 ARCHITECTURE AND SOCIAL JUSTICE
So far, a survey in the present methods of architecture has indicated that architect’s
focus on economic performance appears to have been more successful than focus (or the
absence thereof) on environmental performance. While evidence of economic prosperity was
15
strong, evidence of environmental responsibility was, at best, mixed. In considering the third
node of the triangle, equity, the evidence would seem to indicate that this node has more in
common with the mixed environmental results than the more uniformly positive financial
results.
Certainly, the present climate of economic inequality cannot be blamed entirely on
architects, or even on the built environment. A full investigation of the causes of growing
inequality are outside of the scope of this research. However, it would be sufficient to say that
economic, demographic, and political components confound serious pursuit of how the built
environmental impacts equity and social justice. However, it may be possible to examine some
built-environment-specific criteria to see to what degree equity is valued by architects.
It will certainly come as no surprise to those who have watched rise of neo-liberalism as
a political doctrine that privatization has played an increasing role in our world, and this has
manifest itself in our built environment. In 1996 Privtopia, Evan McKenzie lays out a compelling
argument for the gradual shift from public land use policy to private control of land through the
rise of instruments such as restrictive covenants, homeowners’ associations, private security
and infrastructure, and even physical barriers (e.g. gated communities). For “public” spaces,
where physical gates and doormen are seen as less acceptable, private interests has found
ways to keep control out of public hands, as argued by Klingmann (2014) in her evaluation of
Celebration, Florida:
“The subtle control of communicable space takes on a strategic significance here.
While the proverbial gates have been replaced by a twenty-four-hour security patrol to
create private, branded space that pretends to be public, Celebration is more than just
an example of highly controlled urban environment – it is also an exercise in moral
regulation.”
16
While the rise of privatization of built spaces may represent a threat to equity in the
developed world, there can be little doubt that more serious affronts are found in the
developing world. An example of this can be found in the construction associated with the
World Cup being held in Qatar. The late Zaha Hadid, Pritzker prize winner and star-chitect, was
tasked with designing the signature al-Wakrah Stadium, one of the centerpieces of the event.
When evidence emerged that significant human rights violations were occurring, including the
harm and even death of workers building that stadium, Hadid told the Guardian Newspaper
shortly before her death that “I have nothing to do with the workers.” When further pressed
about the 800 reported deaths, Hadid reaffirmed that “I cannot do anything about because I
have no power to do anything about it. I think it’s a problem everywhere in this world.” (Wilson
et al. in Deamer, 2015)
Architects, for the time being, seem content to throw up their collective arms when
confronted with a growingly less just world, adding that they are only acting as the agents of
others. Perhaps the conflicts faced by architects are best summarized by Worthington (2009),
who writes:
“Today the profession continues to be faced with these paradoxes, and there is
uncertainty about whether its prime role is social concern for the wider interest of
society and the user, or whether it is primarily concerned with the particular interests of
its developer client or success of the practice, now most probably a limited company
whose economic livelihood may rest on the next job from a contractor. The
architectural profession feels alienated. The professions are often no longer head of the
team, being increasingly rejected by the client and user and driven by commercial
imperative.”
17
While architects are licensed ostensibly to protect the “health, safety, and wellbeing” of
the general public, it is clear that architects are either impotent or indifferent to the
requirements of this task. Perhaps this is a result of the changing political and economic
winds. Perhaps it is because that in a post-Pruitt Igoe era, Architect’s no longer feel like the
moral authority in ensuring social equity. Certainly, twentieth century policies of “slum
clearance” and the construction of isolated workforce housing had adverse effects from which
the social cohort of the profession is still reeling.
All of this is not to say that today’s architects are unable to conceive of socially just and
healthful spaces when given the means and reason to do so. Another Pritzker-prize winning
architect, Alejandro Aravena has offered a promising example of what a more equitable
architecture might look like through his incremental housing projects in Chile. In these
subsidized projects, funds are saved by fitting out only one half of a house, leaving the other to
be developed at the discretion of those that dwell within. This affords opportunities for family
growth, small, local businesses, and ultra-fine mixed-use solutions (Aravena, 2011). In this
case, architecture seems to be designed explicitly to help those transitioning from poverty to
better circumstances.
Still more promising work is being done in the research-based corners of the
profession. Sarah Williams Goldhagen’s Welcome to your World has summarized some of the
excellent work being done on the interaction between the built environments and the quality of
life of those who live within it. Still others have focused on the effects of architecture (Ulrich,
2006) and Landscape Architecture (Brown and Corry, 2011), in improving healthcare
outcomes. Yet, to date, there remains no overarching means of balancing the newly
understood benefits of certain architectural strategies with the other requirements of these
buildings. It may be, as research has suggested, that patients with a view of green spaces
18
have a shorter length of stay. However, balancing this wellbeing component with the cost of
glazing and the environmental performance of façade penetrations is not well understood.
Clearly, the picture for architecture as a progenitor of social good is, at best, mixed.
While certain buildings and research appear to have demonstrated an interest in social justice
and health, safety, and well-being, in other cases, prominent architects have washed their
hands of something as basic as the working conditions of those doing the actual building itself.
1.7 THE LACK OF PRESENT BALANCE BETWEEN NODES
When considering the present methods by which architects the trade-offs implicit to the
design process using the Planner’s Triangle, it appears as though, from a sustainability
standpoint, the design professions have fallen out of balance (see Figure 2).
Figure 2: Trade-off's Out of Balance. Adapted from the Planners' Triangle
Of course, as previously discussed, the degree to which this shift may have occurred is
nearly impossible to measure. While a review of the qualitative and semi-objective measures
presented above appears to indicate a profession more concerned with profit that the other
nodes of the triangle, at present, there is no reasonable method to quantify these trade-offs for
projects with any degree of scale. Architecture is also an exceptionally wide discipline. While
19
Figure 2 shows a single point which has drifted in a single direction, a more accurate, if less
legible, diagram would likely show a cloud of points which has drifted disparately from the
center, with a possible higher density of points towards the economic node.
At present, there exists no meaningful method by which to evaluate exactly how a given
practice or, more to the point, a given architectural solution has shifted or to what degree it
succeeds or fails in balancing the competing priorities of sustainable design. With the flaws of
the present method clear, in the following section, a new methodological framework for just
such evaluation will be explored.
20
CHAPTER 2: LITERATURE REVIEW
2.1 OPTIMIZATION
In the previous section, a problem of prioritization was identified. Specifically, evidence
suggested the absence of an empirically rigorous way to quantify the design trade-offs
between environmental, economic, and social sustainability. The notion of the competition
between several competing and contradictory variables, however, is not unique to architecture.
At is root, such a problem is a question of multi-variate optimization. That is, in order to
develop a method of addressing the problem identified infra, one must first seek to understand
methods of optimizing. This chapter explores one such method, the genomic algorithm, and
its potential to address such problems.
2.2 THE GENETIC ALGORITHM
Haupt (2004) defines the Genetic Algorithm as:
“…an optimization and search technique based on the principles of genetics and
natural selection. A GA allows a population composed of many individuals to evolve
under specified section rules to a state that maximizes the “fitness” (i.e. minimizes the
cost function).
Put simply, a minimum seeking algorithm in which a randomly generated sample of “solutions”
to a given problem (in our case, the configuration of a series of buildings) are evaluated by a
cost function (in our case, their social, economic, and environmental impacts), and the most
“fit” solutions are the refined through a series of “generations.” The process mirrors Natural
Selection in that in occurs “through repetitive application of the mutation crossover, inversion,
and selection operators.” (Gupta, 2012)
21
The exact invention of this process is not totally clear. Evolution-based computer simulations
occurred as early as 1954 in the works of Norwegian-Italian mathematician Nils Aall Barricelli
(Gupta, 2012). Australian quantitative geneticist Alex Fraser later, in 1957, published a series of
more widely circulated papers on the topic (Ibid.) Fraser, collaborating with Burnell published
a book on the topic in 1970, with further work done in the intervening decade by Crosby in
1973. Other pioneers include Hans-Joachim Bremmermann, Richard Friedberg, George
Friedman, and Michael Conrad (Ibid.) Where exactly these methods transformed into a
modern GA is not totally clear.
Haupt (2004) dates the creation of the modern genetic algorithm to John Holland in
1975 and the publication of his seminal book Adaptation in Natural and Artificial Systems.
Holland’s student, David Goldberg, used this method to solve a difficult problem involving gas
pipeline control and transmission in his 1989 thesis.
According to Gupta (2012), a genetic algorithm, in its most basic form, requires a
genetic representation of the solution domain (i.e. all possible solutions) and a fitness function
to evaluate the solution domain (i.e. some numerical method of assigning a value for
performance to each solution). Of course, not all problems have solutions which are best
evaluated on a single cost function. It is certainly possible to have a problem in which
optimization on single objective is paramount, e.g. finding the lowest amount of time in which a
task can be completed. It is, however, often the case that solutions are evaluated on myriad of
factors. Instead of merely knowing which solution is the fastest, you may need to know which
solution is the fasted and least expensive, or fastest, least expensive, and more efficient.
22
2.3 ADDRESSING MULTIPLE OBJECTIVES AND THE PARETO SURFACE
The simplest way in which to conduct multi-variate genetic optimization (i.e. with
multiple cost functions) is simply to weight each function according to its importance.
Problems often arise, however, when comparison occurs across multiple objectives which
cannot be simply weighted. As Kramer (2017) notes:
Without a decision for a weight of the objections, it is difficult to solve the optimization
problem. As no unambiguous comparison between solutions that are optimal
concerning at least one objective is possible, solutions of that kind are incomparable.
However, solutions that are worse in all objections are outperformed and from this
perspective useless. However, the challenge in multi-objective optimization is to
approximate a set of solutions that constitute a compromise between all objectives.
This set is also known as the Pareto-set.
Solutions in which the cost function for one objective can be improved without
deteriorating the performance of the solution on a different objective’s cost function are said to
be “dominated” (Ibid). Ergo, for problems with continuous variables and multiple objective
criteria, there are infinite possible set of solutions which together form the “Pareto Front”
(Haupt 2007). Generally, using the sum of weighted cost functions is impractical for finding the
Pareto front, as the necessitates running many possible cost functions weights and is often
computationally infeasible (Ibid.)
Fortunately, there are more sophisticated methods. These include Schaffer’s vector
evaluated GA, devised in 1984 (Ibid), Fronesca and Flemming’s multi-objective GA, devised in
23
1993, and non-dominated sorting GA’s of the first and second type (Ibid). Fortunately, the wide
applicability of these methods has, in recent years, lead to their inclusion in a variety of
software packages.
A visualization of the Pareto-front within a solution space can be found on the left of
Figure 3. As this illustration represents only two cost functions, f1 and f2, it is linear. As shown
in the right side of this figure, for a three-objective optimization, the Pareto front will actually
consist of a surface.
Figure 3: Pareto Front (Left) and Surface (Right) Image adapted from the University of Sheffield https://www.sheffield.ac.uk/acse/staff/peter_fleming/intromo
2.4 OPTIMIZATION IN BUILDINGS AND ARCHITECTURE
While genetic algorithms, both single-objective and multi-objective, can be applied to a wide
range of optimization problems, at issue here is their application towards architecture (in the
bricks and mortar sense of the word). Fortunately, a wide variety of architectural problems
have been explored with the GA to date.
Perhaps most common is the use of GA in optimizing energy performance. This may
because energy performance, being among the quantities metrics on which architecture is
evaluated, lends itself to the construction of straightforward cost functions. The simplest form
24
of environmental modeling is the single-objective variety, in which a single cost function (e.g.
energy use intensity) is evaluated for given variables. Congradac and Kulic (2009), for
example, use variables such as damper position, blower speed, and output temperatures to
optimize CO2 concentration control for the reduction in energy cost. Similarly, Kolokotsa
(2002) uses smart cards to evaluate user-reported human thermal comfort as a cost function by
which various parameters of an existing HVAC systems are optimized. Ooka and Komamura
(2009) examine a Tokyo hospital over a 24-hour period and optimize single-day energy
performance for variables of equipment size and configuration. In these cases, objectives vary
between minimizing energy use and minimizing thermal discomfort, but variables (the
operational parameters of an existing or theoretical HVAC system) are very similar.
In most cases, evaluation on an energy-related variable involves the integration of an
existing hourly energy simulation program. Alajami and Wright (2014) explored this connection
and identified the most important control parameters in single-variable configurations.
More complicated versions of energy optimization often try to account for cost. Znouda
et al (2009) conduct separate GA optimizations for environmental performance and cost of a
small structure in a Mediterranean climate. Variables in this example include window size and
façade orientation. However, this method is limited, as it uses two single-objective GA’s, and is
therefore incapable of considering multiple objectives and the trade-offs therefrom
simultaneously. Li (2017) surveys recent publications and reports that 68.4% of genetic
optimization strategies in published building optimization cases are of a single-objective type,
either basic or improved.
Others solve this issue through the use of multi-objective evaluation criteria in finding a
Pareto-optimal solution set for an architectural energy problem. Wright et al (2002), for
example, use a multi-objective GA with the dual objectives of operating cost and human
thermal comfort and variables including fan speed, volume flow, and air and water velocity
25
within the HVAC system. Similarly, Caldas and Norford (2003) explore the use of a non-
dominated sorting GA to optimize for the multiple criteria. The first optimize simultaneously for
first cost and operational cost of equipment, then for lighting and heating energy used,
considering primarily geometric variables in their effort.
Asadi et al (2014) use energy, occupant comfort, and retrofit costs in a school retrofit
context. Ascione et al (2015) focus more narrowly, evaluating only façade performance in a
Mediterreaan context. Like Asadi et al, they use minimization of energy and maximization of
thermal comfort as their dual objectives, electing not to consider cost. Yu et al (2014) also
consider energy use and occupant comfort, this time for a case of a typical building in China.
Chen and Yang (2017) opts to consider different types of energy, in this case, lighting, cooling,
and heating energy demand, as multiple variants in a high-rise design context.
These latter, multi-objective GA’s demonstrate the capability to find a Pareto-optimal set
of solutions to complex problems. However, their considering is often limited to a given
building or system, and each limit itself to only two (or in Asadi’s case, three) objectives at a
time. Each is similarly limited by having only energy and related variables (e.g. cost of
equipment, occupant thermal comfort) as cost criteria, perhaps because these can be
measured objectively more easily.
In addition to energy, it bears mentioning that structural performance has also been
considered through genetic optimization. Pourzeynali et al (2006) demonstrate that a fuzzy
logic controller combined with a genetic algorithm can be used to control an active tuned mass
damper successfully. Wu and Katta (2009) evaluate how best to reduce the surface to volume
ratio of given shapes, with the applied solution of reducing the total amount of material needed
for structurally viable stadium roof construction. Deutsch (2017) catalogs the use of GA’s by
practitioner Robert Vierlinger, who has used optimization to minimize the structural material
26
necessary for a space-frame-type system and also for a two-dimensional truss given a static
loading condition.
In the case of energy performance and structural performance examples enumerated
above, a wide variety of variables and objectives have been considered, mostly for objective
cost functions which can be more-or-less straightforward in their aims. However, even though
many of these solutions are nominally multi-objective genomic algorithms, they often fail to
consider more than one area of building performance. Some consider cost and structural
integrity, or cost and environmental performance, but rarely do they cross outside of their
discipline. For example, none of the above examples consider both structural and
environmental performance simultaneously. Additionally, all limit themselves to two variables,
preferring the comfort of a two-dimensional solution space and a Pareto-frontier to a three-
dimensional solution space with a Pareto surface.
There have been gestures towards more comprehensive ways of thinking that work
across disciplines. Wang (2004) evaluates both life-cycle cost and life-cycle energy
performance for variables including construction type, aspect ratio, and window-wall ratio. This
example simultaneously considers first costs, operating costs, initial energy performance, and
life cycle performance, expanding the complexity of the model. While only two objectives are
used (lifecycle cost and lifecycle energy performance) the complexity of these two cost
functions hints at the more complex nature of the problem at hand. Wang (2005) even
expands this work through the design of a building, however, it is a relatively simple single-
story office building. More recently, Karaguzel et al (2013) consider both lifecycle energy cost
and typical annual energy uses. While this hints at the possibility for life-cycle and cost-based
evaluation, it fails to account for building first cost and is applied only to a relatively simple
office building.
27
Lin and Gerber (2014) demonstrate an ability to use multi-variate GA’s as a more
complex design tool, employing it earlier in the design process. Their algorithm uses three
cost functions simultaneously, one each for net present value, energy use intensity, and
measure of spatial efficiency which they call “spatial programming compliance score.” Here
we see possibly the most complete gesture towards the complexity of trade-offs in building
design.
Of course, genetic optimization is not the only route by which optimization can be
undertaken with machine-learning methods. Delgram et al (2016) for example, use Particle
Swarm Optimization, an entirely different construct, for multi-objective building energy
performance. Latief et al (2017) use questionnaires and a case study to arrive at conclusions
regarding offsets between upfront cost, operational cost, and energy use. Papantoniou et al
(2014) use several methods simultaneously, combining some genetic algorithm elements with
an artificial neural network and multi-step optimization. Yet, it appears as though the Genetic
Algorithm is the preferred and prevailing method of building optimization.
2.5 OPPORTUNITIES FOR EXPANSION
While the examples above certainly offer compelling evidence of the power of this method, they
are not without their limitations. Two areas which demand further expansion are more inclusive
and comprehensive evaluation criteria, and consideration of problems at a much greater
physical scale.
