Sustainable Cities and Society 74 (2021) 103214

Contents lists available at ScienceDirect

Sustainable Cities and Society

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

Multi-objective urban land use optimization using spatial data: A

systematic review
Md. Mostafizur Rahman a, b, *, György Szabó a
a
Department of Photogrammetry and Geoinformatics, Budapest University of Technology and Economics, Hungary
b
Department of Urban and Regional Planning, Rajshahi University of Engineering and Technology, Bangladesh

A R T I C L E I N F O A B S T R A C T

Keywords: Land use optimization is a promising approach to achieve urban sustainability. Despite the increasing number of
Multi-objective optimization literature on land use optimization, a little investigation is made to systematically review urban land use opti­
Land use mization: its objectives, methodological approaches, and spatial data used etc. This creates room to review the
Spatial compactness
methodological approaches to urban land use optimization. This study systematically reviews 55 articles
Pareto front-based method
Heuristic algorithm
following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) protocol to un­
Preferred reporting items for systematic derstand important aspects of urban land use optimization. We have found that the most common objectives
reviews and meta-analyses (PRISMA) protocol which were used in urban land use optimization are maximization of spatial compactness (16.67 %, n=28) and
maximization of land use compatibility (13.69%, n=23), followed by maximization of land use suitability
(11.90%, n=20). The findings suggest that a) one and only one land use in each cell, b) minimum and maximum
area of certain land use, and c) restriction on specific land use change are the important constraints. This study
also identifies that urban sustainability has been merely touched upon in urban land use optimizations. While
environmental (including ecology) and economic aspects of urban sustainability were included in 46.67% and
43.33% studies respectively, the social aspect (10%, n= 3) was mostly ignored. Our findings also indicate that
there is no generalized method to measure economic, environmental, and social benefits from land use planning.
This study also finds that the Genetic Algorithm (GA) (32.14%, n=18) accounts for a major contribution to solve
the urban land use optimization problem. Based on the findings, this study proposes some recommendations for
further research and practice. The most important of them include a) framing land use optimization objective
functions considering urban sustainability, b) developing a standard method to calculate values of objective
functions, and c) integrating a participatory approach with mathematical optimization to derive more feasible
solutions. These recommendations could be the scope of future research.

1. Introduction competing and even, sometimes, conflicting (Huang & Zhang, 2014).
For example, if residential development occurs in a low-lying area, it
1.1. Background may fulfill the demand for urban housing, but it will create a problem for
urban drainage. Construction of building structures may increase eco­
Land use optimization is an important tool to achieve sustainable nomic benefit, but it will deteriorate the environment and urban health.
urban land use planning, which aims to achieve long-term balanced So, careful land allocation is of paramount importance in land use
urban development through economic prosperity, efficient resources planning. Here comes the concept of land use optimization that allows
use, environmental protection, and social equity (Cao, Huang, Wang, & generating alternative land use scenario from which the decision-maker
Lin, 2012; Cohen, 2017; Ligmann-Zielinska, Church, & Jankowski, choose the best option considering conflicting interest (Cao & Ye, 2013;
2008). It is entrusted with allocating different land uses (e.g., residential Ligmann-Zielinska, Church, & Jankowski, 2008).
land, industrial, commercial, recreational facility, open space, parks, Land use optimization is a branch of spatial optimization that con­
and green land, etc.) in such a way as to derive optimal benefits. (Cao, sists of three essentials elements. These are a) decision variables, b)
Huang, Wang, & Lin, 2012). But, in reality, these objectives are objective functions, and c) constraints (Ligmann-Zielinska, 2017; Tong

* Corresponding author.
E-mail addresses: [email protected] (Md.M. Rahman), [email protected] (G. Szabó).

https://doi.org/10.1016/j.scs.2021.103214
Received 28 October 2020; Received in revised form 25 July 2021; Accepted 26 July 2021
Available online 28 July 2021
2210-6707/© 2021 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Md.M. Rahman and G. Szabó Sustainable Cities and Society 74 (2021) 103214

Fig. 1. Mapping between the decision variable space and the objective function space (Givi & Asadi, 2014).

& Murray, 2012). A multi-objective land use optimization problem can constructed a multi-objective land use optimization model using goal
be formulated as follows (Deb, 2011): programming and a weighted-sum approach supported by a
boundary-based genetic algorithm; Gao et al. (2020) developed an
Minimize or Maximize fm (x), m = 1, 2, ……....M;
improved Non-dominated Genetic Algorithm-II (NSGA-II) to achieve
sustainable urban land use; Zhang, Wang, Cao, He, & Shan, 2019 pro­
Subject to gj (x) ≥ 0, j = 1, 2………….J;
posed a genetic algorithm-based multi-objective optimization (MOO)
hk (x) = 0, k = 1, 2, ……..…K; approach; Li & Ma, 2018 designed a new simulated annealing (SA) al­
gorithm for spatial optimization of land resources. The choice of
x(L) ≤ xi ≤ xi(U) , i = 1, 2………….n methodological approach, objective functions, constraints, and spatial
scale in the data have much impact on the outcome of the study. For
i

Where fm (x) constitute the objective functions; gj (x) and hk (x) are example, Song and D. M. Chen (2018b) performed a comparative
the inequality and equality constraints, respectively. xi is the spatial analysis on three heuristic algorithms, namely Genetic Algorithm (GA),
decision variable; x(L)
i and x(U)
i are the lower and upper bounds of the Simulated Annealing (SA), and Particle Swarm Optimization (PSO), to
decision variable. solve land use optimization problem. They considered three objectives,
Decision variables in the optimization problem can be defined as including a) maximizing land suitability, b) maximizing spatial
quantities whose values can be changed and need to be determined to compactness, and c) minimizing land conversion cost. Their result shows
solve the optimization problem and get the solution(s) (Wang, Zhang, & that the performances of the three algorithms are different on the three
Peng, 2021). The decision variables strongly influence the formulation objectives. So, a clear understanding of their selection is very important.
of objective functions. An objective function simply is a function of A Systematic review of different aspects of land use optimization
decision variables. The objective function is either minimized or maxi­ might offer ground to clearly understand and rethink the optimization
mized to find the optimal values of decision variables. The values of process and foresee associated outcomes. By nature, it selects, identifies,
decision variables are called solutions to the problem. In the context of and assesses related research using explicit methods to collect and
an optimization problem, objective function space is determined by the analyze relevant data and aspects from different studies. It identifies the
decision variable space. For each solution in the decision variable space, differences in approaches and associated outcomes towards a problem
there is a point in the objective space. A mapping is developed between under investigation. It also provides an insightful synthesis of contem­
m-dimensional decision variable space and n-dimensional objective porary evidence related to a specific research problem (Nasir Ahmad,
function space (Givi & Asadi, 2014). The mapping between the decision Mustafa, Muhammad Yusoff, & Didams, 2020). Although there is an
variable space and the objective function space is illustrated in Fig. 1. increasing number of works in urban land use optimization, a limited
In the case of the land use optimization problem, four main decision number of studies investigated land use optimization from different
variables are considered. These four decision variables are type, size, approaches and settings. Therefore, there is a clear need for a systematic
location, and capacity (Huang, Liu, Li, Liang, & He, 2013; Mohammadi, review of urban land use optimization to synthesize the core elements
Nastaran, & Sahebgharani, 2015). By combining these variables, any and to consolidate contemporary evidence of the outcome. As far as the
land use optimization problem tries to find the quantity (size) of specific authors are aware, there is no comprehensive review on this topic. This
land use (type) which needs to be allocated in a particular place (loca­ paper attempts to fill this gap by systematically reviewing and synthe­
tion) in order to optimize a specific objective (e.g., maximize economic sizing contemporary research, and it is expected that this review paper
benefits). Two types of constraints may be contained in the will add value to future research on the topic concerned.
multi-objective optimization problem: a) inequality constraints, b)
equality constraints, and c) lower and upper boundaries constraints.
1.3. Objective and research questions
Detail discussion on constraints has been made in Section 3.2.2.
Against the background described in earlier sections, this paper ex­
1.2. Rationale of this research plores the existing land use optimization to understand its elements,
aspects, and context. To this end, this study will answer the following
Studies on land use optimization are significantly diverse in terms of research questions. The novelty of this study is to investigate the
methods, objectives, and other elements (Mingjie Song and Chen, following research questions, which were not answered in any previous
2018b; Wang, Gao, Liu, & Song, 2010). For example, Cao et al. (2020) study by a systematic investigation.

2
Md.M. Rahman and G. Szabó Sustainable Cities and Society 74 (2021) 103214

Table 1 follow an unstructured review approach to establish a theory on a spe­

Search term, sources, and the corresponding number of articles identified. cific research area (Steward, 2004). For example, some narrative re­
Search Term Number of Articles Search views may illustrate methods and research design, and some may outline
found from Date the historical development of a theory, some may concentrate on eval­
Web of Scopus uating the theory, some may define the scoping of research. Traditional
Science
narratives have several limitations. They only present a broad overview
(land AND (optim* OR Allocat*)) OR (((land 1119 1150 23 July, of a research problem, do not have a predefined protocol, and emphasize
AND zoning) OR (spatial AND planning)) 2020 the author’s intuition and experience, and even they do not answer any
AND optim*)
specific research question (Pae, 2015). Another important limitation in a
narrative review is that this type of review includes the author’s biases,
a) Which objectives are the most important in urban land use and they are not suitable for replication (Robinson & Lowe, 2015).
optimization? On the other hand, systematic reviews, for example, PRISMA pro­
b) Which constraints, approaches, and methods are commonly used to tocol, follow a structured analysis of data and evidence collected from
construct and solve urban land use optimization problems? existing research (Pae, 2015). In the whole review process, strict criteria
c) What are the spatial data and models used in urban land use are defined to include and exclude the available research and to limit the
optimization? author’s bias. Due to the use of a well-defined protocol, it can limit bias
and promote scientific evidence (Zhu, 2020). Considering the limita­
2. Materials and methods tions of traditional narratives, we have used PRISMA protocol in our
study. Based on the PRISMA protocol, the specific method used in this
This study is a presentation of literature on land use optimization study is presented in the following sections.
problems following the Preferred Reporting Items for Systematic Re­
views and Meta-Analyses (PRISMA) protocol (Moher et al., 2009). 2.1. Search strategy
Although PRISMA protocol was developed exclusively for health and
medical-related field, recently, it has been using in many areas, We have searched the Web of Science Core Collection database and
including environment, water, agriculture, land management, and sus­ Scopus database to find the article. We have applied our search term on
tainability research (Cohen, 2017; Higgins, Balint, Liversage, & Winters, “Article Title” only. We have limited our search to journal articles
2018). PRISMA has some advantages, which include a) developing written in English with no time limit. The search term used, and the
straightforward research questions, b) explicit inclusion and exclusion corresponding number of articles identified have been presented in
criteria, c) aiming to identify and evaluate massive relevant scientific Table 1. After the search is finished, we have exported the metadata of
articles, d) charting and tabulating of data, and 5) reporting and sum­ articles as CSV file format from Scopus and an Excel file from the Web of
marizing the results. This approach is also easy to use compared to other Science database. The exported CSV file has been converted into Excel
“traditional narratives.” Traditional narratives in literature review format. We have also identified 22 articles from other sources. All these

Fig. 2. PRISMA flow diagram of the literature search and final inclusion of papers Adapted from (Moher et al., 2009).

