You are on page 1of 15

Journal of Cleaner Production 268 (2020) 121928

Contents lists available at ScienceDirect

Journal of Cleaner Production


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

A decision support framework for the design and operation of


sustainable urban farming systems
Lanyu Li a, Xian Li b, Clive Chong c, Chi-Hwa Wang a, Xiaonan Wang a, *
a
Department of Chemical and Biomolecular Engineering, National University of Singapore, 117585, Singapore
b
NUS Environmental Research Institute, National University of Singapore, 138602, Singapore
c
Paradise Eco Tourism Pte Ltd, Six Battery Road, 049909, Singapore

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

Article history: The increasing population and continuous urbanization make food security prominent in sustainable
Received 10 August 2019 development. It is important to develop economic and resource-efficient farming solutions to meet food
Received in revised form demand. Renewable energy and waste valorization may bring benefits to build sustainable food pro-
17 April 2020
duction systems and facilitate circular economy. This work aims to develop a decision support framework
Accepted 26 April 2020
Available online 10 May 2020
for the stakeholders to quantitatively assess and optimize their urban farming systems for efficient in-
vestment and operation. The proposed framework is based on a holistic system model that considers the
Handling editor: Dr. Govindan Kannan energy and material consumption in vegetable production processes and the economic and environ-
mental performance of urban farming systems. In the multi-dimensional assessment model, the net
Keywords: present value and cradle-to-gate CO2 emission, water consumption, and land occupation of different
Economic assessment configurations of urban farming systems were assessed. In a further development, the assessment model
Environmental assessment was embedded in an optimization framework to identify the optimal system design and operation. The
Optimization optimal crop mix and the corresponding cultivation set points (such as temperature, humidity, irradi-
Precision agriculture
ance, illumination time, and CO2 concentration) for the farming modules were determined via optimi-
Sustainable urban farming
zation. To demonstrate the proposed framework, a case study on the design and operation of a vertical
Waste valorization
farm in Singapore was carried out. The case study examined alternative farming systems with glass-
enclosed vs window-free structural design, grid vs solar photovoltaic (PV) energy supply, and tradi-
tional chemical fertilizers vs food waste compost fertilization. Results showed that plant-factory farming
systems integrated with solar PV and beer-residue-derived fertilizer could be a promising and sustain-
able farming solution for Singapore as a tropical megacity.
© 2020 Elsevier Ltd. All rights reserved.

1. Introduction emissions; and (d) the investment in sustainable agriculture is


insufficient. Today, modern farming has largely enhanced the effi-
Food production is a major environmental burden that accounts ciency of crop production, reflected in the huge increase in crop
for 10%e30% of an individual person’s environmental impact yield and decrease in manpower requirement, by the introduction
(Stoessel et al., 2012). According to the recent report from the Food of farm machinery, use of chemical fertilizers and pesticides,
and Agriculture Organization (FAO, 2018), food production is facing development of high yield varieties, and increased knowledge
several challenges: (a) there is a growing demand for food, espe- about farm management practices (Trautmann et al., 2020). How-
cially the resource-intensive products (driven by increased popu- ever, while the highly mechanized and chemical-oriented farming
lation, income, and urbanization); (b) land and water resources for practice makes agricultural production more efficient and cost-
food production are limited and diminishing in quality; (c) the effective, it does not make the production more sustainable. The
agricultural sector is increasingly threatened by climate change, intensive use of machines, irrigation, and chemicals in modern
while itself is one of the major sources of greenhouse gas farming increases energy and water consumption and leads to
more significant environmental impacts, including potential
degradation of the soil and water resources (Trautmann et al.,
2020). Apart from the environmental impact, the lack of arable
* Corresponding author.
land and the food wastage in the supply chain are also potential
E-mail address: chewxia@nus.edu.sg (X. Wang).

https://doi.org/10.1016/j.jclepro.2020.121928
0959-6526/© 2020 Elsevier Ltd. All rights reserved.
2 L. Li et al. / Journal of Cleaner Production 268 (2020) 121928

problems. In 2019, about 10% of the habitable land was used for system selection and design. However, it can be a challenging task
farming. As it is foreseen that 70% of the population will stay in for farming practitioners to derive the optimal design and opera-
cities by 2050, the expansion of urban territory will lead to further tional plan. To address this problem, this work aims to propose a
shrinkage of arable land (FAO, 2009). The increased distance makes comprehensive and versatile decision support framework, which
food supply from rural farms to cities less preferable as it degrades could serve the simulation, assessment, and optimization purposes
food quality and quantity, introducing wastes and environmental to help decision-makers with the design and operation of sus-
burden. In the meanwhile, the rising demand for animal products tainable urban farming systems.
and the strong growth of bioenergy are shunting away more arable Life cycle assessment (LCA) and life cycle cost analysis (LCCA)
land to the production of animal feed and energy crop, which could are mostly used for system assessment in the literature. These
aggravate food shortages (AGA Webmaster, 2018; FAO, 2017). With methods have been applied to study the performance of the food
these challenges and the problems of the current farming practice, value chain in multiple scales. In process-scale studies, various
the FAO pointed out that new solutions and innovative technolo- forms of farming system designs were investigated. Maucieri et al.
gies for sustainable intensification, efficient use of land and water, (2018) used a life cycle assessment method to quantify the envi-
energy and greenhouse gas reduction are required to move the ronmental impact of a micro aquaponics system for educational
farming practice away from the “business as usual” (BAU) towards purposes. Liaros et al. (2016) conducted a techno-economic eval-
sustainable production (FAO, 2018). uation of urban plant factories using life cycle cost analysis. Oliver-
In the shift to efficient and sustainable agriculture, various Sola et al. (2015), Sanye
-Mengual et al. (2015) and Sanjuan-Delm as
innovative farming systems have been developed. Agriculture is et al. (2018) carried out successive life cycle cost and life cycle
increasingly spreading to towns and cities. Urban farming, a prac- assessment studies on a real rooftop farm in Barcelona, which was
tice of cultivating, processing, and distributing food in or around reconstructed from an open-field rooftop farm to an integrated
the urban area, is an emerging alternative to conventional rural rooftop greenhouse (iRTG) system. Besides, network-scale studies
agriculture. It not only ensures the supply of fresh food to major were also conducted to take into account the interaction of the
consumers but also generates employment, improves supply chain urban farming process with other industrial or ecological systems.
efficiency, recycles urban wastes, creates greenbelts, and Stoessel et al. (2012) carried out detailed assessment of the life
strengthens cities’ resilience to climate change (FAO, 2020). In cycle inventory and carbon and water footprint of 34 fruits and
addition, recycling waste in urban farms further fosters urban vegetables of a large Swiss retailer using an LCA approach with the
symbiosis, helping to build a circular economy in cities and reserves goal of providing decision support to the retailer on purchasing
ecosystems outside cities (Goldstein et al., 2016a, 2016b; Piezer options and improving supply chain management. Piezer et al.
et al., 2019). To fit in the city context, urban agriculture is ex- (2019) developed a combined LCA and ecological network anal-
pected to be small in ecological footprint, flexible, autonomous, and ysis (ENA) approach to assess the sustainability of growing to-
sustainable. A variety of urban farming systems have been derived, matoes in an integrated rooftop greenhouse under the urban
including rooftop farming (cultivating crops on the top of build- energy-food nexus context.
ings), vertical farming (growing crops in vertically stacked layers), In most of the studies, researchers focus on assessing the
and plant factory (an indoor vertical farming system for efficient farming projects with real-world data available. However, in this
quality food production) (Kozai et al., 2019; Sanye -Mengual et al., way, it is not an easy task to make a design or operational adjust-
2015; Shamshiri et al., 2018). Indoor farming grows plants in ment to improve farm performance if the assessment suggests that
cleanroom-like conditions, where the cultivation environment the existing configuration is not sustainable. Prior assessment and
(from temperature, humidity, and CO2 concentration to lighting, optimization on the design and operation of the farm (e.g. in the
water, and nutrients) is controlled. This is also known as the conceptual or design phase of the project) could help decision-
concept of control environment agriculture (CEA). The environ- makers to specify better choices and increase the profitability and
mental condition, such as the LED lights used to replace sunlight, sustainability of their investment. To achieve this, modeling and
can be carefully calibrated to maximize plant photosynthesis and optimization techniques may be required. Actually, several decision
improve crop yield (Kee, 2019). Advancement in sensor technology support systems (DSS) based on modeling and simulation have
enables the monitoring of the environmental conditions in the CEA. been developed to assess farming systems. Benis et al. (2017)
With further development of the wireless sensor network and developed a comprehensive simulation-based decision-support
devices, the concept of Internet of things (IoT) that interrelates workflow which simulates and visualizes food production, water
everything connected to the internet for data transfer without use, and energy use for Building-Integrated Agriculture (BIA). They
human interaction can be adopted to facilitate automation of the covered the site, farm structure, operation, and plant growth
farming process (Kamilaris and Prenafeta-Boldú, 2018; simulation in their model. However, they did not consider incor-
Muangprathub et al., 2019; Wallace, 1994). With the help of these porating optimization methods to facilitate and automate the se-
novel technologies, accurate monitoring and control of urban lection of a better design or operation strategy of the BIA.
farming systems can be achieved for standardized quality control, Optimization techniques have been applied in several studies in the
food sourcing, and keeping track of the environmental impact of management of conventional farms (RaheliNamin et al., 2016)
food production. It is expected that indoor agriculture will continue (Capitanescu et al., 2017) (Udias et al., 2018). But the application of
expanding at nearly 20% annually through 2023. With this prom- optimization in the field of agriculture is still underexploited
ising outlook, Singapore, a highly urbanized modern city-state with (Capitanescu et al., 2017), especially for urban farming systems.
limited arable land, is attracting more research and multinational Still, it is worth noting that there are two relevant studies on urban
investment in urban agriculture (Kee, 2019). farming systems. Brulard et al. (2018) investigated the optimization
However, there is no “one-size fits all” technology. To improve on the production and storage of perishable products, with the
the quantity, quality, and sustainability of agricultural production, objective of maximizing the revenue of the urban farming system.
decision-makers or farmers need to develop site-specific strategies. Johnson et al. (2018) proposed a concept of “personal food com-
For decision-makers, how to quantify the techno-economic and puter” and optimized the plant metabolites in an open-source
environmental impact of an urban farming system and how to controlled environment using surrogate modeling. However, the
determine the choice of crops and the set points for optimal eco- publications so far have not addressed the aforementioned gap of
nomic and environmental performances are two key questions for helping farmers to decide what design of the urban farm system,
L. Li et al. / Journal of Cleaner Production 268 (2020) 121928 3

