Sei sulla pagina 1di 13

JID: EOR

ARTICLE IN PRESS [m5G;January 10, 2020;16:2]


European Journal of Operational Research xxx (xxxx) xxx

Contents lists available at ScienceDirect

European Journal of Operational Research


journal homepage: www.elsevier.com/locate/ejor

Decision Support

Urban land use and transportation planning for climate change


mitigation: A theoretical framework
Benjamin D. Leibowicz
Operations Research and Industrial Engineering, University of Texas at Austin, ETC 5.128D, 204 E. Dean Keeton St., C2200, Austin, TX 78712, United States

a r t i c l e i n f o a b s t r a c t

Article history: Cities account for 75% of global greenhouse gas (GHG) emissions from energy use, and their share is
Received 22 February 2019 increasing due to rapid urbanization. While compact urban forms with public transit are viewed as im-
Accepted 24 December 2019
portant strategies for reducing emissions, environmental benefits must be weighed against the costs of
Available online xxx
public transit infrastructure, road improvements to alleviate congestion in dense urban space, and more
Keywords: expensive housing resulting from land use restrictions. The literature largely lacks a theoretical frame-
OR in environment and climate change work for assessing these tradeoffs. This paper derives analytical insights into urban land use and trans-
Urban planning portation planning for climate change mitigation by formulating a social planner’s utility maximization
Land use problem. The planner chooses the residential densities of urban zones as well as investments in road and
Transportation public transit infrastructures that link these zones to the city center. Road travel is subject to conges-
Policy analysis tion. Any feasible solution must accommodate a fixed total population and ensure that residents of all
zones have the same maximum utility. GHG emissions associated with housing, road travel, and public
transit generate damages. Analytical results show that incorporating GHG damages into urban planning
always leads to an optimal solution with a more compact urban form, and reduces automobile travel in
each zone if a specific condition involving the marginal congestion cost and the marginal effectiveness
of road investment is satisfied. Numerical examples demonstrate that near-optimal emissions reductions
and utility improvements can be achieved via public transit investment and mode shifting even if the
planner inherits and cannot modify a suboptimal land use and road configuration.
© 2019 Elsevier B.V. All rights reserved.

1. Introduction climate solutions at the urban scale that could collectively add up
to an urban mitigation wedge (Creutzig, Baiocchi, Bierkandt, Pich-
The global urban population surpassed the number of rural ler, & Seto, 2015; Creutzig et al., 2018; Marshall, 2008). City gov-
dwellers for the first time in human history in 2008, and we are ernments around the world are heeding the call to action, and are
now a majority urban civilization (Marshall, 2008). The United becoming more prominent climate policy players in the midst of
Nations (2018) projects that rapid and persistent urbanization will uncertainty about the Paris Climate Agreement (Deetjen, Conger,
result in 68% of the global population residing in urban areas in Leibowicz, & Webber, 2018; Solecki et al., 2018). In fact, the C40
2050. Between now and then, cities around the world will gain 2.5 Cities Climate Leadership Group committed to achieving the 1.5◦ C
billion net inhabitants. While urbanization is occurring in all re- goal of the Paris Agreement has expanded to include over 90 cities
gions of the world, 90% of these new urban residents will live in around the world that represent more than 700 million people and
Asia and Africa (United Nations, 2018). a quarter of the global economy (C40 Cities, 2019).
Cities already account for 75% of all global carbon dioxide (CO2 ) Decades of empirical research have confirmed that urban form
emissions from energy use, and their share of greenhouse gas – which essentially encompasses land use patterns and the lay-
(GHG) emissions will continue to rise along with urbanization (Bai outs of transportation infrastructures (Marshall, 2008) – is a major
et al., 2018). There is growing recognition that climate change miti- driver of urban GHG emissions, especially in the residential and
gation must increasingly take place in urban contexts. For instance, transportation sectors. In the residential sector, Ewing and Rong
the Intergovernmental Panel on Climate Change (IPCC) hosted its (2008) find that residents of more sprawling U.S. counties are more
first Cities and Climate Change Science Conference in 2018 (Prieur- likely to live in larger homes and more likely to live in detached
Richard et al., 2018), and researchers have called for demand-side houses, both of which increase energy use and emissions. Glaeser
and Kahn (2010) discover that more geographically concentrated
urban areas in the U.S. consume less electricity per capita, and
E-mail address: bleibowicz@utexas.edu that central city residents tend to consume less electricity than

https://doi.org/10.1016/j.ejor.2019.12.034
0377-2217/© 2019 Elsevier B.V. All rights reserved.

Please cite this article as: B.D. Leibowicz, Urban land use and transportation planning for climate change mitigation: A theoretical frame-
work, European Journal of Operational Research, https://doi.org/10.1016/j.ejor.2019.12.034
JID: EOR
ARTICLE IN PRESS [m5G;January 10, 2020;16:2]

2 B.D. Leibowicz / European Journal of Operational Research xxx (xxxx) xxx

suburban residents within the same metropolitan area. Similarly, generate damages that reduce utility, and thus factor into the
Navamuel, Rubiera Morollón, and Moreno Cuartas (2018) find that planner’s problem.
urban households consume significantly less electricity than small Based on this model formulation, several important properties
municipality and rural households in Spain, and Belaïd (2016) con- of the optimal urban land use and transportation plan are de-
cludes that residential energy consumption in France increases rived analytically. These analytical findings constitute generalizable
moving from urban to suburban to rural settings. In the trans- insights in that they do not stem from any particular functional
portation sector, Newman and Kenworthy (1989) show in their forms or numerical parameterization of the model, but rather are
landmark paper that per-capita gasoline consumption is negatively guaranteed to hold under a fairly standard and minimal set of
related to urban core population density and the proportion of structural assumptions. The main analytical results of this paper
jobs in the city center. The close relationship between compact are asserted in four propositions that characterize the optimal ur-
urban form and low gasoline consumption holds across a sam- ban planning response to climate change. A key element of the
ple of U.S. cities as well as a sample of global cities. This find- optimal strategy is to pursue a more compact urban form with
ing is corroborated by many more recent studies (Karathodorou, more residents living in central locations. Even as the planner
Graham, & Noland, 2010), including analyses that use personal ve- spatially reallocates households across zones to reduce emissions
hicle travel (Marshall, 2008) or transportation GHG emissions per (with some central zones absorbing additional residents), the ab-
capita (Kennedy et al., 2009; Ma, Liu, & Chai, 2015; VandeWeghe & solute level of personal automobile travel should decline in each
Kennedy, 2007) as the dependent variable. The latter specification zone if a specific analytical condition involving the marginal cost
captures another mechanism by which compact urban form low- of congestion and the marginal effectiveness of road capacity in-
ers GHG emissions: in addition to reducing trip distances, it makes vestment is satisfied.
public transit (which tends to be less GHG-intensive than personal Following the theoretical analysis, the model is fully specified
vehicle travel) more viable (Lohrey & Creutzig, 2016; Pan, Shen, & with functional forms and parameter values, and solved numeri-
Zhang, 2009). cally for four scenarios. The scenarios differ in terms of whether
While the empirical literature shows that more compact urban the planner accounts for GHG damages and the degree of flexibil-
forms with public transit reduce GHG emissions, these benefits ity the planner has to optimize various aspects of the urban form.
must be weighed against a variety of costs in order to make sound Results show that even if the planner inherits and cannot mod-
urban planning decisions. Public transit infrastructure is costly ify a land use plan and road infrastructure that are suboptimal
to provide, especially in low-density settings (Lohrey & Creutzig, under climate change, the planner can achieve most of the GHG
2016). To cope with traffic congestion in a denser urban space, road emissions reduction and utility improvement that are achieved in
infrastructure might need to be improved (Çolak, Lima, & González, the unrestricted optimum. This demonstrates how effective public
2016). In addition, realizing a more compact urban form via land transit investment and transportation mode shifting can be as mit-
use restrictions can come at a cost to residents due to higher hous- igation measures, and is encouraging from a practical standpoint.
ing prices (Grout, Jaeger, & Plantinga, 2011; Leibowicz, 2017). Urban The remainder of this article is structured as follows. The lit-
space is a complex system that involves tradeoffs among myriad erature review in Section 2 summarizes the most relevant urban
conflicting objectives, and the literature largely lacks a theoretical planning models from the urban economics and operations re-
framework for guiding urban land use and transportation planning search literatures. The model developed for the present study is
strategies to reduce GHG emissions while limiting adverse impacts. presented in Section 3. In Section 4, theoretical results are ob-
The research community studying cities and climate change tained by analyzing the optimality conditions of the model in a
has increasingly called attention to this research gap. According highly general setting. Numerical examples are solved and visual-
to Acuto, Parnell, and Seto (2018), urban science “remains trapped ized in Section 5 to illustrate the theoretical results from the pre-
in the twentieth-century tradition of the systematic study of indi- vious section and explore the implications of different decision-
vidual cities.” The IPCC Fifth Assessment Report attests that “the making scenarios. Section 6 clarifies limitations of the methodol-
literature on urban form and infrastructure drivers of GHG emis- ogy. Section 7 concludes with a summary of the most important
sions... is dominated by case studies of cities in developed coun- findings and possible directions for future research.
tries” (Seto et al., 2014). As a result, urban planners seeking to re-
duce emissions through land use and transportation planning are 2. Literature review
often forced to extrapolate lessons gleaned from limited case study
experiences from specific places and times, which can lead to inap- 2.1. Urban economics literature
propriate policy prescriptions. This issue is particularly problematic
given that case studies have overwhelmingly taken place in devel- In terms of its simplifying assumptions and emphasis on
oped country cities, but the greatest potential for urban climate analytical tractability, the methodology of this study builds upon
change mitigation exists in developing countries where urbaniza- spatial equilibrium models developed in the urban economics lit-
tion is in its early stages (Seto et al., 2014). Bai et al. (2016) con- erature. This modeling tradition was pioneered by Alonso (1964),
clude that “a systems approach is urgently needed in urban re- Mills (1967), and Muth (1969), who developed the original mono-
search and policy analysis,” one that captures “tradeoffs between centric city model. This framework depicts utility-maximizing
the positive and negative consequences of policy actions.” The households interacting with profit-maximizing housing producers,
present study addresses this gap in the literature. and generates an urban spatial structure by enforcing supply-
This paper derives analytical insights into urban land use demand equilibrium in the housing market and spatial equilibrium
and transportation planning for climate change mitigation by in utility. Its significant impact on urban economic modeling en-
formulating a social planner’s utility maximization problem. The dures because it yields analytical results that constitute theoretical
planner chooses the residential densities of urban zones as well explanations for patterns usually observed in the urban forms
as investments in road and public transit infrastructures that link of most cities, such as housing and population densities that fall
these zones to the city center. Road travel is subject to congestion. with distance from the city center (Brueckner, 1987). While the
Any feasible solution must accommodate a fixed total population, monocentric city model is highly stylized and static, Anas, Arnott,
ensure that residents of all zones have the same maximum utility and Small (1998) confirm that its predictions are reasonably
level, and remain within a total resource budget. GHG emissions close to empirically observed urban population density gradients.
associated with housing, road travel, and public transit travel Nevertheless, some researchers question the applicability of the

