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