3
Md.M. Rahman and G. Szabó Sustainable Cities and Society 74 (2021) 103214

Fig. 3. Country-wise distribution of reviewed studies.

articles’ data have been combined in Excel for further analysis. The duplicate records. Some 865 articles were removed, leaving 1246 arti­
systematic screening procedure, to include the full-text article for re­ cles for the next stage of screening. In the third stage, we have excluded
view, has been described in Section 2.3. the studies if a) they are not directly relevant to our study objective
(skimming in article title and abstract), and b) they do not focus on
urban areas. Thus, 1131 articles are excluded keeping 295 articles for
2.2. Eligibility criteria
full-text evaluation. In the fourth and final stage, we have scanned these
295 articles thoroughly and finally included 55 articles for our review
According to PRISMA protocol, two categories of eligibility criteria
based on the inclusion and exclusion criteria described in Section 2.2.
are recommended for the inclusion and exclusion of research articles.
These are study characteristics (e.g., problem, intervention and study
design, etc.) and report characteristics (e.g., geographical location, years 2.4. Charting and tabulating of data
considered, language, publication status, etc.) (Moher et al., 2009).
Following this guideline, we have included those studies which a) We have created the article database using MS Excel. The article
explicitly focuses on multi-objective land use optimization; b) follows database includes the title of the paper, names of the authors, name of
the mathematical approach of optimization; c) uses spatial data; and d) journal, country of the study, year of publication year, objectives, con­
produce measurable quantitative results. In the case of report charac­ straints, spatial data used, spatial data model, Approach to construct
teristics, we have considered the studies which a) focused only on urban optimization problem, and Method. The list of the selected articles
areas; b) were written in the English language; c) were published in a (n=55), along with relevant attributes, can be found in https://doi.org/
peer-reviewed journal article. Some land use optimization problems 10.17632/vr8w6v6yj8.1. The study location map has been created using
focused on a regional scale; we did not include those studies in our re­ ArcGIS 10.2 software, and all other Fig.s have been prepared in R(R
view because the objectives of regional level land use optimization are Development Core Team, 2011).
somewhat different from those of urban areas. For example, one com­
mon objective for land use optimization on a regional scale is the 3. Results and discussion
maximization of agricultural productivity, but it is not much relevant in
the case of an urban area. This section presents our findings of the study. Firstly, we have
presented the general report characteristics (e.g., geographical location,
publication year, and publication journal) then we have presented our
2.3. Systematic screening of the articles
detailed findings of study characteristics.
The PRISMA flow diagram has been presented in Fig. 2. PRISMA
protocol has proposed a four-stage systematic screening procedure for 3.1. Report characteristics
inclusion and exclusion of the articles for review. In the first stage, based
on the search strategy, a total of 2291 articles have been initially We have compiled 55 papers, global-scale distribution of which
selected from Web of Science, Scopus, and other sources. In the second shows a broader perspective about how urban land use optimization was
stage, we have sorted all the articles in MS Excel and removed the emphasized in different countries in the land use planning process. Our

4
Md.M. Rahman and G. Szabó Sustainable Cities and Society 74 (2021) 103214

the western world, and most of them focus on a regional scale rather
than the urban scale. For example, García, Rosas, García-Ferrer, &
Barrios, 2017 worked on sustainable land use allocation at a subregional
scale in Spain, Bourque et al. (2019) worked on land use optimization in
California to increase agricultural benefits, Uludere Aragon et al. (2017)
focused on perennial bioenergy crop production in land allocation in the
United States, Armsworth et al. (2020) optimized land resources allo­
cation in the southwest United States. The studies did not focus on an
urban area, and there are differences in the optimization objectives
between an urban area and other areas (e.g., agricultural area). Whereas
our study focuses only on the urban area and multi-objective optimi­
zation problems, so we have included those studies only, which fulfilled
the eligibility criteria as discussed in Section 2.2.
Fig. 4 shows the trend of studies on urban land use optimization.
Fig. 4 indicates that before 2002 there were no studies on urban land use
optimization. It can be noted that, although there were many studies on
land use optimization before 2002, those studies did not fulfill our in­
clusion criteria, and hence, we did not include those studies. This review
Fig. 4. Distribution of studies by publication year. only focuses on urban land use optimization, not other land uses (e.g.,
Rural land use optimization, Agricultural land use optimization, etc.).
analysis (Fig. 3) shows that urban land use optimization studies were There is one interesting point that there is an increasing trend in studies
conducted in 7.61% (n=15) countries globally (Considering 197 coun­ on urban land use optimization since 2011.
tries in the world according to UN’s recognition). Out of which, 52.73 % The top 7 journals in which studies on urban land use optimization
(n=29) studies were conducted in different cities in the Peoples Republic have been published are presented in Fig. 5. “International Journal of
of China, while 7.27% (n=4) studies were conducted in Iran and only Geographical Information Science” and “Computers, Environment and
5.45% (n=3) studies were conducted in Bangladesh, Greece, Urban Systems” are two important journals in which most of the studies
Netherlands, and Spain. It is clear from our analysis that the Peoples’ (n=7) were published.
Republic of China emphasized more on optimizing urban land use
compared to other countries in the world. However, the geographical
3.2. Study characteristics
distribution of studies seems disproportionate since most of the studies
are from China, where the representation of the western world is
In section 1.4, we have set our research objective and research
limited. The fact is that there are many studies of land management in
questions. We have set three research questions to understand urban

Fig. 5. Top seven Journal, which published at least two papers with a primary focus on urban land use optimization.

Fig. 6. Ten (10) most frequently used objectives in the urban land use optimization problem.

5
Md.M. Rahman and G. Szabó Sustainable Cities and Society 74 (2021) 103214

land use optimization. In the following sections, we have presented and Handayanto, Tripathi, Kim, & Guha, 2017; Mingjie Song and Chen,
discussed our study findings to answer all the research questions. 2018b). Our findings on land use compatibility are also supported by
many other studies. For example, Pahlavani, Sheikhian, & Bigdeli, 2020
3.2.1. Objective formulation evaluated residential land use compatibilities using the density-based
We have identified about 43 objectives1 from 55 Journal articles, model to understand to what extent other non-residential land uses
which were used in urban land use optimization problems. These ob­ are compatible with residential land uses; Kiani Sadr, Nassiri, Hosseini,
jectives are generally of two types in the case of urban land use opti­ Monavari, & Gharagozlou, 2014 assessed the impact of land use
mization problems: a) additive objectives and b) spatial objectives. compatibility on noise pollution and found that incompatible land uses
Additive objectives are calculated using the representative value or leads to noise pollution; Libby & Sharp, 2003 worked on policy issues to
attribute of each land parcel (referred to as a cell) without considering ensure land use compatibility in rural-urban fringe; Cohn et al. (2005)
the characteristics of land use configuration. On the other hand, spatial studied the impact of land use compatibility on highway noise.
objectives are calculated by considering the characteristics of land-use The third important objective is the maximization of land use suit­
configuration, e.g., the connectedness, contiguousness, or fragmenta­ ability (LUS) which has also been used in many land use optimization
tion of land-uses under the area of interest (Stewart, Janssen, & Van problems. LUS is defined as the degree of fitness of a certain type of land
Herwijnen, 2004). The ten (10) most frequent objectives are presented use to be allocated in a specific land parcel considering preferences,
in Fig. 6. requirements, or predictors of some activities. LUS has been considered
According to Fig. 6, the most frequently used objective is the maxi­ in many land use optimization problems because it informs decision-
mization of spatial compactness. It constitutes about 16.67 % (n=28) of makers about the social, economic, and environmental consequences
the total objectives. This objective has been undertaken in many studies of particular land use choices within land use planning. Due to the
because it is preferred that similar land uses should be in the vicinity of continuous urban sprawling, it creates the burden on the urban envi­
the same land uses to generate high general benefit from land uses. ronment beyond the capacity of the city, which leads to serious conflicts
(Feixue Li et al., 2018). The benefits of spatial compactness in urban among different competing urban land uses, and creates serious envi­
land use are well documented in the literature. These benefits include ronmental problems, including fragmentation of natural resources,
less emission from transport, promotion of walking and bicycling, short deficiency of land resources, and air pollution (Huang, Li, & Zhang,
travel distance, conservation of countryside rural area, encourage public 2019). It is argued that part of these issues arise due to the lack of
transport, low motorized mobility, equitable access to social infra­ suitability among different land uses within the urban area (Akbari,
structure and services, promote public health, efficient utility services, Neamatollahi, & Neamatollahi, 2019). The significance of land use
and revitalization and regeneration of urban core, etc. (Williams, Bur­ suitability in land use optimization is also justified by other research
ton, & Jenks, 2013). Mouratidis (2019) showed that residents of findings. Bagheri, Sulaiman, & Vaghefi, 2012 urged to evaluate the land
compact-city avail higher level of social interaction, personal satisfac­ suitability in a coastal area of Malaysia to provide a suitable room for the
tion, and perceived physical health benefits; Russo & Cirella, 2018 coastal ecosystem; R. Liu, Zhang, Zhang, & Borthwick, 2014 developed
showed that compact city development contributes to achieving urban an Urban Development Land-use Suitability Mapping (UDLSM)
sustainability. Therefore, maximizing spatial compactness was consid­ approach to mapping urban development land-use suitability in Beijing;
ered in many land use optimization problems. Parry, Ganaie, & Sultan Bhat, 2018 attempted to find suitable locations
Since global land is scarce and limited, city authorities are concerned for urban amenities and other services based on GIS-based land suit­
about making the city compact through high-density development, ability analysis using AHP model.
reducing scattered development, and saving land. It is well-recognized Some other objectives used frequently in urban land use optimiza­
that urban sprawl is an urgent problem in cities, especially in devel­ tion are Maximization of economic benefits (7.74%, n=13), Maximiza­
oping countries. On the other hand, compact cities play an important tion of ecological benefits (7.14%, n=12), Minimization of land
role in eliminating urban sprawl, alleviate pollution and promote social conversion cost (7.14%, n=12), Maximization of land value (4.17%,
justice (Huang & Zhang, 2014; Feixue Li et al., 2018). It is argued that a n=7), Maximization of accessibility (2.38%, n=4), and Maximization of
compact city contributes much to achieve overall sustainable urban Ecosystem Service Value (ESV) (2.38%, n=4). Maximization of ecolog­
development. Many governments have formulated various policies to­ ical benefits and maximization of Ecosystem Service Value (ESV) are
wards achieving sustainable urban development through compact cities. similar to each other. Both the objectives were addressed in many land
Thus, consideration of maximizing spatial compactness in many studies use optimization problems since they provide multiple services which
reflects global concern about sustainable urban land use planning. have significance to the health, well-being, livelihood, and survival of
Fig. 6 indicates that the second important objective used in urban humans. Due to urbanization worldwide, human-ecology interaction has
land use optimization is the maximization of land use compatibility fol­ been a major issue of environmental concern. The impact of land use
lowed by maximization of land use suitability, which accounts for 13.69% changes, resulted from human-ecology interaction and urbanization, on
(n=23) and 11.90% (n=20) of the total objectives, respectively. Land ecosystem services is well documented. For example, Su, Li, Hu, Xiao, &
use compatibility can be defined as the situation in which current land Zhang, 2014 assessed the ESV changes of Shanghai, China, from 1994 to
use or activity can co-exist with neighboring land use or activity without 2006 and found that ESV had decreased significantly due to land use
creating any adverse effect. Land use compatibility has been addressed changes caused by urbanization; Estoque & Murayama, 2013 investi­
in many land use optimization problems due to several reasons. First, gated the changes in ESV of a city of Philippines and noticed that there is
land use compatibility contributes to attaining the highest and best use a substantial decrease in annual ESV due to land use changes caused by
of land. The best use of land helps to create economic vitality, sound population growth and urban expansion.
community, and promote social interaction. Secondly, compatible land Maximization of economic benefits has been undertaken in many
uses reduce the physical, social, and economic conflicts that arise due to land use optimization problems since differences in geographical loca­
incompatible land uses. Third, it has significance in achieving social tions, different land uses have a different economic benefit, and even the
interactions, a pleasant and healthier environment, increasing urban same land use in a different location might have a different economic
livability through human-environment interaction (Kai Cao et al., 2020; benefit. Economic benefit from land use has also been considered from
an urban sustainability point of view. Maximizing total economic benefit
from land use is not enough for urban sustainability; equitable spatial
1
This number represents unique objectives. Some objectives were used in distribution is also necessary to reduce social conflict. Land use opti­
multiple studies, thus a total of 168 objectives including duplication were mization plays an important role in maximizing economic benefit while
identified from 55 Journal articles. ensuring equitable distribution of economic production and social