To date, evaluation criteria have been many, however, it appears as though the
simultaneous consideration of the three nodes of the planner’s triangle has yet to take place.
No example readily available in the literature considers economic, environmental, and social
28
sustainability simultaneously. Additionally, each node of the planner’s triangle has room for
development in its own right:
Economic performance has, to date, been measured through first cost,
operational/energy costs, and/or net present value. Yet, economic sustainability
entails more than just these factors. A comprehensive evaluation of economic
sustainability must include the total cost of building, the expected cash flow from rent
or sale of the space during its useful lifetime, and the cost of operations and
maintenance.
Environmental performance has, to date, been measured through energy use intensity,
energy cost, and lifecycle carbon emissions associated with operational energy. Yet,
environmental sustainability entails more than just these factors. A comprehensive
evaluation of environmental sustainability must include embodied energy, operational
energy, and the energy associated with transportation necessary to support building in
a given way.
Social Sustainability has, to date, not been evaluated rigorously using these
optimization methods. This is likely due to the difficulty in quantifying social goals. A
comprehensive evaluation of social sustainability must first solve the problem of
evaluating what variable effect this outcome and how an objective cost function can be
constructed.
Physical scale has also been fairly limited in evaluation through this method. The
examples above work at a variety of scales, from the size of a single truss, to the size of a
room, to the size of a large hospital building. Yet, increasingly, architects are asked to think
about problems in the context of their larger urbanity.
29
In the following section, with deals with methods and outcomes, this thesis will offer a
more comprehensive view of how the method in question may be expanded to work at the
urban scale and to consider the evaluation of buildings more holistically.
30
CHAPTER 3: METHODS
3.1 EXPLORING A NEW METHOD
Chapter One argued that the architectural profession is sorely in need of new methods
by which empirically rigorous design may be accomplished. In Chapter Two, the genetic
algorithm and its application to architecture was considered as a mode of optimization for
solving such multi-faceted problems empirically. Chapter three will lay out a case to which
such an algorithm may be applied. This will be accomplished through finding an appropriate
scale case for intervention and designing algorithmic parameters by which the case will be
evaluated.
3.2 CASE DEFINITION
As shown in the previous section, modeled optimization problems in architecture have
tended to focus on the building scale. In some cases, this was the scale of a large building
(e.g. a hospital) and in others it was a small, theoretical building given few attributes (e.g.
single-story office). Also shown were examples of single component optimization (e.g.
trusses). If one were to envision the built environment as a spectrum stretching from the detail
level to more macro levels, as shown in Figure 4, one would likely consider these scales the
purview of Building Information Modeling. As Deutsch (2017) points out, the software tools
typically employed for optimization at this scale are quite different than the software tools used
for other scales. Software employed at this scale include the Rhino-integrated Grasshopper
and the Revit-integrated Dyanmo, which provide optimization capabilities within a BIM platform
for detail to building scale problems (Ibid.).
At the other end of the spectrum, the sustainability of urban systems has been explored
through the use of urban and regional-scale GIS systems. Deal and Schunk (2004) develop
the Land Use Evolution and Impact Assessment Model to answer questions about
31
environmental sustainability and Land Use policy. Even “below” the urban scale, Steinitz
(2012) framework for Geodesign shows promise in improving urban performance as evaluated
by many stakeholders using disparate criteria for large urban areas varying in size from
university campuses to portions of modern megalopolises.
Figure 4: Levels of Digital Explorations in Architecture: Adapted from Carl Steinitz (2012)
Ergo, enhancing multi-objective urban performance is taking place at the scale of inches in
grasshopper and at the scale of miles in ArchGIS. Yet, one cannot help but notice a gap but
notice a gap between these two strategies, as Architecture seldom evaluated in inches and
almost never quantified in miles.
Indeed, some of the most compelling development opportunities facing cities today live
in the wide gap between civic-scale geodesign and building-scale BIM. This is the realm of
neighborhood-scale optimization. A recent popular press article identified thirteen “mega-
developments,” occurring in just a single city, Chicago, at just one nexus in time (Koziarz
2019). Some of these projects amount to a single building or a handful of towers (e.g. Old Post
Office Redevelopment, Wolf Point Tower Construction). Others appear more visionary, ultra-
long-term proposals of civic transformation, involving multiple supertall buildings (e.g One
Central) or transformation of huge swathes of city (e.g. Former Southworks Site). Yet, several
projects are proposed at the scale of several city blocks: sites to large to consider as a
building, but too small to consider as an entire sector of city. These projects are as follows:
32
Lincoln Yards, The River District, Southbank and Riverline, The 78, Southbridge, and The
Burnham Lakefront.
Figure 5: Six Proposed Mega-Developments in Chicago. Adapted from Google Earth and Photos by Respective Developers
These developments are typical of the large-scale urban infill sites on which mega-
developments are often placed. Formerly, the sites were a Steel Industrial Facility, a
Newspaper Printing Plant, a Railroad Station, a Railyard, a Public Housing Project, and a
Hospital, respectively.
For the purpose of this investigation, consider The 78, a development proposed at the
southwest corner of Clark Street and Roosevelt Road in Chicago. This site was selected from
among its peers as it has a defined and publicly available program for the proposed $7.2 billion
33
dollar project on-site (Baiceanu 2019). This program includes a total of 13 million square feet
of gross floor area (Ori 2019). The site is also relatively adjacent to Downtown Chicago, which
will make finding economic comparisons to existing buildings easier. The 78 will provide the
lab in which it will be possible to test the idea of multi-variate optimization for social, economic,
and environmental sustainability at this scale.
In considering this site, the model will attempt to optimize a set of variables for three
objectives (social, economic, and environmental sustainability) given a set of practical
constraints. These components are to be configured as explained below.
3.3 VARIABLES
Given a site on which to operate, it is now possible to define which variables a model
will consider. It is possible to imagine an infinite number of discrete and continuous variables
associated with the construction of a development of such a massive scale. As the literature
review has shown, just one truss in one building can provide fertile ground for optimization.
However, considering such small building components would be impractical for a
neighborhood scale investigation. Instead, a more broad-strokes mix of program and height
will be more appropriate for the scale in question.
The first variable will concern the number of buildings associated with any given use.
While the proposed project has, to some degree, an established amount of gross floor area for
use, it is unclear how this balance was reached and if this balance is optimal for any given
objective. Possible use types will include Apartment Rental Housing, Condominium For-Sale
Housing, Affordable Apartment Housing, General Office Space, and Hospitality Space. These
five use types were selected as they seem to account for the vast majority of gross floor area in
the type of high-rise development proposed for this site. While it is typical for high-rises to have
some ancillary functions (e.g. Ground Floor Retail, Parking Podium), generally such buildings
34
are classified by their primary use as one of the five aforementioned use types. This variable
will be discrete (as one cannot have a fraction of a building) and will be limited to values
greater than zero in single increments of one. The maximum possible value for each type of
building will be ten, as this allows for fifty possible buildings (five uses times zero to ten
buildings per use) which would saturate the site. The maximum also accounts for the fact that
demand would likely be reduced with each new building. Even if the algorithm were to select
to build, say, forty hotels, this solution would not be appropriate as no market could support
such an abundance of hospitality.
The second variable will concern the optimal height for each type of building. Once
again, this variable will be discrete, as it will be measured in floors rather than in strict vertical
distance, and one story of a building cannot be sub-divided. Once again, an upper and lower
bound for this variable is appropriate. The lower bound considered will be ten stories, as
buildings shorter than this are generally not appropriate for the urban character of the site. The
upper bound will be 100 stories, as building above this height begins to strain credulity as to
economic and structural ability.
Height alone may initially seem an insufficient parameter to adequately describe the
scope of a building. Even a simple rectangular prism has three dimensions to be considered.
Yet, if one examines the common normative theory around each building type, one finds that
certain floorplate depths and lengths are often common to building types. As Residential and
Hotel Uses require relatively narrow floor plates, a typical building width may be on the order of
60’-0”. This allows for two 30’-0” units to be arranged across a double loaded corridor. Such
narrow floor plates allow the residential spaces access to natural light and affords suitable
views of the surroundings. It is possible that an optimal width for such a building may be 29’-0”
or 31’-0”, but determining the exact optimal in width is outside the scope of the present
35
investigation. The selected value allows for two or three individual spans of depths appropriate
to residential and hotel spaces.
Width alone is insufficient to define a floor plate area. Even if given a height in stories, in
order to define a floor plate area, one must also know the length of the building. For apartment
and hotel buildings, a length of 200’-0” is used for this investigation. This round number was
selected as it appears consistent with the typical high-rise residential or hotel floor area of
approximately 12,000 square feet. If one looks at recently completed building stock, one finds
the majority of such buildings have floor plates with similar dimensions to allow for economic
construction in post-tensioned flat-plate concrete. Once again, determining if 11,000 square
feet or 13,000 square feet would be more appropriate is outside the scope of the present
investigation.
If Rental, Condominium, and Affordable Housing are similar to Hotels in that shallow
floor plates are required, then it is also fair to note that in office buildings, generally much
broader floor plates are required. Typical Class-A offices have much longer “leasable spans”
than their residential counterparts. These spans tend to be column-free to allow for optimal
interior organization of a given office tenant’s space. For an office floor plate, an exterior span
is likely to be closer to 40’-0’. Similarly, these buildings are generally not organized around a
single double-loaded corridor, but rather a central circulation core. Ergo, let us assume that
the width of the building will consist of two 40’-0” lease spans to the exterior and a single 40’-0”
span for the interior core, providing a total building width of 120’-0”. Generally, office floor
plates are much larger, with buildings often being constructed of a composite structure
consisting of a cast-in-place concrete core with hot-rolled steel beams and columns and
composite decking. Given this structural system, one may assume a length of approximately
300’-0”, which creates a typical office high-rise footprint of 36,000 gross square feet.
36
As we have defined typical floors for each building type, then finding the gross floor
area of a given building is simply a matter of multiplying the number of stories variable by the
set foot-print size to arrive at a given floor area. Ergo, the variables for our model will be limited
to number of buildings of a given use, and optimal height for each use type. Each permutation
of solution will the be evaluated by a set of three objectives: economic, social, and
environmental. These objective functions are discussed in greater detail below. A summary of
variable structure is shown in Table 1 below
Table 1: Summary of Design Variables
Variable Lower Bound Upper Bound Number of Market-Rate Apartment Buildings 0 Buildings 10 Buildings Number of Condominium Buildings 0 Buildings 10 Buildings Number of Affordable Housing Buildings 0 Buildings 10 Buildings Number of Offices 0 Buildings 10 Buildings Number of Hotels 0 Buildings 10 Buildings Height of Market-Rate Apartment Buildings 10 Floors 100 Floors Height of Condominium Buildings 10 Floors 100 Floors Height of Affordable Housing Buildings 10 Floors 100 Floors Height of Offices 10 Floors 100 Floors Height of Hotels 10 Floors 100 Floors
3.4 ECONOMIC OBJECTIVE
The first of the three objectives considered is the Economic Objective. Put simply, this
is a maximum-seeking function which divides Profit by Construction Cost, endeavoring to
calculate simple return on investment. In reality, the construction timeframe for taller buildings
would necessitate having financing for a greater amount of time. However, a measure which
includes time of construction, such as Internal Rate of Return, would add a confounding
variable. Similarly, to a developer, all return would not be created equally, as the project may
be funded by debt, equity, and levels of mezzanine financing. This simple maximum-seeking
algorithm does not seek to divide the capital stack into its constituent parts and makes no
distinction between return on debt and return on equity.
37
Necessarily, the economic model used is not reflective of actual tall-building pro-forma,
which are typically multi-period models which evaluate cash flows over the process of land
acquisition, design, construction, and stabilization. A single-period simplified model has been
used for two primary reasons. First, many of the variables in the more complicated multi-period
pro-forma would be constant across multiple solutions. For example, any land acquisition cost
added to the “front end” of construction would, in the case of the project selected, not change
between design scenarios. This could have the effect of dampening the model’s
responsiveness to variables. Second, a more complex model may require evaluation input
which is not readily available and is outside the scope of this investigation. For example,
ascertaining how quickly rent rolls in residential buildings become stabilized may represent an
entire research project unto itself and is not sufficiently germane to the effects under
investigation to be exhaustively elucidated.
3.4.1 Building Costs
In considering the parts of this calculation, let us first consider building cost. For our
purposes, base costs were calculated on a per-unit-floor area basis using Marshall Swift Cost
Estimation Calculator Method (CoreLogic 2019). These costs are based on “final costs to the
owner and will include average architects’ and engineers’ fees.” (Ibid.) They include all
material and labor costs, sales taxes, interest on building funds during the period of
construction, normal site preparation including finish, grading, and excavation for foundation
and backfill for structure only, utilities from structure to lot line, and contractor’s overhead and
profit. This method explicitly excludes the cost of assembling lands, land planning costs,
special foundations, discounts and/or bonuses, off-site costs, furnishings and fixtures,
marketing costs, and general contingency (Ibid. 1.3).
Land costs are explicitly excluded from the model, as in the case considered they are
likely consistent across scenarios. Of course, the whole impetus for tall buildings involves the
38
scarcity of land and its intrinsic values in dense urban settings. Further research would be
necessary to incorporate the cost of land were the project to have variable site or footprint
characteristics.
For Residential and Hospitality Buildings, Class B Structure (Cast-in-Place Concrete) is
selected, as it is more economical on a per-unit-square foot basis. For Office Buildings, Class
A Construction (Steel Composite) is utilized. While it is less economical by 1.8%, it is more
commonly observed and suitable for office spans, as discussed infra. For market-rate housing,
Luxury High-Rise value for Good Quality is utilized. For Office, Good Quality is utilized,
intending to imitate Class-A office space. For Affordable Housing, Apartment High-Rise is
substituted for Luxury High-Rise, although Good Quality is still utilized. As dictated by the
calculator method, all values include an approximation of the cost of fire-sprinklers on a per-
unit-floor area basis. As this latter cost is dynamic with building size, a middle-height building
is utilized to avoid separately calculating this value for each potential building height.
However, due to the height of the buildings, calculating merely a square footage cost is
insufficient, as costs are dynamic and affected by building height. Costs tend to rise as the
building height rises due to the increase of structure necessary to ensure rigidity, the cost of
vertical communications, the cost of fire systems, and the cost of plumbing and electrical
systems (Zelazowski 2015). Inversely, groundwork, foundations, and roofing costs tend to
decease with height, as each assembly is being divided over a greater floor area (Ibid.). De
Jong and van Oss (2007) calculate the premium for each ten floors to be 8% on a unit-floor-are
basis. While this study focused on Holland and not Chicago, the model it offers is sufficiently
simple for the calculation method selected. Also, applying the 1.008 multiplier (an 8% premium
divided by ten total floors), a cubic relationship is achieved, rather than the simple linear
relationship recommended by Marshal and Swift. Figure 6 below shows the relationship
39
between height and cost created by this multiplier effect. Figure 7 shows how this impacts the
costs of each Building Type for a ten, forty, seventy, and one hundred story building.
Figure 6: Cost Premium per Height using Multiplier Method
Figure 7: Per-Unit-Floor Area Costs by Building Type and Height
100%
120%
140%
160%
180%
200%
220%
240%
1 11 21 31 41 51 61 71 81 91
Per
SF
Cos
t Mul
tiplie
r
Height in Floors
Cost Premium for Height
$-
$100.00
$200.00
$300.00
$400.00
$500.00
$600.00
$700.00
10 40 70 100
Cos
t per
SF
Height of Building in Floors
Cost/SF by Building Type and Height
Apartment Condominium Affordable Housing Office Hotel
40
Measured against total building cost is net present value. Net present value is
calculated as gross operating income divided by a capitalization rate. The capitalization rate
represents the ratio between a buildings income and value. Capitalization rate is taken from
the Cushman & Wakefield Mid-2017 Capitalization Rates by Property Type survey for
Multifamily and Office Buildings (Cushman & Wakefield, 2017). The capitalization rate for
Central Business District Class-A Office space in Chicago is approximately 6.67%. For
Multifamily Residential (which will be applied to Apartment, Affordable, and Condominium) a
value for Chicago is not provided, so 5.2% is utilized, approximating the mid-points of the
ranges provided for surrounding cities. For Hotels, the CBRE North American Capitalization
Rate Survey is utilized, showing a Central Business District, Full-Service Hotel Capitalization
rate of 8.28% (Baker 2019).
Capitalization Rate is applied to Net Operating Income (NOI). For our purposes, NOI
Consists of the Gross Floor Area multiplied by the efficiency off the floor plate, multiplied by the
occupancy percentage, multiplied by the per-net-square-foot annual revenue. For
condominiums, which are for-sale rather than rental, a simpler model accounting for sale price
rather than annual rent is utilized. Rents and sale prices are also cost-adjusted.
3.4.2 Building Revenues
In calculating the height premium for residential market-rate rentals, three buildings are
considered. In each building, the rental rate, unit size in square feet, and floor level are
gathered from the buildings website. As each of these buildings is complete, it is possible that
data gaps exist in the forms of units which have already been rented. It is also possible that the
units which have not been rented actually represent less desirable price points than those
which have been rented. However, since precise rental rates are generally not publicly
published, this investigation will assume that the advertised rates are an accurate
representation of market conditions.