Please cite this article as: B.D. Leibowicz, Urban land use and transportation planning for climate change mitigation: A theoretical frame-
work, European Journal of Operational Research, https://doi.org/10.1016/j.ejor.2019.12.034
JID: EOR
ARTICLE IN PRESS [m5G;January 10, 2020;16:2]

B.D. Leibowicz / European Journal of Operational Research xxx (xxxx) xxx 3

monocentric model to modern cities whose urban forms are leverage transportation mode shifting to reduce emissions, in ad-
polycentric (Kloosterman & Musterd, 2001). dition to modifying urban land use.
To preserve analytical tractability, the model in this paper
adopts several simplifying assumptions made in the monocen- 2.2. Operations research literature
tric city model. Specifically, it assumes that households are
homogeneous, utility is constant across space, and travel demand There is an extensive operations research literature on urban
is a function of distance from the city center. However, the model planning that dates back at least to the 1960s (Harris, 1961). Con-
formulated for the present study departs from the monocentric siderable work has focused on optimally allocating urban land to
framework by representing urban space discretely rather than con- various uses. These operations research approaches are similar to
tinuously, omitting housing production, and seeking to optimize Borck and Brueckner (2018) in that they view urban land use
urban form rather than characterize its equilibrium. through the lens of a social planner seeking to optimize it. In con-
This analysis contributes to an emerging literature that extends trast to most models in urban economics, they do not treat the ur-
urban spatial equilibrium models to investigate energy and envi- ban form as arising from the strategic interactions of households,
ronmental issues in cities. These studies essentially attribute en- housing producers, and other agents. Benefits and costs must be
ergy consumption and/or GHG emissions to housing and trans- exogenously specified by the modeler. Operations research mod-
portation, and explore how various parameter changes or policies els represent land in terms of discrete zones rather than a con-
influence these outcomes via their effect on urban form. Gaigné, tinuous space, essentially a generalization of approaches in ur-
Riou, and Thisse (2012) and Hirte and Tscharaktschiew (2013) for- ban economics based on two distinct zones (Brueckner & Franco,
mulate analytically tractable frameworks that are necessarily based 2018; Hirte & Tscharaktschiew, 2013). The model developed in
on drastic simplifying assumptions. Gaigné et al. (2012) construct a Section 3 adopts this optimization structure and discrete represen-
two-city model and demonstrate that a more compact urban form tation of space, but incorporates household utility, spatial equilib-
can actually increase transportation GHG emissions if emissions rium, and transportation requirements into its objective and con-
from the inter-city shipping of goods increase more than emissions straints. Optimal land use allocation models correspond closely to
from local commuting decrease. The former mechanism arises be- real-world planning decisions and have been applied in practice to
cause compactness induces higher land rents that incentivize firms provide decision support (Bammi & Bammi, 1979; Brill, Chang, &
to relocate. However, their model assumes that population den- Hopkins, 1982). However, the fact that they must be solved numer-
sity is uniform within each urban area. Hirte and Tscharaktschiew ically limits their ability to provide insights that can be generalized
(2013) conduct a general equilibrium analysis using a spatial model beyond a particular setting and parameterization.
of a city with two zones that includes commuting, endogenous Brotchie (1969) formulates a quadratic programming model
labor supply, carbon emissions, and fuel and power taxes. Their for optimally allocating urban zones to different land uses (e.g.,
simpler two-zone representation of space allows them to analyti- high-density residential, city park, heavy industry). Land uses have
cally derive the optimal power tax rate for electric vehicles. Larson, zone-specific costs and benefits, and the costs and benefits of
Liu, and Yezer (2012) and Leibowicz (2017) construct more spa- land uses in different zones can interact depending on how far
tially detailed extensions of the classic monocentric city model, apart the zones are. Brotchie (1978) extends his earlier model to
but the tradeoff is that they must be solved numerically with spe- incorporate diversity of benefits within a zone as well as between
cific functional forms and parameter values, which limits their gen- zones. This extension tends to disperse land uses more evenly
erality. Larson et al. (2012) simulate how a variety of housing, throughout the city. Bammi and Bammi (1979) build upon these
land use, and transportation policies influence urban energy use foundations by representing the allocation of zones across a plan-
through their effects on urban form. Their model includes traffic ning region to various land uses as a multi-objective optimization
congestion, multiple dwelling types, and detailed calculations of problem. The five objectives they include are: minimize conflict
energy consumption associated with commuting and residences. between different land uses, minimize travel distance of new
Leibowicz (2017) assesses the GHG emissions and welfare impacts trips to the existing transportation network, minimize tax costs,
of urban land-use regulations. He demonstrates that urban growth minimize adverse environmental impact, and minimize capital
boundaries can reduce emissions by increasing urban densities, but costs of community facilities. The model is actually used by the
this comes at a cost to residents because they restrict land supply regional planning commission in DuPage County, Illinois, outside
and raise housing prices. Chicago, to develop a comprehensive land use plan in the face of
Borck and Brueckner (2018) present what is perhaps the most rapid suburban growth. Brill et al. (1982) use this DuPage County
similar analysis in the existing literature to the one described in application to demonstrate the Hop, Skip, and Jump (HSJ) method
this paper. They formulate a social planner’s problem based on to generate attractive, feasible solutions to an optimization prob-
a monocentric city model where housing and commuting gener- lem that are significantly different from one another. They show
ate GHG emissions, which diminish utility. This study similarly that the HSJ method produces multiple “good” land use plans that
adopts a social planner perspective that seeks to optimize urban differ significantly with respect to the distributions of land uses
form rather than merely determine equilibrium, and incorporates across the county, even if tight restrictions are placed on indi-
GHG damages into utility in much the same manner. However, vidual objectives. Gilbert, Holmes, and Rosenthal (1985) construct
Borck and Brueckner (2018) assume that the social planner’s ob- a multi-objective integer programming model to determine the
jective is to minimize total resource consumption, subject to the optimal location and shape of a land development. They test their
requirement that all households achieve a common, minimum util- model on the example of a proposed residential development
ity level. This setup would be difficult to apply in practice because in Tennessee, and experiment with solving the integer program
the minimum utility constraint does not correspond to any real- subproblems using alternative enumeration methods.
world limitation. The model developed in this study assumes that More recent contributions represent the built environment and
the planner’s objective is to maximize the common utility level, transportation infrastructures in impressive detail, but this in-
subject to a budget constraint. This is more consistent with the creases computational and data intensities, makes results even
usual situation where a policymaker attempts to achieve the most more parameterization-specific, and limits the comprehensiveness
desirable results given available resources. This analysis also builds of the model scope. Ying (2015) formulates a mathematical pro-
upon Borck and Brueckner (2018) by incorporating two transporta- gram with equilibrium constraints to maximize social surplus by
tion modes instead of only one. The social planner can therefore simultaneously optimizing the locations of housing supply and

Please cite this article as: B.D. Leibowicz, Urban land use and transportation planning for climate change mitigation: A theoretical frame-
work, European Journal of Operational Research, https://doi.org/10.1016/j.ejor.2019.12.034
JID: EOR
ARTICLE IN PRESS [m5G;January 10, 2020;16:2]