6
Md.M. Rahman and G. Szabó Sustainable Cities and Society 74 (2021) 103214

Fig. 7. Ten most frequently used constraints to the optimization and corresponding number of instances in the literature.

justice in terms of economic benefits from land use allocation. Various economic benefit (Li & Parrott, 2016), while others used land rent
studies have emphasized this issue of equitable distribution of economic theory to calculate economic benefit (Ma & Zhao, 2015). Some re­
benefits from land use. For example, in Kerala, India, the social conflict searchers considered carbon storage as a proxy variable to measure
has been reduced through equitable distribution of economic resources environmental benefit (Yuan, Liu, He, & Liu, 2014), where some re­
and land use rather than only maximizing economic benefits (Basiago, searchers considered ecological suitability (Cao & Ye, 2013). Similarly,
1998). there is no established method to calculate social benefits from land use.
The objective “minimizing land conversion costs” has been included So, there is a huge research scope to work on developing standard
in many land use optimization problems because it will result in methods to measure social, environmental, and economic benefits from
decreasing the development cost and, in turn, will improve overall so­ land use optimization.
cietal and economic benefits. Accessibility indicates how accessible are
the urban services. It is measured in terms of the distance between the 3.2.2. Constraints
two land uses types under concern; the smaller the distance between two There are three types of constraints that are used in multi-objective
land use types, the more is the accessibility between them. Maximization optimization problems. These constraints are a) inequality constraints,
of accessibility has been considered in land use optimization because it b) equality constraints, and c) lower and upper boundaries constraints.
contributes to urban sustainability, enhances the overall quality of In any optimization problem, decision variables are used to define the
urban life, and improves urban livability. Studies suggest that a higher constraints. As mentioned in Section 1.1, there are four main decision
level of accessibility within the city can contribute to decreasing up to variables in land use optimization problems. In which land use types
80% of the CO2 emission motorized transportation (Cao, Huang, Wang, include residential, commercial, water bodies, agricultural, recrea­
& Lin, 2012). Accessibility issue in urban land use planning has also tional, and green space, etc., land use size means the quantity of area of
been emphasized in many studies, including Duranton and Guerra particular land use to be allocated, land use location defines where the
(2016) and Mora, Gilart-Iglesias, Pérez-Del Hoyo, & Andújar-Montoya, specific land use to be allocated, and capacity is the attribute attached to
2017. the land use or grid cell, e.g., the slope of the land or population density
It may be noted that the maximization of social and environmental in a particular area. From the literature, we have identified about 15
benefits was not frequently used in urban land use optimization prob­ unique constraints, excluding repetition. The ten (10) constraints are
lems. One reason for this may be many researchers argued that social presented in Fig. 7.
benefits could be collectively achieved through compatibility, From Fig. 7, it is evident that the constraint “one and only one land
compactness, and accessibility, etc. (Cao, Huang, Wang, & Lin, 2012; Y. use in each cell” was used more frequently (n=53) in urban land use
Liu, Zhang, Zhang, & Borthwick, 2014; Zhang et al., 2016), and envi­ optimization problem followed by “minimum and maximum area of
ronmental benefits can be achieved through ecological and ecosystem certain land use” (n=25) and “restriction on specific land use change”
service value, etc. (Feixue Li et al., 2018; Zhang et al., 2016). Therefore, (n=16). The first constraint indicates that only one land use type can be
consideration of environmental benefits was not directly visible in the allocated in one land unit. But the practical scenario is seen to be
urban land use optimization problem. different. In many cities, there are buildings with multiple uses on
In connection with urban sustainability dimensions, our findings different floors of the same building. Due to computational complexities,
indicate that less attention was drawn to incorporate sustainability di­ most of the authors did not consider multiple uses in the same building
mensions in urban land use optimization; only two studies focused on while optimizing land use allocation. We have identified only two
urban sustainability (Cao, Huang, Wang, & Lin, 2012; Yuan, Liu, He, & studies by Haque & Asami, 2014 and Sharmin, Haque, & Islam, 2019 in
Liu, 2014). Overall, environmental (including ecology) and economic which multiple uses in the same building have been allowed for opti­
aspects were included in 46.67% and 43.33% of studies, respectively, mizing land use allocation. The second most frequently used constraint
but the social aspect (10%, n= 3) was mostly ignored. Our findings also is the minimum and maximum area of certain land use. The minimum
suggest that there is a lack of using common proxy variables and stan­ and maximum limits of specific land use are defined to ensure balanced
dard methods to evaluate economic, environmental, and social benefits development and specific requirements. This range of limits totally de­
from land use optimization. Some researchers used GDP to measure pends on the high-level government, stakeholder’s opinion, and land use

7
Md.M. Rahman and G. Szabó Sustainable Cities and Society 74 (2021) 103214

Fig. 8. Approaches to the construction of Urban land use optimization.

policy of the respective city. For example, the land use policy of China
requires that the minimum area of arable land should not be less than 10
× 109 mu to ensure food security (Y. Liu, Zhang, Zhang, & Borthwick,
2014). Restriction of specific land use changes has also been used as a
constraint in many studies to preserve certain land use categories. This
restriction depends on the city’s land use policy, need, and availability Fig. 9. Illustration of Pareto optimal front and solutions (Yang, Zhu, Shao, &
Chi, 2018).
of specific land use types. For example, local regulations and national
laws of Wuhan, China requires that a) in preservation areas, agricultural
land cannot be converted to urban land, b) water bodies should be kept about 42.86% of studies used the Pareto front-based (n=21) method to
intact, c) urban land cannot be converted to agricultural land and d) land construct urban land use optimization problems, followed by the
with a slope less than 25◦ cannot be converted to arable land (Yuan, Liu, Weighted sum method (36.73%, n=18) and Goal programming
He, & Liu, 2014). Constant total land use has been used as a constraint in (20.41%, n=10). These methods are shortly described in the following
some studies to confirm that total land area cannot be increased or sections.
decreased while optimizing land use allocation. This constraint has
implications regarding the urbanization and expansion of cities. This 3.2.3.1. Pareto front based method. The Pareto approach attempts to
constraint strictly prohibits cities from physical expansion and conver­ find a set of non-dominated solutions. According to this concept, a so­
sion of agricultural land around the cities. Authors considered this lution is said to be Pareto optimal when no further improvement is
constraint in cities where there is no need or scope for further expansion possible in any objective function without degrading at least one of the
of cities. For example, Zhang et al. (2016) used this constraint in their other objective function(s) (Deb, 2011). Fig. 9 illustrates that the solu­
studies conducted in Changsha city of China to ensure that the city will tions on the Pareto front line exhibit different combinations of tradeoffs
not be allowed for further expansion. Allowable change limit of land use among the multiple conflict objectives.
has been defined in some studies to allow increase or decrease of existing Our findings show that many studies have applied the Pareto front-
land uses up to a certain limit for land use. This constraint has impli­ based method to solve land use optimization problems because this
cations in optimizing land use allocation. It is a general tendency for the method can satisfy the demand of multiple stakeholders while keeping
quantity of land uses which produce higher economic benefits to be everyone’s objective optimized. Despite having many positive sides,
increased. For example, in general, commercial land produces higher Pareto-optimality may show inefficient tradeoffs in practical problems,
economic benefits, whereas urban greenspace generates lower economic and even sometimes, it may be impossible to find the Pareto front if the
benefits. Under this situation, if no range is defined, the optimization optimization problem is associated with a large number of decision
system will try to increase commercial land use and decrease urban variables. In the face of this problem, a heuristic algorithm such as
green space as much as possible. But this will not be a feasible solution. particle swarm optimization and genetic algorithms may be efficient to
So, the allowable change limit is defined to control the maximum and find the Pareto front.
minimum area of land use. For example, Sharmin, Haque, & Islam, 2019
set 25% allowable land use change limit in their study to generate 3.2.3.2. Weighted sum method. According to the weighted sum method,
alternative land use allocation for a small mixed-use area of Dhaka city all objective functions are combined into one scalar objective function
in Bangladesh. In some studies, the number of total populations has been assigning specific weight to each objective. Then this scalar optimization
kept constant for the efficient management of the city and to reduce the problem is solved to find the non-dominated solutions. This method can
burden on existing utilities. Because if the number of populations would be expressed by Eq. (1).
have been allowed to increase, it would create a burden on the utilities,
∑
k
transport and other urban services leading to mismanagement of city. min wi fi (x) (1)
Slope range was also used in some studies as constraint to specify the i=1

land use types in specific land in terms of slope. For example, due to the ∑k
Subject to x ∈ F, i=1 wi = 1, where wi ≥ 0
local and national regulations, it was decided that no land of Wuhan,
Nonnegative weights are assigned to each objective function. Since
China can be converted to arable land if it has slope less than 25o (Yuan,
the weighted sum method can generate effective and efficient solutions,
Liu, He, & Liu, 2014).
this method has been applied in many urban land use optimization
problems, including Zhang and Wang (2019), Song and D. M. Chen
3.2.3. Construction of optimization problem
(2018b), Feixue Li et al. (2018), Handayanto, Tripathi, Kim, & Guha,
Construction methods of multi-objective optimization problems are
2017, and Yang, Sun, Peng, Shao, & Chi, 2015. Although the weighted
classified into two categories: Scalarization and Pareto front-based
sum approach is very popular and simple, it has some inherent prob­
method (Cao, Zhang, & Wang, 2019). The most common methods
lems. First, the outcome of the final solution largely depends upon the
under the scalarization technique are a) the Weighted sum method, b)
selection of weights given to each objective function. So, careful selec­
Goal programming, and c) ε-constraint method. We have identified
tion weight is necessary; otherwise, the outcome may not end up with a
three types of construction methods that are used in urban land use
desirable result. Second, in a weighted-sum approach, multiple
optimization and have been presented in Fig. 8. The Fig. shows that

8
Md.M. Rahman and G. Szabó Sustainable Cities and Society 74 (2021) 103214

Fig. 10. Methods to solve the urban land use optimization problem.