41
The three buildings in question are NEMA Chicago, located at 1210 South Indiana
Avenue in Chicago, Essex on the Park, located at 808 South Michigan in Chicago, and The
Cooper Southbank, located at 720 South Wells in Chicago. While all projects are nearer to the
Central Business District than The 78 site, they represent the nearest new-construction
buildings with sufficient height for investigation for which unit prices, floors, and areas are
available. Figure 8 shows the per-unit-floor area rental price by floor for each building. Table 2
summarizes the number of points, y-intercept, and slope of each graph. To ensure each
project is weighted equally despite having an unequal number of data points, linear regression
is performed for each model, with the appropriate values averaged to create a “predictive”
relationship for height vs. residential rent. General Building information is provided by The
Skyscraper Center.
42
Figure 8: Relationship between Unit Height and Rental Price per unit Floor Area for Three Buildings
y = 0.0182x + 3.5215R² = 0.4652
$-
$1.00
$2.00
$3.00
$4.00
$5.00
$6.00
$7.00
0 10 20 30 40 50 60 70 80
Mon
tly R
ent p
er N
et S
F
Floor of Unit
NEMA Available Rents
y = 0.0108x + 3.3113R² = 0.1248
$-
$0.50
$1.00
$1.50
$2.00
$2.50
$3.00
$3.50
$4.00
$4.50
0 5 10 15 20 25 30 35
Mon
thly
Ren
t per
Net
SF
Floor of Unit
Cooper Available Rents
y = 0.0175x + 3.055R² = 0.3345
$-
$1.00
$2.00
$3.00
$4.00
$5.00
$6.00
0 10 20 30 40 50 60
Mon
thly
Ren
t per
Net
SF
Floor of Unit
Essex Available Rents
43
Table 2: Summary of Buildings and Regression Date for Residential Rental Model
Project Name Total Floors Number of Units Available Line Slope Y-Intercept
NEMA Chicago
81 Floors 121 of 800 Units 0.0182 3.52
Cooper Southbank
29 Floors 53 of 473 Units 0.0108 3.31
Essex at the Park
57 Floors 94 of 479 Units 0.0175 3.34
A similar technique can be utilized for residential for-sale units (Condominiums). In this
case, it is the per-unit-square foot advertised sales price which can be compared to the floor
height and area of any given unit. As condominiums often involve pre-sales, two under-
construction buildings were added to one existing building. The former use initial sale prices
while the latter uses units available for re-sale. While re-sale may be less accurate, due to the
limited number of condominium projects of suitable height which have been constructed in
recent years, it is necessary to utilize this data also. The projects selected are 1000 Michigan
located at 1000 South Michigan in Chicago, Legacy at Millenium Park, located at 60 East
Monroe Street in Chicago, and Vista Tower Chicago, located at 345 East Wacker in Chicago.
Figure 9 shows the regression analysis by building. Table 3 summarizes the collected data
used for the predictive model.
44
Figure 9: Relationship between Unit Height and Sale Price per unit floor area for Three Buildings
y = 12.89x + 447.98R² = 0.7762
$-
$200.00
$400.00
$600.00
$800.00
$1,000.00
$1,200.00
$1,400.00
$1,600.00
0 20 40 60 80
Sal
e P
rice
per
SF
Floor of Unit
1000M Price by Floor
y = 6.0716x + 377.09R² = 0.4101
$- $100.00 $200.00 $300.00 $400.00 $500.00 $600.00 $700.00 $800.00 $900.00
0 20 40 60 80
Sal
e P
rice
per
SF
Floor of Unit
Legacy Price by Floor
y = 14.594x + 509.24R² = 0.8607
$- $200.00 $400.00 $600.00 $800.00
$1,000.00 $1,200.00 $1,400.00 $1,600.00 $1,800.00 $2,000.00
0 20 40 60 80 100
Sal
e P
rice
per
SF
Floor of Unit
Vista Price byFloor
45
Table 3: Summary of Building and Regression Data for Residential For-Sale Model
Project Name Total Floors Number of Units Counted Line Slope Y-Intercept
1000 Michigan 74 Floors 31 Units 12.89 447.98
Legacy at Millennium Park
72 Floors 16 Units 6.07 337.09
Vista Tower Chicago
101 Floors 22 Units 14.36 509.24
For office rental rates, a suitable number of advertised suites of comparable sizes in
Class-A Buildings was not publicly available for review. While some prices were listed publicly
for some buildings, the sizes of the units, ages of the buildings, and other confounding
variables made calculation with the method used for residential rents and sale prices
impractical for office rents. Instead, for base prices, a media report from 2019 indicates that
office rental rates are now “over $40 per square foot” (Roeder 2019). This is consistent with
office listing site squarefoot.com’s assessment of Chicago Rentals, which indicates that Class-
A space in the West Loop, Central Loop, East Loop, River North, Fulton Market, and North
Michigan Avenue sectors are $51, $45, $43, $50, $53, and $44, respectively. For simplicity, we
will use a base price of $50.00 per net square foot.
As for premium for height, Barton and Watts (2013) note that while value is added per
floor for residential projects, the relationship in offices is less clear, opining that “There can be
little doubt that a tall office building attracts a premium of some nature, but articulating what
that is, and how it relates to particular levels, would seem to be less than straightforward.”
Koster et al (2011) similarly note that “To what extent taller building receive rent premiums, and
more specifically, whether there is a reputation premium, is unknown.” Their study, conducted
in Amsterdam, Brussels, and Utrect, indicates that “firms are willing to pay about 4% for a 10%
46
increase in building height.” Further, they find that a “landmark effect” for buildings at least six
times the average height is “about 17.5%” (Ibid.).
Clearly, some premium for height is warranted, however the precise magnitude remains
elusive. For the purpose of this investigation, it will be assumed that as height of the building
increases (measured by the height of the average floor, or the total height divided by two), the
total rent for all floors of the building will increase 0.725%. This accounts for some modest rent
premium while conceding that the premium is likely substantially less than that observed in
residential construction.
As with Offices, the relationship between the value of Hotel Rooms and their respective
height is unclear. Anecdotal evidence suggests that higher-floor rooms are more valuable and
command some sort of premium, but this effect could be due to the tendency to locate suites
or deluxe rooms on upper floors. In any case, while the existence of some premium for height
seems likely, there appears to be no support for such a notion in the literature. Therefore, for
the purpose of this exercise, a hotel room located on the tenth floor and one on the 80th floor
will be considered to command equal rent.
For affordable housing, rent does not appear to be a factor of market forces. Rather,
the Chicago Affordable Housing ordinance considers housing to be affordable if a household
earning 60% of the area’s median income can pay rent with 30% of the household’s gross
income (City of Chicago, 2014). In Chicago, 60% of Area Median Income is $41,041.80 for a
family of four. 30% of this amount is $12,312.54 per year, or $1,026.05 per month. In market-
rate rentals, the average unit size for units observed as part of the height investigation
conducted above indicated an average rental unit size of 972 square feet. This suggests a per-
unit-square foot affordable rent of approximately $1.05 per square foot per month.
Revenue for Hotel Rooms is calculated using published values for Chicago’s Average
Daily Rate and Occupancy for 2018. These values are $210.64 and 75.2%, respectively (Miller
47
2018). Multiplying these values provides RevPar, a per-key, per-night revenue expectation for
each hotel room. The number of rooms, which is to be multiplied by RevPar and 365 days per
year, is calculated by dividing the net square footage of the proposed hotel by the average
hotel room size times a multiplier to account for non-room spaces. While the average hotel
room size in the United States is approximately 330 square feet (Castillo 2015), this figure does
not include other spaces, such as meeting, lobby, amenity, and back-of-house spaces, which
still represent net square footage. Therefore, the net floor area in square feet of each
hospitality building is divided by 450 to achieve a reasonable estimate of the total number of
rooms. This figure does not include net-to-gross losses, which are discussed in greater detail
below.
3.4.3 Building Efficiency Loss and Vacancy Loss
The model must also account for the loss of efficiency implicit to tall buildings. As
building height climbs, a greater amount of space is needed for structure and vertical
circulation. Barton and Watts (2013) quote Langdon in providing values for buildings ranging
from 0-50 stories and efficiencies ranging from 81-84% for “short” buildings to 75-79% for taller
buildings. Sev and Ozgen (2009) add another ten data points by measuring the floor plate
efficiencies of ten tall buildings around the world with floor counts varying from 69-110 and
efficiencies ranging from 60% to 77%. Considered together, these studies provide fifteen data
points upon which linear regression can be performed. The implied efficiency relationship is
shown in Figure 10 below. It is worth noting that both surveys consider Office Buildings, which
have different structural and vertical circulation requirements from the other building types
considered. However, due to lack of efficiency and height data for other building types, the
office relationship will be utilized, creating a possible source of error. It is also worth noting that
the values in question are on a per-floor-plate basis and do not consider mechanical spaces,
48
lobby spaces, etc. For the purpose of this investigation, these spaces are assumed to impact
all buildings equally.
Figure 10: Floor Count vs. Floor Plate Efficiency with data from Barton and Watts (2013) and Sev and Ozgen (2009)
Of course, even converting gross floor area to net floor area using an efficiency factor
does not result in a number of square feet which will ultimately be rentable or salable, as some
space will be vacant as a structural cost of doing business. Vacancy for hotel rents has been
discussed above. Vacancy for for-sale units is considered to be a risk to the owner and not the
developer, and thus is not considered. Vacancy rates for Office and Rental Properties,
however, must be accounted for. For the purpose of office vacancy, resources from early 2019
indicate that cumulatively Class-A office vacancy in Chicago’s Downtown Market are
approximately 13% (Ecker 2019). Apartment vacancy can be estimated through the Census
American Community Survey program, which most recently reports an apartment vacancy rate
for Chicago of 6.18%. Ergo, for the purpose of this exercise, offices will be said to be 87%
occupied while apartments will be said to be 94% occupied.
3.4.4 Operating Expenses
y = -0.0014x + 0.8145R² = 0.4913
0.5
0.55
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0 20 40 60 80 100 120
Height vs. Floor Efficiency
49
In order to devise a Net Operating Income, one must subtract expenses from revenue.
Of course, building expenses are complicated, multi-faceted, and vary by building type, size,
and management. Ergo, this study will apply a simple per-unit-floor area or per-unit expense
model, depending on type, to devise operating expenses. For for-sale residential
(condominium) buildings, ongoing expenses are considered to be the purview of the Home
Owner’s Association and therefore are not considered. Hotel operating expenses are
calculated on a per-key basis using the 2017 value for U.S. Limited Service Hotels annual
expenses of $21,390 per operating room (Mellen 2019). Rental and Office expenses are
calculated on per-unit-floor-area basis. For Apartments, the National Apartment Association
Survey data suggests annual operating expenses of $9.12 per square foot, excluding Capital
Expenditures (Munger and Smith, 2019). The Building Owners and Managers Association
reports per-square-foot operating annual expenditures for Office Buildings at $12.47 per square
foot, including Cleaning, Utilities, Fixed Expenses, Parking Expenses, Roads and Grounds
Maintenance, Building Repair and Maintenance, and Real Estate Taxes (BOMA 2018).
As a matter of necessity, operating expenses are simplified to reflect information that is
readily available and directly affected by building square footage. It is not intended to be an
exhaustive catalog of the way in which buildings effect the profitability of organizations. It may
be, for example, that the greatest “operating expense” for any building is in fact the staff that
work within the building. It is possible that changing architectural variables could change staff
retention or other dependent variables not considered that have a greater impact on
organizational profitability than the simple maintenance costs observed here. The relationship
between these broader economic variables is important to understand, but it outside the scope
of the current investigation.
3.4.5 Economic Cost Function
Considering all above factors, the cost function for each building type is as follows:
50
Maximize RR = ((NOI / Cap) – IC) / IC
Where: RR = Rate of Return
NOI = Net Operating Income
Cap = Capitalization Rate
IC = Initial Cost
For rental projects, Net Operating income is further defined as follows:
NOI = SF x Eff x Rent x Occ (SF x Opp)
Where: SF = Gross Floor Area
Eff = Floor Plate Efficiency
Rent = Revenue per net square foot of floor area
Occ = Occupancy Percentage
Opp = Operating Expenses per gross square foot of floor area
For For-Sale Residential (Condominium), no capitalization rate or net present value
calculation is necessary. Instead, the sale price is simply divided by the initial cost to find the
rate of return.
3.5 ENVIRONMENTAL OBJECTIVE
Stated plainly, the goal of the environmental objective is to measure life-cycle
greenhouse gas emissions associated with operational energy, embodied energy, and
transportation energy for any given solution. While it is certain that environmental objectives
outside of the reduction of greenhouse gas emissions exist (e.g. water conservation, air
quality), for the purpose of this investigation, the “common thread” in measuring each type of
environmental impacts will be tons of carbon dioxide equivalent gases emitted over a fifty year
time-frame.
51
The environmental objective is to minimize the total emission. It does not consider what
alternatives these buildings may be replacing. For example, it is possible that by adding
occupants in a dense urban setting removes those occupants from more carbon-intensive
means of living, thereby providing a net environmental benefit. However, comparing the
relative carbon intensities at different densities in this manner is beyond the scope of this
investigation, which seeks only to minimize net lifecycle carbon emissions. For the purpose of
this exercise, adding one more square foot or one more occupant will always raise carbon
output. This raises the question of whether or not the model will simply try to minimize the total
gross floor area, rather than selecting for the most appropriate environmental option. This
exigency is dealt with using the penalty functions explained in Section 3.7 below.
While a lifecycle model is used for the environmental objective, the economic and social
objectives rely on a simpler “snapshot” in time and do not consider the entire lifetime of the
buildings. This is the case primarily because current and future emission can be discussed in
the same terms (i.e. tons of greenhouse gasses) while talking about economic or social
performance longitudinally is more difficult. For example, considering the lifecycle economic
impact would require knowledge of the discount rate to account for the time-value of money,
the escalation rate for construction expense, and changes in consumer demand over time.
Such lifecycle analysis may be fertile ground for future research, but here it is not considered.
3.5.1 Operational Energy
Operational energy is approximated using the New Construction Commercial Reference
Buildings offered by the Department of Energy’s Energy Efficiency and Renewable Energy
Office (2012). These reference buildings offer Energy Plus input files for ASHRAE 90.1
compliant buildings of sixteen types. For the purpose of this exercise, the most recently
updated reference buildings, dated November 13, 2012, were utilized.
52
Building simulations are provided for each ASHRAE Climate Zone. For this problem,
Climate Zone 5A, the zone corresponding to Chicago, was selected. The models chosen for
residential projects was the “Midrise Apartment” model, with a gross floor area of 3,135 square
meters. Office was simulated with the “Large Office” model, which measured 46,320 square
meters. Hospitality used the “Large Hotel” Model, encompassing 11,345 square meters of
gross floor area. In all cases (except perhaps the Large Office), the square footages of the
reference buildings are likely smaller than the solutions proposed. This can be partially
accounted for by using a per-unit-floor-area technique for calculating total operational energy.
However, as smaller buildings may expend less energy on certain elements (e.g. vertical
circulation), use of the reference buildings remains a possible source of some modest error.
Simulation results are provided in terms of kilowatt hours of electricity and therms of
natural gas for each model. While this is provided in Site Energy, a site-to-source factor is also
provided, which is 3.546 for Electricity and 1.092 for Natural Gas (Ibid.). Carbon equivalencies
for both Electrical Generation and Natural Gas are calculated using the Environmental
Protection Agencies Greenhouse Gases equivalency calculator (EPA 2019). A total emissions
per gross square feet of floor area by energy source is shown in Figure 11. Multiplying the
gross floor area by the above per-unit-floor-area emissions for a span of fifty years yields the
total carbon dioxide equivalent emissions for the lifetime of the building.
53
Figure 11: Greenhouse Gas Emissions per Unit Floor Area for Subject Building Types
3.5.2 Embodied Energy
In addition to the operational energy used by each building, there exists some
embodied energy associated with the construction of the building and the support
infrastructure surrounding the building. For the purpose of this investigation, embodied energy
will be limited to the embodied energy of the building’s themselves and will not include
surrounding infrastructure. While a greater density of buildings will likely require more
infrastructure, some base level of road and utility access will be required regardless of the
height of the building. Determining the marginal increase in infrastructural embodied energy
associated with each modality is outside the scope of this research. Likewise, some material is
likely to be replaced as part of periodic capital expenditures or ongoing operations and
maintenance within the building within the 50-year time horizon considered. Once again,
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
Apartment Condo Affordable Office Hotel
CO
2E E
mis
sios
per
Yea
r pe
r S
F Fl
oor
Are
aPer SF GHG Emissions by Building Type
Electricity Gas
54
determining the percentage of replacement and the embodied energy of these materials is
outside the scope of this investigation.
To simplify the embodied energy calculation, an adjusted per-unit-floor-area metric is
used. Thiel et al (2013) compare the embodied energy of a net-zero building to five existing
case studies. For the purpose of this study, the five case study buildings are of greater interest
than the comparison itself. The uses of the five buildings considered range from Office to
Laboratory and Residential Space. Their heights vary from three to 38 floors, with only a single
building being greater than six floors in height. Their structural systems and envelope materials
are also varied, providing a useful average. Evaluation of these precedents shows a
distribution centered around 350 kg of carbon dioxide equivalent emissions for each square
meter (Ibid.). Converted to imperial units, this suggests as “baseline” figure of 0.325 tons of
CO2E per gross square foot.
As explained above, taller buildings require a disproportionate amount of materials to
account for exponentially growing lateral loads and the requirements of vertical circulation.