4 B.D. Leibowicz / European Journal of Operational Research xxx (xxxx) xxx

congestion prices in the (existing and fixed) road network. Sim- function. Considering possible alternatives, focusing on a miti-
ilar to the present analysis, Ying’s model features a discrete set gation objective in isolation would neglect tradeoffs with other
of zones that the planner develops with housing. The model is urban priorities such as consumption. Maximizing utility subject
computationally intensive, and whether the global optimum can to a maximum emissions constraint would require this limit to
be found depends on the initial setting of variables. Hammad, be specified exogenously, and fail to incentivize further emissions
Akbarnezhad, and Rey (2017) formulate a multi-objective facility reductions.
location problem to balance concerns about noise pollution with Several basic assumptions from the classic spatial equilibrium
travel times on the underlying road network. Their model is a formulations (Alonso, 1964; Mills, 1967; Muth, 1969) underlie the
bilevel problem with facility location decisions in the upper level model. The city is assumed to be monocentric, in that all resi-
and route choices in the lower level. They apply this framework to dents commute to and from the city center, and thus a household’s
a district of Sydney, Australia and use Benders decomposition to fa- transportation costs are proportional to the distance between its
cilitate solutions. Hua, Hovestadt, Tang, and Li (2019) create a very location and the center. All households are taken to be homoge-
detailed integer program that optimizes the built environment of neous, with the same preferences. The requirement of spatial equi-
an urban district including the layouts, geometries, and positions of librium in utility is a key underlying assumption of the model. Ac-
individual structures. Constraints relate to minimum sunlight ex- cording to this assumption, the utility levels of households at all
posure and other high-spatial-resolution issues. The authors apply locations in the city should be equal; otherwise, some households
their model using data from Shanghai, but due to the computa- would be incentivized to move to different locations, and the spa-
tional complexity, they are only able to develop optimal plans for tial distribution of the population would not be in equilibrium.
parcels that are less than 50,0 0 0 meter2 (roughly 12 acres) in area. The city contains a discrete set of residential zones, denoted I.
It is worth briefly noting that there is a growing literature Each zone i ∈ I is parameterized by its area ai and distance from
on sustainable urban design and operations enabled by increasing the city center xi . The planner must determine how many house-
computational power and real-time data availability. Some exam- holds will reside in each zone, as well as the numbers of them
ples in the literature focus on urban land use planning (Caparros- who travel via car or via public transit. Let Nir and Nit denote the
Midwood, Barr, & Dawson, 2015; Li & Parrott, 2016; Stewart & numbers of households in zone i who travel by road and by public
Janssen, 2014), energy systems (Cajot et al., 2017a; Cajot, Schüler, transit, respectively.1 The combined sum of Nir and Nit over I must
Peter, Koch, & Maréchal, 2017b), transportation networks and logis- equal the total, fixed urban population n.
tics (Savelsbergh & Van Woensel, 2016; Taniguchi, Thompson, & Ya- The residential density Di within a zone is assumed to
mada, 2016), and sustainability indicators (Attardi, Cerreta, Sanni- be uniform, so that each household in zone i consumes Qi =
candro, & Torre, 2018). Compared to the present paper, these stud- ai (Nir + Nit )−1 units of land for housing. The utility of a household
ies employ approaches that are far more spatially detailed, compu- in this zone is given by the quasilinear function Wi = V (Qi ) + Ci −
tationally demanding, and data-intensive. They are well suited to γ E. Quasilinear preferences are a common and analytically con-
addressing operational problems requiring fine temporal and spa- venient assumption in the literature (Brueckner & Franco, 2018;
tial resolution, but they are not analytically tractable, must often Gaigné et al., 2012). The function V satisfies V > 0 and V < 0, so
be solved using heuristics, and make it difficult to generalize in- that it is increasing and concave. Ci represents all non-land con-
sights to different contexts. Therefore, this paper complements the sumption, and is the numeraire good with a cost normalized to
literature on data-driven urban modeling by developing an analyt- unity. The final γ E term accounts for the damages associated with
ically tractable framework that yields theoretical insights for incor- total urban area GHG emissions E, which will be defined below.
porating climate change mitigation into urban planning. Coefficient γ is thus the marginal damage associated with one unit
of emissions, assumed to be a constant. To ensure spatial equilib-
3. Model formulation rium in utility across zones, it must be the case that Wi = U ∀i ∈ I.
Road transportation for a household in zone i incurs a cost
The optimization model formulated here adopts the perspective xi T(Nir , Ri ), where T is the per-unit-distance variable cost of road
of a social planner whose objective is to maximize the common travel. Road transportation is subject to congestion. The more
utility level U of all households in a city, subject to a total budget households there are commuting by road from zone i, the slower
b. To accomplish this objective, the planner must choose how to the traffic moves, and the higher the variable cost of road travel
distribute the population across the urban space, the numbers of becomes. To capture congestion, the function T satisfies T1 > 0 and
residents who travel by car and by public transit, investments in T11 > 0, so that congestion effects grow in severity on the margin.2
road and public transit infrastructures, and consumption bundles The social planner must decide how much road capacity Ri to build
for all residents. The social planner accounts for GHG emissions between zone i and the city center. Per unit distance, road infras-
from residences and transportation, which generate damages and tructure capacity has a constant marginal cost r, but increasing
diminish utility. road capacity alleviates congestion and thus reduces the variable
Adopting a social planner perspective allows for consideration cost of road travel. Road capacity investment is assumed to have
of a wide range of factors that reflect on the desirability of an
urban land use and transportation plan for the city as a whole, in-
cluding those which are not fully internalized by individual house-
holds. This inclusive framing is important for addressing negative 1
Note that the formulation is flexible enough to represent the case where a
externalities such as GHG emissions and traffic congestion, since household can split its transportation usage between the two modes. In such a
case, Nir (Nit ) is interpreted as the total road (public transit) usage by households
the social planner faces the full costs they impose on the city.
in zone i, where each household’s total transportation usage is normalized to one.
Practically, the social planner perspective is popular in the liter- The mode shares in zone i remain the same as in the case where each household is
ature since it is required to determine socially optimal strategies assigned to a single mode: Nir /(Nir + Nit ) for road and Nit /(Nir + Nit ) for public tran-
in complex environments (Borck & Brueckner, 2018; Gaigné et al., sit. In other words, whether half the households use only road and the other half
2012; Hammad et al., 2017). By incorporating GHG emissions as a use only public transit, or every household uses road and public transit equally, the
formulation is equally valid.
damage term in the utility maximization objective, the model cap- 2
In this notation, a subscript k denotes a partial derivative of the function with
tures the benefits of reducing emissions in addition to the costs. respect to its kth argument. For example, T1 = ∂∂NT , T11 = ∂∂NT2 , T2 = ∂∂RT , and T22 =
2

ir i
Optimal solutions will thus balance climate change mitigation with ∂2T .
ir

∂ R2i
the more traditional consumption inputs to the household utility

Please cite this article as: B.D. Leibowicz, Urban land use and transportation planning for climate change mitigation: A theoretical frame-
work, European Journal of Operational Research, https://doi.org/10.1016/j.ejor.2019.12.034
JID: EOR
ARTICLE IN PRESS [m5G;January 10, 2020;16:2]

B.D. Leibowicz / European Journal of Operational Research xxx (xxxx) xxx 5

declining marginal effectiveness. In mathematical terms, T2 < 0 and uses other than urban housing (e.g., agriculture). However, the land
T22 > 0. area used for urban housing is fixed in this model, so there is no
By contrast, public transportation for a household in zone i is need to account for it in the optimization problem. The total emis-
not subject to congestion, and it incurs a constant user cost pxi . In sions of the urban area are specified by Eq. (1e), which includes
the absence of congestion effects, the optimal public transit infras- terms for emissions from road usage, public transit usage, and res-
tructure capacity will simply be equal to Nit , the number of public idential energy. Lastly, Eq. (1f) imposes non-negativity constraints
transit users. Per unit distance, public transit infrastructure capac- on all variables except for total emissions E and utility U. Non-
ity has a constant marginal cost s. From the social planner’s per- negativity of E is guaranteed by the non-negativity constraints in
spective, then, each additional public transit user in zone i costs Eq. (1f) combined with Eq. (1e). The social planner maximizes U,
( p + s )xi in user and infrastructure expenditures. and as utility it has no direct physical interpretation, so there is no
Residences, road transportation, and public transportation all need to enforce non-negativity.
generate GHG emissions, which contribute to climate change
and reduce utility. Each road user and public transit user in 4. Theoretical analysis
zone i produces er xi and et xi units of emissions, respectively. The
marginal emissions factors er and et satisfy er > et , so that road In this section, analytical insights into urban land use and trans-
transportation is more emissions-intensive than public transit use. portation planning for climate change mitigation are obtained by
A household who consumes Qi units of land generates B(Qi ) units writing down the Lagrangian of the optimization problem and an-
of residential emissions, due to energy use for heating, cooling, alyzing the first-order optimality conditions. These optimality con-
appliances, and so on. It is assumed that B > 0 and B < 0. These ditions reveal important properties of the optimal land use and
properties indicate that larger homes on more land consume more transportation plan, and ensuing comparative statics show how
energy, but that increases in energy consumption are diminishing the optimal plan changes as key model inputs are varied. These
on the margin.3 Total urban area residential and transportation insights are highly general in that they are consequences of the
emissions over all zones are denoted E. fundamental structure of the problem. Only minimal assumptions
The mathematical formulation of the social planner’s utility about the functional forms contained in the model are required.
maximization problem is The Lagrangian of the social planner’s optimization problem
is
max
Nir ,Nit ,Ci ,Ri ,E,U
U (1a)  

L = U + λ1 n − (Nir + Nit )

s.t. (Nir + Nit )= n (1b) 
i
   

i + λ2i U − V ai (Nir + Nit )−1 + Ci − γ E
  i
 
V ai (Nir + Nit )−1 + Ci − γ E = U, i∈I (1c) 
+ λ3 b− {(Nir + Nit )Ci +xi [T (Nir , Ri )Nir + rRi + ( p + s )Nit ]}
i
  
{(Nir + Nit )Ci + xi [T (Nir , Ri )Nir + rRi + ( p + s )Nit ]}= b (1d)   
i + λ4 E − xi (er Nir + et Nit ) + eq (Nir + Nit )B ai (Nir + Nit )
−1
.
   i
xi (er Nir + et Nit ) + eq (Nir + Nit )B ai (Nir + Nit )−1 =E (1e) (2)
i
This Lagrangian yields the following first-order conditions by tak-
ing its partial derivatives with respect to all the decision vari-
Nir ≥ 0, Nit ≥ 0, Ci ≥ 0, Ri ≥ 0, i ∈ I. (1f) ables:
In this formulation, Eq. (1b) ensures that the total number of ∂L 
=1+ λ2 j = 0, (3a)
households in the city matches the exogenously fixed urban pop- ∂U j
ulation n. In other words, every household must be assigned to
a residential zone. It also guarantees that every household’s trans-
∂L
portation demand is satisfied by the road and public transit modes. = − λ1 + λ2i ai (Nir + Nit )−2V  − λ3 [Ci + xi (Nir T1 + T )]
Eq. (1c) specifies the utility function, and requires that households ∂ Nir

in all zones have the same utility level U. This is the spatial equi- − λ4 er xi + eq (B − ai (Nir + Nit )−1 B ) = 0, i ∈ I, (3b)
librium condition commonly assumed in the urban modeling liter-
ature (Brueckner, 1987; Larson et al., 2012; Leibowicz, 2017). If util-
ity were not the same in all zones, then some households would ∂L
= − λ1 + λ2i ai (Nir + Nit )−2V  − λ3 [Ci + xi ( p + s )]
be incentivized to move, and thus the spatial distribution would ∂ Nit

not be stable. The argument of the land component of utility V − λ4 et xi + eq (B − ai (Nir + Nit )−1 B ) = 0, i ∈ I, (3c)
is equivalent to Qi , the land consumption per household, equal to
the area of the zone divided by the number of households in it.
Eq. (1d) is the budget constraint. The social planner incurs costs ∂L
= − λ2i − λ3 (Nir + Nit ) = 0, i ∈ I, (3d)
for non-land consumption, road and public transit usage, and road ∂ Ci
and public transit infrastructure capacity investments. In princi-
ple, there is also a cost for land since it potentially has value for ∂L
= − λ3 xi [Nir T2 + r ] = 0, i ∈ I, (3e)
∂ Ri
3
Another way to see this is that as homes consume less and less land, there are
additional energy efficiency gains due to multi-story and multi-family structures. In
∂L 
addition, as the size of a home increases, the numbers of certain appliances (e.g., = λ4 + γ λ2 j = 0. (3f)
dishwashers, refrigerators) do not increase proportionally. ∂E j

Please cite this article as: B.D. Leibowicz, Urban land use and transportation planning for climate change mitigation: A theoretical frame-
work, European Journal of Operational Research, https://doi.org/10.1016/j.ejor.2019.12.034
JID: EOR
ARTICLE IN PRESS [m5G;January 10, 2020;16:2]