objectives are converted to a single objective, and, in most cases, a solved by this method (Copado-Méndez, Pozo, Guillén-Gosálbez, &
single-objective algorithm finds a solution satisfying the first-order Jiménez, 2016).
optimality criterion; the additional test may be required to obtain the
optimal solution (Deb, 2011). Third, there are many multi-objective 3.2.4. Methods to solve the optimization problem
optimization problems that are non-convex, but the weighted sum Different methods are used to solve land use optimization problems.
method does not work with non-convex problems. Fourth, sometimes The decision variables play an important role in determining the solving
multiple solutions generated by the Pareto front may be weakly domi­ approach to multi-objective optimization problems. If the decision
nated by each other; in this case, the weighted-sum approach may fail to variables are discrete, then an exact approach (e.g., Linear Program­
identify some true Pareto front solutions. ming, Integer Programming) is preferred; if the decision variables are
continuous, then the heuristic approach is the most preferred approach.
3.2.3.3. Goal programming. Goal programming is considered as an If the optimization problem contains both discrete and continuous de­
extension of linear programming where the goal for each objective cision variables, mixed-integer methods are used to solve the problem
function is set to be achieved. The commonly used formula to formulate (Panwar, Jha, & Srivastava, 2018; Pappas et al., 2021). In the case of the
the multi-objective optimization problem using goal programming with exact approach, linear programming can only be used if the decision
reference points can be expressed as follows (Li & Parrott, 2016): variables, objective functions, and constraints are all linear (Wang &
Fang, 2018). The graphical method can be used to solve linear pro­
∑Q [ ]p
MinF =
fq (x) − Iq
(2) gramming problems if there are only two variables. If any one of them is
q=1
λq − Iq not linear, then linear programming cannot be used to solve the opti­
mization problem. In our case, there are only three studies in which the
Where F is the overall objective function, Q is the number of objec­ decision variables, objective function, and constraints are linear. These
tives, q is the index of objectives, fq (x) is the objective function with studies used the linear programming method to solve land use optimi­
index q, Iq is the possible ideal value of the objective fq (x), λq is the worst zation problems. In other cases, the decision variables are non-linear,
value of the objective fq (x), p is a penalty coefficient for the objective and the objective functions are continuous, so heuristic approaches
violation increase, value for p is 2 to be considered to be appropriate were used to solve the optimization problem. Fig. 10 describes the ten
(Stewart, 1991). The goal programming approach has been applied in (10) most frequently used optimization techniques in urban land use
many optimization problems, including urban land use optimization optimization problems. Our findings suggest that, the most popular
problems because this method is very suitable to produce technique is the Genetic Algorithm (GA) (n=18), followed by Non­
non-dominated solutions considering stakeholder preferences. dominated Sorting Genetic Algorithm II (NSGA-II) (n=12), Particle
Swarm Optimization (PSO)(n=7), and Simulated Annealing (SA)(n=5).
3.2.3.4. ε-constraint method. Besides the weighted sum method, the We have identified two categories of solving strategies towards land
ε-constraint method is considered as the best approach to solve multi- use optimization problems: exact and heuristic methods. Exact methods
objective optimization problems. In this approach, having p objective evaluate all feasible solutions and find a set of unique non-dominated
functions, only one objective is kept as it is, and other p-1 objectives are solutions. The best solution(s) are guaranteed by exact methods (Lig­
transformed into constraints. This is known as the ε-constraint problem. mann-Zielinska, Church, & Jankowski, 2008; Tong & Murray, 2012).
Then the ε-constraint problems are integrated with heuristics algorithms Some popular exact methods are Enumeration, Linear Programming,
to solve the optimization problems (Chiandussi, Codegone, Ferrero, & Integer Programming with Branch & Bound method, and Dynamic
Varesio, 2012). Thus, the problem: Programming (Rath & Gutjahr, 2014). As shown in Fig. 10, only linear
min
( ) programming was used in urban land use optimization problems (n=3).
xϵX f1 (x), ……..fp (x)
Linear programming technique optimizes linear objective function
Can be substituted by the ε-constraint problem as follows: subject to some linear equality and inequality constraints. Fig. 10 de­
min
picts that the exact approach towards urban land use optimization is not
fj (x)
xϵX common practice by the researchers. The limited use of exact methods in
Where, fk (x) ≤ εk k = 1, ………p, k ∕=j solving spatial optimization problems is primarily due to the location
Although we did not identify the application of the ε-constraint component, which makes land use optimization different from other
method in our cases, nevertheless, many optimization problems were optimization problems. Land use optimization is not a simple allocation

9
Md.M. Rahman and G. Szabó Sustainable Cities and Society 74 (2021) 103214

Table 2 Modelling land use optimization problems is also complex due to

Reason behind using different optimization methods by different authors. geographic and topological relationships and properties of land parcels
Name of Methods The reason behind using the methods such as distance, adjacency, connectivity, containment, and shape (Tong
& Murray, 2012). Adjacency refers to what extent two land units are
Genetic Algorithm (GA) Robustness and efficiency (Haque & Asami, 2014,
Haque & Asami, 2011; Huang & Zhang, 2014; Li neighbors, whereas connectivity indicates the ability to reach one
& Parrott, 2016; Porta et al., 2013; Mingjie Song location from another location in terms of impedance. Containment
and Chen, 2018b; Zhang, Wang, Cao, He, & Shan, signifies whether one land unit is completely within another land unit.
2019); ability to search in complex, large and We have identified several reasons why exact methods are seen as
poorly-understood search spaces (Huang &
Zhang, 2014; Li & Parrott, 2016; Zhang, Wang,
limited in case of land use optimization problems. Many researchers
Cao, He, & Shan, 2019); Natural selection argued that exact methods are seen to be efficient in case of single
towards global optimization (Cao & Ye, 2013; objective or multiple objectives having no conflicts (Mingjie Song and
Gong, Liu, & Chen, 2012; Li & Ma, 2018; Y. Liu, Chen, 2018b). But in the real world, land use optimization problems
Zhang, Zhang, & Borthwick, 2014; Zhang, Wang,
consist of multiple objectives having conflicting interests among the
Cao, He, & Shan, 2019; Zhang & Huang, 2015)
Nondominated Sorting Genetic Derivation of Non-dominated solutions ( stakeholders. Another drawback of the exact method is that they pro­
Algorithm II (NSGA-II) Karakostas, 2017; Lubida, Veysipanah, Pilesjo, & duce only a single optimum solution. But in most cases, the single best
Mansourian, 2019; Mohammadi, Nastaran, & solution may not be the feasible solution for the decision-maker because
Sahebgharani, 2015; Schwaab et al., 2018; multiple stakeholders with conflicting interests are involved in land use
Sharmin, Haque, & Islam, 2019; Shaygan,
Alimohammadi, Mansourian, Govara, & Kalami,
decisions (Li & Ma, 2018; Mingjie Song and Chen, 2018a). The exact
2014; Mingjie Song and Chen, 2018a, 2018b); method shows good performance when optimization problems are
intelligent ranking of the Pareto solutions ( simple. But land use optimization problems are complex due to it’s
Lubida, Veysipanah, Pilesjo, & Mansourian, spatial component. So exact methods are infeasible in solving land use
2019; Sharmin, Haque, & Islam, 2019; Mingjie
optimization problems. Exact methods are suitable when there is a
Song and Chen, 2018b); Diversity preservation in
the solutions (Kai Cao et al., 2011; Karakostas, limited number of land units to be allocated. But if the number of land
2015; Lubida, Veysipanah, Pilesjo, & units is large, then it becomes very difficult for an exact method to
Mansourian, 2019; Wang, Zhang, & Wang, generate a feasible solution. For example, an exact method like integer
2019); programming become infeasible if there are more than 2000 or 3000
Particle Swarm Optimization Less computational time (Handayanto, Tripathi,
land units need to be allocated (Stewart, Janssen, & Van Herwijnen,
(PSO) Kim, & Guha, 2017); To ease Continuous
optimization problem (Handayanto, Tripathi, 2004). Although exact methods explicitly guarantee an optimal solution
Kim, & Guha, 2017; Mingjie Song and Chen, but can be much slower when there exist multiple conflicting objectives.
2018b); Rapid convergence (Feixue Li et al., Besides, exact methods find it very difficult to solve continuous opti­
2018; Li & Parrott, 2016; Yaolin Liu, Li, Shi,
mization problems since the feasible solution region is non-convex
Huang, & Liu, 2012; Sahebgharani, 2016);
flexibility and simplicity of operators (Feixue Li (Aerts, Eisinger, Heuvelink, & Stewart, 2003). Due to the above rea­
et al., 2018; Yaolin Liu, Li, Shi, Huang, & Liu, sons, exact methods seem to be limited in land use optimization
2012; Yang, Sun, Peng, Shao, & Chi, 2015); problems.
improved adaptability of search-space (Feixue Li In contrast to exact methods, the heuristic approach utilizes the rule
et al., 2018; Yaolin Liu, Li, Shi, Huang, & Liu,
of thumb or best practice to produce a set of near-optimal solutions. Our
2012);
Simulated annealing (SA) Rapid convergence (Mingjie Song and Chen, findings illustrate that the most frequently used heuristic to solve urban
2018b); Improved performance (Mingjie Song land use optimization problems are GA (32.14%), NSGA-II (21.42%),
and Chen, 2018b); can control solution PSO (12.50%), and SA (8.93%). In GA, a set of candidate solutions or
acceptance probability (Santé-Riveira,
populations are generated randomly in an iterative process. The set of
Boullón-Magán, Crecente-Maseda, &
Miranda-Barrós, 2008); can derive the ideal
populations in each iteration is called generation; the fitness value of the
alternative solution (Caparros-Midwood, Barr, & populations is evaluated in each generation, and the best-fitted values
Dawson, 2015; Li & Ma, 2018); to ease are selected for the next generation, and this process continues until the
computational complexity (Yl Liu, Li, Shi, Huang, best solutions are found. Detailed procedures of GA can be found in
& Liu, 2012)
many kinds of literature (Nandi & Deb, 2016). Table 2 summarizes the
Linear Programming guarantee an optimal solution (Yl Liu, Li, Shi,
Huang, & Liu, 2012) reasons behind using different optimization methods by respective au­
Ant Colony Optimization (ACO) Better efficiency and effectiveness (Li & Parrott, thors. We have identified three main reasons for which GA was used in
2016); minimum search path cost (X. Liu, Li, Shi, so many land use optimization problems. First, GA is robust and efficient
Huang, & Liu, 2012)
in handling varieties of multi-objective optimization problems to find
Multi-agent System (MAS) To integrate local simulation to global
optimization (Yuan, Liu, He, & Liu, 2014; Zhang
the global optimum. Second, GA has the ability to search in complex,
et al., 2016) large, and poorly understood search spaces to find the fittest solution.
BFGA Efficient to solve large problems (Cao, Huang, Third, it follows a natural selection procedure towards global optimi­
Wang, & Lin, 2012; Kai Cao et al., 2020) zation. This is an important criterion for choosing GA in many studies
AIS Specially developed for multi-objective
because land use allocation is an outcome of many competing factors
optimization problem (Ma & Zhao, 2015); better
performance in spatial data (Huang, Liu, Li, and interests. Despite the many advantages of GA, it has two challenges
Liang, & He, 2013) in land use optimization problems. Firstly, land use optimization is a
Artificial Bee Colony (ABC) low computational cost (Yang, Sun, Peng, Shao, & spatial optimization problem, and the land use is represented by a raster
Chi, 2015); to increase genetic diversity (Yang, cell where the decision variable is represented by each cell, making
Zhu, Shao, & Chi, 2018)
millions of decision variables in the whole optimization process. It
makes very difficult to find the optimum solution. Secondly, genes used
regarding how much land units would be allocated to each category; in the land use optimization process are non-computable because these
rather, it also entails where to allocate. Due to this property of genes are character and only used to mark the land use type. So, a
location-allocation, land use optimization becomes more complex common type of crossover is not applicable in this case, although
compared to other optimization problems. This complexity disfavor land crossover is the main operator in the selection process. Therefore, this is
use optimization problems to be solved by exact methods. a challenging task to design crossover in traditional GA to optimize land
use.