This “premium for height” phenomenon must also be accounted for in embodied energy.
While heating and cooling of one square foot of space on the tenth and ninetieth floor of a
given building require roughly the same amount of operational energy, the energy required to
build these square feet varies substantially. Foraboschi et al (2013) calculate the embodied
energy of buildings of 20-70 floors in height using different structural systems to estimate the
premium for height for embodied energy. Their paper also includes an investigation of the
impact of some lightweight floor systems. For the purpose of this paper, the most interesting
results are those for steel-concrete and reinforced concrete buildings. One of the three
lightweight systems considered is also included in the calculations below. A summary of
findings is presented in Figure 12.
55
Figure 12: Embodied Energy Premium For Height, Calculated based on Values from Forabachi et al (2013)
Performing linear regression on five height cases in this study provides a road-map on
how to scale embodied energy per unit height. It is worth noting that the referenced work
includes only the columns, core, primary and secondary beams, and floors. It is likely that
including only these structural elements represents a “worst case scenario” for premium for
height, as one might expect finish and envelope materials to have a lesser premium or no
premium at all. Unfortunately, information on premium for height for non-structural materials is
not well defined in the literature.
3.5.3 Transportation Energy
Of the three types of energy considered herein, transportation is the most difficult to
quantify and requires a myriad of assumptions about transportation use. A meaningful effort to
quantify transportation may begin with an effort to number quantity of people using each type
of building and assigning a an amount of transportation energy to each occupant.
y = 0.0098x + 0.7616R² = 0.933
0%
20%
40%
60%
80%
100%
120%
140%
160%
0 10 20 30 40 50 60 70 80
% In
crea
se in
EE
per
uni
t flo
or a
rea
Number of floors
Structural Embodied Energy Premium for Height
56
For residential projects, occupancy is based on the investigation of unit sizes and
locations conducted in Section 3.4 above. This investigation showed an average rental unit
size of 972 net square feet. Condominium (for sale) units were found to be much larger. These
units measured, on average, 2173 net square feet per unit. No distinction was made between
affordable and market rate rental unit sizes. To calculate occupancy, the net floor area of each
proposed building was multiplied by the occupancy of the building and then divided by the unit
sizes above. Each unit was said to have 2.6 occupants, the national average U.S. Household
size (Fry 2019). It is possible that multi-family households have a different number of people or
that regional variation exists in this number, but accounting for this variance is outside the
scope of the present investigation. It is also possible (perhaps even likely) that some share of
such urban multi-family units are only partially occupied, being second homes or peid-á-terres
for the wealthy. For the purpose of this exercise, all units are assumed to be occupied full time.
For offices, the Commercial Real Estate Development Association tracks the number of
net rentable square feet per office employee, finding most recently that the average office has
181 net square feet per employee (Ponsen 2019). For hotels, the number of rooms is
estimated by the same method as in the revenue model, by dividing the total gross square
footage by 450 square feet. This is then multiplied by 1.6 persons for room. This accounts for
both a guest in the occupied room and staff, which is estimated at 0.5 to 1.0 persons per room
(Vallen and Vallen 2018). The cumulative net floor area for each occupant (not accounting for
vacancy loss) is shown in Figure 13.
57
Figure 13: Net Floor Area per Occupant, not accounting for vacancy loss
Given the ability to calculate the number of occupants, the next step is calculating how
much of each occupant’s transportation energy should be dedicated to the proposed
development. For office, 40 hours/week is a safe assumption. Given the total of 168 hours in
each week, office users transportation emissions are to be assigned at a rate of 24%. Similarly,
Residential use fraction is estimated at 121 hours per week (72%) and hospitality use is
estimated at 84 hours per week (50%). Multiplying per capita fuel use by these usage fraction
ensures that only a portion of the fuel consumed by the average user is attributed to the
development. Stated another way, if a given person works in the development, but does not
live there, only 24% of his/her fuel usage will be accounted for, corresponding to the fraction of
his/her time estimated to be spent at the development. Figure 14 shows time usage fractions
for each of the subject building types.
305
843
305
181
281
0 100 200 300 400 500 600 700 800 900
Apartment
Condo
Affordable
Office
Hotel
Net SF per Occupant by Use Type
58
Figure 14: Use Fraction multiplies applied to fuel consumption of each building type
According to the US Department of Transportation (2019), per capita fuel consumption
in Chicago, IL measures 206 gallons of gasoline per person per year and 37.42 gallons of
diesel fuel per person per year. The Environmental Protection Agency (2019) greenhouse gas
calculator reports that each gallon of gasoline burned releases 0.008887 metric tons of carbon
dioxide while each gallon of diesel fuel burned releases 0.010180 metric tons of carbon
dioxide.
3.5.6 Environmental Cost Function
The total environmental impact of the project can be given by the following equation:
Minimize GHGTOTAL = GHGOPP + GHGEE + GHGTRANS
Where:
GHGTOTAL is total Greenhouse Gas Emissions
GHGOPP is total Operational Energy
0%
10%
20%
30%
40%
50%
60%
70%
80%
Apartment Condo Affordable Office Hotel
Percentage of Transportation Use for Each Type
59
GHGEE is total embodied energy
GHGTRANS is total Transportation Energy
Each of the three terms represents the summation of the total of all buildings in a given
solution.
3.6 SOCIAL OBJECTIVE
While energy and economics can be evaluated in bottom line terms (in greenhouse gas
emissions and rate-of-return, respectively), it is much more difficult to discuss the abstract
notion of social good in terms of a single, bottom-line value. Social good tends to be
amorphous and difficult to assign a singular, objective value.
Social benefit may also be realized through design at a scale which is not feasible to
measure in the current model. It could be, for example, that having common spaces more
readily visible would enhance social interaction, but the variables considered in the model are
agnostic to this quality. Even where the scale is appropriate, the direct social benefits are not
always directly measurable. For example, it may be that having taller buildings which result in
leaving more of the site “open” for recreation or public use would be the most equitable
outcome. However, the relationship between open space preservation and social equity is
difficult to quantify. It is unclear, for example, what portion of land should be reserved for
public use and if this percentage would relate linearly to beneficial social outcomes.
Complicating matters further is that the known determinants of social sustainability do
not necessarily impact the variables in the neighborhood-scale model. Kamail and Hewage
(2015) identify twelve social “Sustainability Performance Indicators,” yet several of these
indicators are not of suitable scale for considering here. For example, the indicator “cultural
and heritage conservation” would not appear to apply to the vacant site considered in this
investigation, at least not in as much as it would affect height and use choices. “Safety and
60
security” are likewise difficult to relate to the height and use variables, as it appears possible to
have safety and security in any arrangement selected.
Maleki et al (2019) evaluate the social sustainability specifically for high-rise buildings.
In addition to safety and security, they identify sense of belonging to place, comfort, and
aesthetics as criteria of social sustainability. Each criteria is then assigned further
determinants. Determinants for comfort, for example, include thermal comfort, user flexibility,
healthcare, noise pollution, natural lighting, and indoor air quality. While it may be easier to
achieve thermal comfort or natural lighting with a given density of buildings, these variables
would seem to be more suited to building-scale investigation. Healthcare availability and noise
pollution seem unlikely to be factors which would be impacted by the number, height, and type
of building selected. Perhaps a more complete analysis could identify connections between
building massing and, for example, noise pollution, but such investigation is outside the scope
of the current problem.
Stender and Walter (2019) develop a social sustainability matrix by which to evaluate
existing projects on three categories: accessibility, social cohesion, and participatory process.
One again, ideas like inclusion and participation may be integral to social sustainability, but
their connection to the selection of building types and heights is, at best, poorly understood
and at worst, totally intractable. The Framework for Strategic Sustainable Development seeks
to measure social well being and at least claims to have some measure of external validity
(Broman and Robert 2015). In addition to some ecological measures, this framework evaluates
“structural obstacles to health, influence, competence, impartiality, and meaning-making.”
Removing obstacles to learning, free speech, safe working conditions, and the creation of
common meaning would seem to be laudable goals. However, their aim appears to be more a
question of social structure than of building massing.
61
Murphy (2012) review eight bodies of literature contributing to “the social pillar of
sustainable development.” These include United Nations and European Union Policy
Documents, Multi-lateral Sustainable Development Indicators, Social Sustainability Literature,
Green Social Policy Literature, and Environmental Justice Literature. Murphy proposes four
“organizing dimensions” common to much of the literature: equity, awareness for sustainability,
participation, and social cohesion. These dimensions seem likely to be impacted to some
degree by density and building use, but Murphy’s investigation appears focused on policy
implications rather than on elucidating design objectives for the built environment.
More interest in urban forms, perhaps, is shown by Eizenberg and Jabareen (2016) who
identify Urban Forms as one of four major aspects of social sustainability, along with Equity,
Safety, and Eco-Prosumption. The paper recommends urban forms which promote “a sense
of community, place attachment, a sense of safety, and healthy communities.” While these
objectives may seem nebulous, several concrete steps are provided, including compactness,
density, mix of uses, and passive solar properties. However, little is done to validate the
connection of these steps to social outcomes. While it is argued that “these typologies
compose the structure of desired sustainable urban form,” too little empirical evaluation of
these metrics is provided to establish a relationship between these concrete recommendations
and optimizing neighborhood-scale problems.
Some determinants of social sustainability do consider the height and use variables in
subject model, but are otherwise unsuitable for measuring social sustainability meaningfully.
Du et al (2017) compare life satisfaction (“one of several key measures of social sustainability”)
between low and high-density urban settings, finding that high-rise residents reported higher
life satisfaction when controlling for demographic variables. However, the type of housing
explained less than 5% of the observed variance. Furthermore, even controlling for
demographic variables, such a small margin could easily be explained by another confounding
62
variable, such as access to urban amenities or more robust urban employment prospects. An
argument that greater height leads to greater satisfaction does not appear to be supported by
the literature.
Al-Kodmany (2018) also appears interested explicitly in the social sustainability of tall
building developments, even relying on the economic, social, and environmental framework for
defining sustainability. This framework, to its ultimate credit, identifies several means of
improving social sustainability in tall buildings including improving resistance to terror attacks,
minimizing disparity in the quality of life of residences, and even cleaning windows to allow for
visual connection to the public realm. It is certainly possible that height and mixes of uses
affect resiliency and public connectedness. However, it is also possible to argue that buildings
of any of the proposed heights and types can offer seismic and fire safety. Similarly, buildings
in any configuration considered in the subject optimization algorithm can have clean windows.
Perhaps the most valid means of evaluating the social sustainability ramifications of the
chosen design objectives is not offered by any singular framework, but rather draws on the
recurrent themes in a number of frameworks. The following section will argue that Housing
Affordability, Employment Opportunity, and Tax Base generation are three measurable ways of
assigning value to a given solution within the proposed framework. These three determinants
are not selected arbitrarily, but rather represent recurring themes in the literature of social
sustainability evaluation.
3.6.1 Housing Affordability
Of the social criteria considered, perhaps the most directly supported in the literature is
the call for some component of affordable housing, by this or another name. Kamali and
Hewage (2015) cite affordability among their Social Sustainability Performance Indicators
(although not explicitly affordable housing). Stender and Walter (2018) explicitly include
affordable housing in the accessibility portion of their Social Sustainability Framework. The
63
provision of shelter is cataloged among signs of the development of equity in Murphy’s (2012)
analysis, which also notes that housing provision is among the Social “Themes” of the United
Nations Commission for Sustainable Development. Eizenbern and Jabareen (2017) suggest
inclusion of a greater diversity of housing types and incomes promotes Social Sustainability.
Al-Kodmany (2018) cautions against the creation of tall buildings solely as “mansions in the
sky,” which are said to “reinforce social and racial segregations.”
The question then is not whether or not affordable housing is important to Social
Sustainability, but rather how it can be measured convincingly in the present model.
Enhancements to the Affordable Requirements Ordinance (City of Chicago, 2014) provide a
useful frame for defining affordable housing for rental units as “affordable for households
earning up to 60% of the Area Median Income… for a term of 30 years.” Given the high-rise
mix is likely to consist of one through three-bedroom apartments, HUD’s Maximum Affordable
Rent for 2019 for the City of Chicago suggests monthly rent of $928 for a one-bedroom, $1,110
for a two-bedroom, and $1,281 for a three-bedroom (where tenant pays utilities) (HUD 2019). If
one assumes a precise unit mix is unknown, then taking 30% (the fraction of net income
allowable for housing cost) of 60% of Chicago’s Area Median Income (as suggested by the
ARO), yields a monthly rent of $1,026.25. Dividing this by our average rental unit size of 972
net square feet, this suggests a per-square-foot, per-year rent of $12.67 (in line with HUD
figures).
Considered conceptually, some percentage of the 13,000,000 square feet allotted for
development can be spent on affordable housing units, with the bounds being 0% (no
affordable housing component) and 100% (entire development consists of affordable housing).
Ergo, if one multiplies the gross square footage of the affordable units provided by the
efficiency factor of height of the building purposed, and divides the results by the total number
of possible affordable units (given by dividing the total development gross square footage by
64
the average unit size), one may arrive at a percentage of total development expended on
affordable housing. Given the support for affordable housing in the literature relative to the
other two metrics discussed below, provision of such housing is to be weighted doubly in
calculating the social sustainability score.
3.6.2 Employment
More abstract than the direct provision of affordable housing is the notion that socially
sustainable development should offer prospects for employment, thus contributing directly to
the economic well-being of the host city. Kamali and Hewage (2015) suggest that contribution
to the local economy is a key Social Sustainability Performance indicator. Broman and Robert
(2015) indicate that social sustainability requires that people “are not subject to structural
obstacles to competence [including] obstacles for education and developing competence
individually and together.” These goals, while abstract, can certainly be read to suggest the
importance of employment.
Others are more direct in connecting employment to social sustainability. Murphy’s
(2012) literature review of the Pillar of Social Sustainability finds numerous suggestions of the
importance of employment opportunities. Stender and Walter (2018) include employment
along with education amongst their Accessibility categories in their social sustainability
framework. Eizenbern and Jabareen (2017) identify a connection to employment, although
they see it as a “non-physical” component of social sustainability.
Measuring prospects for employment could be accomplished by simply counting the
number of full-time-equivalent jobs generated by each building. However, a simple count
would be mathematically difficult to compare to the percentage values given for the other two
measures of social sustainability (affordable housing above and tax base below). Therefore, as
with affordable housing, a “percentage of possible total jobs” is suggested. It then becomes
necessary to define how many “jobs per square foot” are offered by each building type.
65
Staffing for multi-family buildings appears to vary widely, with the precise number of
staff based on the total number of units in the building and the level or service provided to each
unit. For rental properties, a very rough ratio of one staff person per fifty rental units is
suggested by some contemporary accounts (Champagne 2012) to account for maintenance,
leasing, and site management personnel. This ratio seems suitable for market-rate-rental and
affordable properties, where the number of staff can be calculated by dividing the net square
footage of the building by the average unit size by 50 units per staff person to find a total
number of staff. For-sale multi-family will still requirement management and maintenance, but
will likely not require ongoing leasing personnel. Therefore, for the sake of simplicity, the
model will use a smaller ratio of 75 units per staff. Along with the much higher average unit
size, this makes for-sale residential buildings much less employment intensive on a per-unit-
floor-area basis.
As stated above in calculating Office Census, Ponsen (2017) suggests a ratio of 181 net
square foot of office space to one office employee. Ergo, the number of office employees per
net square footage of office space can be found by simple division. This turns out to be the
most employment-intensive type of use considered here. Using the value of 450 net square
feet per hotel key calculated above, a number of hotel rooms per building area can be
calculated. Vallen and Vallen (2018) suggest a room-to-staff ratio of 0.625 for select service
hotels. This means the number of employees for hotels can be calculated by dividing the net
floor area by the floor area per room and multiplying by 0.625. Figure 15 shows the net floor
area per employee for each building type.
66
Figure 15: Employees for a 1,000,000 NSF Building of Each Use Type
Using the above figures, a number of employees can be calculated for each building
use case. As shown in Figure 15, Office use is the most intense case. Ergo, office use
represents the maximum potential employment. To yield a percentage value, consider that the
maximum possible employment for the development would be if all buildings were 100% office
buildings. Such an arrangement would yield a total employment figure of approximately
71,823 employees. All other employment configurations can be measured as a percentage of
this total possible. By measuring as a percentage, this value can be compared directly to the
values for percentage of possible affordable housing and percentage of possible tax base.
3.6.3 Tax Base
Moving from the concrete to the abstract, it can be seen that affordable housing and
employment represent key measures of socially sustainable development. Still more abstract,
though, is the idea of generating suitable property tax base. While the benefits of jobs and
20.58
6.14
20.58
5524.86
3555.56
1.00 10.00 100.00 1000.00 10000.00
Apartment
Condo
Affordable
Office
Hotel
Employees (log scale)
Employees per 1,000,000 Net SF by Use Type
67
affordable homes can be seen immediately as social goods, the idea of developing more value
for taxation is at least a second-order social good. That is to say, the immediate consequence
of taxation is not particularly socially beneficial, but the schools, transportation network, and
redistributive policies that taxes pay for can be of immense social benefit.
Of the Social Sustainability Performance Indicators identified by Kamali and Hewage
(2015), many appear immediately connected to benefits typically funded through taxation.