6 B.D. Leibowicz / European Journal of Operational Research xxx (xxxx) xxx

∂ 2Wi
Using Eq. (3a), (3d), and (3f) to solve for all but one of the = V  (Qi ) − γ ∂ E2 = V  (Qi ) − γ eq (Nir + Nit )B (Qi ).
2
to find that
Lagrange multipliers, it is found that λ2i = −(Nir + Nit )n−1 ∀i ∈ I, ∂ Qi2 ∂ Qi
λ3 = n−1 , and λ4 = γ . By plugging these expressions back into the Therefore, in zone i, the utility function is concave iff V  (Qi ) <
first-order conditions, it is possible to obtain analytical insights γ e (N + N )B (Q ). Assumption 2 only holds if this condition is
q ir it i
from the resulting system. satisfied in all zones and for any feasible spatial allocation of
First, note that Eq. (3e) implies that the urban area population, which is equivalent to requiring that
V (Qi ) < γ eq nB (Qi ). Since it was assumed earlier that V < 0 and
−Nir T2 = r ∀i ∈ I. (4) B < 0, the assumption essentially says that the indirect effect of
This rule determines the optimal road capacity investment Ri for home size on utility via damages from residential emissions is too
each zone. It intuitively states that the social planner should in- weak to reverse the concavity of utility with respect to home size
vest in road capacity until the marginal benefit of adding capacity stemming from its direct effect captured in V. This is almost surely
becomes equal to the marginal cost. The marginal benefit is the the case in practice, as the direct effect of home size on utility
number of households in the zone who travel by road (Nir ) times likely dominates.
the negative change in their road travel cost with respect to road
capacity (T2 ). The marginal cost is simply r. Proposition 1 (Home size and population density gradients). In
The optimal balance between road and public transit users can the optimal solution, home size increases monotonically, and popula-
be examined by setting Eq. (3b) and (3c) equal to one another, tion density decreases monotonically, with distance from the city cen-
∂Q ∂D
then eliminating like terms, to yield the condition ter. Mathematically, ∂ x i > 0 and ∂ x i < 0.
i i

(Nir T1 + T ) + nγ er = ( p + s ) + nγ et ∀i ∈ I. (5) −1
Proof. With Qi = ai (Nir + Nit ) , and substituting U + γ E − V (Qi )
The left-hand side is the marginal social cost (per unit distance) of for Ci based on the definition of the utility function, Eq. (3c) can
an additional household in zone i traveling by road. This household be re-expressed as
directly incurs the road user cost T, and the Nir T1 term captures
the congestion externality that this household imposes on all other i ≡ λ1 n + U + γ E + xi ( p + s + nγ et ) + V  (Qi )Qi − V (Qi )

road users in zone i. The nγ er term represents the social cost of + nγ e q B ( Q i ) − Q i B  ( Q i ) = 0 . (6)
the GHG emissions associated with road travel. The right-hand side
is analogously the marginal social cost of an additional household di
This condition must hold at any distance xi , meaning that dxi
= 0.
in zone i traveling by public transit. At the optimum, the marginal
Taking this total derivative,
social costs of road and public transit households in each zone
must be in equilibrium. Note that the derivative of the left-hand di ∂ Qi 

side of Eq. (5) with respect to Nir is Nir T11 + 2T1 > 0, based on the = p + s + nγ et + Q V (Qi ) − nγ eq B (Qi ) = 0. (7)
dxi ∂ xi i
previously stated assumptions about T. Therefore, if Eq. (5) yields
an interior solution for Nir as a function of Ri , then this solution is ∂Q
Rearranging to isolate ∂ x i yields
i
unique.

Assumption 1. Every zone in the city has some households who


∂ Qi −( p + s + nγ et )
= . (8)
travel by road and some households who travel by public transit. ∂ xi Qi [V  (Qi ) − nγ eq B (Qi )]
Mathematically, Eq. (5) yields an interior solution for Nir .
The numerator is clearly negative, and Qi is positive, so the sign of
∂ Qi
This assumption is required to facilitate the theoretical analysis. ∂ xi hinges on the sign of the bracketed term in the denominator.
In the context of the present model, this assumption would not Invoking Assumption 2’s implication that V (Q ) < γ e nB (Q ), the
i q i
hold if, for instance, public transit were very expensive, leading to ∂Q
bracketed term is negative, and therefore ∂ x i > 0. Since Di = Qi−1 ,
a corner solution where everyone travels by road. In the less an- i
∂D
alytically tractable but perhaps more realistic setting where pub- it immediately follows that ∂ x i < 0. 
i
lic transit provision entails economies of scale, the possibility of a
corner solution dissipates. In any case, Assumption 1 is assumed to The structural properties asserted in Proposition 1 are rather in-
hold throughout this analysis.4 tuitive. Home size and population density gradients that are posi-
Proposition 1 confirms that the optimal spatial allocation of the tive and negative, respectively, moving outward from the city cen-
urban area population exhibits some intuitive properties. Before ter are typical patterns observed in most urban areas, and these
establishing them, one additional assumption is made about the same results are found using the standard monocentric city model
utility function, as its implications will be invoked in the proof of (Brueckner, 1987). Nevertheless, it is reassuring that these prop-
Proposition 1. erties also emerge from the present framework, with its discrete
representation of space, multiple transportation modes, and incor-
Assumption 2. The full utility function Wi = V (Qi ) + Ci − γ E is
poration of damages associated with GHG emissions.
concave with respect to Qi in all zones and for any feasible
Proposition 2 affirms some of the major theoretical results of
spatial allocation of the urban area population. Mathematically,
this paper, describing how the optimal degree of residential cen-
∂ 2Wi tralization responds to several key model parameters.
< 0 ∀ i ∈ I, Nir , Nit | 0 ≤ Nir + Nit ≤ n .
∂ Qi2
Proposition 2 (Residential centralization response). The optimal
To establish the implication of this assumption that will be used home size and population density gradients become steeper if GHG
in the proof of Proposition 1, differentiate the utility function twice damages are more severe, public transit is more costly, public transit is
more emissions-intensive, or residences are more emissions-intensive.
∂2Q ∂2Q ∂2Q ∂2Q
4
Note that Assumption 1 is not required to solve the model numerically, as the Mathematically, ∂ x ∂ γi > 0, ∂ x ∂ ( p+
i
s)
> 0, ∂ x ∂ ei > 0, and ∂ x ∂ ei > 0
i i i t i q
optimal solution to a particular problem instance could include zero road or public
(the opposite results for Di immediately follow).
transit investment. However, it greatly facilitates the theoretical analysis, and is not
practically limiting because the modeler could choose the function T(Nir , Ri ) so that
Eq. (5) yields an interior solution with positive but negligible investment in one of Proof. Taking cross-partial derivatives of the expression for the
the transportation options. home size gradient in Eq. (8) yields

Please cite this article as: B.D. Leibowicz, Urban land use and transportation planning for climate change mitigation: A theoretical frame-
work, European Journal of Operational Research, https://doi.org/10.1016/j.ejor.2019.12.034
JID: EOR
ARTICLE IN PRESS [m5G;January 10, 2020;16:2]

B.D. Leibowicz / European Journal of Operational Research xxx (xxxx) xxx 7

∂ 2 Qi hold under a change in γ , so by taking the total derivative while


∂ xi ∂ γ holding all other parameters constant,


net V  (Qi ) −nγ eq B (Qi ) +neq B (Qi )( p+s+nγ et ) ∂ Nir ∂ Ri
= − Qi−1 >0 , (2T1 + Nir T11 ) + (T2 + Nir T12 ) = −n(er − et ). (11)
∂γ ∂γ
[V  (Qi ) −nγ eq B (Qi )]
2

(9a) Eqs. (10) and (11) comprise a system of two linear equations for
∂N ∂R
the partial derivatives ∂γir and ∂γi . Solving this system yields the
results
∂ 2 Qi
−1
= −Qi−1 V  (Qi ) − nγ eq B (Qi ) > 0, (9b) ∂ Nir n(er − et )Nir T22
∂ xi ∂ ( p + s ) = , (12a)
∂γ (T2 + Nir T12 )2 − Nir T22 (2T1 + Nir T11 )
∂ 2 Qi
−1
= −nγ Qi−1 V  (Qi ) − nγ eq B (Qi ) > 0, (9c) ∂ Ri −n(er − et )(T2 + Nir T12 )
∂ xi ∂ et = . (12b)
∂γ (T2 + Nir T12 )2 − Nir T22 (2T1 + Nir T11 )
∂ 2 Qi The number of road users decreases in each zone as GHG dam-
∂N
ages become more severe when ∂γir < 0, and Eq. (12a) offers some
∂ xi ∂ eq

−2 neat intuition about the properties of the road user cost function
= −nγ B (Qi )( p + s + nγ et )Qi−1 V  (Qi ) − nγ eq B (Qi ) > 0. T that make this condition hold. Recall that the standard assump-
(9d) tions about the derivatives of T introduced in Section 3 are T1 > 0,
T11 > 0, T2 < 0, and T22 > 0, and that er > et . Therefore, the numera-
 tor of Eq. (12a) is positive, the squared term in the denominator is
positive, and the second term in the denominator is negative. As a
These findings provide theoretical support for urban land use ∂N
policies that view environmental sustainability as at least a par- result, the sign of ∂γir depends on the relative magnitudes of the
tial justification for pursuing a more compact urban form (Creutzig two terms in the denominator.
∂N
et al., 2018; Deetjen et al., 2018). If urban activities are more Specifically, ∂γir < 0 if and only if Nir T22 (2T1 + Nir T11 ) >
emissions-intensive, or GHG damages are more severe, then the (T2 + Nir T12 )2 . This condition intuitively says that the number of
social planner should shift households’ residential locations toward road users in each zone is favored to decrease as GHG damages
the city center. As the cost of public transit rises, the optimal res- become more severe when the marginal cost of traffic congestion
idential centralization response becomes stronger. Increasing the is high (T1 is large), when the marginal cost of traffic congestion
cost of public transit makes shifting transportation modes from is steeply increasing (T11 is large), when the marginal effective-
road to public transit a less attractive option for reducing GHG ness of road capacity investment is low (|T2 | is small), or when
emissions. As a result, the residential centralization strategy is now the marginal effectiveness of road capacity investment is steeply
less costly in relative terms, so the planner should pursue it more diminishing (T22 is large). If the condition holds in every zone and
intensively. ∂ Nir
∂γ < 0 ∀i ∈ I, then total public transit usage in the urban area is
Now that Propositions 1 and 2 have established some proper- ∂(