10
Md.M. Rahman and G. Szabó Sustainable Cities and Society 74 (2021) 103214

Fig. 11. Common Spatial data used in urban land use optimization

The second most frequently used method to solve urban land use use planning. Preparation of a good land use plan requires as much data
optimization is NSGA-II. In contrast to GA, NSGA-II follows an elitist and information (Rahman & Szabó, 2020). Most importantly, these data
principle to select the best fittest population. A set of non-dominated include transportation and road network, travel behavior and pattern,
solutions are selected in such a way that no further improvement is physical feature, population density, socio-economic condition, soil
possible without losing the value of any objective. As it is seen from characteristics, land cover, water, and environment, etc. The availability
Table 2, there are several reasons for using NSGA-II by many authors. of spatial data, data structure, accuracy, level of detail greatly affects the
Derivation of non-dominated solutions is the most important among solution strategies and final output.
them. Table 2 also suggests that NSGA-II can preserve the diversity We have identified fifteen (15) types of major spatial data used in the
among solutions and can intelligently rank the non-dominated solutions urban land use optimization process. The spatial data includes the sta­
to ease the decision-maker to select the best option. Successful appli­ tistical data, which were converted to spatial data adding the corre­
cation of NSGA-II in urban land use optimization is frequent and can be sponding location component. The spatial data used in the urban land
found in many kinds of literature (Gao et al., 2020; Mohammadi, Nas­ use optimization process have been summarized in Fig. 11. The spatial
taran, & Sahebgharani, 2015; Shaygan, Alimohammadi, Mansourian, data has been represented as point, line, and area features. Needless to
Govara, & Kalami, 2014). However, NSGA-II also has some limitations. say, the land use data is the prime data that was used in every study.
Sometimes NSGA-II fails to find well-diversified non-dominated solu­ Many studies have classified land use data differently. Different city
tions because it may lose its selection pressure while evaluating fitness authorities and researchers classified land use differently, considering
function. This happens due to the generation of the higher portion of the study purposes. For example, Cao, Zhang, & Wang, 2019 used six cat­
initial population from non-dominated solutions (He & Yen, 2016). In egories of land use, including agriculture, forest, grassland, water
addition, if there are higher distances among parent solutions, they are bodies, built-up land, and unused land, whereas Handayanto, Tripathi,
likely to produce offspring far beyond the true solutions. Kim, & Guha, 2017 used ten (10) types of land use including commercial
PSO has been applied in many studies to solve land use optimization buildings, industrial buildings, elementary schools, middle schools,
problems (Feixue Li et al., 2018; Masoomi, Mesgari, & Hamrah, 2013). colleges, medical facilities, sports areas, parks, low-density residential
Table 2 indicates the reasons behind the use of PSO in many studies. PSO areas, and high-density residential areas. To prepare land use map,
can perform complex problem with less computational cost and time. mainly satellite images and existing secondary sources were used in
Some authors used PSO due to its rapid convergence and improved many studies. Population data was also frequently used in urban land
adaptability of search space. PSO has the advantage due to its flexibility use optimization. Two applications of population data were mostly
and simplicity of its operators. Although PSO showed a better perfor­ observed in land use optimization: demand calculation and maximum or
mance compared to GA, it results in lower accuracy. If a small initial minimum limit for which any objective to be optimized. In most of the
weight is used, this problem may be solved, but it may also cause to cases, Population data was seen to be measured as population density.
lower particle’s searching distance. SA is another popular heuristic al­ Another important spatial data that was used frequently is the road
gorithm for land use optimization. When it becomes important to find network. The road network data was mainly used to measure spatial
approximate global optima compared to precise local optima, SA is a access to different urban facilities. Gross domestic product (GDP) and
preferable alternative to other search techniques (Goffe, Ferrier, & land value were also used in many land use optimization studies, which
Rogers, 1994; Suppapitnarm, Seffen, Parks, & Clarkson, 2000). Table 2 focus on the maximization of economic benefits and landowner’s gain.
indicates that the most important characteristic of SA for which it has GDP was presented as per unit of area or block. Digital Elevation Model
been used by many authors is its ability to control the acceptance (DEM) and Slope were mainly used to calculate land suitability and
probability of the solutions. In SA, each solution is attributed to tem­ agricultural productivity (Feixue Li et al., 2018; W Zheng, Ke, Xiao, &
perature. The temperature between the two nearest solutions decides Zhou, 2019). Ecosystem Service Value (ESV), Climate and Weather data
the acceptance probability. This strategy helps to find the fittest solu­ were used to measure the environmental benefits (Zhang et al., 2016,
tions from a large search space. Eventually, due to this property, SA 2019; Weiwei Zheng, Ke, Xiao, & Zhou, 2019).
eases the computational complexity of problems having multiple ob­
jectives and large search space. 3.2.5.2. Spatial data model. The selection of spatial data model, either
vector or raster, can affect the construction of the overall model and
3.2.5. Spatial Data in urban land use optimization results. Fig. 12 shows the type of popular spatial data models in spatial
optimization. Fig. 12 shows that about 80% (n=44) of studies used the
3.2.5.1. Spatial data type. Spatial data is an essential element in land raster data model while only six studies used the vector model.

11
Md.M. Rahman and G. Szabó Sustainable Cities and Society 74 (2021) 103214

example, environmental benefits from land uses have been calculated by

using carbon storage as a proxy variable (Yuan, Liu, He, & Liu, 2014),
valuing natural resources and ecology (Cao, Huang, Wang, & Lin, 2012),
quantifying ecological suitability (Cao & Ye, 2013), etc. So, there is a
clear research gap in using proxy variables and framing a common
approach towards evaluating sustainability dimensions in relation to
land use optimization. Third, from our reviewed articles (n=55), we did
not find any established method to calculate social benefits. Researchers
used the different concepts to measure social benefits without similarity.
For example, Zhang et al. (2016) used social security service value as the
indicator of social benefits, Yuan, Liu, He, & Liu, 2014 used spatial
compactness, and Cao, Huang, Wang, & Lin, 2012 used accessibility,
compatibility, and compactness as indicators of social benefits. This gap
Fig. 12. Spatial data model used to design urban land use optimiza­ may be fulfilled by conducting research on developing methods to
tion problem measure social benefits from land use. Fourth, most of the studies have a
constraint that “one and only land use” should be allocated in each plot,
However, in the two studies, it was not clearly mentioned the data model which might not be practical in many cities having mixed uses in the
used. same building. We have identified only two studies (Haque & Asami,
There are advantages and disadvantages both in raster and vector 2014; Sharmin, Haque, & Islam, 2019) that allow multiple land uses
data structure in the land use optimization model. Land use optimization (more than one use in one building) in the same plot. This necessitates to
problems can be easily formed using raster-based representation. The development of efficient strategies to incorporate multiple land uses in
most revealed advantage of a raster data model is that the land uses can the building while developing an optimization model. Fifth, based on
be easily encoded, and the representation of spatial relationships is very our understanding from the reviewed papers, we have identified that
straightforward. Raster-based model is preferable and can be found in urban land use optimization has been constructed mathematically. This
many studies, including Gao et al. (2020), Zhang, Wang, Cao, He, & approach results in a quantitative allocation of land use considering
Shan, 2019, Yang, Zhu, Shao, & Chi, 2018, Li & Ma, 2018, and Feixue Li multiple conflicting objectives and do not integrate qualitative value as
et al. (2018). However, the raster model sometimes may become desired by the local public. For example, the value of conservation of
impractical because it may lead to multiple land use categories allocated wetland and green space may not only be evaluated quantitatively. At
to a single land category and a single land use plot to multiple categories. the same time, land use decision is strongly influenced by stakeholders
In addition, raster-based representation requires more units and space opinions, but hardly any method can incorporate stakeholders opinion
compared to vector-based representation for a similar feature. Land use to construct optimization problem. Owing to this limitation, only
optimization problems using raster data also take unrealistically higher quantitative optimization is insufficient to meet the public agreement.
computation time. This makes the clear need for developing an approach to incorporate
In contrast to the raster-based model, the vector-based model is more stakeholders’ opinions integrated with the mathematical optimization
intuitive and matches real-world land use planning. Since, in the vector process. This knowledge gap may be filled up by doing further research
data model, spatial features are represented by coordinates through on coupling public participation with land use optimization.
points, lines, and polygons; real-world spatial entities are best repre­
sented with less deviation. Thus, the accuracy and computational effi­ 5. Recommendations
ciency can be improved by using the vector data model in land use
optimization. As compared to the raster data model, in vector format, Based on the findings from our study, we are proposing the following
the number of spatial units can also be decreased. The use of vector- recommendations to be incorporated urban land use optimization. We
based urban land use optimization can be found in Cao et al. (2020), strongly argue that these new ideas will improve the existing approach
and Handayanto, Tripathi, Kim, & Guha, 2017. In the vector-based to land use optimization.
model, there are also some problems. For instance, the spatial rela­ Selection of objectives: In the past decades, land use optimization
tionship is somewhat difficult and sophisticated. Another problem with has been investigated towards achieving various objectives (Cao, Zhang,
using vector format is related to spatial units. In a vector model, a large & Wang, 2019; Haque & Asami, 2011; Karakostas, 2015), and these
area may not be subdivided into smaller spatial units since it will create objectives are diverse and contextual. Fig. 6 indicates that some objec­
computational complexities. tives show clear importance over others in land use optimization prob­
lems. Maximization of spatial compactness and maximization of land
4. Knowledge gaps and future scope of research use compatibility have great significance in land use planning and urban
sustainability. So, we recommend considering these two objectives with
In this study, we have systematically reviewed 55 articles that other location-specific objectives in future land use optimization. Our
concentrate on urban land use optimization. This provides plausible findings suggest that sustainability dimensions have been poorly
ground to identify the knowledge gaps in urban land use optimization. addressed in urban land use optimization problems. Hence, we recom­
Firstly, although there exists diversification in optimization objectives, mend that objective functions should be framed considering three core
sustainability was not the main focus in urban land use optimization, pillars of sustainability. There may be many sub-objectives within core
although sustainable land use is an important consideration in urban sustainability dimensions. For example, optimizing urban green space
land use planning. We have identified only two studies (Cao, Huang, (Kai Cao et al., 2020), minimizing pollution (Huang & Zhang, 2014),
Wang, & Lin, 2012; Yuan, Liu, He, & Liu, 2014) that focused on sus­ optimizing ecosystem service value (Weiwei Zheng, Ke, Xiao, & Zhou,
tainability, including all three aspects; other studies considered only one 2019) fall within the objective function of the environmental domain.
or two aspects of sustainability. It may be noted that only 10% (n=3) of Even, there may be conflicting objectives as discussed in section 1. So,
studies included social aspects of sustainability. To fill this knowledge before the selection of sub-objectives stakeholder consultation work­
gap, future research could focus on sustainable urban land use. Sec­ shop should be arranged to understand the trade-offs among different
ondly, although sustainability dimensions have been considered objectives.
partially in many studies, there is a lack of consensus in using proxy Calculating value (benefit) of objective functions: Our discussion in
variables to calculate the benefits of sustainability dimensions. For section 3.2.1 illustrates that there is no common agreement to calculate