Specifically, the health of occupants, influence on location social development, cultural and
heritage conservation, safety and security, and neighborhood accessibility and amenities are
directly connected to taxes. Landmarks Ordinances, Police Patrols, Parks, and Public Transit
access all are funded, to some degree, through local taxation. The security against crime, fire,
and earthquake recommends by Maleki et al (2019) can be seen as analogues for well-funded
police departments, fire departments, and code enforcement officials. Amenities, Connection,
and Safety are keys to the Social Cohesion component of Stender and Walter’s (2018) Social
Sustainability Formwork. Once again, it seems a key provided of physical access, physical
safety, and neighborhood-scale amenities is the local government. The Framework for
Strategic Sustainable Development (Broman and Robert 2015) promotes removal of obstacles
to free speech, education, and partial treatment. Such justice and education outcomes are
inexorably tied to locally-funded government. The participation and social cohesion axes
identified by Murphy (2012) and the safety and equity axes identified by Eizenbern and
Jabareen (2016) appear similarly connected to the public sector.
If taxes are seen as a net social benefit, then one must ask what type of physical
development will generate the most revenue. For the purpose of this investigation, property
taxes are considered, as they provide the sort of local funding for the services enumerated
above. Obviously, income taxes connected to employment and capital gains taxes connected
to real estate development may also be impacted. As the connection between such taxes and
68
local services is less clear, they are considered outside the scope of this investigation.
Similarly, the calculation of property taxes is fairly opaque, will local millages varying by block
and the assessed value of a given property being calculated via a complicated formula
involving revenue, expenses, and exemptions. For the purpose of this model, the assessed
value of a building on which taxes will be levied will be said to be identical to the net present
value of the building for simplicity. Similarly, while individual owned housing units, multi-family
properties, and office properties may all be taxed at different rates in different areas of the city,
the reported average general tax rate of 6.786% will be utilized.
Once again, for the mathematical necessity of expressing the potential taxation as a
value between zero and one (or as a percentage), the ratio of the tax liability of a given solution
will be compared to the largest possible tax which could be collected, operating on the thinking
that any solution will offer between 0 and 100 percent of the maximum possible tax benefit. By
“flexing” the model across a number of different scenarios, the maximum possible net present
value per square foot was found to be $573.98. Therefore, the maximum possible tax collected
is given by multiplying the prevailing tax rate by this net present value per square foot times the
13,000,000 total square feet.
3.6.4 Social Cost Function
The calculated social sustainability factor of the project can be given by the following
equation:
Maximize SSF = ((2 x RAFF) + REMP + RTAX) / 4
Where:
SSF is the calculated Social Sustainability Factor
RAFF is the ratio of Affordable Housing in a given solution to the maximum
possible amount
REMP is the ratio of Employment Provided to the maximum possible amount
69
RTAX is value of Tax Collected to the maximum possible amount
The algorithm will seek to maximize this value. That is, the algorhitm will favor values
nearer to one and farther from zero, with zero and one being the minimum and maximum
possible values respectively.
3.7 CONSTRAINTS
The variable outcomes of the model are constrained in two ways. First, the variable
values themselves must meet certain criteria. Second, and more complexly, there exist penalty
functions on each variable which seek to “move” the solution set towards 13,000,000 square
feet as solutions are generated.
The simplest way in which the possible outcomes are constrained is via the “outside
brackets” on the variable values themselves. Each building type can have only zero to ten
buildings. Each building must be at least ten stories tall and a maximum of 100-storys tall.
These values are chosen to ensure that density suitable to the urban context is achieved and
that possible solutions are not outside the realm of possibility. Both variables are treated as
discrete. That is, it is impossible to have one half of one floor or one third of one building, so
generated solutions feature only whole numbers.
Because the model seeks to provide outcomes which are closest to 13,000,000 square
feet of gross floor area (the proposed square footage of The 78 Development), penalty
functions are introduced to each variable. Put simply, solutions in the initial population are
based on a random number generator. Thus, while they will fall inside the ranges provided,
there is no way to ensure that they sum to the correct number of total square footage. The
penalty function applied to each cost function serves to “nudge” each successive generation
towards the desired number of square feet by apply a proportional “handicap” to its cost
function scores.
70
For the environmental cost function, for each square foot greater than 13,500,000
square feet or less than 12,500,000 square feet, a penalty of 10 tons of carbon is imposed.
This is intended to “direct” the evolving scenarios towards the appropriate square footage.
Therefore, if a proposed solution contained 14,000,000 square feet and generated 20,000,000
million tons of carbon dioxide equivalents, it would be adjusted by the difference between 14
and 13.5 million times ten, assigning this solution a value of 35,000,000 million tons of carbon
dioxide equivalents. A similar principle is used for the Sustainability Score cost function, albeit
of a much lower magnitude. In these cases, each square foot subtracts 1.0 x 10-7 from the cost
function outcome. The same value is applied to the return on investment cost function in the
economic model.
These values are calibrated specifically for the problem at hand. Using penalty
functions which are too large causes the model to converge on solutions with the correct floor
area without sufficient regard to their performance in the cost function categories. Using
penalty functions too small does not penalize solutions with too small or too large of floor areas
enough, allowing them to potentially become first-rank solutions.
3.8 MODEL SCENARIOS
In the following chapter, data will be provided for four scenarios. The parameters for
each of these scenarios is expounded upon in the first portion of Chapter four below. Three
single-objective scenarios will evaluate a breadth of solutions based on each of the cost
functions enumerated above (one each for Social, Environmental, and Economic
sustainability). Settings for the single-objective runs are shown in Table 4 below.
71
Table 4: Single-Objective Genetic Algorithm Settings
Setting Value Algorithm Standard Genetic Algorithm Seeds and Repetition Random Seed (Mersenne Twister) Termination Criteria Perform 50,000 Function Evaluations Initial Population 500 Selection Method Tournament Selection with Tournament Size 10 Crossover Method Double Crossover with a Probability of 0.7 Mutation Method Jump (Probability 0.1) and Creep (Probability 0.75) with Elitism of
1 The final, multi-objective scenario incorporates each of the three cost functions into a single
run.
Table 5: Multi-Objective Genetic Algorithm Settings
Setting Value Algorithm Non-Dominated Sorting Genetic Algorithm III Seeds and Repetition Random Seed (Mersenne Twister) Termination Criteria Perform 1,000,000 Function Evaluations Initial Population 500 Crossover Probability 0.5 Mutation Probability 0.5 Mutation Rate 0.02 Mutation Sigma 0.1
An understanding of the general operation of the model and the underlying cost
functions will help to contextualize results. A brief discussions of the limitations of this method
is included with Conclusions in Chapter Five.
72
CHAPTER 4: RESULTS
4.1 MODEL PARAMETERS
In the following chapter, data will be provided for four scenarios. These scenarios
consist of three single-objective genetic algorithm runs and one multi-objective genetic
algorithm run. The three single-objective problems represent one each of Economic,
Environmental, and Social Performance. The final multi-objective run represents all three
variables considered simultaneously.
For the single-objective runs, a standard genetic algorithm is used. Due to the number
of initial solutions that will be unsuitable due to having too great or too small of a floor area, a
large initial population size of 500 is utilized. The model uses tournament selection with a
tournament size of ten. The model relies on mutation to avoid early convergence, selecting the
“correct” winner of the tournament 100% of the time. An aggressive crossover probability of
0.7 is used to ensure a breadth of novel solutions. Both ‘Jump’ and ‘Creep’ mutations are
utilized with a probability of 0.1 and 0.75 respectively, allowing mutations to both change
solutions incrementally and in more substantive ways. This should allow for searching of
“nearby” solutions while also prevent convergence at a relative minimum rather than a global
minimum. The model is set to stop after 50,000 function evaluations, or approximately 141s
generations.
For the multi-objective run, a NSGA-III Algorithm is selected. The robust initial
population of 500 members is once again utilized. Crossover probability and mutation
probability are set to 0.5, meaning that in each successive generation, one half of the
population will consist of crossovers of the best solutions from the previous generation and one
half of the total population will receive some mutation. The mutation rate is 0.02, meaning that
two percent of each “mutants” “genes” will be changed. Sigma, which controls the mutation
73
magnitude, is set to 0.1. The model is set to run for the number of total calculations
corresponding to 100 generations.
Results are centered around four scenarios. Scenario A through C are single-objective
scenarios evaluating optimum performance for economic, environmental, and social cost
functions respectively. Scenario D is a multi-objective scenario evaluating all three cost
functions simultaneously. The results for each of the four scenarios is preceded by some pre-
scenario information. This consists of additional information of interest about individual
buildings rather than neighborhood scale implementation. This latter information was gleaned
from the model without any application of optimization algorithms.
4.2 PRE-SCENARIO INFORMATION
Even prior to the execution of any optimization, it is possible to use a more “brute force”
approach to interrogate the model for certain questions about individual buildings. For
example, while the model will search for a combination of buildings and heights which will best
fulfill the given scenario, it may be of interest to investigate the relationships between individual
buildings and their height to better understand the more complex scenarios. Three such
questions are answered below:
4.2.1 Relationship Between Building Height and Profit.
The first pre-scenario evaluates the economic portion of the model to ascertain at what
height buildings generate the most profit on a per square footage basis. This question
balances the rent or sale price premium for height vs. the reduced floor plate efficiency and
increased costs associated with building taller buildings. A summary of profitability per
building for heights at ten floor increments is provided in Figure 16 below.
74
Figure 16: Per Unit Floor Area Profit per Building by Type and Height
As can be seen in this figure, it appears Apartment and Condominium Buildings have
the tallest “optimum height” for the model as presently constructed. This logically follows the
parameters on which the model is based, where apartment rent and condominium sale price
consider height premiums. That is, apartments and condominiums on taller floors are more
valuable, which tends to offset the increase in price for taller buildings to some degree. A
similar effect is visible in Office Prices, where some rent premium is commanded, but the
premium is less as a percentage of construction costs than that of the market-rate residential
uses. Premium for height is not considered in the model for hotel rooms or affordable housing,
as the relationship is unclear in the former case and in the latter case, rent is based entirely on
income and does not vary by floor height. The model appears to suggest an optimal
profitability for market-rate residential projects between 30 and 50 floors in height. Certainly,
within Chicago, there has been no shortage of such construction. However, the model also
suggests that shorter Offices and Hotels are preferable from a profit perspective. This notion is
$(400.00)
$(300.00)
$(200.00)
$(100.00)
$‐
$100.00
$200.00
$300.00
10 20 30 40 50 60 70 80 90 100
Profit/SF by Height for Five Building Types
Apartment
Condo
Affordable
Office
Hotel
75
less supported by recent building, which has included numerous hotel and office projects of
heights greater than ten stories.
4.2.2 Relationship Between Building Height and Return on Investment
Of course, the cost function by which economic outcomes are evaluated is not simple
profit, but rather return on investment. Profit alone does not consider how much money must
be invested to access economic gains. Certainly an investor would rather invest ten dollars for
a two dollar return rather than investing one hundred dollars for a three dollar return, despite
the fact that net profit is greater in the second case.
Figure 17 shows the relationship between return on investment and height for each
building type. Once again, this accounts for the premium rents and for-sale prices paid for
higher floors, which is weighed against the greater cost and reduced efficiency of such
building.
Figure 17: Return On Investment by Building Type and Height
‐150.0%
‐100.0%
‐50.0%
0.0%
50.0%
100.0%
10 20 30 40 50 60 70 80 90 100
ROI by Height for Five Building Types
Apartment
Condo
Affordable
Office
Hotel
76
When one considers the increased amount of money which must be paid “up front” to
construct such tall buildings, the profitability advantage previously observed in market rate
residential projects of greater height appears to evaporate entirely. While greater net profit may
be made on a market-rate apartment tower of forty floors than twenty floors on a per-unit-floor-
area basis, the initial investment in the forty-story building is so much higher, it outweighs the
modest gains in profitability expected.
Were the above information 100% true, it is difficult to envision a world in which any
high-rise would ever be built for profit motive. This suggests that some of the variables driving
high-rise construction are not totally accounted for in the model. Such variables which are not
considered may include land cost, the cost of infrastructure development, and the prestige and
brand awareness associated with taller buildings.
4.2.3 Relationship Between Greenhouse Gas Emissions and Building Height
Just as the model makes it possible to evaluate individual buildings economic
performance based on height, so to is it possible to evaluate economic performance. The net
greenhouse gas emissions produced by each building type and height are show in Figure 18.
This figure accounts for the embodied, operational, and transportation energy as explained in
the previous chapter.
As one may expect, greenhouse gas emissions appear to show a linear relationship
with building height. As transportation, embodied, and operational energy were evaluated on a
per-unit-floor area basis, it stands to reason that the greater the floor area, the greater the
measurement of such emissions. A more interesting way to evaluate the relationship between
height and environmental performance may be on a per-unit-floor area basis. Figure 19
provides such an evaluation.
77
Figure 18: Total Greenhouse Gas Emissions by Building Type and Height
Figure 19: Per Unit Floor Area Greenhouse Gas Emissions by Building Type and Height
0
1000000
2000000
3000000
4000000
5000000
6000000
7000000
10 20 30 40 50 60 70 80 90 100
Tota
lLife
cycl
e G
HG
(M
T C
02E
)
Building Height
Lifecycle GHG Emissions by Building Height and Type
Apartment Condo Affordable Office Hotel
0
1
2
3
4
5
6
10 20 30 40 50 60 70 80 90 100
Life
cycl
e G
HG
Em
issi
ons
per
Net
SF
(MT
C02
E)
Building Height
Lifecycle GHG Emissions per Net SF by Building Height and Type
Apartment Condo Affordable Office Hotel
78
Figure 19 appears to show that even accounting for the obvious fact that
buildings with greater floor area under roof will use greater amounts of energy, height still
comes at a premium in terms of energy consumption. This stands to reason, given that the
taller the building, the less efficient the floor plate. Spaces for hallways and vertical circulation
must still be heated, cooled, and illuminated. Thus, the fact that more such spaces are
included in taller structure makes the environmental performance worse on such building per
unit of rentable floor space. Note that each residential modality conforms so closely that their
representative lines “cover up” one another.
This simple investigation is not intended to imply that taller buildings are intrinsically
less sustainable (although that may be the case). Rather, it is intended to illuminate what the
model does not consider. For example, the construction of tall buildings may make the city
denser, allowing for more efficient transportation networks and reducing the embodied energy
of infrastructure. Any such gains are not captured in the model and may provide fertile ground
for further investigation.
4.3 SCENARIO A: ECONOMIC OPTIMUM
The first scenario in which the use of an evolutionary algorithm is undertaken involves a
single-objective optimization for economic performance. The parameters for both model and
cost function are explained in Chapter 3 above. The average and best fitness of the solution
pool is shown in Figure 20. The performance of the “best ever” solution for each generation in
shown in Figure 21. This represents only the portion of the algorithm which was necessary to
find the best solution, plus ten generations for context. In reality, the model was allowed to run
for several more generations after convergence had occurred to see if any creep or jump
mutations might slightly improve the solution. In this run, the best solution was found in
generation 107.
79
Figure 20: Best and Average Solutions by Generation for Single-Objective Economic Model
Figure 21: Best Solution by Generation for Single-Objective Economic Model
-1.400
-1.200
-1.000
-0.800
-0.600
-0.400
-0.200
0.000
0.200
0.400
0.600
1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 100
103
106
Rat
e of
Ret
urn
(Hig
her
is B
ette
r)
Generation
Economic Performance by Generation
Best Average
0.200
0.250
0.300
0.350
0.400
0.450
0.500
0.550
1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 100
103
106
Rat
e of
Ret
urn
(Hig
her
is B
ette
r)
Generation
Best Economic Solution
80
As shown in the figures above, while the solutions generated as part of the initial
population were not terribly fit, the model seems to have worked correctly in selecting fitter and
fitter outcomes, without converging prematurely on the preferred solution. The optimum
solution had a cost function value of 0.503. Since the optimum solution is between 12,500,000
square feet and 13,500,000 square feet, the cost function is not penalized and thus its value of
0.503 corresponds to a return on investment of 50.3%. The “optimum” solution selected is
shown in Figure 22, where one bar represents one building and the height of each cluster of
bars represents the optimal building height.
Figure 22: Optimum Solution for Single-Objective Economic Model
Perhaps unsurprisingly, the optimum economic solution contains no affordable
housing, as the rent for such housing is considerably lower than other building modalities. The
inclusion of Hotel and Office buildings of lesser height is also unsurprising, given that lower
heights are where returns are maximized for these typologies where rent is less height sensitive
or not height sensitive at all. That the apartment buildings are taller is also explicable as a
0
10
20
30
40
50
60
70
80
90
100
Apartment Condo Affordable Office Hotel
Heigh
t (Floors)
Building Use
Economic Factors Only
81
matter of return. As shown in Figure 22, while Hotel and Office building return is maximized at
lower heights, this is less true for apartment buildings, which are economical to build taller due
to the rent premiums commanded for taller units. A single condominium building is included,
but this is likely only because the model built near the maximum ten apartment buildings and
adding height to these buildings would harm their economic prospects due to increasing costs
and lowering efficiencies. Since the model seeks to reach 13,000,000 square feet, a single
condominium building is added.
4.4 SCENARIO B: ENVIRONMENTAL OPTIMUM
The second scenario considered is the single-objective optimization for environmental
performance. As explained in Chapter 3, environmental performance in this case means the
reduction of greenhouse gas to the lowest levels allowable while maintaining a square footage
in the desired range. Figure 23 shows the average and best solutions through the generations
which were necessary to find the best solution (plus ten generations for context). Figure 24
shows the evolution of the best solution alone over this timeframe. Solutions are measured in
lifecycle tons of carbon-dioxide equivalent emissions.