Nit )
i∈I
ties of the optimal land use plan, Proposition 3 states important guaranteed to increase, ∂γ > 0.
results about the effects of GHG emissions on the optimal use of The proof to show that the same condition is necessary and suf-
the transportation modes. ∂ Nir
ficient for the result ∂ (e −e < 0 proceeds exactly the same way,
r t)

Proposition 3 (Transportation mode use). If GHG damages become since γ and (er − et ) only appear in Eq. (5) as a product term. 
more severe, then the number of households in each zone who travel The aforementioned properties of the road user cost function
∂N
by road in the optimal solution declines (i.e., ∂γir < 0) if and only T that cause the optimal number of road users to decline in
if Nir T22 (2T1 + Nir T11 ) > (T2 + Nir T12 )2 . Intuitively, this condition in- each zone as GHG damages become more severe, or as the dif-
volving the properties of the road user cost function T is favored to ference in road and public transit emissions intensities increases,
hold when the marginal cost of traffic congestion is high (T1 is large), share similar intuition. The residential centralization response from
when the marginal cost of traffic congestion is steeply increasing (T11 Proposition 2 means that these changes in parameter values will
is large), when the marginal effectiveness of road capacity investment result in more households residing in central zones close to the
is low (|T2 | is small), or when the marginal effectiveness of road ca- city center. The properties of T mentioned in Proposition 3 all cor-
pacity investment is steeply diminishing (T22 is large). The same con- respond to a setting where continuing to add road users to these
dition is necessary and sufficient for the result that the number of zones would increase road user costs significantly, and where ad-
households who travel by road in the optimal solution declines with ditional road capacity investment is not a very effective solution.
the difference between road and public transit emissions intensities As a result, in such a setting, the benefit of reducing emissions by
∂ Nir shifting households from road to public transit (which tends to re-
(i.e., for ∂ (e −e
r ) < 0).
t
duce Nir ) outweighs the increase in the total number of households
Proof. Eq. (4) can be expressed as Nir T2 + r = 0. This optimality in the zone (which tends to raise Nir ), leading to a net decrease in
condition must continue to hold under a change in γ , even as the the number of households who travel by road there. If the condi-
planner’s optimal decisions adjust in response to the parameter. tion Nir T22 (2T1 + Nir T11 ) > (T2 + Nir T12 )2 fails to hold, then the op-
This requirement that the optimal values of the variables change in posite is true and the latter effect outweighs the former. However,
such a way that the condition remains valid is equivalent to stipu- the properties of T that support the condition being true seem rea-
lating that its total derivative must equal zero, sonably likely to hold in most real-world settings. The literature
consistently suggests that the cost of traffic congestion is high, that
∂ Nir ∂ Ri
(T2 + Nir T12 ) + Nir T22 = 0. (10) small increases in traffic volume often have significant effects on
∂γ ∂γ travel time, and that the effectiveness of expanding road capacity
Similarly, Eq. (5) can be expressed as (Nir T1 + T ) − ( p + s ) + diminishes rapidly on the margin (Arnott & Small, 1994; Mogridge,
nγ (er − et ) = 0. This optimality condition must also continue to 1997; Piyapong, Noland, & Graham, 2012).

Please cite this article as: B.D. Leibowicz, Urban land use and transportation planning for climate change mitigation: A theoretical frame-
work, European Journal of Operational Research, https://doi.org/10.1016/j.ejor.2019.12.034
JID: EOR
ARTICLE IN PRESS [m5G;January 10, 2020;16:2]

8 B.D. Leibowicz / European Journal of Operational Research xxx (xxxx) xxx

Though less important, Eq. (12b) implies that the direction of Repeating the proof of Proposition 1 using Eq. (13c) instead of
the road infrastructure capacity response to γ depends not only on Eq. (3c) results in the expression
the sign of its denominator – which is the same as the denomina-
∂ Qi − f  (xi )( p + s + nγ et )
tor of Eq. (12a) and whose sign thus depends on the same condi- = . (14)
tion – but also on the sign of (T2 + Nir T12 ) in the numerator, which
∂ xi Qi [V  (Qi ) − nγ eq B (Qi )]
is ambiguous in general. This term is actually the derivative of the The presence of f (xi ) > 0 in the numerator does not change the fact
marginal benefit of road capacity investment on the left-hand side ∂Q
that ∂ x i > 0, by the same reasoning as in the original proof. There-
∂R i
of Eq. (4) with respect to Nir . Therefore, the sign of ∂γi depends on fore, Proposition 1 continues to hold.
whether the marginal benefits of expanding road capacity increase Using Eq. (14) instead of Eq. (8) to prove Proposition 2 yields
or decrease as more travelers use the road. the same expressions for the four cross-partial derivatives in
The final proposition of this section confirms that all of the Eq. (9a)–(9d), except all of them are multiplied by f(xi ) > 0. Incor-
theoretical results derived above continue to hold when one porating this positive coefficient does not change the signs of the
of the more restrictive assumptions of the model is relaxed. cross-partial derivatives, so Proposition 2 continues to hold.
Assumption 3 describes this generalization, and Proposition 4 for- The proof of Proposition 3 leverages Eqs. (4) and (5), which
malizes the outcome. remain valid as argued above. Therefore, its proof under the
relaxation in Assumption 3 is identical to the original, and
Assumption 3. Instead of household travel demand being propor-
Proposition 3 continues to hold. 
tional to the distance from the city center, as has been the case
up to this point, household travel demand is now allowed to be While the model remains necessarily stylized and features some
any monotonically increasing function of distance. Mathematically, simplifying assumptions, Proposition 4 is encouraging. It demon-
each household in zone i has a travel demand f(xi ), where f > 0. strates that none of the theoretical results derived in this section
relies on the rather rigid assumption that household travel demand
Proposition 4 (Travel demand relaxation). All of the theoretical re-
is proportional to distance from the city center, which is inherited
sults derived above remain valid under the relaxation in Assumption
from classic monocentric city model formulations. Assuming more
3. That is, Eqs. (4) and (5), as well as Propositions 1–3, continue to
generally that travel demand increases the further away from the
hold as long as household travel demand increases monotonically with
center a household resides likely encompasses the vast majority of
distance from the city center.
all real-world urban travel demand patterns.
Proof. Under Assumption 3, the first-order conditions become The four propositions proven above add a theoretical dimen-
∂L  sion to the evidence supporting the use of compact urban form
=1+ λ2 j = 0, (13a) and public transit investment for climate change mitigation. Larson
∂U j et al. (2012) and Leibowicz (2017) provide support for these urban
land use and transportation approaches based on numerical sim-
∂L
= − λ1 + λ2i ai (Nir + Nit )−2V  − λ3 [Ci + f (xi )(Nir T1 + T )] ulations, but the propositions in this section are far more general
∂ Nir in that they are derived analytically, and are not conditional on

− λ4 er f (xi ) + eq (B − ai (Nir + Nit )−1 B ) = 0, i ∈ I, specific functional forms and parameter values. Borck and Brueck-
ner (2018) analytically derive optimal energy taxation rates in an
(13b) urban setting with GHG damages, but they turn to numerical sim-
ulations to explore the implications for urban form. Unlike these
∂L previous model-based studies on urban planning and climate mit-
= − λ1 + λ2i ai (Nir + Nit )−2V  − λ3 [Ci + f (xi )( p + s )] igation, the model developed in this paper incorporates two trans-
∂ Nit

portation modes. Therefore, the propositions above shed new light
− λ4 et f (xi ) + eq (B − ai (Nir + Nit )−1 B ) = 0, i ∈ I, on transportation mode shifting as a component of urban miti-
(13c) gation strategies by elucidating the conditions under which auto-
mobile usage should decrease. These analytical results also reflect
greater coverage of urban GHG emissions than those obtained by
∂L Gaigné et al. (2012), who focus only on transportation emissions.
= −λ2i − λ3 (Nir + Nit ) = 0, i ∈ I, (13d)
∂ Ci In the broadest sense, the analysis in this section has provided
general theoretical support for urban policy approaches that view
∂L environmental sustainability as a motivation to encourage compact
= −λ3 f (xi )[Nir T2 + r ] = 0, i ∈ I, (13e)
∂ Ri urban forms, and clarified the settings in which automobile travel
should be reduced in absolute terms. It has also demonstrated how
∂L  urban climate change mitigation strategies in the residential and
= λ4 + γ λ2 j = 0. (13f)
∂E j
transportation sectors are closely intertwined, thereby offering a
strong argument in favor of integrated land use and transportation
Eq. (13a), (13d), and (13f) are the same as Eq. (3a), (3d), and planning.
(3f) from the original problem, because they do not depend on xi .
Therefore, λ2i ∀i ∈ I, λ3 , and λ4 , whose values are determined by 5. Numerical examples
these three first-order conditions, retain their same expressions.
Replacing the original Eq. (3e) with Eq. (13e) still yields Eq. (4) as In this section, the model is fully specified with functional
the rule for optimal road capacity investment. Similarly, setting forms and parameter values, then solved numerically. These nu-
Eq. (13b) equal to Eq. (13c) and eliminating like terms still yields merical examples are designed to visualize the behaviors of the
Eq. (5) as the rule for balancing road and public transit users, since model, illustrate the theoretical insights derived in the previous
all the f(xi ) coefficients will cancel just as the xi coefficients did in section, and explore the implications of varying degrees of flex-
the original analysis. Since all transportation capacity investment ibility to determine the urban land use and transportation plan.
and user costs are specified in per-unit-distance terms, it is unsur- It is important to clarify up front that the parameter values are
prising that Eqs. (4) and (5) continue to hold. essentially arbitrary and unitless, and they do not represent any

Please cite this article as: B.D. Leibowicz, Urban land use and transportation planning for climate change mitigation: A theoretical frame-
work, European Journal of Operational Research, https://doi.org/10.1016/j.ejor.2019.12.034
JID: EOR
ARTICLE IN PRESS [m5G;January 10, 2020;16:2]

B.D. Leibowicz / European Journal of Operational Research xxx (xxxx) xxx 9

Table 1
Model parameterization for the numerical examples.