12
Md.M. Rahman and G. Szabó Sustainable Cities and Society 74 (2021) 103214

the economic, environmental, and social benefits from land use opti­ be used if rapid convergence is a prime factor and if the optimizer in­
mization. This creates a definite problem to measure the degree to which tends to control the solution acceptance probability. However, we have
a particular city has attained or to be attained the progress towards identified that GA and NSGA-II perform well in spatial objectives
sustainability. So, more work is needed to define proxy variables and to (Annepu, Subbaiah, & Kandukuri, 2012; K Cao et al., 2020, ), whereas
develop standard methods to measure social, environmental, and eco­ SA and PSO perform well in additive objectives (F Li et al., 2018; Li &
nomic benefits from land use. Based on our findings and understanding, Ma, 2018; Mingjie Song and Chen, 2018b). Based on these findings, we
we recommend that land use compatibility, land use suitability, spatial recommend that GA and NSGA-II may be used where the number of
compactness, and spatial access to basic services and urban facilities spatial objectives is greater than additive objectives, and in alternative
could be a measure of social benefits; four groups of ecosystem services cases, SA and PSO can be used preferably. If the number of spatial and
(e.g., provisioning services, supporting services, regulating services and additive objectives are equal, then both groups can be tested to select the
cultural services) could be a measure of environmental benefits; and algorithm that produces a good result.
GDP and land development cost could be the measure of economic Using spatial data model: Although the raster data model is most
benefits from land use. Our justification for proposing the above mea­ frequent, there is a limited discussion to define the neighborhood within
sures of social benefit is that social benefit from land use indicates equal the raster data model. Definition of neighborhood plays an important
access to basic services and urban facilities, where land use compati­ role in calculating the objective function (Yuan, Liu, He, & Liu, 2014).
bility, suitability, spatial compactness, and spatial access are the Based on the literature, we have identified two types of neighborhoods:
important measures to social equity (Bryan et al., 2015). a) Moore neighborhood and b) Von Neumann neighborhood as shown
Defining constraints: Existing urban development policy, stake­ inFig. 13.
holder groups, government regulations, and future expectations affect In a two-dimensional raster data model, the Moore neighborhood is
the selection of constraints (Niu et al., 2013; Shi, Zhan, Yuan, Wu, & Li, defined as a central cell surrounded by the neighboring eight cells,
2015; M Song and Chen, 2018; Xia, Liu, Liu, He, & Hong, 2014). For where the Von Neumann neighborhood is composed of a central grid cell
example, K Cao et al. (2020) considered green space as a conservation surrounded by four adjacent cells (Karakostas, 2017, 2016). Moore
area in their study and set residential area to be 50% of the total area. neighborhood takes higher computational time but produces a more
Similarly, conservation of wetland or minimum green space as con­ accurate result. On the other hand, using the Von Neumann neighbor­
straints could be different in different cities. Some cities may impose hood is straightforward but produces a less accurate result in some cases
conversion of agricultural land (Deelstra & Girardet, 2000) while others (Karakostas, 2015). We recommend that in the land use optimization
may not; some cities may want to increase residential area (Cook, Hall, problem, the Moore neighborhood should be used because, in the Moore
& Larson, 2012) while others may want to increase commercial area. So, neighborhood, the central cell is influenced by adjacent eight cells
the selection of constraints is completely area-specific. Based on the compared to four cells in the Von Neumann neighborhood. So, the use of
findings, our suggestion is to study government regulation, urban the Moore neighborhood will produce a more rational value of the
development policy and to conduct stakeholder workshop to select objective function.
appropriate constraints. We have also identified that most of the studies
considered only single use in one land unit due to computational com­ 6. Limitations and implications of this research
plexities. But the practical scenario is different; in the same building,
there may be different uses on different floors. But this issue was not Our study has added fruitful insights into urban land use optimiza­
considered. Hence, we recommend considering multiple uses in the tion. Nevertheless, it has some limitations. Our review is based on the
same building in future studies. peer-reviewed articles published in the English language only. It can be
Constructing optimization problem: Our discussion in section 3.2.3 mentioned that most of the studies were conducted in China, but we did
suggests that each method to construct an optimization problem has its not include any article published in the Chinese language. This is likely
advantages and disadvantages. For example, although Pareto-optimality that articles published in other languages might contain valuable in­
may show inefficient trade-offs in the case of a large number of decision formation. Another limitation is that we have only included articles that
variables (Mingjie Song and Chen, 2018a). Another problem with the focused on urban land use optimization; surely, there are many articles
Pareto front may be the generation of an infinite number of solutions that focused on rural land use optimization, agricultural land use opti­
from which true solutions need to be discovered(Gao et al., 2020). To mization, which might be a valuable source of insights. We recommend
resolve the problem, mathematical optimization techniques having local that future studies can be conducted, including those variations, to un­
convergence properties may be integrated with evolutionary derstand the differences. We also did not compare the performance of
multi-objective optimization. On the other hand, the weighted sum different methods because for performance evaluation; it is required to
approach largely depends upon the selection of weight (Shaygan, Ali­ apply different methods on the same study area with the same dataset,
mohammadi, Mansourian, Govara, & Kalami, 2014). Wrong selection of which requires a separate study. We have just mentioned different
weight of objective function may result in impractical solutions. To methods relating to land use optimization (e.g., how to calculate
remove this problem, the selection of weight of objective function different objectives, how to construct the problem, and solution
through stakeholder consultation and public participation. After methods) along with some brief discussion. Elaboration of those
resolving the above two problems, we believe that a combination of methods might be helpful for the readers to understand the detailed
Pareto-optimality and weighted may improve the identification of true procedure, but we did not elaborate on those methods because this was
solutions. not the scope of our study.
Solving optimization problem: As discussed in section 3.2.4, some Despite the limitations above, this study has some academic and
algorithms may work better in a certain setting, but the same algorithm empirical implications. To the best of our knowledge, this review is the
may show poor performance in another setting (Yaolin Liu, Li, Shi, first of its kind considering the most important aspects of urban land use
Huang, & Liu, 2012). So, it is not possible to tell which algorithm will optimization. This paper offers some recommendations on selecting
produce the best result without investigating the problem setting under objective functions, calculating the value of objective functions, defining
consideration. Based the table 2 and analysis of the findings, we constraints, constructing and solving the optimization problem, and
recommend that GA can be used if the search space is large, complex, spatial data model. The future researcher may find indicative guidance
and is poorly-understood search spaces; NSGA-II can be used in a situ­ from those suggestions. This will also help policymakers to decide
ation where multiple non-dominated solutions are required by realistic objective functions to address urban sustainability.
decision-makers rather than single best solution; PSO can be used if
computational time is considered as important selection factor; SA can

13
Md.M. Rahman and G. Szabó Sustainable Cities and Society 74 (2021) 103214

Acknowledgment

This research work is an output of Doctoral study at the Department

of Photogrammetry and Geoinformatics, Budapest University of Tech­
nology and Economics. The authors thank the Ministry of Foreign Affairs
and Trade and the Tempus Public Foundation of the Hungarian Gov­
ernment for providing the opportunity to study Ph.D. with the Stipen­
dium Hungaricum Scholarship.

Reference

Aerts, J. C. J. H., Eisinger, E., Heuvelink, G. B. M., & Stewart, T. J. (2003). Using linear
integer programming for multi-site land-use allocation. Geographical Analysis, 35,
148–169. https://doi.org/10.1111/j.1538-4632.2003.tb01106.x
Akbari, M., Neamatollahi, E., & Neamatollahi, P. (2019). Evaluating land suitability for
Fig. 13. (a) Von Neumann neighborhood. (b) Moore neighborhood. spatial planning in arid regions of eastern Iran using fuzzy logic and multi-criteria
analysis. Ecological Indicators, 98, 587–598. https://doi.org/10.1016/j.
ecolind.2018.11.035
7. Conclusion Annepu, G., Subbaiah, K. V, & Kandukuri, N. R. (2012). Genetic algorithm approach to a
multiobjective land allocation model: A case study. International Journal of
Agricultural and Environmental Information Systems, 3, 86–99. https://doi.org/
Based on our findings, this study concludes that the most common 10.4018/jaeis.2012070106
objectives in urban land use optimization are maximization of spatial Armsworth, P. R., Benefield, A. E., Dilkina, B., Fovargue, R., Jackson, H. B., Le
compactness, maximization of land use compatibility followed by Bouille, D., & Nolte, C. (2020). Allocating resources for land protection using
continuous optimization: biodiversity conservation in the United States. Ecological
maximization of land use suitability, while frequently used constraints Applications, 30(6). https://doi.org/10.1002/eap.2118
are a) one and only one land use in each cell, b) minimum and maximum Bagheri, M., Sulaiman, W. N. A., & Vaghefi, N. (2012). Land use suitability analysis using
area of particular land use, and c) restriction on specific land use change. multi criteria decision analysis method for coastal management and planning: A case
study of Malaysia. Journal of Environmental Science and Technology, 5, 364–372.
This study also clarifies that sustainability in urban land use optimiza­
https://doi.org/10.3923/jest.2012.364.372
tion is merely touched upon while no generalized method was estab­ Basiago, A. D. (1998). Economic, social, and environmental sustainability in
lished to measure economic, social, and environmental benefits from development theory and urban planning practice. Environmentalist, 19, 145–161.
land use planning. This study also finds that Pareto based method is https://doi.org/10.1023/A:1006697118620
Bourque, K., Schiller, A., Loyola Angosto, C., McPhail, L., Bagnasco, W., Ayres, A., &
more popular to construct urban land use optimization problems while Larsen, A. (2019). Balancing agricultural production, groundwater management,
the Genetic Algorithm (GA) accounts for the major contribution to solve and biodiversity goals: A multi-benefit optimization model of agriculture in Kern
the optimization problem. This study also recognizes that spatial data is County, California. Science of the Total Environment, 370, 865–875. https://doi.org/
10.1016/j.scitotenv.2019.03.197
an indispensable part of formulating urban land use optimization Bryan, B. A., Crossman, N. D., Nolan, M., Li, J., Navarro, J., & Connor, J. D. (2015). Land
problems where the raster data model is preferable to design urban land use efficiency: Anticipating future demand for land-sector greenhouse gas emissions
use optimization problems. Based on the findings of the study, we have abatement and managing trade-offs with agriculture, water, and biodiversity. Global
Change Biology, 21, 4098–4114. https://doi.org/10.1111/gcb.13020
proposed some recommendations for further studies. Most importantly, Cao, K., Zhang, W., & Wang, T. (2019). Spatio-temporal land use multi-objective
future studies should focus on urban sustainability while formulating optimization: A case study in Central China. Trans. GIS., 23(4), 726–744. https://doi.
objective functions, and a common method should be developed to org/10.1111/tgis.12535
Cao, Kai, Batty, M., Huang, B., Liu, Y., Yu, L., & Chen, J. (2011). Spatial multi-objective
measure the values of objective functions. We also recommended that land use optimization: Extensions to the non-dominated sorting genetic algorithm-II.
the participatory approach should be integrated with mathematical International Journal of Geographical Information Science, 25, 1949–1969. https://doi.
optimization to derive acceptable solutions in land use allocation. We org/10.1080/13658816.2011.570269
Cao, K., Huang, B., Wang, S., & Lin, H. (2012). Sustainable land use optimization using
think that our proposed recommendations can remove the existing
Boundary-based Fast Genetic Algorithm. Computers, Environment and Urban Systems,
shortcomings and improve the overall approach towards land use opti­ 36, 257–269. https://doi.org/10.1016/j.compenvurbsys.2011.08.001
mization and help to achieve urban sustainability. According to the best Cao, K, Liu, M., Wang, S., Liu, M., Zhang, W., Meng, Q., & Huang, B. (2020). Spatial
of our knowledge, this study is the first of its kind. No previous study multi-objective land use optimization toward livability based on boundary-based
genetic algorithm: A case study in Singapore. ISPRS International Journal of Geo-
comprehensively discussed and synthesized the different aspects of land Information, 9(1), Article 40. https://doi.org/10.3390/ijgi9010040
use optimization. In this review, different aspects of land use allocation Cao, K., & Ye, X. (2013). Coarse-grained parallel genetic algorithm applied to a vector
have been explored, and future research direction has been indicated based land use allocation optimization problem: The case study of Tongzhou
Newtown, Beijing, China. Stochastic Environmental Research and Risk Assessment, 27
based on our findings. Thus, we strongly believe that this research is a (5), 1133–1142. https://doi.org/10.1007/s00477-012-0649-y
novel work and has fulfilled the previous research gap on this topic. So, Caparros-Midwood, D., Barr, S., & Dawson, R. (2015). Optimised spatial planning to
researchers in this field are expected to get benefit from this review meet long term urban sustainability objectives. Computers, Environment and Urban
Systems, 54, 154–164. https://doi.org/10.1016/j.compenvurbsys.2015.08.003
paper by understanding the overall idea of urban land use optimization, Chiandussi, G., Codegone, M., Ferrero, S., & Varesio, F. E. (2012). Comparison of multi-
its current state, and future research scope. objective optimization methodologies for engineering applications. Computers and
Mathematics with Applications, 63(5), 912–942. https://doi.org/10.1016/j.
camwa.2011.11.057
Funding Cohen, M. (2017). A Systematic Review of Urban Sustainability Assessment Literature.
Sustainability, 9(11), Article 2048. https://doi.org/10.3390/su9112048
This research did not receive any specific grant from funding Cohn, L. F., Harris, R. A., Shu, N., & Li, W. (2005). Highway Noise and Land Use
Compatibility. Journal of Urban Planning and Development. https://doi.org/10.1061/
agencies in the public, commercial, or not-for-profit sectors.
(asce)0733-9488(2005)131:3(125)
Cook, E. M., Hall, S. J., & Larson, K. L. (2012). Residential landscapes as social-ecological
Declaration of Competing Interest systems: A synthesis of multi-scalar interactions between people and their home
environment. Urban Ecosystem, 15, 19–52. https://doi.org/10.1007/s11252-011-
0197-0
The authors declare that they have no known competing financial Copado-Méndez, P. J., Pozo, C., Guillén-Gosálbez, G., & Jiménez, L. (2016). Enhancing
interests or personal relationships that could have appeared to influence the ε-constraint method through the use of objective reduction and random
the work reported in this paper. sequences: Application to environmental problems. Computers & Chemical
Engineering, 87, 36–48. https://doi.org/10.1016/j.compchemeng.2015.12.016
Deb, K. (2011). Multi-objective optimisation using evolutionary algorithms: An
introduction. Multi-objective evolutionary optimisation for product design and
manufacturing. https://doi.org/10.1007/978-0-85729-652-8_1