what crop to be planted, and what operating conditions would help simulation tool to understand the effect of different settings on the
them yield the highest economic benefit while achieving good farm production; (c) an assessment tool to evaluate the economic
environmental sustainability. To fill this research gap, a holistic and environmental performance of various farming scenarios; and
decision-support framework integrating modeling, multi- (d) a holistic decision support system to recommend the optimal
dimensional sustainability assessment, and optimization, which is production decisions to the users. The methodologies involved in
in line with the essence of precision agriculture research (i.e. each individual part of the framework is discussed in the following
optimizing returns on inputs while preserving resources) subsections.
(McBratney et al., 2005), was proposed in this study. It was devel-
oped to aid the design and production scheduling of the farm with
the following contributions to the field: 2.1. Database

 It helps farmers and stakeholders develop their design and Data collection is an essential step to carry out the simulation,
operation strategies with handy farm-wide input-output anal- assessment, and optimization of urban farming systems. The pro-
ysis, economic and environmental evaluation, and information posed framework in this paper requires five major categories of
of the optimum system design and operation under different parameters which are listed below:
scenarios.
 The proposed decision support system, which provides solu-  Site and climate data: temperature profile and humidity infor-
tions to scheduling production of urban farms, can provide mation of the selected location.
guidance for farm-wide control. By linking the DSS to online  Market data: market demand and selling price for different
database captured by IoT technologies and robotic machines, it crops.
is promising to achieve a high level of automation in the field of  Environmental analysis data: the elementary consumption and
urban agriculture, and quickly adopt economically and envi- emission for the production of each unit of energy or material
ronmentally sustainable production practices according to the inlet.
instant information without human intervention.  Farm data: available land area, dimension and layout of the
farm, structural design, and culture method.
A case study based on data of Singapore was carried out to  Crop data: characteristic temperature, humidity, daylight inte-
showcase the modeling, assessment, and optimization of urban gral, light intensity, photoperiod, and CO2 concentration for crop
farming systems using the proposed framework. In the case study, growth, and basic yield information.
three sets of scenarios were examined: (a) the scenarios where the
farming system was designed as modular plant factory with no These data were collected and stored in the database to support
window, glass-enclosed greenhouses, and open field were decision making using the framework.
compared to find out the effect of greenhouse structure on the
system operation and performance; (b) the scenarios where the 2.2. System modeling
farming system was powered by the grid and solar PV were
compared to find out the power supply method that was more With input data from the database, the crop yield and mass and
profitable and less resource/carbon-intensive; (c) the scenarios energy flows around the farming system were estimated using the
where the crops were fertilized by chemical fertilizers and beer- integrated system models as follows.
residue-derived compost were compared to study the economic
and environmental potential of waste-to-resource with the goal of
achieving circular economy. To illustrate this work, the remaining 2.2.1. Crop yield model
paper is structured as follows: Section 2 presents the methodology This work aims to use a versatile model to predict the crop yield
including the proposed framework and the assessment and opti- using the reaction mechanisms in photosynthesis. Plants convert
mization models; Section 3 gives the background about the case solar energy into biomass through photosynthesis. One way to
study and the scenario settings; Section 4 presents the results and estimate the yield of plants is based on the efficiency of the
discussion; and Section 5 shows the conclusion of this study. A list photosynthesis process, which is affected by temperature, CO2
of the symbols used in this study is also available in the Supple- concentration, and light intensity for most of the plant species
mentary Information (Part F). (Boote et al., 1991; Lawlor, 2001). As illustrated by Fig. 2, each plant
has an optimal growing temperature at which its yield is optimized.
2. Methodology Either a higher or lower temperature diminishes the rate of
photosynthesis. On the other hand, this rate increases with CO2
The proposed decision support framework for the urban concentration and light intensity until the saturation point is hit
farming system is illustrated in Fig. 1. and stays constant afterward.
In the framework, there are three main modules, which are Therefore, the total mass of production m, which depends on the
Module I - Modeling of urban farming system, Module II - Multi- maximum yield the crop can achieve under the optimum condi-
dimensional sustainability assessment, and Module III - Optimiza- tions Ymax , the planting area A, and the turns of crop production in a
tion. Data were extracted from databases and fed into Module I to year ncrop;max, is discounted when the operating condition is not at
model the material and energy flows around the boundary of the its optimum (Eqn. (1)). The efficiency of photosynthesis was
farm. The mass and energy balances were further tabulated as the defined as the ratio of actual rate of photosynthesis r and the rate at
input for the economic and environmental assessment in Module II. optimum condition Rmax . The efficiency of photosynthesis caused
Furthermore, in Module III, decision support at a higher level was by the deviation of temperature, CO2 concentration, and light in-
provided by optimization, which integrated all the modules for tensity from the optimum were estimated by Eqns. (2)e(4) (Erwin
iterative calculation using a searching algorithm that identifies the et al., 2017; Yan and Hunt, 1999). An overall efficiency hyield was
best design and operation of the farming system. Based on these determined by the smallest value among hT , hCO2 , and hL since the
modules, this framework can provide users with four levels of de- overall rate of photosynthesis is determined by the rate limiting
cision support: (a) a database to look up for information; (b) a factor (Eqn. (5)).
4 L. Li et al. / Journal of Cleaner Production 268 (2020) 121928

Fig. 1. Decision support framework for the urban farming system.