82
Figure 23: Best and Average Solutions by Generation for Single-Objective Environmental Model
Figure 24: Best Solution by Generation for Single-Objective Environmental Model
0
20,000,000
40,000,000
60,000,000
80,000,000
100,000,000
120,000,000
140,000,000
160,000,000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27LIfe
cycl
e G
HG
Em
issi
ons
(Low
er is
Bet
ter)
Generation
Environmental Performance by Generation
Best Average
13,000,000
13,500,000
14,000,000
14,500,000
15,000,000
15,500,000
16,000,000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27
Life
cycl
e G
HG
Em
issi
ons
(Low
er is
Bet
ter)
Axis Title
Best Environmental Solution
83
Unlike the Economic Model, which took 97 generations to find the best outcome, the
best outcome found by the Environmental Model was located in only the 17th generation and
was not improved upon in subsequent generations. It is unclear if this is due to a more well
calibrated penalty function or if it simply shows luck in the pool of initial solutions generated. In
any case, the lowest greenhouse gas emissions by a qualifying model were found to be
13,394,795 MT of CO2E. This optimum solution is shown in Figure 25.
Figure 25: Optimum Solution for Single-Objective Environmental Model
Once again, the result here is somewhat unsurprising given what is known about the
model. As Figure 25 above shows, residential uses have the lowest per-unit-floor area lifecycle
energy consumption when compared to office and hotel uses. The preference for
Condominium over Apartment and Affordable Residences is explained by the larger footprint of
the units in each condominium. In these buildings, there are fewer persons per square foot,
less the transportation emissions, which are associated with the number of people in the
building, are reduced. Since condominiums offer the mix of low operational energy and fewer
0
10
20
30
40
50
60
70
80
90
100
Apartment Condominium Affordable Office Hotel
Heigh
t (Floors)
Building Use
Environmental Factors Only
84
traveling occupants, their square footage is almost maximized. Since this model is not
concerned with the costs associated with greater height, much taller buildings are selected.
The embodied energy of taller condominium buildings appears to be overwhelmed by their
other “energy saving” characteristic, as seen by the model.
Market-rate apartments and Affordable Housing should be somewhat interchangeable
in terms of energy use associated, so the models preference for the former appears to be a
matter of chance. The inclusion of a single, minimum height office building is also unexpected,
given the energy use intensity of these buildings. It is possible that the model here has found a
very good solution, but that a solution without ten stories of office building is actually the
solution with the overall minimum cost function result.
4.5 SCENARIO C: SOCIAL OPTIMUM
The third scenario executed is a single-objective case which looks for the optimum
configuration resulting in the highest “social score,” as defined in Chapter 3 above. Figure 26
shows the average and best solutions over the course of the generations necessary to find
optimum. Figure 27 shows only the best outcome over the course of the subject generations.
As with the other single-objective models, ten additional generations are included for context.
85
Figure 26: Best and Average Solutions for Single-Objective Social Model
Figure 27: Best Solution for Single-Objective Social Model
-0.700
-0.500
-0.300
-0.100
0.100
0.300
0.500
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43
Soc
ial S
usta
inab
iltiy
Fac
tor
(Hig
her
is B
ette
r)
Generation
Social Performance by Generation
Best Average
0.320
0.325
0.330
0.335
0.340
0.345
0.350
0.355
0.360
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43
Soc
ial S
usta
inab
ility
Fac
tor
(Hig
her
is B
ette
r)
Generation
Best Social Solution
86
In this model, the number of generations necessary to find the best solution located is
somewhere between the possible “lucky guess” of the Environmental Model and the more
laborious process of the Economic Model. The best social score was found in the 34th
generation of the algorithm run shown above. This solution had a cost function value of 0.354,
corresponding to a social score of 35.4%. This solution is shown in Figure 28.
Figure 28: Optimum Solution for Single-Objective Social Model
Given the focus of the Social Score on providing Affordable Housing, it is unsurprising
that the majority of buildings in this configuration are Affordable Apartments. Given the
importance of jobs, the inclusion of offices, which are the most job-intensive of the modalities
considered, also stands to reason. The provision of a single, very tall apartment tower is a little
bit more of a mystery.
Apartment buildings have tended to have the highest net present value, thus generation
the highest taxes. Yet, because taxation is based on income net of costs and expenses, it
87
seems as though several shorter apartment buildings would have performed better. Given the
declining efficiency of taller buildings, it is possible the model is trying to generation as few
market-rate units as possible in an effort to avoid reducing the percentage of housing that is
affordable. Since the model considers the percentage of total housing units that are affordable
rather than the absolute number of units, an aversion to construction of market-rate units would
be expected.
The fact that this model does not appear averse to tall buildings is unsurprising, as it
does not consider the cost of generating such buildings. The inefficiency of taller building floor
plates may be the only factor limiting the model from selecting even fewer even taller buildings.
4.6 SCENARIO D: MULTI-OBJECTIVE OPTIMUMS
The final scenario involves the simultaneous optimization for all three variables:
economic, environmental, and social, as explained in Chapter 3 above. From the population
size of 500 individuals, generation 100 offers 161 novel, first-rank solutions. This means that as
of this generation, 161 solutions are non-dominated and can not be improved for any one
objective without harming their performance on another objective.
The data are difficult to visualize in a single manner, as each represents multiple axes.
With that said, it may make sense to consider each variable individually, then consider the
relationship between two variables, and finally visualize all three variables simultaneously. To
consider each variable individually, Figure 29 provides a histogram for the 161 novel solutions
according to their evaluation by each of the three cost functions. These data have not been
normalized, but the environmental performance has been reduced by 10-7 to allow for legibility.
Each set of solutions is split into even deciles.
88
Figure 29: Histograms showing Distributions of First Rank Solutions by Objective Cost Function
89
Generally, the model appears to have found non-dominated solutions which maximize
economic performance, as the first histogram above shows a distribution weighted towards
high return on investment. The same can be said of environmental performance, with the
carbon impacts of many of the solutions minimized. The least skewed distribution belongs to
the social objective, where the highest social scores were achieved in relatively few solutions
relative to higher economic and lower environmental scores.
This phenomenon may suggest that it is more difficult to harmonize social objectives
with economic and environmental objectives, which are more easily harmonized with one
another. If one normalizes the entire data set to values between one and zero (with one
representing the best outcome and zero the worst for each cost function), a cursory look at the
data suggests that this may be correct. Of the 161 solutions, the highest performing social
solution has the worst performance for both economic and environmental objectives. This was
not true for the best economic and environmental scoring outcomes. The best economic
outcome in the data set had moderate environmental and social performance. Likewise, the
best environmental outcome in the dataset has poor economic performance but fairly strong
social performance.
This relationship can be seen more clearly by examining the variables two-at-a-time to
identify the correlation between the best performing solutions. Figure 30 shows the
relationship between normalized environmental and economic scores. Figure 31 shows the
relationship between normalized economic and social scores. Figure 32 shows the
relationship between environmental and social scores.
90
Figure 30: Non-Dominated Solution Set by Economic and Environmental Performance
Evaluation of the solution set on only economic and environmental variables shows that
the variables may be somewhat correlated, but that the correlation does not appear terribly
strong. While few solutions have poor scores in both economic and environmental
performance, solutions with strong performance in one category may have only moderate
performance or even weak performance in another. Given the right-weighted distributions seen
in the histograms above, the trade-off between environmental and economic performance does
not appear as straightforward as the other comparisons below. Some solutions were able to
perform well in both economic and environmental score.
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Nor
mal
ized
Env
ironm
enta
l Sco
re
Normalized Economic Score
1st Rank Solutions: Economic vs. Environmental
0 - 20% Social 21 - 40% Social 41 - 60% Social 61 - 80% Social 81 - 100% Social
91
Figure 31: Non-Dominated Solution Set by Economic and Social Performance
A much stronger correlation is shown between economic and social scores. Solutions
with strong economic performance tend to have poor social performance. Likewise, solutions
with strong social performance tend to have poor economic performance. The absence of any
data points in the upper right of the graph suggests that no solution was able to harmonize
economic and social performance terribly well. This suggests that the trade-offs between
economic and social performance are more direct that the trade-offs between economic and
environmental performance.
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Nor
mal
ized
Soc
ial S
core
Normalized Economic Score
1st Rank Solutions: Economic vs. Social
0 - 20% Env. 21 - 40% Env. 41 - 60% Env. 61 - 80% Env. 81 - 100% Env.
92
Figure 32: Non-Dominated Solution Set by Environmental and Social Performance
The relationship between social and environmental performance appears to be neither
as strongly correlated as that between economic and social performance nor as loosely
correlated as that between economic and environmental performance. While the bareness of
the lower left quadrant in the graph does suggest that few solutions had poor performance on
both categories, some solutions appear to have performed fairly well on both categories. This
would suggest that environmental and social performance are not mutually exclusive.
However, some trade-off between environmental and social performance is observed.
Finally, it is possible to visualize all three performance criteria at once. Figure 33
generates a plane for economic and environmental performance from which prisms are
projected upwards. The height of these prisms represents social performance.
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Nor
mal
ized
Soc
ial S
core
Normalized Environmental Score
1st Rank Solutions: Environmental vs. Social
0 - 20% Econ 21 - 40% Econ 41 - 60% Econ 61 - 80% Econ 81 - 100% Econ
93
Figure 33: Normalized Economic, Environmental, and Social Performance of Non-Dominated Solutions
The solutions appear to form a “surface” as would be expected for a Pareto-optimal set.
Observations drawn from examining any two variables at once appear to be reinforced by
seeing this whole point cloud/surface simultaneously. First, economic and environmental
performance do not appear mutually exclusive, with many solutions performing well on both
economic and environmental grounds. Second, economic and environmental performance
appear to have some trade-off, but this did not prevent some environmentally strong solutions
from having adequate-to-good social performance. Third, economic and social performance
appear to have the strongest negative correlation, with good performance in both categories
appearing to be mutually exclusive.
With an understanding of which variables appear to affect one another most closely, the
largest question remaining is what the individual solutions look like. Aggregated information
about the solutions can help to draw some conclusions regarding the way in which the model
was trending. Such population information is provided in Table 6.
94
Table 6: Population Information for First Rank Solutions
The non-dominated solutions appear to have selected from a wide range of possible
building number combinations, especially for residential projects, which show the greatest
degree of variation between total building counts. For offices and hotels, the variation is less,
with no solution opting for more than five offices or three hotels. The average number of hotels
is also noteworthy. While the range of outcomes was between zero and three, the average of
only 0.04 suggests that most solutions had no hotel use at all. Hotels may be hampered by
providing relatively poor environmental performance and relatively few jobs. Offices have
stronger employment numbers, but their weak environmental and economic performance likely
contributed to their exclusion as a major part of many solutions.
By number, the highest average and median belong to apartments, then
condominiums, and finally affordable housing. Apartments offer strong economic performance
and ecological performance that is better than offices or hotels. Condominiums are a slightly
weaker economic prospect but have superior environmental performance due to their lower
number of occupants. Affordable housing was likely selected due solely to its impact on social
variables.
In terms of building height, the tallest building of any solution was a 97-story apartment
building. Generally, apartments had the highest average and median heights, possibly
because of rent premium paid for heights in residential projects. It is also worth noting that
market-rate residential had the highest standard deviation, suggesting a wide-spread between
the tallness of these buildings between non-dominated solutions. Affordable housing tended to
Apart.
Number
Condo
Number
Afford.
Number
Office
Number
Hotel
Number
Apart.
Height
Condo
Height
Afford.
Height
Office
Height
Hotel
Height
Minimum 1 1 0 0 0 46 35 28 20 19
Maximum 9 8 8 5 3 97 72 53 59 42
Average 7.78 4.23 2.76 2.04 0.04 80.94 45.65 38.61 25.39 36.23
Median 8 4 2 2 0 85 45 39 24 36
StdDv 1.15 1.83 1.99 1.36 0.33 13.22 6.43 5.67 5.00 3.52
95
be lesser height, likely because rents are not responsive to taller units. Hotels and offices built
only in the most economic (i.e. least expensive) cost ranges.
Of course, information about the population is insufficiently granular to understand any
given solution. It may also be of interest to examine several “middle of the road” solutions to
see their composition. For this exercise, three solutions were selected. These scenarios are
not necessarily preferable to any other scenarios. Aside from meeting the criteria outlined
below, their selection from the pool of non-dominated solutions is a matter of chance. In all
three solutions, no objective had a normalized score of less than 0.42. Solution A had
Economic, Social, and Environmental scores of 0.60, 0.42, and 0.74 respectively. Solution B
had scores of 0.79, 0.52, and 0.53. Finally, Solution C had scores of 0.51, 0.84, and 0.54.
Each solution managed strong performance of at least 0.74 in its strongest category. Those
strengths were Environmental for Solution A, Economic for Solution B, and Social for Solution
C. The solutions are shown in Figures 34, 35, and 36 respectively. These are not intended to
be an exhaustive view of the solution space, but only to represent how three points, selected
with some empirical criteria but otherwise randomly, “look” in terms of model outcomes.
96
Figure 34: Solution A Building Types and Heights
Figure 35: Solution B Building Types and Heights
0
10
20
30
40
50
60
70
80
90
100
Apartment Condo Affordable Office Hotel
Heigh
t (Floors)
Building Use
Solution A
0
10
20
30
40
50
60
70
80
90
100
Apartment Condo Affordable Office Hotel
Heigh
t (Floors)
Building Use
Solution B
97
Figure 36: Solution C Building Types and Heights
Each solution is emblematic of the greater solution set in several ways. First, no
solution relies on hotels, which were rarely used within the body of solutions. Second, the
tallest buildings are residential, with the tallest among those being apartments. Third, offices
are used in varying degrees but are not as tall as residential buildings. The number and height
of residences likely owes to their economic strength and rent premiums for height. The offices
have a lower premium for height, and thus are typically built in a more cost effective range.
At least for these “middle-of-the-road” solutions, buildings have a fairly substantial
variety in type within each solution. The number of buildings between solutions is somewhat
consistent, with the greatest difference being the inclusion of more Offices in Solution C.
Between solutions, the heights of each building type are remarkably uniform.
Of course, this represents only three of the 161 possible non-dominated solutions from
the 100th generation of the model run. More variety may be captured by selecting other
0
10
20
30
40
50
60
70
80
90
100
Apartment Condominium Affordable Office Hotel
Heigh
t (Floors)
Building Use
Solution C
98
solution candidates, especially if less care was given to ensuring the selected candidates
performed at least moderately well on each of the three objectives.
4.7 SCENARIO VISUALIZATIONS
A simple number and type of buildings, while of importance to the programming and
design process, is not in and of itself a design. To contextualize one proposed solution and
demonstrate how such information may be realized on the actual site in question, so
visualizations have been conducted. Figures 37 and 38 show two such visualizations.
Figure 37: View of Scenario A from south along Wells/Wentworth Connector
99
Figure 38: View of Scenario A from northwest along the Chicago River
100
CHAPTER 5: CONCLUSIONS
The course of this investigation began with identifying a problem: the lack of balance in
the present architectural discourse and profession. It continued by examining one empirical
method of resolving the kind of competing objectives faced by architects: the multi-variate
evolutionary algorithm. It then proposed a specific, neighborhood-scale problem and a
method by which such an algorithm might by applied. Next, it explored the results produced
by the algorithm. Finally, it must offer some evaluation of the success or failure of this
endeavor.
5.1 OUTCOME EVALUATION
Just as we began with the Planners Triangle as guide to sustainable balance in Chapter
1, we may now return to this diagrammatic notion as a means of evaluating the method put
forward. Figure 45 shows the Planner’s Triangle updated with the single-variable optimization
solutions envisioned in Scenarios A through C and the multi-variable optimization solutions
reviewed in Scenario D. While this is not intended to be a literal placement of where each
solution might graphically fall, it does offer a means of understanding what each solution to the
problem considered has sought to do.
While the changing nature and economic realities of architectural practice may have
forced the architect to place a disproportionate amount of importance on profitability, the
method identified in this investigation offers one possible solution as to how the architect’s
objectives may be realigned. By quantifying the desired objectives, understanding the trade-
offs between those objectives, and modeling a means of empirically balancing competing
demands, this investigation has sought to demonstrate that balance itself is possible with the
right tools.
101
This investigation was applied to a new scale, considering more than a building but less
than a city. This scale was attacked through the coopting of the mega development, which
offers the opportunity to make neighborhood-scale decisions in a relatively narrow timeframe.
The proposed method seems suitable to the scale selected, given the limitations discussed
below.
Figure 39: Optimization Solutions on the Planner's Triangle
102
5.2 LIMITATIONS
The proposed method unfortunately does not offer a panacea for all that ails
architecture. It is, after all, only as good as the information which is used to make the
underlying decisions. Should the variables change from the number and height of buildings to
the orientation or massing of buildings, it would be difficult to make such a mid-course
correction. Likewise, if rents should drop from (for example) $50.00 per square foot to $45.00
per square foot, the utility of the model may very well be compromised. Clearly defined inputs
and means of evaluation appear to be required for this method. It seems as though, in the day-
to-day trenches of architectural practice, such clear-cut problems may rarely present
themselves.
The method also relies on a willingness to use empirical means to solve problems.