Parameter Description Value

n Total urban area population 100


b Budget 2000
x1 Distance from inner zone to city center 2
x2 Distance from outer zone to city center 3
a1 Area of inner zone 4
a2 Area of outer zone 4
r Road capacity investment cost 1
s Public transit capacity investment cost 2
p Public transit user cost per unit distance 0.001
er Emissions intensity of road travel 0.01
et Emissions intensity of public transit travel 0.002
eq Emissions intensity of residences 0.001
γ Damage coefficient for emissions 2
ρ1 Road user cost function coefficient 0.003
ρ2 Road user cost function exponent on Nir 1.8
ρ3 Road user cost function exponent on Ri -0.2
1 Land utility function coefficient 3
β1 Residential energy use function coefficient 2
β2 Residential energy use function exponent 0.5
Fig. 1. Optimal spatial allocations of the population between the inner and outer
zones in the four scenarios. Since the inner and outer zones are assumed to have
the same area, the population densities are proportional to the heights of these
particular real-world setting. Application of the model to a real- bars. Home sizes are inversely proportional to their heights. (For interpretation of
world sustainable urban planning problem is left for future work. the references to color in this figure, the reader is referred to the web version of
this article.)
5.1. Parameterization and scenarios

Fully specifying the model requires functional forms for the respect to climate change, but cannot modify the underlying
land component of utility V(Qi ), road user cost T(Nir , Ri ), and land use pattern.
residential energy use B(Qi ). Parameter values must be restricted Scenario (d): The planner sets γ = 2 and optimizes public
to ensure that these functions conform to the desired properties transit infrastructure investments and transportation mode
stated in Section 3. The land component of utility is assumed to shares, but the number of households in each zone and the
ρ ρ
have the form V (Qi ) = 1 log Qi , where 1 > 0. T (Nir , Ri ) = ρ1 Nir 2 Ri 3 road capacity from each zone to the city center are fixed
is the chosen form for road user cost, where ρ 1 > 0, ρ 2 > 1, and to their values from the optimal solution to scenario (a). In
ρ 3 < 0. Residential energy use is assumed to follow the form other words, the planner can make optimal public transit in-
β vestment and transportation mode share decisions with re-
B(Qi ) = β1 Qi 2 , where β 1 > 0 and 0 < β 2 < 1. The form of the road
user cost is adopted from Brueckner and Franco (2018), whereas spect to climate change, but cannot modify the underlying
the land component of utility and residential energy use func- land use pattern or road infrastructure.
tions are merely simple forms chosen to fit the previously stated
The solution to scenario (b) is the optimal land use and trans-
assumptions.
portation plan under the true climate damage coefficient γ = 2.
The numerical examples presented here all represent an urban
Scenario (a) is included to assess how poorly a plan developed
area with two zones, an inner zone and an outer zone. Table 1
without considering the effects of climate change would perform
lists the parameter values assumed and includes descriptions of all
due to the excessive GHG emissions it generates relative to the true
model parameters. The model is implemented in the General Alge-
optimum. Whereas scenarios (a) and (b) are greenfield situations
braic Modeling System (GAMS)5 and solved using the IPOPT solver.
where the planner has the freedom to design the urban form from
Optimal urban land use and transportation plans are deter-
scratch, scenarios (c) and (d) are brownfield situations where cer-
mined for four scenarios. They differ in terms of whether GHG
tain aspects of the urban form already exist and cannot be altered
damages are considered in the planning process and the degree
to address climate change. Scenarios (c) and (d) both assume that
of flexibility the planner has to optimize the urban form. The four
the spatial allocation of the population is locked in to the optimal
scenarios are defined as follows:
spatial allocation from scenario (a), and scenario (d) also inherits
Scenario (a): The planner sets γ = 0, and therefore does not ac- the road infrastructure from scenario (a). The results of scenarios
count for damages caused by GHG emissions. The land use (c) and (d) thus demonstrate how much of the gap in performance
and transportation plan that is optimal for γ = 0 will not be between the scenario (a) and (b) plans can be reduced by optimiz-
optimal under the true damage coefficient value γ = 2. ing certain elements of the urban form for climate change, but not
Scenario (b): The planner sets γ = 2 and determines the optimal others.
land use and transportation plan in the presence of climate
damages. 5.2. Scenario results
Scenario (c): The planner sets γ = 2 and optimizes transporta-
tion infrastructure investments and mode shares, but the Fig. 1 shows how the population is optimally allocated between
number of households in each zone is fixed to its value the inner and outer zones in each scenario. It also indicates the
from the optimal solution to scenario (a). In other words, numbers of households in each zone who travel by road and by
the planner can make optimal transportation decisions with public transit. Since the inner and outer zones are assumed to have
the same area, their population densities are proportional to the
5
The GAMS model code is available on GitHub under an open source license. The
heights of the bars, while the amounts of land that each house-
GitHub repository can be found at https://github.com/bleibowicz/urban_climate_ hold consumes are inversely proportional to the bar heights. In
mitigation. line with Proposition 1, the inner zone always features a higher

Please cite this article as: B.D. Leibowicz, Urban land use and transportation planning for climate change mitigation: A theoretical frame-
work, European Journal of Operational Research, https://doi.org/10.1016/j.ejor.2019.12.034
JID: EOR
ARTICLE IN PRESS [m5G;January 10, 2020;16:2]

10 B.D. Leibowicz / European Journal of Operational Research xxx (xxxx) xxx

Table 2 than the difference in emissions for two reasons. First, GHG dam-
Utility and GHG emissions results.
ages are only one component of utility, which also depends on
Scenario (a) (b) (c) (d) land and non-land consumption. Second, the more compact ur-
Utility (Objective Value) 21.027 21.681 21.657 21.530 ban form and additional public transit investment required to re-
Total GHG Emissions 2.328 1.630 1.656 1.761 duce total GHG emissions in scenario (b) come at a cost, borne by
households through reductions in the other inputs to utility. Nev-
ertheless, the 3% lower utility in scenario (a) compared to scenario
(b) demonstrates the welfare losses that can arise due to failing to
population density than the outer zone. Moving from scenario (a)
account for climate change in urban planning.
where γ = 0 to scenario (b) where γ = 2, the optimal land use
In Table 2, it is very interesting that the planner can achieve
pattern shifts households from the outer zone to the inner zone.
outcomes fairly close to the optimum from scenario (b) even in
This is an illustration of the residential centralization response to
the brownfield scenarios (c) and (d) where significant elements of
more severe GHG damages from Proposition 2. Consistent with
the urban form cannot be modified from their scenario (a) con-
Proposition 3 and its case where the necessary and sufficient con-
figurations. In terms of reducing total GHG emissions from their
dition holds, the number of households traveling by road from
scenario (a) level, scenarios (c) and (d) respectively achieve 96%
each zone is lower in scenario (b), which accounts for GHG dam-
and 81% of the reduction that scenario (b) achieves. In terms of
ages, than in scenario (a), which does not consider them. This is
raising utility from its scenario (a) level, scenarios (c) and (d) re-
true in the inner zone even though it absorbs more households in
spectively achieve 96% and 77% of the improvement that scenario
scenario (b), as the increase in its population is more than offset
(b) achieves. From a practical standpoint, these results are very en-
by the increase in its public transit use.
couraging. They suggest that urban planners can effectively lever-
The spatial allocations of the population between the two zones
age transportation planning for climate change mitigation by in-
in scenarios (c) and (d) are constrained to be the same as that from
vesting in public transit to enable mode shift in this direction, even
scenario (a), so the heights of their bars are identical in Fig. 1. Rel-
if suboptimal land use patterns and road infrastructures are inher-
ative to the true greenfield optimum from scenario (b), the optimal
ited and cannot be modified.
solutions in scenarios (c) and (d) feature too many households in
The focus of this paper is undoubtedly on the analytical find-
the outer zone and too few in the inner zone because the plan-
ings and urban planning insights that can be derived from the
ner cannot leverage the residential centralization response. In sce-
model, rather than on data-intensive applications or computational
nario (c), the planner is able to optimize all transportation system
algorithms. Nevertheless, Appendix A examines the computational
variables with respect to the true GHG damage coefficient γ = 2.
scalability of the model as the number of zones increases.
Therefore, public transit mode shares are much higher in scenario
(c) than in scenario (a), and road mode shares are much lower. Sce-
6. Limitations
nario (d) exhibits a similar mode shift from road to public transit,
but the shift is slightly weaker in this case because the road infras-
Before summarizing the conclusions of this study, it is help-
tructure is fixed to its configuration from the scenario (a) solution.
ful to first clarify its primary limitations. The model formulation
The road infrastructure optimized for scenario (a) where γ = 0 is
is intentionally parsimonious in order to allow for analytical re-
oversized relative to the optimum under γ = 26 , but its fixed exis-
sults. These findings provide broad insights into urban land use
tence in scenario (d) reduces the variable cost of road travel and
and transportation planning for climate change mitigation that do
thereby weakens the optimal mode shift to public transit. This
not rely on specific functional forms and parameter settings. This
effect is visible in Fig. 1 as the bars for scenario (d) containing
style of modeling inevitably necessitates making simplifying as-
slightly more pink and less green than the bars for scenario (c).
sumptions and omitting many detailed factors which are important
Fig. 2 displays essentially the same information as Fig. 1, but
aspects of urban forms. As a result, the model must be viewed as
visualizes the optimal land use and transportation plans in each
a stylized representation of reality.
scenario via a two-dimensional representation of urban space. The
The social planner perspective brings all decision variables un-
colors of the zones indicate their population densities according to
der the control of a single optimizing agent. In reality, some of
the color scale appearing to the right of each subplot. The thick-
these decisions are made by individual households, such as trans-
nesses of the pink and green lines linking the zones to the city
portation mode choice. The model represents urban form only
center are proportional to the numbers of households traveling by
in terms of land used for housing and transportation infrastruc-
road and public transit, respectively. While Fig. 2 leads to the same
tures. It therefore does not include workplaces, schools, hospitals,
insights as Fig. 1, its spatial format is well suited to communicating
or other land uses that occupy space and could influence the ur-
results to urban policymakers and providing decision support.
ban form. The model only captures GHG emissions from the res-
Table 2 reports utility (objective value) and total GHG emis-
idential and passenger transportation sectors, thus omitting the
sions in each scenario. These are two important, high-level sum-
contributions of industrial, commercial, and other sectors to ur-
mary outcomes which were not depicted in the figures above. Util-
ban emissions. On the level of an individual household, residen-
ity is lowest in scenario (a) because the planner does not properly
tial and transportation emissions in the model depend only on
account for damages associated with GHG emissions when deter-
urban form variables, as in Gaigné et al. (2012), Larson et al.
mining the urban land use and transportation plan. The resulting
(2012), Leibowicz (2017), and Borck and Brueckner (2018). In re-
plan is optimal with respect to γ = 0, but not with respect to the
ality, household emissions are influenced by myriad other factors
true damage coefficient value γ = 2. Compared to the true opti-
that are not represented in the model, including income, technol-
mum from scenario (b), the scenario (a) plan generates 43% greater
ogy adoption, lifestyle, and behavioral choices.
GHG emissions, which yield a roughly 3% reduction in utility. Be-
The model inherently assumes a simple monocentric urban lay-
tween the two scenarios, the difference in utility is much smaller
out that is not reflective of the many real cities whose layouts are
polycentric or based on other geometries. Proposition 4 did show
6
that the key analytical results continue to hold when the classic
Specifically, total road infrastructure investment is only 0.411 in scenarios (b)
and (c) where it is optimized for γ = 2, compared to 6.857 in scenario (a) where
assumption that travel is strictly proportional to distance from the
it is optimized for γ = 0. Scenario (d) inherits the scenario (a) road infrastructure, center is relaxed. Furthermore, the formulation could accommo-
which is now oversized. date a polycentric layout simply by setting the value of each xi