Fig. 2. Illustration of the response of the rate of photosynthesis to temperature, light intensity, and CO2 concentration (Boote et al., 1991; Lawlor, 2001).

  
hCO2 ¼ 1  exp  kCO2 , CCO2  CCO2 ;0 (3)
m ¼ Ymax ,A,ncrop;max ,hyield (1)
  
hL ¼ 1  exp  kI , IPAR  IPAR;0 (4)
    Topt
r Tmax  T T Tmax Topt
hT ¼ ¼ , (2)
Rmax Tmax  Topt Topt hyield ¼ minðhT ; hCO2 ; hL Þ (5)
L. Li et al. / Journal of Cleaner Production 268 (2020) 121928 5

2.2.2. Energy model outdoor environment in  C.


Energy consumption in indoor urban farming is mainly The energy rates referred to the ventilation are calculated by:
composed of two-fold aspects: (1) electricity consumed by the
heating, ventilation and air-conditioning (HVAC) devices, and (2) Qven;in ¼ Hven;in m_ ven;in (8)
electricity consumed by lighting for plant growth.
Qven;out ¼ Hven;out m_ ven;out (9)
2.2.2.1. Energy for the HVAC devices. As depicted in Fig. 3, the en-
ergy consumption model of the HVAC used in vertical farming is
where Hven;in and Hven;out are the specific enthalpy of the fresh air
similar to that of conventional buildings, associated with the real-
and exhaust gas, respectively. m_ ven;in and m_ ven;out are the mass flow
time and complicated parameters that mainly embrace solar posi-
rates of the air supply and exhaust gas, respectively. The enthalpy of
tion and irradiance, dry-bulb temperature, relative humidity (RH),
the wet air is expressed as:
and internal gains from persons and appliances.
Considering the fact that different plants require different pa- 
Hven;out ¼ 1:005Tindoor þ ð2501 þ 1:86Tindoor Þdindoor
rameters of the indoor temperature and humidity, a general and (10)
Hven;in ¼ 1:005Tair þ ð2501 þ 1:86Tair Þdair
simplified model e that is acceptable by various plants e is
developed herein. In this work, we aim to include fogging and air-
where dindoor is the humidity ratio of indoor air. dair is the humidity
con cooling to investigate their performance in the module of
ratio of fresh air in kg/kg (dry air).
vertical farming. A rated cooling capacity of the HVAC unit was
Qvap;gain is composed of two-fold factors: (1) the vapor gain from
determined by the maximum cooling load of the module in an
the fogging device and (2) the vapor gain from the evaporation of
entire year, while the maximum cooling load was calculated by the
the plants, which can be calculated by:
energy conservation of the module of vertical farming. The energy
conservation of the vertical-farming module is expressed as: 
_
Qvap;gain ¼ HV ðm _
eva þ mfog (11)
Qappliance;gain þ Qsolar;gain þ Qperson;gain þ Qven;in þ Qvap;gain
þ Qinfiltration where HV is the specific enthalpy of vapor. m_ eva and m_ fog are the
¼ Qven;out þ Qloss þ Qcooling (6) vapor flow rate of the evaporation of the plants and the fogging
device, respectively. The specific enthalpy of the vapor and
where Qappliance;gain , Qsolar;gain , and Qperson;gain are the energy gains Qinfiltration can be calculated by:
contributed by the appliances, solar radiation, and persons inner
the indoor environment of the module, respectively. Qven;in repre- HV ¼ 2501 þ 1:86Tvap (12)
sents the energy input of the fresh air supplied by the ventilation
device. Qvap;gain is the energy input of the vapor (from both the 0:6Hven;in rair V
fogging device and the evaporation of the plants). Qven;out denotes Qinfiltration ¼ Hven;in m_ infil ¼ (13)
3600
the energy loss of the exhaust gas. Qcooling is the cooling load of the
indoor environment of vertical farming. Qinfiltration refers to the where Tvap is the temperature of vapor. rair is the density of the air
leakage of air through unsealed gaps. in the outdoor environment. V is the indoor volume.
Qloss is defined as the sensible heat rate from the indoor envi- Solar heat gain Qsolar;gain is composed of two aspects:
ronment to the outdoor environment, and can be expressed as:
Qsolar;gain ¼ QF þ QU (14)
Qloss ¼ Uvalue  A  ðTindoor  Tair Þ (7)
with
where Qloss is the heat gained from the outside environment in kW,
Uvalue is the overall heat transfer coefficient in W m2  C 1, A is the
QF ¼ AF UF ðTwall  Tindoor Þ (15)
surface area of greenhouse in m2, Tindoor is the set point of the in-
door air temperature in  C, Tair is the dry-bulb temperature of the

Fig. 3. Schematic of a module of vertical farming compositing of electric power supply, HVAC and fogging system.
6 L. Li et al. / Journal of Cleaner Production 268 (2020) 121928

2.2.3.1. Water. Water consumption mainly relates to the evapora-


QU ¼ AU UU ðTwindow  Tindoor Þ (16) tion of the crop. It is commonly estimated using the Blaney-Criddle
Equation (Brouwer and Heibloem, 1986):
where AF is the surface area of the wall. UF and UU are the overall
heat transfer coefficients of the wall and window, respectively. Twall ETcrop ¼ Kc pð0:46Tmean þ 8Þ (23)
is the outer surface temperature of the wall. AU is the surface area of
windows. Twindow is the surface temperature of the window. where ETcrop is the crop evapotranspiration in mm/day, Kc is the
The wall temperature can be calculated by energy conservation: crop factor specified for each species, Tmean is the mean daily
  temperature ( C), and p is the mean daily percentage of annual
ISRG aAap ¼ QF þ sεF AF FF Twall
4 4
 Tsky þ UF AF ðTwall  Tair Þ daytime hours dependent on latitude. Therefore, the water con-
sumption can be estimated based on crop transpiration and area of
(17)
plantation A as shown below:
where ISRG is solar irradiance. a denotes the absorptivity of the
UH2 O ¼ ETcrop A (24)
outer surface of the wall. Aap is the aperture area of a module of
vertical farming. s is StefaneBoltzmann constant. εF is the thermal
emissivity of the outer surface of the wall. FF is the view factor of
the thermal radiation for the wall. Tsky is the sky temperature, 2.2.3.2. Carbon dioxide. The concentration of CO2 in air (CCO2 ;air ) is
which can be calculated by: about 400 parts per million (ppm) (NASA, 2019). As CO2 is
consumed by photosynthesis, supplemental CO2 may be needed to
1:5
Tsky ¼ 0:0552Tair (18) maintain the CO2 concentration for indoor farming systems (CCO2 ).
The amount of CO2 required for enrichment depends the mass
The window temperature can be calculated by:
flowrate of CO2 leaving the system and the total photoperiod:
 
NU Ns ISRG AU ¼ QU þ sεU AU FU Twindow
4 4
 Tsky þ UU AU ðTwindow m_ ven;out    
UCO2 ¼ CCO2  CCO2 ;air rCO2 tsun þ tlight (25)
 Tair Þ rven;out
(19)
where m_ ven;out is the mass flowrate of vented gas. rven;out and rCO2
are the density of the vented gas and carbon dioxide.
where NU is the glazing factor (solar heat gain coefficient). Ns is the
shading factor. FU is the view factor of the thermal radiation for the
window. 2.2.3.3. Seed. The number of seeds required was calculated based
Finally, for a conventional air-source electric compression on the mass of crop production mðiÞ, the mass of a crop mper head ,
chiller, the hourly electricity consumption is estimated by: and the seed yield yseed (ratio of survival crops to the total seed
sown):
Qcooling
UT ¼ (20) m
COP Useed ¼ (26)
mper head  yseed
where COP is the coefficient of performance of a heat pump,
refrigerator, or an air conditioning system.

2.2.3.4. Fertilizer and pesticide. The amount of fertilizer and pesti-


2.2.2.2. Energy for lighting. Plants are sensitive to the light spec- cide required for crop production was estimated by the average
trum of 400e700 nm, i.e. the photosynthetically active radiation quantity of nutrient and pesticide applied to the field for each
(PAR). IPAR , measured in mmol/m2/s, is the number of photons that hectare of plantation in common practice:
reach a plant for each second, while daily light integral (DLI), X
measured in mol/m2/day, is the cumulative photon flux density Ufertilizer;k , xf ;k;l  mf ;l A,ncrop;max (27)
throughout a day (Benis et al., 2017). The DLI can be met by both k

natural and artificial light as shown below:


Upesticide ¼ munit; pest A,ncrop;max (28)
DLI ¼ IPAR;sun tsun þ IPAR;light tlight (21)
where Ufertilizer;k is the fertilizer consumption. Subscript k denotes
where IPAR;sun and IPAR;light are the number of photons that reach the the different types of fertilizer that provide the necessary nutrients
plant from the sun and artificial light. tsun and tlight are the illumi- for plant growth. xf ;k;l is the mass fraction of nutrient l in fertilizer k.
nation time of sunlight or artificial light, respectively. Lighting from mf ;l is the mass of nutrient l required for each unit hectare of land in
artificial light consumes energy. Energy used for lighting can be each crop cycle. Upesticide is pesticide consumption and munit; pest is
estimated by the number of lights required, the rated power of the mass of fertilizer required for each unit hectare of land in each
lights Plight , and the photoperiod of artificial lighting tlight : crop cycle.