Even without the encumbrance of complex algorithms, it seems that many design decisions
made today do not use the best information available. Even where progress has been made
towards solving design problems, this progress seems to be often overlooked in pursuit of
more familiar ways of building.
The narrowness of the problem considered also qualifies as a limitation of this method.
While a neighborhood-scale development has untold millions of design decisions which must
be made, here focus was placed only on the limited problem of building use, quantity, and
height. It is likely that an actual design would require the addition of many more variables,
which may cause the model to become more convoluted. While the ability to consider multiple
competing objectives is this methods strength, its complexity relative to the application of more
normative theory may be a weakness, at least as far as implementation is concerned.
The “black box” nature of algorithmic design may also be a limitation in a field in which
assigning a reason to one’s decision-making is highly valued. While the conventional design
103
process allows for projects to be traced through numerous intermittent steps as design goals
change and evolve, this method produces solutions somewhat opaquely.
Finally, unlike single-variable optimization, this multi-variable method intrinsically
produces a multitude of solutions. While the problem considered herein was comprised of
discrete, bounded variables and thus had a finite number of solutions in the Pareto Set, design
problems with less constrained or continuous variables will offer a theoretically infinite number
of solutions. Therefore, this method can be used only to “narrow the field” of possible
solutions. Selecting from among the solutions identified as appropriate remains a design
problem which is outside the scope of this investigation.
It is also worth noting that the diverse body of solutions proposed likely have impacts to
which the model is totally agnostic. Consider that one solution may propose fifteen buildings
while the other proposes sixteen. If both solutions emit an equal amount of carbon, the model
will express no environmental preference. Yet, it may very well be that the elimination of an
entire building footprint allows for some other important environmental goal, such as the
inclusion of on-site power generation or greater carbon sequestration in on-site plantings.
Because the methods employed are intentionally novel, little considering has been
given to previous work regarding form-based codes, solar zoning codes, and other evaluations
of tall building massing impacts on cityscapes. A more complex model may be able to
incorporate some of the body of research which evaluates outcomes based on tall building
massing, but as presently constructed, the model is not sufficiently detailed to incorporate this
information.
5.3 APPLICABILITY TO DESIGNERS
A survey of the methodology of this investigation may imply that the analysis things
more in “developer terms” than “architectural terms.” After all, the questions asked, how many
104
buildings and what height, are often conceived of in the development and programming
process and may or may not be within the purview of the architect to change.
However, the polemical exercise of balancing the competing objectives is very
architectural. While the literal questions of height may be a forgone conclusion due to zoning
or programming requirements, the idea of evaluating solutions not by aesthetic whims or by
performance in a singular criterion appears to have merit for architectural design. The greatest
value of this investigation may be less about the explicit answers to the questions and more
about the method in which the questions are posed.
5.4 FURTHER RESEARCH
While the author has some modest hope that preceding pages have contained some
answers to difficult questions, it is undoubtable that just as many or more new questions have
been raised as a result of this exercise. Around every corner in which one decision was made,
another decision could have yielded a different result, potentially radically changing outcomes.
First, while evidence is offered to suggest that the architect’s goals have become
somewhat misaligned, scarce research exists on the change in architectural priorities over
time. If the discipline really is changing for the worse, it is likely insufficient to evaluate such a
change anecdotally. More research is required to understand the magnitude of such a change
as well as its social, economic, and political progenitors.
The investigation relied on the concept of premium for height in cost, structural material,
and rent. Yet, while the concept of premium for height has existed at least as long as the high-
rise itself, the specific relationships between height and variables have been poorly defined.
More research is necessary to better understand the trade-offs between tall and small in order
to understand the trade-offs between competing objectives.
105
The external validity of the method in question also requires further investigation. While
a single problem was selected here, it is unclear if the model that was built could be used to
solve different problems or if each novel problem would necessitate the building of an entirely
new model. In practice, it seems as though “short cuts” would have to be developed to
simplify the model building process by incorporating variables and objectives common to
many projects.
More research is also needed in understanding the longitudinal variability in these
design problems. While timeline of construction was omitted for simplicity, the time of
construction and occupation likely has tremendous effects on the cost of construction and
other financial outcomes. In a broader sense, neighborhoods are not built overnight. It is
entirely within the realm of possibility that The 78, the example analyzed in this investigation,
may take multiple decades to build. Over the course of such a time period, objectives and
even variables may change substantially as new concerns become more important and old
modalities fade.
The model does not address “weighting” of variables as presently constructed. That is
to say, each outcome is treated as equally important and the algorithm is agnostic to outcomes
that perform better one on one variable than the other. It may be true that this is rarely the case
in design practice, where certain variables are often more important than others. Other
prospective variables, such as life safety, may not be suitable for compromise at all. While
these “make or break” variables may be more accurately treated as constraints, the exercise of
assigning weights may be instructive. Further research would be necessary to explore how the
model would operate under different weighting scenarios, and which weighting scenarios
would be most applicable to design practice.
Finally, while this model sought to predict outcomes for design objectives for given
solutions, in reality any such exercise would need to be validated with post-occupancy review.
106
Whether the solutions that our computers deem optimal actually performed optimally remains
somewhat of an open question, not only in regards to this investigation, but more broadly in
the field of computational design and construction.
107
BIBLIOGRAPHY
AH&LA. (2015). Lodging Industry Trends 2015. American Hotel & Lodging Association.
AIA. (n.d.). Where we stand: Professional licensure - AIA. Retrieved March 14, 2018, from https://www.aia.org/resources/174971-where-we-stand-professional-licensure-?editing=true
AIA. (2018). Architectural Billings Index (ABI) January 2018 (p. 2). American Institute of Architects.
Alajmi, A., & Wright, J. (2014). Selecting the most efficient genetic algorithm sets in solving unconstrained building optimization problem. International Journal of Sustainable Built Environment, 3(1), 18–26. https://doi.org/10.1016/j.ijsbe.2014.07.003
Al-Kodmany, K. (2018). The Sustainability of Tall Building Developments: A Conceptual Framework. Buildings, 8(1), 7. https://doi.org/10.3390/buildings8010007
Anderson, C. (2008). The long tail: Why the future of business is selling less of more. HarperCollins e-books. http://search.ebscohost.com/login.aspx?direct=true&scope=site&db=nlebk&db=nlabk&AN=251939
Asadi, E., Silva, M. G. da, Antunes, C. H., Dias, L., & Glicksman, L. (2014). Multi-objective optimization for building retrofit: A model using genetic algorithm and artificial neural network and an application. Energy and Buildings, 81, 444–456. https://doi.org/10.1016/j.enbuild.2014.06.009
Ascione, F., Bianco, N., De Masi, R., Mauro, G., & Vanoli, G. (2015). Design of the Building Envelope: A Novel Multi-Objective Approach for the Optimization of Energy Performance and Thermal Comfort. Sustainability, 7(8), 10809–10836. https://doi.org/10.3390/su70810809
Baiceanu, R. (2019, October 14). Chicago’s The 78 Gets in Gear. Multi-Housing News. https://www.multihousingnews.com/post/chicagos-the-78-gets-in-gear/
Baker, K., Chu, J., & Riskus, J. (2014). Designing the Construction Future (p. 28). American Institute of Architects.
Baker, M. (2019, March 1). For cap rates, stability the driving force. RE Journals. https://www.rejournals.com/for-cap-rates,-stability-the-driving-force-20190301
Barton, J., & Watts, S. (2013). Office vs. Residential: The Economics of Building Tall. CTBUH Journal, 2013(II), 38–43.
Bashier, F. (2014). Reflections on architectural design education: The return of rationalism in the studio. Frontiers of Architectural Research, 3(4), 424–430. https://doi.org/10.1016/j.foar.2014.08.004
BLS. (n.d.). Architects: Occupational Outlook Handbook: : U.S. Bureau of Labor Statistics. Retrieved March 16, 2018, from https://www.bls.gov/ooh/architecture-and-engineering/architects.htm
108
BOMA. (2018, September 18). BOMA International’s Office and Industrial Benchmarking Reports Released. Building Owners and Managers Association International. https://www.boma.org/BOMA/Research-Resources/3-BOMA-Spaces/Newsroom/PR91818.aspx
Broman, G. I., & Robèrt, K.-H. (2017). A framework for strategic sustainable development. Journal of Cleaner Production, 140, 17–31. https://doi.org/10.1016/j.jclepro.2015.10.121
Bruegmann, R. (2005). Sprawl: A compact history. University of Chicago Press.
Caldas, L. G., & Norford, L. K. (2003). Genetic Algorithms for Optimization of Building Envelopes and the Design and Control of HVAC Systems. Journal of Solar Energy Engineering, 125(3), 343. https://doi.org/10.1115/1.1591803
Campbell, S. (1999). Green cities, growing cities, just cities. Urban Planning and the Contra.
Castillo, M. (2015, November 3). Tiny hotels: Check in, then squeeze in. CNBC. https://www.cnbc.com/2015/11/03/tiny-hotels-check-in-then-squeeze-in.html
Champagne, P. (2012, August 19). Apartment complex site staffing. Wren Investment Group LLC. http://wreninvestment.com/2012/08/apartment-complex-site-staffing/
Chen, X., & Yang, H. (2017). A multi-stage optimization of passively designed high-rise residential buildings in multiple building operation scenarios. Applied Energy, 206, 541–557. https://doi.org/10.1016/j.apenergy.2017.08.204
City of Chicago. (2014). Affordable Requirements Ordinance: Proposed Enhancements. City of Chicago. https://www.chicago.gov/content/dam/city/depts/dcd/general/housing/ARO_Proposed_Enhancements_Dec_2014_Web_Final.pdf
City of Chicago. (2019). Income and Rent Limits. https://www.chicago.gov/content/dam/city/depts/doh/general/2019_Income_and_Rent_Limits.pdf
Clark, N., Price, B., British Council for Offices, & Council on Tall Buildings and Urban Habitat (Eds.). (2016). Tall buildings: A strategic design guide (2nd edition). RIBA Publishing.
Coello Coello, C. A., & Lechuga, M. S. (2002). MOPSO: A proposal for multiple objective particle swarm optimization. Proceedings of the 2002 Congress on Evolutionary Computation. CEC’02 (Cat. No.02TH8600), 2, 1051–1056. https://doi.org/10.1109/CEC.2002.1004388
Conceição António, C. A., Monteiro, J. B., & Afonso, C. F. (2014). Optimal topology of urban buildings for maximization of annual solar irradiation availability using a genetic algorithm. Applied Thermal Engineering, 73(1), 424–437. https://doi.org/10.1016/j.applthermaleng.2014.08.007
Congradac, V., & Kulic, F. (2009). HVAC system optimization with CO2 concentration control using genetic algorithms. Energy and Buildings, 41(5), 571–577. https://doi.org/10.1016/j.enbuild.2008.12.004
109
Cuff, D. (1999). The political paradoxes of practice: Political economy of local and global architecture. Arq: Architectural Research Quarterly, 3(1), 77–88.
Das, I., & Dennis, J. E. (1998). Normal-Boundary Intersection: A New Method for Generating the Pareto Surface in Nonlinear Multicriteria Optimization Problems. SIAM Journal on Optimization, 8(3), 631–657. https://doi.org/10.1137/S1052623496307510
de S. Motta, R., Afonso, S. M. B., & Lyra, P. R. M. (2012). A modified NBI and NC method for the solution of N-multiobjective optimization problems. Structural and Multidisciplinary Optimization, 46(2), 239–259. https://doi.org/10.1007/s00158-011-0729-5
Deal, B., Kim, J. H., & Chakraborty, A. (2009). Growth management and sustainable transport: Do growth management policies promote transit use? Journal of Public Transportation, 12(4), 2.
Deal, B., & Schunk, D. (2004). Spatial dynamic modeling and urban land use transformation: A simulation approach to assessing the costs of urban sprawl. Ecological Economics, 51(1–2), 79–95. https://doi.org/10.1016/j.ecolecon.2004.04.008
Deamer, P. (Ed.). (2015). The architect as worker: Immaterial labor, the creative class, and the politics of design. Bloomsbury Academic.
Deb, K., Pratap, A., Agarwal, S., & Meyarivan, T. (2002). A fast and elitist multiobjective genetic algorithm: NSGA-II. IEEE Transactions on Evolutionary Computation, 6(2), 182–197. https://doi.org/10.1109/4235.996017
Delgarm, N., Sajadi, B., Kowsary, F., & Delgarm, S. (2016). Multi-objective optimization of the building energy performance: A simulation-based approach by means of particle swarm optimization (PSO). Applied Energy, 170, 293–303. https://doi.org/10.1016/j.apenergy.2016.02.141
Deutsch, R. (2017). Convergence: The redesign of design. http://proquest.safaribooksonline.com/?fpi=9781119256212
Dixit, M. K., Fernández-Solís, J. L., Lavy, S., & Culp, C. H. (2010). Identification of parameters for embodied energy measurement: A literature review. Energy and Buildings, 42(8), 1238–1247. https://doi.org/10.1016/j.enbuild.2010.02.016
Doyle, S., & Senske, N. (2016). Between design and digital: Bridging the gaps in architectural education.
Drew, C., Fernandez-Nova, K., & Fanning, K. (2015). The Environmental Impact of Tall vs. Small: A Comparitive Study. International Journal of High-Rise Buildings, 4(2), 109–116.
Du, P., Wood, A., Ditchman, N., & Stephens, B. (2017a). Life Satisfaction of Downtown High-Rise vs. Suburban Low-Rise Living: A Chicago Case Study. Sustainability, 9(6), 1052. https://doi.org/10.3390/su9061052
Du, P., Wood, A., Ditchman, N., & Stephens, B. (2017b). Life Satisfaction of Downtown High-Rise vs. Suburban Low-Rise Living: A Chicago Case Study. Sustainability, 9(6), 1052. https://doi.org/10.3390/su9061052
110
Ecker, D. (2019, April 9). ’We’re off to a great start in 2019: A burst of move-ins brought down the downtown office vancancy rate after three straight quarters of increases. Crains Chicago Business. https://www.chicagobusiness.com/commercial-real-estate/were-great-start-2019
Eizenberg, E., & Jabareen, Y. (2017). Social Sustainability: A New Conceptual Framework. Sustainability, 9(1), 68. https://doi.org/10.3390/su9010068
Elkington, J. (1998). Cannibals with forks: The triple bottom line of 21st century business. Capstone.
Erfani, T., & Utyuzhnikov, S. V. (2011). Directed search domain: A method for even generation of the Pareto frontier in multiobjective optimization. Engineering Optimization, 43(5), 467–484. https://doi.org/10.1080/0305215X.2010.497185
Fischhoff, B., & Beyth, R. (1975). I knew it would happen. Organizational Behavior and Human Performance, 13(1), 1–16. https://doi.org/10.1016/0030-5073(75)90002-1
Florida, R. L. (2014). Rise of the creative class—Revisited (Paperback of the revised edition). Basic Books.
Foraboschi, P., Mercanzin, M., & Trabucco, D. (2014a). Sustainable structural design of tall buildings based on embodied energy. Energy and Buildings, 68, 254–269. https://doi.org/10.1016/j.enbuild.2013.09.003
Foraboschi, P., Mercanzin, M., & Trabucco, D. (2014b). Sustainable structural design of tall buildings based on embodied energy. Energy and Buildings, 68, 254–269. https://doi.org/10.1016/j.enbuild.2013.09.003
Fry, R. (2019, October 1). The numer of people in the average U.S. household is going up for the first time in over 160 years. Pew Research. https://www.pewresearch.org/fact-tank/2019/10/01/the-number-of-people-in-the-average-u-s-household-is-going-up-for-the-first-time-in-over-160-years/
Goldhagen, S. W. (2017). Welcome to your world: How the built environment shapes our lives (First edition). Harper, an imprint of HarperCollinsPublishers.
Google. (n.d.). Google Ngram Viewer. Retrieved March 14, 2018, from https://books.google.com/ngrams/graph?content=architect&year_start=1900&year_end=2018&corpus=15&smoothing=3&share=&direct_url=t1%3B%2Carchitect%3B%2Cc0
Hammond, G. P., & Jones, C. I. (2008). Embodied energy and carbon in construction materials. Proceedings of the Institution of Civil Engineers - Energy, 161(2), 87–98. https://doi.org/10.1680/ener.2008.161.2.87
Haupt, R. L., & Haupt, S. E. (2004). Practical genetic algorithms (2nd ed). John Wiley.
Hill, J. (2003). Actions of architecture: Architects and creative users. Routledge.
Huberman, N., & Pearlmutter, D. (2008). A life-cycle energy analysis of building materials in the Negev desert. Energy and Buildings, 40(5), 837–848. https://doi.org/10.1016/j.enbuild.2007.06.002
111
Institute of Electrical and Electronics Engineers (Ed.). (2002). Proceedings of the 2002 Congress on Evolutionary Computation: CEC’02, May 12-17, 2002, Honolulu, Hawaii. IEEE.
Jyoti, N. G. (2012). Genetic Algorithms: A Problem Solving Approach. Proceedings of the National Conference on Trends and Advances in Mechanical Engineering, 940–944.
Kamali, M., & Hewage, K. (2015). Performance indicators for sustainability assessment of buildings. Proceedings of the International Construciton Speciality Conference of the Canadian Society for Civil Engineering, 8–10.