14
Md.M. Rahman and G. Szabó Sustainable Cities and Society 74 (2021) 103214

Deelstra, T., & Girardet, H. (2000). Urban agriculture and sustainable cities. Growing Journal of Geographical Information Science, 22(6), 601–622. https://doi.org/
cities, growing food: urban agriculture on the policy agenda (pp. 43–65). Feldafing, 10.1080/13658810701587495
Germany: Deutsche Stiftung fur Internationale Entwicklung (DSE), Zentralstelle fur Liu, R., Zhang, K., Zhang, Z., & Borthwick, A. G. L. (2014). Land-use suitability analysis
Ernahrung und Landwirtschaft. for urban development in Beijing. Journal of Environmental Management, 145,
Duranton, G., & Guerra, E. (2016). Urban accessibility: Balancing land use and 170–179. https://doi.org/10.1016/j.jenvman.2014.06.020
transportation. The Geography of Transport Systems. Liu, X., Li, X., Shi, X., Huang, K., & Liu, Y. (2012). A multi-type ant colony optimization
Estoque, R. C., & Murayama, Y. (2013). Landscape pattern and ecosystem service value (MACO) method for optimal land use allocation in large areas. International Journal
changes: Implications for environmental sustainability planning for the rapidly of Geographical Information Science, 26(7), 1325–1343. https://doi.org/10.1080/
urbanizing summer capital of the Philippines. Landscape and Urban Planning, 116, 13658816.2011.635594
60–72. https://doi.org/10.1016/j.landurbplan.2013.04.008 Lubida, A., Veysipanah, M., Pilesjo, P., & Mansourian, A. (2019). Land-use planning for
Gao, P., Wang, H., Cushman, S. A., Cheng, C., Song, C., & Ye, S. (2020). Sustainable land- sustainable urban development in africa: A spatial and multi-objective optimization
use optimization using NSGA-II: theoretical and experimental comparisons of approach. Geodesy and Cartography, 45(1), 1–15. https://doi.org/10.3846/
improved algorithms. Landscape Ecology, 36, 1877–1892. https://doi.org/10.1007/ gac.2019.6691
s10980-020-01051-3 Ma, X., & Zhao, X. (2015). Land use allocation based on a multi-objective artificial
García, G. A., Rosas, E. P., García-Ferrer, A., & Barrios, P. M. (2017). Multi-objective immune optimization model: An application in Anlu county, China. Sustain, 7(11),
spatial optimization: Sustainable land use allocation at sub-regional scale. Sustain, 9 15632–15651. https://doi.org/10.3390/su71115632
(6), Article 927. https://doi.org/10.3390/su9060927 Masoomi, Z., Mesgari, M. S., & Hamrah, M. (2013). Allocation of urban land uses by
Givi, M. K. B, & Asadi, P. (2014). Artificial neural network and optimization. Advances in Multi-Objective Particle Swarm Optimization algorithm. International Journal of
Friction-Stir Welding and Processing (pp. 543–599). UK: Elsevier. Geographical Information Science, 27(3), 542–566. https://doi.org/10.1080/
Goffe, W. L., Ferrier, G. D., & Rogers, J. (1994). Global optimization of statistical 13658816.2012.698016
functions with simulated annealing. Journal of Economics, 60(1–2), 65–99. https:// Mohammadi, M., Nastaran, M., & Sahebgharani, A. (2015). Sustainable spatial land use
doi.org/10.1016/0304-4076(94)90038-8 optimization through non-dominated sorting genetic algorithm-II (NSGA-II): (Case
Gong, J., Liu, Y., & Chen, W. (2012). Optimal land use allocation of urban fringe in Study: Baboldasht District of Isfahan). Indian Journal of Science and Technology, 8(3),
Guangzhou. Journal of Geographical Sciences, 22(1), 179–191. https://doi.org/ 1–12. https://doi.org/10.17485/ijst/2015/v8iS3/60700
10.1007/s11442-012-0920-7 Moher, D., Liberati, A., Tetzlaff, J., Altman, D. G., Altman, D., Antes, G., Atkins, D.,
Handayanto, R. T., Tripathi, N. K., Kim, S. M, & Guha, S. (2017). Achieving a Sustainable Barbour, V., Barrowman, N., Berlin, J. A., Clark, J., Clarke, M., Cook, D.,
Urban Form through Land Use Optimisation: Insights from Bekasi City’s Land-Use D’Amico, R., Deeks, J. J., Devereaux, P. J., Dickersin, K., Egger, M., Ernst, E.,
Plan (2010–2030). Sustainability, 9(2), Article 221. https://doi.org/10.3390/ Gøtzsche, P. C., Grimshaw, J., Guyatt, G., Higgins, J., Ioannidis, J. P. A., Kleijnen, J.,
su9020221 Lang, T., Magrini, N., McNamee, D., Moja, L., Mulrow, C., Napoli, M., Oxman, A.,
Haque, A., & Asami, Y. (2011). Optimizing urban land-use allocation: Case study of Pham, B., Rennie, D., Sampson, M., Schulz, K. F., Shekelle, P. G., Tovey, D., &
Dhanmondi Residential Area, Dhaka, Bangladesh. Environment and Planning B: Tugwell, P. (2009). Preferred reporting items for systematic reviews and meta-
Planning and Design, 38(3), 388–410. https://doi.org/10.1068/b35041 analyses: The PRISMA statement. PLoS Med. https://doi.org/10.1371/journal.
Haque, A., & Asami, Y. (2014). Optimizing urban land use allocation for planners and pmed.1000097
real estate developers. Computers, Environment and Urban Systems, 64, 57–69. Mora, H., Gilart-Iglesias, V., Pérez-Del Hoyo, R., & Andújar-Montoya, M. D. (2017).
https://doi.org/10.1016/j.compenvurbsys.2014.04.004 A comprehensive system for monitoring urban accessibility in smart cities. Sensors,
He, Z., & Yen, G. G. (2016). Many-objective evolutionary algorithm: Objective space 17(8), Article 1834. https://doi.org/10.3390/s17081834
reduction and diversity improvement. IEEE Transactions on Evolutionary Computation, Mouratidis, K. (2019). Compact city, urban sprawl, and subjective well-being. Cities, 92,
20(1), 145–160. https://doi.org/10.1109/TEVC.2015.2433266 261–272. https://doi.org/10.1016/j.cities.2019.04.013
Higgins, D., Balint, T., Liversage, H., & Winters, P. (2018). Investigating the impacts of Nandi, A. K., & Deb, K. (2016). Multi-Objective Evolutionary Algorithms: Application in
increased rural land tenure security: A systematic review of the evidence. J. Rural Designing Particle Reinforced Mould Materials. In S. Datta, & J. P. Davim (Eds.),
Stud., 61, 34–62. https://doi.org/10.1016/j.jrurstud.2018.05.001 Computational Approaches to Materials Design: Theoretical and Practical Aspects (pp.
Huang, B., & Zhang, W. (2014). Sustainable land-use planning for a Downtown Lake Area 301–345). IGI Global.
in Central China: Multiobjective optimization approach aided by urban growth Nasir Ahmad, N. S. B., Mustafa, F. B., Muhammad Yusoff, S.@. Y., & Didams, G. (2020).
modeling. Journal of Urban Planning and Development, 140(2). https://doi.org/ A systematic review of soil erosion control practices on the agricultural land in Asia.
10.1061/(ASCE)UP.1943-5444.0000186 International Soil and Water Conservation Research, 8(2), 103–115. https://doi.org/
Huang, H., Li, Q., & Zhang, Y. (2019). Urban residential land suitability analysis 10.1016/j.iswcr.2020.04.001
combining remote sensing and social sensing data: A case study in Beijing. China. Niu, C.-Y., Alamusa, Zong, Q., Luo, Y.-M., Oshid, T., Sun, G.-F., & Liu, Q. (2013).
Sustain., 11(8), Article 2255. https://doi.org/10.3390/su11082255 Allocation patterns of above- and below-ground biomass of Caragana microphylla in
Huang, K., Liu, X., Li, X., Liang, J., & He, S. (2013). An improved artificial immune Horqin Sandy Land, North China. Chinese Journal of Ecology, 32(8), 1980–1986.
system for seeking the Pareto front of land-use allocation problem in large areas. Pae, C. U. (2015). Why systematic review rather than narrative review? Psychiatry
International Journal of Geographical Information Science, 27(5), 922–946. https://doi. Investig., 12(3), 417–419. https://doi.org/10.4306/pi.2015.12.3.417
org/10.1080/13658816.2012.730147 Pahlavani, P., Sheikhian, H., & Bigdeli, B. (2020). Evaluation of residential land use
Karakostas, S. (2015). Multi-objective optimization in spatial planning: Improving the compatibilities using a density-based IOWA operator and an ANFIS-based model: A
effectiveness of multi-objective evolutionary algorithms (non-dominated sorting case study of Tehran, Iran. Land Use Policy, 90, Article 104364. https://doi.org/
genetic algorithm II). Engineering Optimization, 47(5), 601–621. https://doi.org/ 10.1016/j.landusepol.2019.104364
10.1080/0305215X.2014.908870 Panwar, D., Jha, M., & Srivastava, N. (2018). Optimization of risk and return using fuzzy
Karakostas, S. M. (2017). Bridging the gap between multi-objective optimization and multiobjective linear programming. Advances in Fuzzy Systems, 2018, Article
spatial planning: a new post-processing methodology capturing the optimum 4279236. https://doi.org/10.1155/2018/4279236
allocation of land uses against established transportation infrastructure. Pappas, I., Avraamidou, S., Katz, J., Burnak, B., Beykal, B., Türkay, M., &
Transportation Planning and Technology, 40(3), 305–326. https://doi.org/10.1080/ Pistikopoulos, E. N. (2021). Multiobjective Optimization of Mixed-Integer Linear
03081060.2017.1283157 Programming Problems: A Multiparametric Optimization Approach. Industrial &
Karakostas, S. M. (2016). Land-use planning via enhanced multi-objective evolutionary Engineering Chemistry Research, 60(23), 8493–8503. https://doi.org/10.1021/acs.
algorithms: optimizing the land value of major Greenfield initiatives. Journal of Land iecr.1c01175
Use Science, 11(5), 595–617. https://doi.org/10.1080/1747423X.2016.1223187 Parry, J. A., Ganaie, S. A., & Sultan Bhat, M. (2018). GIS based land suitability analysis
Li, Feixue, Gong, Y., Cai, L., Sun, C., Chen, Y., Liu, Y., & Jiang, P. (2018). Sustainable using AHP model for urban services planning in Srinagar and Jammu urban centers
land-use allocation: A multiobjective particle swarm optimization model and of J&K, India. Journal of Urban Management, 7(2), 46–56. https://doi.org/10.1016/j.
application in Changzhou. China. Journal of Urban Planning and Development, 144(2). jum.2018.05.002
https://doi.org/10.1061/(ASCE)UP.1943-5444.0000425 Porta, J., Parapar, J., Doallo, R., Rivera, F. F., Santé, I., & Crecente, R. (2013). High
Kiani Sadr, M., Nassiri, P., Hosseini, S. M., Monavari, M., & Gharagozlou, A. (2014). performance genetic algorithm for land use planning. Computers, Environment and
Assessment of land use compatibility and noise pollution at Imam Khomeini Urban Systems, 37, 45–58. https://doi.org/10.1016/j.compenvurbsys.2012.05.003
International Airport. Journal of Air Transport Management, 34, 49–56. https://doi. Development Core, R, & Team, R. (2011). R: A language and environment for statistical
org/10.1016/j.jairtraman.2013.07.009 computing. R Foundation for Statistical Computing. https://doi.org/10.1007/978-3-
Li, X., & Ma, X. (2018). An improved simulated annealing algorithm for interactive multi- 540-74686-7
objective land resource spatial allocation. Ecological Complexity, 36, 184–195. Rahman, M. M., & Szabó, G. (2020). National spatial data infrastructure (NSDI) of
https://doi.org/10.1016/j.ecocom.2018.08.008 Bangladesh – Development, progress and way forward. The ISPRS Annals of the
Li, X., & Parrott, L. (2016). An improved Genetic Algorithm for spatial optimization of Photogrammetry, Remote Sensing and Spatial Information, V-4-2020, 131–138. https://
multi-objective and multi-site land use allocation. Computers, Environment and Urban doi.org/10.5194/isprs-annals-v-4-2020-131-2020
Systems, 59, 184–194. https://doi.org/10.1016/j.compenvurbsys.2016.07.002 Rath, S., & Gutjahr, W. J. (2014). A math-heuristic for the warehouse location-routing
Libby, L. W., & Sharp, J. S. (2003). Land-use compatibility, change, and policy at the problem in disaster relief. Computers & Operations Research, 42, 25–39. https://doi.
rural-urban fringe: Insights from social capital. American Journal of Agricultural org/10.1016/j.cor.2011.07.016
Economics, 85(5), 1194–1200. https://doi.org/10.1111/j.0092-5853.2003.00529.x Robinson, P., & Lowe, J. (2015). Literature reviews vs systematic reviews. Australian and
Ligmann-Zielinska, A. (2017). Spatial optimization. International encyclopedia of New Zealand Journal of Public Health, 39(2), 103. https://doi.org/10.1111/1753-
geography: people, the earth. Environment and Technology. https://doi.org/10.1002/ 6405.12393
9781118786352.wbieg0156 Russo, A., & Cirella, G. T. (2018). Modern compact cities: How much greenery do we
Ligmann-Zielinska, A., Church, R., & Jankowski, P. (2008). Spatial optimization as a need? International Journal of Environmental Research and Public Health, 15(10),
generative technique for sustainable multiobjective land-use allocation. International Article 2180. https://doi.org/10.3390/ijerph15102180