UL ¼ Nlight Plight tlight (22) 2.3. Multi-dimensional sustainability assessment

Based on the mass and energy balance given by Module I, the


material and energy flows around the farm gate can be calculated
2.2.3. Material flow model for the assessment of the economic and environmental assessment.
Apart from energy consumption, vegetable production also re- The net present value (NPV), return on investment (ROI) and
quires material inputs such as water, CO2, seed, fertilizer, and payback period were used as the economic indicators, while the
pesticide, which will be discussed in this section. CO2 emission, water consumption, and land occupation were
L. Li et al. / Journal of Cleaner Production 268 (2020) 121928 7

evaluated as the environmental indicators. As shown in Fig. 4, an The equations for the economic and environmental assessment are
agricultural product goes through the production, packing and provided in the following subsections.
product treatment, transport and distribution, retail, consumer, and
recycle or disposal processes throughout its life cycle. Considering
from a farm operator’s perspective, the economic and environ- 2.3.1. Economic indicators
mental indicators were evaluated at a cradle-to-gate level, i.e., The net present value NPV is a cumulative sum of the discounted
counting the economic and environmental burdens from resource cash flow in each year t, which depends on the revenue Revenuet ,
extraction (cradle) to the end of the vegetable production process operating expenditure OPEXt , and the capital expenditure CAPEXt of
(gate). For the economic assessment, the capital investment for the year and the interest rate ir:
construction and equipment as well as the operating cost of energy
and material utilization, labor, and land were considered. For the ,
X
N
environmental assessment, the environmental burdens generated NPV ¼ CFt ð1 þ irÞt (29)
from HVAC, lighting, CO2 enrichment, irrigation, pest control, fer- t¼0
tilizer application, and seeding processes were taken into account.

Fig. 4. Scope of the multi-dimensional sustainability assessment.


8 L. Li et al. / Journal of Cleaner Production 268 (2020) 121928

estimated based on inlet energy and material utilization and their


CFt ¼ Revenuet  OPEXt  CAPEXt (30) background emission or abatement per service unit.
X X 
Revenuet ¼ pcrop;i;t mt;i (31) Water consumption ¼ UH2 O þ uiH2 O;j * Uj (37)
i j

X
OPEXt ¼ FCt þ Cj (32) X 
Land occupation ¼ Uland þ uiland;j * Uj (38)
j
j

FC ¼ Labourt þ Rentt (33)  


CO2 emission ¼ UCO2  UCO2 fix
X 0 1
Cj ¼ pj Uj (34) X X
i þ @ uiCO2 ;j * Uj  uiCO2 fix;j * Uj A (39)
j j
where CFt is the cash flow in year t, pcrop is the unit selling price of
crop i in year t, mt;i is the mass of production of crop i in year t. FCt , where Uj refers to material and energy consumption as mentioned
Labourt , and Rentt are the fixed cost, labor cost, and land rental cost in the economic analysis. Specifically, UH2 O , Uland , UCO2 , and UCO2 fix
in year t. Cj , pj , and Uj are the cost, unit cost, and the amount of are the water consumption, land area occupation, CO2 emission,
energy and material consumption (j ¼ electricity, H2O, CO2, seed, and CO2 fixation of the within the foreground farm system. uiH2 O;j ,
fertilizer, pesticide). uiland;j , uiCO2 ;j , and uiCO2 fix;j are the water consumption, and land
While NPV shows how much economic benefit can be gained occupation, CO2 emission, and CO2 emission fixation for the pro-
from a project, return on investment (ROI) and payback period are duction of each unit of energy or material inlet.
other two standard metrics that indicate the investment perfor-
mance of a project from other aspects (Phillips et al., 2014). ROI is a
calculation of the net project benefit divided by investment. In this
study, short-term ROI, calculated by the annual net project benefit 2.4. Optimization model
divided by investment, was used to show the speed of return on
investment (PVsell, 2019). On the other hand, payback time, the Optimization was the last step and the highest level of the de-
inverse of short-term ROI, was used to depict how long it takes for cision support framework. It takes in the output from the previous
an investment to pay for itself. steps discussed in Section 2.1-2.3. As shown in Eqn. (40)e(43), the
optimization can be carried out to optimize either the economic
Revenuet  Costt profit NPV or the environmental indicators, or optimize the both
ROI ¼ (35) economic and environmental objectives at the same time. The
Investment
economic and environmental analysis are integrated by the over-
Investment arching framework in the former case, while the two objectives are
Payback time ¼ (36) integrated by optimization in the latter option (Miah et al., 2017).
Revenuet  Costt
The searching for the optimal solutions was done by varying the
decision variables: (1) growing area Ai allocated to crop i, (2)
temperature Ti, (3) humidity RHi, (4) light intensity IPAR,i, and (5)
CO2 concentration CCO2 ;i of the growing environment if controlled
2.3.2. Environmental indicators environment agriculture was used. When the growing environ-
As aforementioned, water and land resource consumption and ment cannot be controlled, the only decision variable would be the
climate change effect are major concerns in food production. The growing area Ai allocated to crop i.
cradle-to-gate water consumption, land occupation, and CO2 It was assumed that all the vegetables produced can be sold as
emission were assessed as the environmental indicators in this case long as the production does not exceed the demand (Eqn. (44)).
study. The equations for these specific environmental indicators are Additionally, the number of crops that could be planted was limited
shown in Eqn. (37)-(39). The cradle-to-gate water consumption by the available planting area (Eqn. (45)). Therefore, the optimi-
was calculated as a sum of the water utilized in the farming process zation model was mathematically formulated as follows:
and the water consumption incurred by the production of the en-
ergy and material flows delivered to the farm gate. The latter was a X
N X .
product sum of water consumed for per service unit of input energy max NPV ¼ CFt;i ð1 þ irÞt (40)
or material and the amount of energy or material required for farm t¼0 i
operation. Similarly, the cradle-to-gate land occupation was esti-
mated by the adding the land area utilized by the farm itself Uland and/or min.
and the land area utilized to produce the material and energy flows X 
delivered to the farm gate. On the other hand, the cradle-to-gate Water consumption ¼ UH2 O þ uiH2 O;j * Uj (41)
CO2 emission was estimated as a net effect of the CO2 emission j
and the emission abatement achieved via biogenic fixation. CO2
enrichment was identified as the major source of carbon emission. X 
Land occupation ¼ Uland þ uiland;j * Uj (42)
The CO2 fixation in crop production was estimated based on the
j
mass and carbon content of the vegetables. The CO2 emission and
fixation introduced by the production of the farm inflows were
L. Li et al. / Journal of Cleaner Production 268 (2020) 121928 9