Please cite this article as: B.D. Leibowicz, Urban land use and transportation planning for climate change mitigation: A theoretical frame-
work, European Journal of Operational Research, https://doi.org/10.1016/j.ejor.2019.12.034
JID: EOR
ARTICLE IN PRESS [m5G;January 10, 2020;16:2]

B.D. Leibowicz / European Journal of Operational Research xxx (xxxx) xxx 11

Fig. 2. Optimal urban land use and transportation plans in the four scenarios. The colors of the two zones indicate their population densities. The widths of the pink
and green lines linking the zones to the city center are proportional to the numbers of households traveling by road and public transit, respectively. Note that the two-
dimensional positions of the zones and their rectangular shapes are arbitrarily chosen for the purpose of illustration, since only their distances from the city center and land
areas feature in the model. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

parameter to equal the distance from zone i to the nearest urban monotonically, with distance from the city center. These are rather
center. In reality, the desirability of an urban land use and trans- intuitive structural properties that can also be derived from the
portation plan depends on countless details that are not captured standard monocentric city model (Brueckner, 1987). Nevertheless,
in this analytical model designed to yield broad insights, such as it is reassuring that they also emerge from the present framework
sunlight gain, visual quality, noise, and so forth. with its discrete representation of space, multiple transportation
modes, and incorporation of damages caused by GHG emissions.
Proposition 2 asserted that the optimal home size and population
7. Conclusions
density gradients become steeper if GHG emissions produce more
severe damages, residential and public transit emissions intensities
This paper has developed and analyzed a theoretical model of
are higher, or public transit is more costly. Therefore, climate dam-
urban land use and transportation planning for climate change
ages justify a higher degree of residential centralization, especially
mitigation. The model adopts the perspective of a social planner
if public transit is costly and thus a relatively less attractive mit-
who aims to maximize the common utility level of households
igation strategy. Proposition 3 revealed that if GHG damages are
in a city by spatially allocating the population to a set of resi-
more severe, or the emissions intensity advantage of public tran-
dential zones, investing in road and public transit infrastructures,
sit is greater, then the optimal amount of automobile travel de-
and determining transportation mode shares. GHG emissions asso-
clines in each zone if and only if a specific condition involving
ciated with residences, road travel, and public transit travel gener-
the properties of the road user cost function is satisfied. It says
ate damages that reduce utility. The major advantage of this frame-
that this is likely to be the case if the marginal cost of conges-
work is that it permits theoretical analysis that yields analytical
tion is sufficiently high and rising, or the marginal effectiveness
insights for environmentally sustainable urban planning.
of road capacity investment is sufficiently low and diminishing.
The main theoretical results were summarized in four propo-
Proposition 4 confirmed that all of the preceding theoretical results
sitions. Proposition 1 showed that, in the optimal solution, home
continue to hold as long as household travel demand increases
size increases montonically, and population density decreases

Please cite this article as: B.D. Leibowicz, Urban land use and transportation planning for climate change mitigation: A theoretical frame-
work, European Journal of Operational Research, https://doi.org/10.1016/j.ejor.2019.12.034
JID: EOR
ARTICLE IN PRESS [m5G;January 10, 2020;16:2]

12 B.D. Leibowicz / European Journal of Operational Research xxx (xxxx) xxx

monotonically with distance from the city center. This is a gen- Table A.1
Scaling of the problem size and solution time as the number of zones increases.
eralization of the original formulation, where travel demand was
implicitly assumed to be proportional to distance. Number of zones Number of Number of Solution time (s)
Following the theoretical analysis, the model was fully specified variables constraints
with functional forms and parameter values, and solved numeri- 2 10 5 0.057
cally for four scenarios. The scenarios differ in terms of whether 10 42 13 0.063
the planner accounts for GHG damages and the degree of flexibil- 100 402 103 0.141
1000 4002 1003 0.929
ity the planner has to optimize various aspects of the urban form.
10,000 40,002 1003 5.628
Results show that even if the planner inherits and cannot mod-
ify a land use plan and road infrastructure that are suboptimal
under climate change, the planner can achieve most of the GHG
emissions reduction and utility improvement that are achieved in algorithms. Nevertheless, it is worthwhile to briefly examine the
the unrestricted optimum. This demonstrates how effective public computational scalability of the model as the number of zones in-
transit investment and transportation mode shifting can be as mit- creases. The scenarios in Section 5 feature only two zones for ease
igation measures, and is encouraging from a practical standpoint. of illustration and to elucidate the key insights, but in reality, an
In the broadest sense, this paper has provided theoretical sup- urban area would likely be divided into many more zones for land
port for urban policy approaches that view environmental sustain- use and transportation planning purposes.
ability as a motivation to encourage compact urban forms, and First, it is instructive to assess how the problem size increases
clarified the conditions under which the absolute level of auto- with η, chosen to denote the number of zones. The model incorpo-
mobile usage should be reduced in each zone. It has also demon- rates 4η + 1 variables and η + 3 constraints. Therefore, the problem
strated how urban climate change mitigation strategies in the res- size scales linearly with the number of zones and does not suffer
idential and transportation sectors are closely intertwined, thereby from the curse of dimensionality.
offering a strong argument in favor of integrated land use and Some computational tests are conducted to confirm that the so-
transportation planning. lution time remains manageable as the number of zones increases
Future model extensions could enhance its capabilities and to reflect the degrees of spatial resolution likely to characterize
open up new analysis possibilities. The framework could represent real-world applications. As previously mentioned, the model is im-
transportation infrastructure networks that link zones to one an- plemented in GAMS and solved using the IPOPT solver.7 The tests
other in addition to the city center. Allowing the social planner to are run on a Lenovo ThinkPad X1 Carbon (6th Gen) laptop with an
affect building energy intensities would incorporate an additional Intel Core i5-8250U processor and 8 GB of RAM.
urban emissions reduction strategy. Investment in road and public The computational test results are shown in Table A.1. Deter-
transit infrastructures might reflect fixed costs and economies of mining optimal solutions for small instances like the two-zone ex-
scale rather than constant marginal costs. The current scope focus- amples in Section 5 takes under one-tenth of a second. Even with
ing on energy demand could be expanded to incorporate decisions thousands of zones, optimal solutions are achieved in less than
related to urban energy supply. A dynamic version of the model ten seconds. Since each zone would correspond to a spatial unit
could be constructed in which the planner gradually evolves the ranging from a city block up to a neighborhood in a real plan-
urban form from an initial state by making decisions over multi- ning application, the number of zones is unlikely to be greater than
ple time periods. In each period, there would be a limit on how the larger instances considered in Table A.1. While this theoretical
many households the planner can relocate to different zones or model designed for analytical insight has limitations in terms of
shift to the other transportation mode, or on how much the plan- practical application (covered in Section 6), its computational com-
ner can incrementally invest in transportation infrastructure. This plexity would not be a decisive problem.
dynamic extension would be well suited to explore the implica-
tions of behavioral rigidities and how they can be overcome. For References
example, the constraints on the allowable household relocations or
Acuto, M., Parnell, S., & Seto, K. C. (2018). Building a global urban science. Nature
transportation mode shifts in each period could be tightened to re- Sustainability, 1, 2–4.
flect households who resist these changes in lifestyle. Constraints Alonso, W. (1964). Location and land use. Cambridge, MA: Harvard University Press.
in later periods could depend endogenously on patterns realized Anas, A., Arnott, R., & Small, K. A. (1998). Urban spatial structure. Journal of Eco-
nomic Literature, 36(3), 1426–1464.
in earlier periods to incorporate peer effects that make transporta- Arnott, R., & Small, K. (1994). The economics of traffic congestion. American Scientist,
tion and lifestyle choices path-dependent (McCoy & Lyons, 2014; 82(5), 446–455.
Morrison & Lawell, 2016). These are all fruitful directions for fu- Attardi, R., Cerreta, M., Sannicandro, V., & Torre, C. M. (2018). Non-compensatory
composite indicators for the evaluation of urban planning policy: The Land-Use
ture research, but any model of such a multifaceted problem must Policy Efficiency Index (LUPEI). European Journal of Operational Research, 264(2),
carefully balance complexity with the ability to obtain generaliz- 491–507.
able insights. Bai, X., Dawson, R. J., Ürge-Vorsatz, D., Delgado, G. C., Salisu Barau, A., Dhakal, S., . . .
Schultz, S. (2018). Six research priorities for cities and climate change. Nature,
555(7694), 23–25.
Acknowledgments Bai, X., Surveyer, A., Elmqvist, T., Gatzweiler, F. W., Güneralp, B., Parnell, S., . . .
Webb, R. (2016). Defining and advancing a systems approach for sustainable
cities. Current Opinion in Environmental Sustainability, 23, 69–78.
While conducting this study, the author received funding from
Bammi, D., & Bammi, D. (1979). Development of a comprehensive land use plan
the UT Energy Institute for its Energy Infrastructure of the Future by means of a multiple objective mathematical programming model. Interfaces,
project. Its support was greatly appreciated. Three anonymous re- 9(2), 50–63.
viewers provided excellent comments and suggestions that helped Belaïd, F. (2016). Understanding the spectrum of domestic energy consumption: Em-
pirical evidence from France. Energy Policy, 92, 220–233.
significantly improve this paper. Borck, R., & Brueckner, J. K. (2018). Optimal energy taxation in cities. Journal of the
Association of Environmental and Resource Economists, 5(2), 481–516.
Appendix A. Computational scalability
7
Note that the model was also solved successfully using other NLP solvers in-
The focus of this paper is undoubtedly on the analytical find- cluding CONOPT and COUENNE to verify that they reach identical solutions. The
ings and urban planning insights that can be derived from the results in Table A.1 were obtained using IPOPT because its solution times were the
model, rather than on data-intensive applications or computational shortest.