Karaguzel, O. T., Zhang, R., & Lam, K. P. (2014). Coupling of whole-building energy simulation and multi-dimensional numerical optimization for minimizing the life cycle costs of office buildings. Building Simulation, 7(2), 111–121. https://doi.org/10.1007/s12273-013-0128-5
Katz, B. (Ed.). (2000). Reflections on regionalism: Bruce Katz, editor ; [foreword by Al Gore]. Brookings Institution Press.
Kennedy, J., & Eberhart, R. (1995). Particle swarm optimization. Proceedings of ICNN’95 - International Conference on Neural Networks, 4, 1942–1948. https://doi.org/10.1109/ICNN.1995.488968
Kirkpatrick, S., Gelatt, C. D., & Vecchi, M. P. (1983). Optimization by Simulated Annealing. Science, 220(4598), 671–680. https://doi.org/10.1126/science.220.4598.671
Klingmann, A. (2010). Brandscapes: Architecture in the experience economy. MIT Press.
Knowles, J., & Corne, D. (1999). The pareto archived evolution strategy: A new baseline algorithm for pareto multi-objective optimisation. Congress on Evolutionary Computation, 1, 98–105.
Knox, P. L. (2008). Metroburbia, USA. Rutgers University Press.
Knox, P. L. (2011). Cities and design. Routledge.
Kolokotsa, D., Stavrakakis, G. S., Kalaitzakis, K., & Agoris, D. (2002). Genetic algorithms optimized fuzzy controller for the indoor environmental management in buildings implemented using PLC and local operating networks. Engineering Applications of Artificial Intelligence, 15(5), 417–428. https://doi.org/10.1016/S0952-1976(02)00090-8
Koster, H. R. A., van Ommerren, J. N., & Rietveld, P. (2011). Is the Sky the Limit? An Analysis of High-Rise Office Buildings (Discussion Paper No. 86). Spatial Economics Research Center.
Koziarz, J. (2019, October 28). 13 megadevelopments that will transform Chicago. Curbed Chicago. https://chicago.curbed.com/maps/chicago-developments-lincoln-yards-78-one-central
Kramer, O. (2017). Genetic algorithm essentials. Springer.
Kruger, J., & Dunning, D. (1999). Unskilled and unaware of it: How difficulties in recognizing one’s own incompetence lead to inflated self-assessments. Journal of Personality and Social Psychology, 77(6), 1121–1134. https://doi.org/10.1037/0022-3514.77.6.1121
112
Kucukvar, M., & Tatari, O. (2013). Towards a triple bottom-line sustainability assessment of the U.S. construction industry. The International Journal of Life Cycle Assessment, 18(5), 958–972. https://doi.org/10.1007/s11367-013-0545-9
Kunstler, J. H. (1993). The geography of nowhere: The rise and decline of America’s man-made landscape. Simon & Schuster.
Lamont, G. B., & Association for Computing Machinery (Eds.). (2002). Applied computing 2002: Proceddings [i.e. proceedings] of the 2002 ACM Symposium on Applied Computing, Universidad Carlos III De Madrid, Madrid, Spain, March 11-14, 2002. Association for Computing Machinery.
Latief, Y., Berawi, M. A., Basten, V., Budiman, R., & Riswanto. (2017). Premium cost optimization of operational and maintenance of green building in Indonesia using life cycle assessment method. 020007. https://doi.org/10.1063/1.4985452
Li, T., Shao, G., Zuo, W., & Huang, S. (2017). Genetic Algorithm for Building Optimization: State-of-the-Art Survey. Proceedings of the 9th International Conference on Machine Learning and Computing - ICMLC 2017, 205–210. https://doi.org/10.1145/3055635.3056591
Lin, S.-H., & Gerber, D. J. (2014). Evolutionary energy performance feedback for design: Multidisciplinary design optimization and performance boundaries for design decision support. Energy and Buildings, 84, 426–441. https://doi.org/10.1016/j.enbuild.2014.08.034
Liu, C. H., Rosenthal, S. S., & Strange, W. C. (2018). The vertical city: Rent gradients, spatial structure, and agglomeration economies. Journal of Urban Economics, 106, 101–122. https://doi.org/10.1016/j.jue.2018.04.001
LLNL. (2017). Estimated U.S. Energy Consumption in 2016: 97.3 Quads. Lawrence Livermore National Labs.
Maleki, B., Casanovas Rubio, M. d M., Hosseini, S. M. A., & de la Fuente Antequera, A. (2019). Multi-Criteria Decision Making in the Social Sustainability Assessment of High-Rise Residential Buildings. IOP Conference Series: Earth and Environmental Science, 290, 012054. https://doi.org/10.1088/1755-1315/290/1/012054
Mason, K., Duggan, J., & Howley, E. (2017). Multi-objective dynamic economic emission dispatch using particle swarm optimisation variants. Neurocomputing, 270, 188–197. https://doi.org/10.1016/j.neucom.2017.03.086
McHarg, I. L. (1971). Design with nature. Doubleday : for the American Museum of Natural History.
Meadows, J. (2019, June 24). North Side Homeowners Seet Property Tax Bills Rise by 11 Percent. Patch. https://patch.com/illinois/chicago/north-side-homeowners-see-property-tax-bills-rise-11-percent
Mellen, S. (2019). Hotel Cap Rates Hold Steady—Values Under Pressure. HVS.
Messac, A., & Mattson, C. A. (2004). Normal Constraint Method with Guarantee of Even Representation of the Complete Pareto Frontier. AIAA Journal, 42(10), 2101–2111.
113
Miller, B. (2019, January 23). Chicago hotel industry enjoys a good 2018. Chicago Business Journal. https://www.bizjournals.com/chicago/news/2019/01/23/chicago-hotel-industry-enjoys-a-good-2018.html
Modi, S. (2014). Improving the Social Sustainability of High-Rises. CTBUH Journal, 2014(I), 24–30.
MSCI. (2018a). MSCI World Real Estate Index. Morgan Stanley Capital International.
MSCI. (2018b). MSCI World Real Estate Management and Development Index (USD). Morgan Stanley Capital International.
Mueller-Gritschneder, D., Graeb, H., & Schlichtmann, U. (2009). A Successive Approach to Compute the Bounded Pareto Front of Practical Multiobjective Optimization Problems. SIAM Journal on Optimization, 20(2), 915–934. https://doi.org/10.1137/080729013
Mufti, A. A., & Bakht, B. (2002a). Fazlur Khan (1929–1982): Reflections on his life and works. Canadian Journal of Civil Engineering, 29, 238–245. https://doi.org/10.1139/l01-092
Mufti, A. A., & Bakht, B. (2002b). Fazlur Khan (1929–1982): Reflections on his life and works. Canadian Journal of Civil Engineering, 29, 238–245. https://doi.org/10.1139/l01-092
Munger, P., & Smith, L. (2019, September). 2019 NAA Survey of Operating Income & Expenses in Rental Apartment Communities. UNITS Magazine. https://www.naahq.org/news-publications/units/september-2019/article/2019-naa-survey-operating-income-expenses-rental
Murphy, K. (2012). The social pillar of sustainable development: A literature review and framework for policy analysis. Sustainability: Science, Practice and Policy, 8(1), 15–29. https://doi.org/10.1080/15487733.2012.11908081
NAA. (2017). Adding value in the age of amenities wars: Enhancing value through amenities & rehabs (p. 18). National Apartment Association.
NCARB. (2017, July 11). NBTN 2017 State of Licensure. NCARB - National Council of Architectural Registration Boards. https://www.ncarb.org/nbtn2017/state-of-licensure
Newsham, G. R., Mancini, S., & Birt, B. J. (2009). Do LEED-certified buildings save energy? Yes, but…. Energy and Buildings, 41(8), 897–905. https://doi.org/10.1016/j.enbuild.2009.03.014
Nicol, D., & Pilling, S. (Eds.). (2000). Changing architectural education: Towards a new professionalism. E & FN Spon.
Oliver, J. G. J., Janssens-Maenhout, G., Mantean, M., & Peters, J. A. H. W. (2016). Trends in global CO2 emissions: 2016 report (No. 2315). PBL Netherlands Enveiornmental Assessment Agency.
Ooka, R., & Komamura, K. (2009). Optimal design method for building energy systems using genetic algorithms. Building and Environment, 44(7), 1538–1544. https://doi.org/10.1016/j.buildenv.2008.07.006
Ori, R. (2019, April 10). Lincoln Yards and The 78: A look at what’s planned. Chicago Tribune. https://www.chicagotribune.com/business/ct-biz-lincoln-yards-the-78-details-20190410-story.html
114
Orr, D. W. (2011). Hope is an imperative: The essential David Orr. Island Press.
Oyedele, L. O. (2010). Sustaining architects’ and engineers’ motivation in design firms: An investigation of critical success factors. Engineering, Construction and Architectural Management, 17(2), 180–196. https://doi.org/10.1108/09699981011024687
Pachauri, R. K., Allen, M. R., Barros, V. R., Broome, J., Cramer, W., Christ, R., Church, J. A., Clarke, L., Dahe, Q., & Dasgupta, P. (2014). Climate change 2014: Synthesis report. Contribution of Working Groups I, II and III to the fifth assessment report of the Intergovernmental Panel on Climate Change. IPCC.
Papantoniou, S., Kolokotsa, D., & Kalaitzakis, K. (2015). Building optimization and control algorithms implemented in existing BEMS using a web based energy management and control system. Energy and Buildings, 98, 45–55. https://doi.org/10.1016/j.enbuild.2014.10.083
Pietsch, M. (2015). Peer reviewed proceedings of digital landscape architecture 2015 at Anhalt University of Applied Sciences: International Digital Landscape Architecture (DLA) Conference on Information Technologies in Landscape Architecture, held in June 4 - 6, 2015 at the Anhalt University of Applied Sciences Campus in Dessau, Germany (E. Buhmann, Ed.). Wichmann.
Ponsen, A. (2017, Fall). Trends in Square Feet per Office Employee: An Update. Development Magazine. http://naiop.org/en/Magazine/2017/Fall-2017
Pourzeynali, S., Lavasani, H. H., & Modarayi, A. H. (2007). Active control of high rise building structures using fuzzy logic and genetic algorithms. Engineering Structures, 29(3), 346–357. https://doi.org/10.1016/j.engstruct.2006.04.015
Roeder, D. (2019, April 24). Downtown Chicago office leasing accelerates, report finds. Chicago Sun Times. https://chicago.suntimes.com/business/2019/4/24/18620728/downtown-chicago-office-leasing-accelerates-report-finds
Samuelson, H., Claussnitzer, S., Goyal, A., Chen, Y., & Romo-Castillo, A. (2016). Parametric energy simulation in early design: High-rise residential buildings in urban contexts. Building and Environment, 101, 19–31. https://doi.org/10.1016/j.buildenv.2016.02.018
Sang, K., Ison, S., Dainty, A., & Powell, A. (2009). Anticipatory socialisation amongst architects: A qualitative examination. Education + Training, 51(4), 309–321. https://doi.org/10.1108/00400910910964584
Sang, K. J. C., Ison, S. G., & Dainty, A. R. J. (2009). The job satisfaction of UK architects and relationships with work-life balance and turnover intentions. Engineering, Construction and Architectural Management, 16(3), 288–300. https://doi.org/10.1108/09699980910951681
Sartori, I., & Hestnes, A. G. (2007a). Energy use in the life cycle of conventional and low-energy buildings: A review article. Energy and Buildings, 39(3), 249–257. https://doi.org/10.1016/j.enbuild.2006.07.001
Sartori, I., & Hestnes, A. G. (2007b). Energy use in the life cycle of conventional and low-energy buildings: A review article. Energy and Buildings, 39(3), 249–257. https://doi.org/10.1016/j.enbuild.2006.07.001
115
Scalzitti, J., & Herrerra, F. (2019). 2018 Cook County Tax Rates Released. Office of Cook County Clerk Karen A. Yarbrough.
Schoenberger, E. (1988). From Fordism to flexible accumulation: Technology, competitive strategies, and international location. Environment and Planning D: Society and Space, 6(3), 245–262.
Sev, A., & Özgen, A. (2009). Space Efficiency In High-Rise Office Buildings. METU Journal of the Faculty of Architecture, 26(2), 69–89. https://doi.org/10.4305/METU.JFA.2009.2.4
Shi, X., Tian, Z., Chen, W., Si, B., & Jin, X. (2016). A review on building energy efficient design optimization rom the perspective of architects. Renewable and Sustainable Energy Reviews, 65, 872–884. https://doi.org/10.1016/j.rser.2016.07.050
Steinitz, C. (2012). A framework for geodesign: Changing geography by design ; [the people of the place, design professions, geographic sciences, information technologies] (1. ed). Esri Press.
Stender, M., & Walter, A. (2019). The role of social sustainability in building assessment. Building Research & Information, 47(5), 598–610. https://doi.org/10.1080/09613218.2018.1468057
Swann, W. B., & Read, S. J. (1981). Self-verification processes: How we sustain our self-conceptions. Journal of Experimental Social Psychology, 17(4), 351–372. https://doi.org/10.1016/0022-1031(81)90043-3
Thiel, C., Campion, N., Landis, A., Jones, A., Schaefer, L., & Bilec, M. (2013). A Materials Life Cycle Assessment of a Net-Zero Energy Building. Energies, 6(2), 1125–1141. https://doi.org/10.3390/en6021125
Thiele, L., Zitzler, E., & Laumanns, M. (2001). SPEA2: Improving the strength pareto evolutionary algorithm. ETH Zurich. https://doi.org/10.3929/ethz-a-004284029
Trends for US degrees earned in architecture and higher education overall. (n.d.). Retrieved March 14, 2018, from http://www.acsa-arch.org/images/default-source/data/acsa_atlas_4.jpg?sfvrsn=6
Tversky, A., & Kahneman, D. (1973). Availability: A heuristic for judging frequency and probability. Cognitive Psychology, 5(2), 207–232. https://doi.org/10.1016/0010-0285(73)90033-9
Tzonis, A. (2014). A framework for architectural education. Frontiers of Architectural Research, 3(4), 477–479. https://doi.org/10.1016/j.foar.2014.10.001
Vallen, G. K., & Vallen, J. J. (2018). Check-in check-out: Managing hotel operations (Tenth edition). Pearson.
Wang, W., Rivard, H., & Zmeureanu, R. (2005). An object-oriented framework for simulation-based green building design optimization with genetic algorithms. Advanced Engineering Informatics, 19(1), 5–23. https://doi.org/10.1016/j.aei.2005.03.002
Wang, W., Zmeureanu, R., & Rivard, H. (2005). Applying multi-objective genetic algorithms in green building design optimization. Building and Environment, 40(11), 1512–1525. https://doi.org/10.1016/j.buildenv.2004.11.017
116
Ward, J. (2018, October 9). Chicago office rents high record high: Report. The Real Deal: Chicago Real Estate News. https://therealdeal.com/chicago/2018/10/09/chicago-office-rents-hit-record-high-report/
Wason, P. C. (1968). Reasoning about a rule. Quarterly Journal of Experimental Psychology, 20(3), 273–281. https://doi.org/10.1080/14640746808400161
Wright, J. A., Loosemore, H. A., & Farmani, R. (2002). Optimization of building thermal design and control by multi-criterion genetic algorithm. Energy and Buildings, 34(9), 959–972.
Wu, M. H., Ng, T. S., & Skitmore, M. R. (2016). Sustainable building envelope design by considering energy cost and occupant satisfaction. Energy for Sustainable Development, 31, 118–129. https://doi.org/10.1016/j.esd.2015.12.003
Wu, X., Kumar, V., Ross Quinlan, J., Ghosh, J., Yang, Q., Motoda, H., McLachlan, G. J., Ng, A., Liu, B., Yu, P. S., Zhou, Z.-H., Steinbach, M., Hand, D. J., & Steinberg, D. (2008). Top 10 algorithms in data mining. Knowledge and Information Systems, 14(1), 1–37. https://doi.org/10.1007/s10115-007-0114-2
Wu, Z. Y., & Katta, P. (2009). Applying genetic algorithm to geometry design optimization (p. 47). Bently Systems, Inc.
Xu, W., Chong, A., Karaguzel, O. T., & Lam, K. P. (2016). Improving evolutionary algorithm performance for integer type multi-objective building system design optimization. Energy and Buildings, 127, 714–729. https://doi.org/10.1016/j.enbuild.2016.06.043
Yang, X., Zhao, L., Bruse, M., & Meng, Q. (2012). An integrated simulation method for building energy performance assessment in urban environments. Energy and Buildings, 54, 243–251. https://doi.org/10.1016/j.enbuild.2012.07.042
Yu, W., Li, B., Jia, H., Zhang, M., & Wang, D. (2015). Application of multi-objective genetic algorithm to optimize energy efficiency and thermal comfort in building design. Energy and Buildings, 88, 135–143. https://doi.org/10.1016/j.enbuild.2014.11.063
Żelazowski, K. (2015). Optimal Height Of Land Development – An Economic Perspective. Real Estate Management and Valuation, 23(1), 15–23. https://doi.org/10.1515/remav-2015-0002
Zitzler, E., Thiele, L., & Deb, K. (2000). Comparison of Multiobjective Evolutionary Algorithms: Empirical Results. Evolutionary Computation, 8(2), 173–195.
Znouda, E., Ghrab-Morcos, N., & Hadj-Alouane, A. (2007). Optimization of Mediterranean building design using genetic algorithms. Energy and Buildings, 39(2), 148–153. https://doi.org/10.1016/j.enbuild.2005.11.015