  framework would return the economically optimum crop choice


CO2 emission ¼ UCO2  UCO2 fix and operation condition and evaluation on multiple sustainability
0 1 indices. However, various factors including the type of farming
X X system, structural design of the farm, the source of energy supply,
þ @ uiCO2 ;j * Uj  uiCO2 fix;j * Uj A (43)
and the type of fertilizer could influence the economic and envi-
j j
ronmental performance of the system. Using the proposed
decision-support framework, different system configurations can
s.t.
be assessed and optimized in a standard manner. Scenario analysis
mt;i  Demandt;i (44) can be easily carried out to identify the most economically and
environmentally favored system design among all constructed
X scenarios. In this study, six different scenarios were constructed for
Ai  TA (45) the scenario analysis, as shown in Table 1.
i
The base case (BC) was a plant factory built from standard
containers with no windows. In this scenario, a two-story layout
where subscript i stands for the type of crops, subscript j refers to
was adopted, while electricity was provided from the grid, and
the material and flows, t represents the year of operation, mt;i is the
nutrient supplied by chemical fertilizer. Glass-enclosed greenhouse
mass of crop i produced in year t, Demandt;i is the market demand
(GEG) represented a glass-enclosed greenhouse with electricity
for crop i in year t, Ai is the planting area allocated to crop i and TA
provided from the grid and nutrient supplied by chemical fertil-
refers to the total available planting area.
izers. Open field (OF) represented the conventional farming prac-
tice where the vegetables are planted horizontally on the ground
2.5. Implementation without environmental control. Different from the base case, the
renewable energy (RE) scenario was powered by solar PV instead of
Finally, the modeling and optimization framework was imple- the grid. On the other hand, the waste valorization (WV) scenario
mented in the General Algebraic Modeling System (GAMS) soft- was different from the base case by the type of fertilizer used.
ware. For each scenario, the mixed integer nonlinear programming Instead of using chemical fertilizer, this scenario used organic fer-
(MINLP) problem was solved using the BARON solver, with solving tilizer derived from beer residue. The crops produced using this
time within 0.016 s. organic fertilizer could be sold at a higher price as organic products
since no chemical fertilizers, pesticides, growth regulators, bioen-
3. Case study gineering methods, or ionizing radiation was used during produc-
tion. Finally, as a two-pronged approach, renewable energy & waste
To demonstrate the proposed assessment and optimization valorization (REWF) was a two-level plant factory with electricity
framework, we carried out a case study on a vertical farm to be built supplied via solar PV and nutrient supply by beer-residue-derived
in Kranji, Singapore. The case study was based in Singapore since it fertilizer. The properties of the artificial lighting equipment and
is a suitable place to testbed urban farming. On the one hand, the fertilizers considered in this scenario analysis are provided in
BAU farming practice cannot be widely deployed since Singapore is Tables S4 and S5 of the Supplementary Information.
a highly urbanized country with 5.6 million populations but only Till this point, the information of the infrastructure of the farm,
0.8% of arable land. On the other hand, the Singapore government the meteorology condition of the surrounding environment, and
aims to become the urban agriculture and aquaculture technology the choice of crops for the scenario analysis have been set up. The
hub in the region, and has been investing more to boost the agri- corresponding parameters are also provided in the Supplementary
cultural sector for innovative technologies (Tan, 2019). The ultimate Information as aforementioned. In addition to these parameters,
goal is to reduce the reliance on food imports, provide seasonal the market data and environmental data used to calculate the NPV
production, and at the same time reduce the carbon footprint. and the cradle-to-gate inventories are required in the multi-
Located just one degree north of the equator, Singapore has dimensional assessment. The unit investment cost for the infra-
abundant rainfall, high humidity, hot and uniform temperature, structure and equipment as well as the costs related to farm
and long hours of sunlight. These climate variables do not have operation can be found in Table S6 and Table S7 in the Supple-
large variations throughout the year and the annual averages of the mentary Information. The unit price of seeds, selling price of crops,
meteorology parameters are summarized in Table S1 in the Sup- and the unit cost for fertilizers can be found in Table S8 and Table S9
plementary Information. in the Supplementary Information. The amount of water con-
The farm has a footprint of 500 m2 with a dimension of sumption, and land occupation, CO2 emission, and CO2 emission
25 m  50 m. Due to the scare land area, a multi-tier vertical fixation for the production of each unit of energy or material inlet
farming structure becomes more space-efficient. It was assumed were obtained from the Ecoinvent database version 3.5 (Wernet
that the walls were insulted by heat resistant material to reduce et al., 2016).
heat loss and solar heat gain. Key information of the infrastructural
settings of the farm is available from Table S3 of the Supplementary 4. Results and discussion
Information. In Singapore, more than 80% of the leafy vegetables
are produced from soil cultivation (AVA, 2017). The studied farm As a demonstration of the proposed framework, the net present
also adopted a soil-based vertical farming system with soil recir- value was maximized in the optimization. The optimally selected
culation. We defined lettuce, spinach, cabbage, broccoli, Cai Xin, Nai crops and their cultivation conditions returned by the optimization
Bai, Xian Cai, and Kang Kong, a mixture of western and eastern are summarized for different scenarios as shown in Table 2.
vegetables, as alternative candidates to be grown in the farm as Detailed information about the farm-wide energy and material
they are the most popular vegetables consumed in Singapore and consumption of these scenarios are also provided in the Supple-
suitable to be produced in various urban farming systems. The mentary Information (Part E).
major properties of the different vegetables can be found in The optimization returns Cai Xin and Xian Cai as the optimum
Table S2 in the Supplementary Information. crops to be grown for all scenarios. The OF scenario can be used as a
With a specific farm design, the proposed decision-support benchmark for comparison as it shows the temperature, relative
10 L. Li et al. / Journal of Cleaner Production 268 (2020) 121928

Table 1
Scenarios under investigation.

Scenarios Structure Energy source Fertilizer type

Greenhouse Plant factory Open field Grid Solar PV Chemical fertilizer Beer-residue-derived fertilizer

Base case (BC) C C C


Glass-enclosed greenhouse (GEG) C C C
Open field (OF) C C C
Renewable energy (RE) C C C
Waste valorization (WV) C C C
Renewable energy & waste valorization (REWV) C C C

Table 2
Optimal operating conditions for different scenarios.

Cai Xin Xian Cai

X T RH CO2 PAR hyield x T RH CO2 PAR hyield


Num. of rooms 
( C) e (ppm) (mmol/m /s)
2
(0e1) Num. of rooms 
( C) e (ppm) (mmol/m /s)
2
(0e1)

BC 20 28 80.3% 503 619 0.77 20 28 80.3% 698 857 0.88


GEG 20 28 80.3% 618 0 0.84 20 28 80.3% 834 1024 0.92
OF 250a 28 80.0% 400 0 0.68 250a 28 80.0% 400 0 0.68
RE 20 28 80.3% 503 619 0.77 20 28 80.3% 698 857 0.88
WV 20 28 80.3% 503 619 0.77 20 28 80.3% 698 857 0.88
REWV 20 28 80.3% 503 619 0.77 20 28 80.3% 698 857 0.88
a
Unit in m.2.

the efficiency of photosynthesis is 87% of the maximum when the


temperature and CO2 concentration are maintained around the
ones for optimum yield. Under the atmospheric CO2 concentration,
the efficiency of photosynthesis is 68% of the optimum when the
temperature and lighting are maintained around the ones for op-
timum yield. Therefore, CO2 concentration is the limiting factor that
affects the overall efficiency of photosynthesis for the OF scenario.
For the plant factories used in Scenario BC, RE, WV, and REWV, no
natural lighting is available as this system was designed with no
window to minimize heat transfer through windows.
For optimized NPV, the CO2 concentration is enriched to
503 ppm with artificial lighting supplied at a PAR of 619 mmol/m2/s
for Cai Xin and 698 ppm and 857 mmol/m2/s for Xian Cai. At the
optimum set of operating conditions, the efficiency of crop pro-
duction for Cain Xin and Xian Cai is 0.77 and 0.88, respectively.
Natural lighting with 824 mmol/m2/s natural PAR is available for the
GEG scenario. For Cai Xin, no artificial lighting is required for the
optimal operation. With an enrichment of the CO2 concentration to
618 ppm, the optimum NPV is obtained for this scenario and the
overall photosynthetic efficiency is as high as 0.84. For Xian Cai,
Fig. 5. The cumulative cash flow of different scenarios. extra lighting is provided by artificial lighting at a PAR of
1024 mmol/m2/s in addition to the natural lighting for 8 h. The CO2
concentration is enriched to 834 ppm and the overall photosyn-
humidity (RH), CO2 concentration, PAR, and photosynthetic effi- thesis efficiency is as high as 0.92. With these sets of operating
ciency when the crops are cultivated in the outside environment. conditions, the efficiency of photosynthesis for each type of crop in
The optimum temperature for all scenarios for both of the crops is GEG turns out to be the highest among all scenarios. Moreover,
the same as the one assumed for the outside environment (28  C) with a higher photosynthesis efficiency and a higher unit area yield,
and several degrees higher than the temperature that optimizes the Xian Cai yields more mass production than Cai Xin. Under the
yield of each type of crop, which is 25  C for Cai Xin and 27  C for optimized operating conditions, the annual production is 21.5 tons/
Xian Cai. This temperature is maintained as it saves cooling demand year (±0.9%) for Cai Xin and 54.5 tons/year (±0.4%) for Xian Cai for
without causing significant discounts in the rate of photosynthesis the considered scenarios. This may make it the most economically
for the crops. The optimum humidity of the indoor farming facil- preferable choice for production. However, not all the space is
ities is 80.3%, which is slightly higher than the RH of the outside allocated for Xian Cai due to the constraint of market demand.
environment. No extra humidification from the fogging system is With the optimized operating conditions, the evaluation of the
required and the 0.3% rise in RH in the indoor environment is corresponding economic and environmental performance for the
caused by the transpiration of plants. The CO2 concentration of the six scenarios was returned by the sustainability assessment module
outside environment is 400 ppm and the PAR outdoors is 824 mmol/ of the proposed framework. In terms of economic performance,
m2/s. The PAR of for both of the crops is zero as it was assumed that several economic indexes were used, including NPV, ROI, and
no artificial lighting was provided to OF. Under the natural lighting, payback time. A cumulative NPV curve is presented in Fig. 5. The
L. Li et al. / Journal of Cleaner Production 268 (2020) 121928 11

Table 3
Key economic indexes of different scenarios.