15
Md.M. Rahman and G. Szabó Sustainable Cities and Society 74 (2021) 103214

Sahebgharani, A. (2016). Multi-objective land use optimization through parallel particle Chinese Geographical Science, 20(2), 176–183. https://doi.org/10.1007/s11769-010-
swarm algorithm: Case study Baboldasht district of Isfahan, Iran. Journal of Urban 0176-z
and Environmental Engineering, 10(1), 42–49. https://doi.org/10.4090/juee.2016. Wang, L., & Fang, M. (2018). Robust optimization model for uncertain multiobjective
v3n1.042049 linear programs. Journal of Inequalities and Applications. , Article 22. https://doi.org/
Santé-Riveira, I., Boullón-Magán, M., Crecente-Maseda, R., & Miranda-Barrós, D. (2008). 10.1186/s13660-018-1612-3
Algorithm based on simulated annealing for land-use allocation. Computers & Wang, S. D., Zhang, H. B., & Wang, X. C. (2019). Simulating land use structure
Geosciences, 34(3), 259–268. https://doi.org/10.1016/j.cageo.2007.03.014 optimization based on an improved multi-objective differential evolution algorithm.
Schwaab, J., Deb, K., Goodman, E., Lautenbach, S., van Strien, M. J., & Grêt-Regamey, A. Polish Journal of Environmental Studies, 28(2), 887–899. https://doi.org/10.15244/
(2018). Improving the performance of genetic algorithms for land-use allocation pjoes/85222
problems. International Journal of Geographical Information Science, 32(5), 907–930. Wang, Y., Zhang, X., & Peng, P. (2021). Spatio-Temporal Changes of Land-Use/Land
https://doi.org/10.1080/13658816.2017.1419249 Cover Change and the Effects on Ecosystem Service Values in Derong County, China,
Sharmin, N., Haque, A., & Islam, M. M. (2019). Generating alternative land-use from 1992–2018. Sustainability, 13(2), Article 827. https://doi.org/10.3390/
allocation for mixed use areas: Multi-objective optimization approach. Geographical su13020827
Analysis, 51, 448–474. https://doi.org/10.1111/gean.12181 Williams, K., Burton, E., & Jenks, M. (2013). Achieving Sustainable Urban Form: An
Shaygan, M., Alimohammadi, A., Mansourian, A., Govara, Z. S., & Kalami, S. M. (2014). Introduction. Achieving Sustainable Urban Form (1st). London: Routledge. https://doi.
Spatial multi-objective optimization approach for land use allocation using NSGA-II. org/10.4324/9780203827925
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 7(3), Xia, Y., Liu, D., Liu, Y., He, J., & Hong, X. (2014). Alternative zoning scenarios for
906–916. https://doi.org/10.1109/JSTARS.2013.2280697 regional sustainable land use controls in China: A knowledge-based multiobjective
Shi, C., Zhan, J., Yuan, Y., Wu, F., & Li, Z. (2015). Land use zoning for conserving optimisation model. International Journal of Environmental Research and Public Health,
ecosystem services under the impact of climate change: A case study in the middle 11(9), 8839–8866.
reaches of the Heihe River Basin. Advances in Meteorology, 2015, Article 496942. Yang, L., Sun, X., Peng, L., Shao, J., & Chi, T. (2015). An improved artificial bee colony
Song, M, & Chen, D. (2018). A comparison of three heuristic optimization algorithms for algorithm for optimal land-use allocation. International Journal of Geographical
solving the multi-objective land allocation (MOLA) problem. Ann. GIS. Information Science, 29(8), 1470–1489. https://doi.org/10.1080/
Song, Mingjie, & Chen, D. (2018a). An improved knowledge-informed NSGA-II for multi- 13658816.2015.1012512
objective land allocation (MOLA). Geo-Spatial Inf. Sci.. https://doi.org/10.1080/ Yang, L., Zhu, A., Shao, J., & Chi, T. (2018). A knowledge-informed and pareto-based
10095020.2018.1489576 artificial bee colony optimization algorithm for multi-objective land-use allocation.
Song, Mingjie, & Chen, D. M. (2018b). A comparison of three heuristic optimization ISPRS International Journal of Geo-Information, 7(2), Article 63. https://doi.org/
algorithms for solving the multi-objective land allocation (MOLA) problem. Ann. GIS.. 10.3390/ijgi7020063
https://doi.org/10.1080/19475683.2018.1424736 Yuan, M., Liu, Y., He, J., & Liu, D. (2014). Regional land-use allocation using a coupled
Steward, B. (2004). Writing a literature review. British Journal of Occupational Therapy, MAS and GA model: From local simulation to global optimization, a case study in
67(11), 495–500. https://doi.org/10.1177/030802260406701105 Caidian District, Wuhan, China. Cartography and Geographic Information Science, 41
Stewart, T. J. (1991). A multi-criteria decision support system for r&d project selection. (4), 363–378. https://doi.org/10.1080/15230406.2014.931251
Journal of the Operational Research Society, 42(1), 17–26. https://doi.org/10.1057/ Zhang, H., Zeng, Y., Jin, X., Shu, B., Zhou, Y., & Yang, X. (2016). Simulating multi-
jors.1991.3 objective land use optimization allocation using Multi-agent system-A case study in
Stewart, T. J., Janssen, R., & Van Herwijnen, M. (2004). A genetic algorithm approach to Changsha, China. Ecological Modelling, 320, 334–347. https://doi.org/10.1016/j.
multiobjective land use planning. Computers & Operations Research, 31(14), ecolmodel.2015.10.017
2293–2313. https://doi.org/10.1016/S0305-0548(03)00188-6 Zhang, W., & Huang, B. (2015). Land use optimization for a rapidly urbanizing city with
Su, S., Li, D., Hu, Y., Xiao, R., & Zhang, Y. (2014). Spatially non-stationary response of regard to local climate change: Shenzhen as a case study. Journal of Urban Planning
ecosystem service value changes to urbanization in Shanghai, China. Ecological and Development, 141(1). https://doi.org/10.1061/(ASCE)UP.1943-5444.0000200
Indicators, 45, 332–339. https://doi.org/10.1016/j.ecolind.2014.04.031 Zhang, W., Wang, H., Cao, K., He, S., & Shan, L. (2019). Ecological conservation- and
Suppapitnarm, A., Seffen, K. A., Parks, G. T., & Clarkson, P. J. (2000). A Simulated economic development-based multiobjective land-use optimization: Case study of a
annealing algorithm for multiobjective optimization. Engineering Optimization, 33(1), rapidly developing city in central. China. Journal of Urban Planning and Development,
59–85. https://doi.org/10.1080/03052150008940911 145(1). https://doi.org/10.1061/(ASCE)UP.1943-5444.0000481
Tong, D., & Murray, A. T. (2012). Spatial optimization in geography. Annals of the Zheng, W, Ke, X., Xiao, B., & Zhou, T. (2019). Optimising land use allocation to balance
Association of American Geographers, 102(6), 1290–1309. https://doi.org/10.1080/ ecosystem services and economic benefits - A case study in Wuhan, China. Journal of
00045608.2012.685044 Environmental Management, 248. https://doi.org/10.1016/j.jenvman.2019.109306
Uludere Aragon, N., Wagner, M., Wang, M., Broadbent, A. M., Parker, N., & Zhu, X. (2020). Introduction to Systematic Review and Meta-Analysis. Handbook of Meta-
Georgescu, M. (2017). Sustainable land management for bioenergy crops. Energy Analysis (1st). 90 High Holborn, London: Routledge. https://doi.org/10.4324/
Procedia, 125, 379–388. https://doi.org/10.1016/j.egypro.2017.08.063 9780429052675-16
Wang, H., Gao, Y., Liu, Q., & Song, J. (2010). Land use allocation based on interval multi-
objective linear programming model: a case study of Pi county in sichuan province.

16