Please cite this article as: B.D. Leibowicz, Urban land use and transportation planning for climate change mitigation: A theoretical frame-
work, European Journal of Operational Research, https://doi.org/10.1016/j.ejor.2019.12.034
JID: EOR
ARTICLE IN PRESS [m5G;January 10, 2020;16:2]

B.D. Leibowicz / European Journal of Operational Research xxx (xxxx) xxx 13

Brill, E. D., Chang, S.-Y., & Hopkins, L. D. (1982). Modeling to generate alternatives: Leibowicz, B. D. (2017). Effects of urban land-use regulations on greenhouse gas
The HSJ approach and an illustration using a problem in land use planning. emissions. Cities, 70, 135–152.
Management Science, 28(3), 221–235. Li, X., & Parrott, L. (2016). An improved genetic algorithm for spatial optimization of
Brotchie, J. (1969). A general planning model. Management Science, 16(3), 265–266. multi-objective and multi-site land use allocation. Computers, Environment and
Brotchie, J. F. (1978). A new approach to urban modelling. Management Science, Urban Systems, 59, 184–194.
24(16), 1753–1758. Lohrey, S., & Creutzig, F. (2016). A ‘sustainability window’ of urban form. Transporta-
Brueckner, J. K. (1987). The structure of urban equilibria: A unified treatment of the tion Research Part D: Transport and Environment, 45, 96–111.
muth-mills model. In E. Mills (Ed.), Handbook of regional and urban economics: Ma, J., Liu, Z., & Chai, Y. (2015). The impact of urban form on CO2 emission from
2 (pp. 821–845). Amsterdam: North-Holland. work and non-work trips: The case of Beijing, China. Habitat International, 47,
Brueckner, J. K., & Franco, S. F. (2018). Employer-paid parking, mode choice, and 1–10.
suburbanization. Journal of Urban Economics, 104(December 2017), 35–46. Marshall, J. D. (2008). Energy-efficient urban form. Environmental Science & Technol-
C40 Cities (2019). C40 Cities. http://www.c40.org/. ogy, 42(9), 3133–3137.
Cajot, S., Peter, M., Bahu, J. M., Guignet, F., Koch, A., & Maréchal, F. (2017a). Ob- McCoy, D., & Lyons, S. (2014). Consumer preferences and the influence of networks
stacles in energy planning at the urban scale. Sustainable Cities and Society, 30, in electric vehicle diffusion: An agent-based microsimulation in ireland. Energy
223–236. Research & Social Science, 3, 89–101.
Cajot, S., Schüler, N., Peter, M., Koch, A., & Maréchal, F. (2017b). Interactive optimiza- Mills, E. S. (1967). An aggregative model of resource allocation in a metropolitan
tion for the planning of urban systems. Energy Procedia, 122, 445–450. area. American Economic Review, 57, 197–210.
Caparros-Midwood, D., Barr, S., & Dawson, R. (2015). Optimised spatial planning to Mogridge, M. J. H. (1997). The self-defeating nature of urban road capacity policy: A
meet long term urban sustainability objectives. Computers, Environment and Ur- review of theories, disputes and available evidence. Transport Policy, 4(1), 5–23.
ban Systems, 54, 154–164. Morrison, G. M., & Lawell, C. Y. C. L. (2016). Driving in force: The influence of work-
Çolak, S., Lima, A., & González, M. C. (2016). Understanding congested travel in ur- place peers on commuting decisions on U.S. Military bases. Journal of Economic
ban areas. Nature Communications, 7, 1–8. Behavior & Organization, 125, 22–40.
Creutzig, F., Baiocchi, G., Bierkandt, R., Pichler, P.-P., & Seto, K. C. (2015). Global Muth, R. (1969). Cities and housing. Baltimore: Johns Hopkins University Press.
typology of urban energy use and potentials for an urbanization mitigation Navamuel, E. L., Rubiera Morollón, F., & Moreno Cuartas, B. (2018). Energy consump-
wedge. Proceedings of the National Academy of Sciences, 112(20), 6283–6288. tion and urban sprawl: Evidence for the Spanish case. Journal of Cleaner Produc-
Creutzig, F., Roy, J., Lamb, W. F., Azevedo, I. M., de Bruin, W. B., Dalkmann, H., . . . We- tion, 172, 3479–3486.
ber, E. U. (2018). Towards demand-side solutions for mitigating climate change. Newman, P. W., & Kenworthy, J. R. (1989). Gasoline consumption and cities: A com-
Nature Climate Change, 8, 268–271. parison of U.S. cities with a global survey. Journal of the American Planning As-
Deetjen, T. A., Conger, J. P., Leibowicz, B. D., & Webber, M. E. (2018). Review of cli- sociation, 55(1), 24–37.
mate action plans in 29 major U.S. cities: Comparing current policies to research Pan, H., Shen, Q., & Zhang, M. (2009). Influence of urban form on travel behaviour
recommendations. Sustainable Cities and Society, 41, 711–727. in four neighbourhoods of Shanghai. Urban Studies, 46(2), 275–294.
Ewing, R., & Rong, F. (2008). The impact of urban form on US residential energy Piyapong, J., Noland, R. B., & Graham, D. J. (2012). Marginal productivity of expand-
use. Housing Policy Debate, 19(1), 1–30. ing highway capacity. Journal of Transport Economics and Policy, 46(3), 337–347.
Gaigné, C., Riou, S., & Thisse, J.-F. (2012). Are compact cities environmentally Prieur-Richard, A.-H., Walsh, B., Craig, M., Melamed, M. L., Colbert, L., Pathak, M.,
friendly? Journal of Urban Economics, 72(2–3), 123–136. Connors, S., Bai, X., Barau, A., Bulkeley, H., Cleugh, H., Cohen, M., Colenbrander,
Gilbert, K. C., Holmes, D. D., & Rosenthal, R. E. (1985). A multiobjective discrete S., Dodman, D., Dhakal, S., Dawson, R., Espey, J., Greenwalt, J., Kurian, P., Lee,
optimization model for land allocation. Management Science, 31(12), 1509–1522. B., Leonardsen, L., Masson-Delmotte, V., Munshi, D., Okem, A., Delgado Ramos,
Glaeser, E. L., & Kahn, M. E. (2010). The greenness of cities: Carbon dioxide emis- G. C., Sanchez Rodriguez, R., Roberts, D., Rosenzweig, C., Schultz, S., Seto, K.,
sions and urban development. Journal of Urban Economics, 67(3), 404–418. Solecki, W., van Staden, M., & Ürge-Vorsatz, D. (2018). Extended version: Global
Grout, C. A., Jaeger, W. K., & Plantinga, A. J. (2011). Land-use regulations and prop- Research and Action Agenda on Cities and Climate Change Science.
erty values in Portland, Oregon: A regression discontinuity design approach. Re- Savelsbergh, M., & Van Woensel, T. (2016). 50th anniversary invited article city lo-
gional Science and Urban Economics, 41(2), 98–107. gistics: Challenges and opportunities. Transportation Science, 50(2), 579–590.
Hammad, A. W., Akbarnezhad, A., & Rey, D. (2017). Sustainable urban facility lo- Seto, K., Dhakal, S., Bigio, A., Blanco, H., Delgado, G., Dewar, D., . . . Ra-
cation: Minimising noise pollution and network congestion. Transportation Re- maswami, A. (2014). Human Settlements, Infrastructure, and Spatial Planning. In
search Part E: Logistics and Transportation Review, 107, 38–59. O. Edenhofer, R. Pichs-Madruga, Y. Sokona, E. Farahani, S. Kadner, K. Seyboth, . . .
Harris, B. (1961). Some Problems in the Theory of Intra-Urban Location. Operations J. C. Minx (Eds.), Climate change 2014: Mitigation of climate change. Contribution
Research, 9(5), 695–721. of working group III to the Fifth assessment report of the intergovernmental panel
Hirte, G., & Tscharaktschiew, S. (2013). The optimal subsidy on electric vehicles in on climate change. Cambridge, UK and New York, NY, USA: Cambridge University
German metropolitan areas: A spatial general equilibrium analysis. Energy Eco- Press.
nomics, 40, 515–528. Solecki, W., Rosenzweig, C., Dhakal, S., Roberts, D., Barau, A. S., Schultz, S., &
Hua, H., Hovestadt, L., Tang, P., & Li, B. (2019). Integer programming for urban de- Ürge-Vorsatz, D. (2018). City transformations in a 1.5 C warmer world. Nature
sign. European Journal of Operational Research, 274(3), 1125–1137. Climate Change, 8(3), 177–181.
Karathodorou, N., Graham, D. J., & Noland, R. B. (2010). Estimating the effect of ur- Stewart, T. J., & Janssen, R. (2014). A multiobjective GIS-based land use planning
ban density on fuel demand. Energy Economics, 32(1), 86–92. algorithm. Computers, Environment and Urban Systems, 46, 25–34.
Kennedy, C., Steinberger, J., Gasson, B., Hansen, Y., Hillman, T., Havranek, M., . . . Taniguchi, E., Thompson, R. G., & Yamada, T. (2016). New opportunities and chal-
Mendez, G. V. (2009). Greenhouse gas emissions from global cities. Environmen- lenges for city logistics. Transportation Research Procedia, 12(June 2015), 5–13.
tal Science & Technology, 43(19), 7297–7302. United Nations (2018). World Urbanization Prospects: The 2018 Revision.
Kloosterman, R. C., & Musterd, S. (2001). The polycentric urban region: Towards a VandeWeghe, J. R., & Kennedy, C. (2007). A spatial analysis of residential greenhouse
research agenda. Urban Studies, 38(4), 623–633. gas emissions in the Toronto census metropolitan area. Journal of Industrial Ecol-
Larson, W., Liu, F., & Yezer, A. (2012). Energy footprint of the city: Effects of ur- ogy, 11(2), 133–144.
ban land use and transportation policies. Journal of Urban Economics, 72(2–3), Ying, J. Q. (2015). Optimization for multiclass residential location models with con-
147–159. gestible transportation networks. Transportation Science, 49(3), 452–471.

Please cite this article as: B.D. Leibowicz, Urban land use and transportation planning for climate change mitigation: A theoretical frame-
work, European Journal of Operational Research, https://doi.org/10.1016/j.ejor.2019.12.034

Potrebbero piacerti anche