BC GEG OF RE WV REWV

ROI 72% 33% 916.79% 71% 136% 131%


Payback period (y) 1.39 3.03 0.11 1.41 0.73 0.76
NPV (USD) 3,334,928 2,705,979 3,001,990 3,524,805 6,576,471 6,766,242

supply and beer-residue-derived fertilizer (WV), plant factory with


PV and chemical fertilizer (RE), plant factory utilizing grid elec-
tricity and chemical fertilizer (BC), and open field (OF). On the
contrary, glass-enclosed greenhouse (GEG) is the least profitable
option. The lowest initial investment is found in OF (20,160 USD)
and the highest found in GEG (614,171 USD). The annual investment
for the plant factories is moderate with a value of 308,840 USD for
plant factories without solar PV installation (BC and WV) and a
value of 332,026 USD for the ones with solar PVs (RE and REWV).
The slopes show that WV and REWV gain a larger annual profit
compared with other scenarios. OF has the smallest annual cash
flow, followed by GEG, BC, and RE. In terms of return on investment,
the lower bound for most vegetables is 40% while the higher bound
can be as high as 300% for some crops (Cordero, 2018). OF has a high
ROI as it has high revenue with little initial investment. As OF has
the fastest return on investment, it has the shortest payback time
and the investment can be recovered within 2 months. Installing
solar PV slightly prolongs the payback period, but it brings more
total benefit than purchasing electricity from the grid. Using a
waste valorization approach for fertilizer supply increases the re-
turn of investment and shortens the payback period. The invest-
ment can be paid back in 8e9 months when beer-residue-derived
fertilizer is used to produce vegetables. The plant factory structure
turns out to be a more profitable option than a glass-enclosed
greenhouse structure and open field. While glass greenhouse may
need 3 years for recovery, using the plant factory may only need
less than 1.5 years.
The high profitability of the waste valorization scenarios (WV
and REWV) mainly attributes to the higher benefit generated by the
sales of organic products. Because in the parameters input step, it
was assumed that the products cultivated without the use of
chemical fertilizers and pesticides sold at the selling price for
organic products, which is higher than the normal selling price for
products grown with chemical fertilizer and pesticide inputs. The
finding suggests that food waste recycling as fertilizer in farms is
not only a measure to reduce urban food waste but also a promising
strategy to boost the profit of the business. Moreover, the use of
renewable solar energy and the plant factory structure could bring
more economic benefit than open-field agriculture in the long run
(more than 8 years in this case). On the other hand, the greenhouse
structure is not favorable for investment for a tropical climate.
As for the environmental performance, the cradle-to-gate water
consumption, CO2 emission, and land occupation for various sce-
narios were evaluated. The results are shown in Fig. 6. Stacked bar
Fig. 6. Water consumption, CO2 emission, and land occupation of different scenarios.
charts were used to present the total environmental footprint and a
further breakdown of the impact caused by the vegetable produc-
key economic indexes of different scenarios are also given in tion process itself and the impact caused by the upstream processes
Table 3. including electricity production, water production, fertilizer pro-
In Fig. 5, the intercept at year 0 reflects the investment cost for duction, CO2 production, and seed production.
the project, the intercept at year 20 is the total profit gained According to Fig. 6, vegetable production accounts for the major
throughout 20 years of operation, the slope of the NPV curves in- water consumption for most of the scenarios, followed by elec-
dicates the annual cash flow, and the intercept with the x-axis tricity generation. RE and REWV have less water consumption than
represents the payback time. Comparing the six scenarios, with a other scenarios since they utilize solar energy which requires less
project-lifespan of 20 years, plant factory with solar PV and beer- water compared to grid electricity supplied from the natural gas
residue-derived fertilizer (REWV) generates the largest profit based power plant. OF also has less cradle-to-gate water con-
among the six scenarios, followed by plant factory with grid power sumption even though it requires more water in the crop
12 L. Li et al. / Journal of Cleaner Production 268 (2020) 121928

Table 4
Optimal economic and environmental performance of different scenarios based on the functional unit of 1 kg of crop production.

Scenario NPV (USD/kg) Ranking Water consumption (m3/kg) Ranking CO2 emission(kg CO2-eq/kg) Ranking Land occupation (m2/kg) Ranking

BC 3.07 5 0.016 2 1.44 5 0.014 2


GEG 2.45 6 0.022 3 2.82 6 0.016 3
OF 3.51 3 0.013 1 0.13 1 0.016 3
RE 3.25 4 0.013 1 0.22 3 0.040 5
WV 6.07 2 0.016 2 1.43 4 0.013 1
REWV 6.24 1 0.013 1 0.21 2 0.038 4

production process compared with the others. This is because the As a summary of the economic and environmental evaluations, a
increase of water use in irrigation is compensated by the reduced ranking of the scenarios based on NPV, water consumption, CO2
contribution from energy consumption as it does not require any emission, and land occupation is shown in Table 4 (1 indicate the
artificial lighting and cooling. Contrarily, GEG has the highest highest preference while 6 is the least preferred). A demonstration
cradle-to-gate water consumption mainly due to its high electricity of the breakdown of the key contributors that affect the economic
consumption caused by the high cooling demand. This high cooling and environmental performance is shown in Fig. 7, using the most
demand can be traced back to the high solar heat gain or green- economic scenario (REWV) as an example.
house effect of this type of glass structure. REWV has the best economic performance among all the sce-
CO2 emission is dominantly affected by upstream electricity narios. Its water consumption and carbon emission are also within
production. Open field has the least CO2 emission - a net negative the top two lowest. However, it has a large land occupation caused
value, since it has no electricity consumption while the atmo- by the installation of the solar PV system. As one possible solution
spheric CO2 is fixed by photosynthesis during vegetable production. inspired by the concept of vertical farming, a vertical layout of solar
However, the net CO2 emission of the indoor scenarios are still panels can also be considered to deal with the problem of ineffi-
positive since the carbon emissions from energy consumption in cient land use for solar power utilization. Following REWV, WV has
these scenarios are too large to be offset. GEG has the largest CO2 the second largest NPV, second smallest water consumption, and
emission among all since it has a large cooling demand as afore- minimum land occupation. However, its CO2 emission is not the
mentioned. Comparing BC, RE, and WV, it shows that replacing the smallest, which is larger than the open-field agriculture and plant
fossil-fuel energy source by renewable energy significantly helps to factory with renewable integration. Open-field farming ranks the
reduce the CO2 emission but replacing chemical fertilizer by waste- 3rd in cost efficiency. Its water consumption and CO2 emission are
derived fertilizer may not have a significant effect. also the smallest. However, it is not as flexible and space-efficient as
Land occupation is mainly raised by the vegetable production plant factories which can be easily stacked up to increase unit area
activity and sometimes due to the utilization of solar energy. RE and production yield. The remaining options are not as recommended
REWV have the largest land occupation as a large space is required as the top three since they are less attractive in economic and
to install solar panels to meet the electricity demand. Compared environmental sustainability. Particularly, a greenhouse structure is
with BC, WV has a smaller land occupation since producing fertil- not recommended for farming in Singapore and other tropical cities
izer from food waste has less land occupation than the production since a lot of heat from solar radiation will be trapped in the
of chemical fertilizer. Comparing BC, GEG, and OF, it seems that greenhouse, which is not suitable for the cultivation of most of the
varying the structure of the farm does not have a large impact on crops and would consume a lot of energy if cooling is applied.
the land footprint required to produce each unit of the crop. When considering crop cultivation in tropical weather, although
However, the modular plant factory is more flexible than GEG and indoor farming systems can overcome the technical obstacles in
OF and several modules can be stacked up to increase the yield of planting seasonal crops in the tropical area, it may not be an
each unit of land in times. economically or environmentally sustainable choice. The most
Overall, the result indicates that replacing the energy source basic scenario BC will be used as an example for illustration. The
from fossil-fuel based grid electricity by renewable solar energy NPV for the production in the BC-type plant factory of the eight
will help reduce water consumption and carbon emission. How- types of crops considered in this study is shown in Fig. 8. The OF
ever, it has a disadvantage of increased land occupation. Consid- scenario was also included as a benchmark. Lettuce, broccoli,
ering the pros and cons of indoor and outdoor farming, indoor spinach, and cabbage are typical cool-climate crops. Therefore, as
farming has less water consumption, CO2 emission, and land shown by the OF curve in Fig. 8, it is not generating any profit when
occupation than the open field for onsite vegetable production. they are planted outdoors in such a hot climate as Singapore.
However, the total water consumption and carbon emission for the However, as indicated by the solid curve, it is not economically
open-field agriculture are less than an indoor design taking into viable to do it using the plant factory option, either. Only planting
account the impact of the production of input material upstream. Asian crops Cai Xin, Xian Cai, and Nai Bai are economically viable. It
Further comparing the two types of indoor farming structure, the is noted that Kang Kong, which is also an Asian crop, does not show
plant factory seems to be a better indoor farming option than the apparent economic competitiveness. This owes to its low unit
glass-enclosed greenhouse for a tropical climate. By good insulation production yield and relatively cheap selling price.
of walls and elimination of windows, a plant factory receives less Since the current BC scenario uses a two-story layout, it is un-
heat gain from solar radiation and result in less cooling demand for certain whether the number of stacked stories will affect the situ-
crop cultivation in hot weather. Although artificial lighting is ation. As a follow-up discussion, the single-story and three-story
necessary for a plant factory, the plant factory option still consumes scenarios are evaluated. By comparing the BC scenarios, it can be
less total energy than a greenhouse owing to the saving in cooling. found that adding more levels by stacking up more plant-factory
Lastly, replacing chemical fertilizers by beer-residue-derived fer- modules will increase the NPV of planting Xian Cai, Cai Xin, Nai
tilizer does not seem to have a significant impact on water con- Bai, and lettuce, while further decreasing the profitability of Broc-
sumption and CO2 emission, but it incurs less land occupation and coli, Spinach, Kang Kong, and Cabbage. It is worth noting that the
brings greater profit potential as aforementioned. plantation of lettuce in Singapore can be changed from an
L. Li et al. / Journal of Cleaner Production 268 (2020) 121928 13

Fig. 7. Breakdown of the annual cash flow, water consumption, CO2 emission, and land occupation for REWV.

provide convenient decision support for the stakeholders to design


economically viable farming systems and assess their environ-
mental performance. The case study demonstrated the applicability
of the proposed framework in evaluating different farming system
configurations and finding out the best system design by standard
comparison. The sustainability in the economic and environmental
performance of the different scenarios was quantified using net
present value, water consumption, CO2 emission, and land occu-
pation indexes. The optimum crop choice and the growing condi-
tions of a Singapore-based farming system for six different
scenarios were identified using the proposed framework.
It was found that Cai Xin and Xian Cai are the most suitable and
economically attractive crops to be planted in Singapore. The most
profitable system design for a vertical farm in Singapore among all
the studied alternatives was REWV e the plant factory design with
no window, complete artificial lighting, PV system, and beer-
residue-derived fertilizer. This design also has the smallest water
Fig. 8. Net present value of an open field and multi-story plant factories for the eight
consumption and moderate CO2 emission. Besides, the results also
types of crops. bring some implications in pursuing sustainable urban farming in
general tropical cities. Firstly, the greenhouse structure is not rec-
ommended for farming in Singapore and other tropical cities since
uneconomic production to a profitable option by adding the a lot of heat from solar radiation will be trapped in the greenhouse,
number of vertical layers of the farm. which is not suitable for the cultivation of most of the crops and
would consume a lot of energy if cooling is applied. On the other
5. Conclusion hand, using a plant factory modular design as the farm structure
may bring more net present value and occupies less land than an
In this work, we proposed a framework for the simulation, open field. Secondly, using renewable solar PV systems for
assessment, and optimization of urban farming systems in order to
14 L. Li et al. / Journal of Cleaner Production 268 (2020) 121928

electricity supply has a longer payback time, but it may bring more Boote, K.J., Loomis, R.S., Crop Science Society of America Division C-2, American
Society of Agronomy Division A-3, 1991. Modeling Crop Photosynthesis - from
net economic benefit than using electricity generated from fossil
Biochemistry to Canopy. Crop Science Society of America.
fuels in the places where solar radiation is abundant, which is Brouwer, C., Heibloem, M., 1986. Chapter 3: crop water needs. Irrig. Water Manag.
usually the case for tropical cities. Using solar PV also has a lower Irrig. Water Needs.
water consumption and CO2 emission than using electricity from Brulard, N., Cung, V.D., Catusse, N., Dutrieux, C., 2018. An integrated sizing and
planning problem in designing diverse vegetable farming systems. Int. J. Prod.
the grid. Although a solar PV system may require a large space for Res. 7543 https://doi.org/10.1080/00207543.2018.1498985.
receiving solar radiation, a vertical layout of the PV panels like the Capitanescu, F., Marvuglia, A., Navarrete Gutie rrez, T., Benetto, E., 2017. Multi-stage
vertical farm can be considered as a possible solution to reduce the farm management optimization under environmental and crop rotation con-
straints. J. Clean. Prod. 147, 197e205. https://doi.org/10.1016/
land area required. Thirdly, using beer-residue-derived fertilizer is a j.jclepro.2017.01.076.
measure of waste valorization that practices the circular economy. Cordero, T., 2018. Crowdfunding: a new way of investing in agribusiness | Money |
It brings more net present value and results in a shorter payback GMA News. Online [WWW Document]. https://www.gmanetwork.com/news/
money/personalfinance/641371/crowdfunding-a-new-way-of-investing-in-
period compared with using chemical fertilizers, meanwhile agribusiness/story/ (accessed 2.7.20).
reducing wastes in cities, and may have a smaller land occupation Erwin, J.E., Erwin, J., Gesick, E., 2017. Photosynthetic Responses of Swiss Chard , Kale
than using chemical fertilizers. , and Spinach Cultivars to Irradiance and Carbon Dioxide Concentration. https://
doi.org/10.21273/HORTSCI11799-17.
On the whole, the proposed framework can be a convenient Fao, 2009. Global Agriculture towards 2050, High Level Expert Forum-How to Feed
quantitative tool for the design, assessment, and comparison of the the World 2050 https://doi.org/http://www.fao.org/fileadmin/templates/wsfs/
urban farming systems. Based on the current multi-dimensional docs/Issues_papers/HLEF2050_Global_Agriculture.pdf.
Fao, 2017. The Future of Food and Agriculture: Trends and Challenges.
sustainability assessment and optimization, this framework can
Fao, 2018. The Future of Food and Agriculture: Alternative Pathways to 2050. Roam.
be further empowered by the internet of things, machine learning https://doi.org/10.4161/chan.4.6.12871.
methods, and automatic control systems to truly realize precision Fao, 2020. Urban Agriculture [WWW Document]. http://www.fao.org/urban-
agriculture for higher profitability and cleaner production. agriculture/en/ (accessed 2.9.20).
Goldstein, B., Hauschild, M., Ferna ndez, J., Birkved, M., 2016a. Testing the environ-
mental performance of urban agriculture as a food supply in northern climates.
Declaration of competing interest J. Clean. Prod. 135, 984e994. https://doi.org/10.1016/j.jclepro.2016.07.004.
Goldstein, B., Hauschild, M., Ferna ndez, J., Birkved, M., 2016b. Urban versus con-
ventional agriculture, taxonomy of resource profiles: a review. Agron. Sustain.
The authors declare that they have no known competing Dev. 36, 1e19. https://doi.org/10.1007/s13593-015-0348-4.
financial interests or personal relationships that could have Johnson, A.J., Elliot, M., de la Parra, J., Timothy, L.S., Miikkulainen, R., B. Harper, C.,
appeared to influence the work reported in this paper. 2018. Flavor-Cyber-Agriculture: Optimization of Plant Metabolites in an Open-
Source Control Environment through Surrogate Modeling. https://doi.org/
10.1371/journal.pone.0213918.
CRediT authorship contribution statement Kamilaris, A., Prenafeta-Boldú, F.X., 2018. Deep learning in agriculture: a survey.
Comput. Electron. Agric. 147, 70e90. https://doi.org/10.1016/
j.compag.2018.02.016.
Lanyu Li: Conceptualization, Methodology, Software, Valida- Kee, A.N., 2019. SingaporeeThe Next Great Place for Urban Agriculture [WWW
tion, Investigation, Writing - original draft, Writing - review & Document]. https://www.enterprisesg.gov.sg/blog/agritech/singapore-the-
editing, Visualization. Xian Li: Methodology, Software, Validation, next-great-place-for-urban-agriculture?utm_source¼website&utm_
medium¼blog&utm_campaign¼IndoorþAg-Conþ2019 (accessed 2.2.20).
Writing - original draft, Writing - review & editing, Visualization. Kozai, T., Niu, G., Takagaki, M., 2019. Plant Factory: an Indoor Vertical Farming
Clive Chong: Conceptualization, Validation, Resources. Chi-Hwa System for Efficient Quality Food Production. Academic press.
Wang: Conceptualization, Supervision, Funding acquisition. Lawlor, D.W., 2001. Photosynthesis. Bios.
Liaros, S., Botsis, K., Xydis, G., 2016. Technoeconomic evaluation of urban plant
Xiaonan Wang: Conceptualization, Methodology, Writing - review factories: the case of basil (Ocimum basilicum). Sci. Total Environ. 554e555,
& editing, Supervision, Funding acquisition. 218e227. https://doi.org/10.1016/j.scitotenv.2016.02.174.
Maucieri, C., Forchino, A.A., Nicoletto, C., Junge, R., Pastres, R., Sambo, P., Borin, M.,
2018. Life cycle assessment of a micro aquaponic system for educational pur-
Acknowledgment poses built using recovered material. J. Clean. Prod. 172, 3119e3127. https://
doi.org/10.1016/j.jclepro.2017.11.097.
This work was supported by the National Research Foundation, McBratney, A., Whelan, B., Ancev, T., Bouma, J., 2005. Future directions of precision
agriculture. Precis. Agric. 6, 7e23.
Prime Minister’s Office, Singapore under its Campus for Research Miah, J.H., Koh, S.C.L., Stone, D., 2017. A hybridised framework combining integrated
Excellence and Technological Enterprise (CREATE) program (Grant methods for environmental Life Cycle Assessment and Life Cycle Costing.
Number R-706-000-103-281 and R-706-001-102-281) and the J. Clean. Prod. 168, 846e866.
Muangprathub, J., Boonnam, N., Kajornkasirat, S., Lekbangpong, N.,
Singapore RIE2020 Advanced Manufacturing and Engineering
Wanichsombat, A., Nillaor, P., 2019. IoT and agriculture data analysis for smart
(AME) IAF-PP grant “Cyber-physical production system (CPPS) to- farm. Comput. Electron. Agric. 156, 467e474. https://doi.org/10.1016/
wards contextual and intelligent response” by the Agency for Sci- j.compag.2018.12.011.
Nasa, 2019. Carbon Dioxide | Vital Signs e Climate Change: Vital Signs of the Planet
ence, Technology and Research under Grant No. A19C1a0018.
[WWW Document]. https://climate.nasa.gov/vital-signs/carbon-dioxide/
(accessed 7.30.19).
Appendix A. Supplementary data Oliver-Sola , J., Gianquinto, G., Montero, J.I., Sanye -Mengual, E., Rieradevall, J.,
Orsini, F., 2015. Techniques and crops for efficient rooftop gardens in Bologna,
Italy. Agron. Sustain. Dev. 35, 1477e1488. https://doi.org/10.1007/s13593-015-
Supplementary data to this article can be found online at 0331-0.
https://doi.org/10.1016/j.jclepro.2020.121928. Phillips, J.J., Phillips, P., Pulliam, A., 2014. The ROI methodology: a tool to measure
and improve. In: Measuring ROI in Environment, Health, and Safety. Wiley
Online Books, pp. 73e101. https://doi.org/10.1002/9781118899281.ch4.
References Piezer, K., Petit-boix, A., Sanjuan-delm as, D., Briese, E., Celik, I., Rieradevall, J.,
Gabarrell, X., Josa, A., Apul, D., 2019. Science of the Total Environment Ecological
Aga Webmaster, 2018. More Fuel for the Food/Feed Debate [WWW Document]. network analysis of growing tomatoes in an urban rooftop greenhouse. Sci.
http://www.fao.org/ag/againfo/home/en/news_archive/2017_More_Fuel_for_ Total Environ. 651, 1495e1504. https://doi.org/10.1016/j.scitotenv.2018.09.293.
the_Food_Feed.html (accessed 3.29.20). PVsell, 2019. What does ROI/IRR/NPV/LCOE/Payback mean? [WWW Document].
Ava, 2017. Farming in Singapore. Agri-Food Vet. Auth. Singapore [WWW Docu- https://pvsell.zendesk.com/hc/en-us/articles/223284248-What-does-ROI-IRR-
ment]. https://www.ava.gov.sg/explore-by-sections/farms/land-farms/farming- NPV-LCOE-Payback-mean-#roi-means-return-on-investment (accessed 2.6.20).
in-singapore (accessed 3.28.19). RaheliNamin, B., Mortazavi, S., Salmanmahiny, A., 2016. Optimizing cultivation of
~o, P., 2017. Development of a simulation-based decision
Benis, K., Reinhart, C., Ferra agricultural products using socio-economic and environmental scenarios. En-
support workflow for the implementation of Building-Integrated Agriculture viron. Monit. Assess. 188 https://doi.org/10.1007/s10661-016-5599-2.
(BIA) in urban contexts. J. Clean. Prod. 147, 589e602. https://doi.org/10.1016/ Sanjuan-Delma s, D., Llorach-Massana, P., Nadal, A., Ercilla-Montserrat, M.,
j.jclepro.2017.01.130. Mun ~ oz, P., Montero, J.I., Josa, A., Gabarrell, X., Rieradevall, J., 2018.
L. Li et al. / Journal of Cleaner Production 268 (2020) 121928 15

Environmental assessment of an integrated rooftop greenhouse for food pro- singapore-aiming-to-become-regions-urban-agriculture-technology-hub-dr-


duction in cities. J. Clean. Prod. 177, 326e337. https://doi.org/10.1016/ koh-poh-koon (accessed 3.28.19).
j.jclepro.2017.12.147. Trautmann, N.M., Porter, K.S., Wagenet, R.J., 2020. Modern Agriculture: its Effects on
Sanye-Mengual, E., Oliver-Sola , J., Montero, J.I., Rieradevall, J., 2015. An environ- the Environment [WWW Document]. Cornell Univ. URL. http://psep.cce.cornell.
mental and economic life cycle assessment of rooftop greenhouse (RTG) edu/facts-slides-self/facts/mod-ag-grw85.aspx (accessed 2.8.20).
implementation in Barcelona, Spain. Assessing new forms of urban agriculture Udias, A., Pastori, M., Dondeynaz, C., Carmona Moreno, C., Ali, A., Cattaneo, L.,
from the greenhouse structure to the final product level. Int. J. Life Cycle Assess. Cano, J., 2018. A decision support tool to enhance agricultural growth in the
20, 350e366. https://doi.org/10.1007/s11367-014-0836-9. Mekrou river basin (West Africa). Comput. Electron. Agric. 154, 467e481.
Shamshiri, R.R., Kalantari, F., Ting, K.C., Thorp, K.R., Hameed, I.A., Weltzien, C., https://doi.org/10.1016/j.compag.2018.09.037.
Ahmad, D., Shad, Z., 2018. Advances in greenhouse automation and controlled Wallace, A., 1994. High-precision agriculture is an excellent tool for conservation of
environment agriculture: a transition to plant factories and urban agriculture. natural resources. Commun. Soil Sci. Plant Anal. 25, 45e49. https://doi.org/
Int. J. Agric. Biol. Eng. 11, 1e22. https://doi.org/10.25165/j.ijabe.20181101.3210. 10.1080/00103629409369002.
Stoessel, F., Juraske, R., Pfister, S., Hellweg, S., 2012. Life cycle inventory and carbon Wernet, G., Bauer, C., Steubing, B., Reinhard, J., Moreno-Ruiz, E., Weidema, B., 2016.
and water foodprint of fruits and vegetables: application to a swiss retailer. The ecoinvent database version 3 (part I): overview and methodology. Int. J. Life
Environ. Sci. Technol. 46, 3253e3262. https://doi.org/10.1021/es2030577. Cycle Assess. 21, 1218e1230.
Tan, S.-A., 2019. Singapore Aiming to Become Region’s Urban Agriculture Technol- Yan, W., Hunt, L.A., 1999. An equation for modelling the temperature response of
ogy Hub: koh poh koon, Business News & Top Stories - the Straits Times [WWW plants using only the cardinal temperatures. Ann. Bot. 84, 607e614. https://
Document]. The Straits Times. https://www.straitstimes.com/business/ doi.org/10.1006/anbo.1999.0955.

You might also like