Accepted Manuscript

A Feasibility Study for the Locations of Waste Transfer Stations in Urban Centers: A
Case Study on the City of Nashik, India

Vinay Yadav, Subhankar Karmakar, A.K. Dikshit, Shivkumar Vanjari

PII: S0959-6526(16)30077-4
DOI: 10.1016/j.jclepro.2016.03.017
Reference: JCLP 6866

To appear in: Journal of Cleaner Production

Received Date: 12 August 2015

Revised Date: 3 January 2016
Accepted Date: 3 March 2016

Please cite this article as: Yadav V, Karmakar S, Dikshit AK, Vanjari S, A Feasibility Study for the
Locations of Waste Transfer Stations in Urban Centers: A Case Study on the City of Nashik, India,
Journal of Cleaner Production (2016), doi: 10.1016/j.jclepro.2016.03.017.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT

A Feasibility Study for the Locations of Waste Transfer

Stations in Urban Centers: A Case Study on the City
of Nashik, India

PT
Vinay Yadava , Subhankar Karmakara,b,c,∗, A. K. Dikshita,c , Shivkumar
Vanjarid

RI
a Centre for Environmental Science and Engineering, Indian Institute of Technology
Bombay, Mumbai-400076, India
b Inter Disciplinary Program on Climate Change, Indian Institute of Technology Bombay,

SC
Mumbai-400076, India
c Centre for Urban Science and Engineering, Indian Institute of Technology Bombay,

Mumbai-400076, India
d Nashik Municipal Corporation, Nashik-422002, India

U
Abstract AN
Rapid population growth and high rate of urbanization make municipal solid
M
waste management a challenging task for municipalities due to the inadequacy
of infrastructural services. Taking this into consideration, this paper proposes
an approach for finding economically optimal locations and feasibility demon-
D

stration of a municipal solid waste management infrastructure unit, i.e. waste

TE

transfer stations. This analytical approach consists of two basic elements: (i) a
mathematical model to optimize overall cost for municipal solid waste manage-
ment; and (ii) geographical information system tools to create a data inventory
EP

for the mathematical model. The novelty of the proposed model includes het-
erogeneity in data compilation, on-road measurements of distances and strategic
allocation of transfer stations. Also, to assess waste load estimation more pre-
C

cisely, analyzing the impacts of heterogeneity in data and on-road distances

on model, three distinct cases have been considered. Further, the developed
AC

approach has been demonstrated on the city of Nashik, India. The mathemat-

∗ Correspondingauthor
Email address: skarmakar@iitb.ac.in (Subhankar Karmakar)
URL: http://www.cese.iitb.ac.in/people/facinfo.php?id=skarmakar (Subhankar
Karmakar), http://www.cese.iitb.ac.in/people/facinfo.php?id=dikshit (A. K. Dikshit)

Preprint submitted to Journal of Cleaner Production March 20, 2016

 ACCEPTED MANUSCRIPT

ical model for Nashik city is written in a mathematical programming language

AMPL and the solutions have been obtained using KNITRO solver. Optimiza-

PT
tion model selects three best locations out of eight potential locations with the
optimum capacity of 200 TPD. Furthermore, the assessment of model’s behavior
under uncertainty has been performed with a comprehensive sensitivity anal-

RI
ysis for further improvements. The integrity of all components in this study
produces a generic framework for helping municipal officials and researchers

SC
for development of a municipal solid waste management system with transfer
stations in an economically optimal manner.
Keywords: Environmentally sound technologies, Facility location problem,

U
Mixed integer nonlinear programming, Municipal solid waste management,
Transfer stations, TSS model.
AN
1. Introduction
M
In last few decades, the rate of urbanization in the world has increased sig-
nificantly. The level of urbanization was 30% in 1950, 54% in 2014 and has
D

projected to be 66% by 2050 (United Nations, 2014). In India, for instance,

as per censuses of 1901, 2001 and 2011 population residing in urban centers
TE

was 11.4%, 28.53% and 31.16% respectively (Datta and Pranati, 2006; Census
of India, 2011). The main consequence of rapid urbanization is the stunning
growth of population density and hence increased amount of municipal solid
EP

waste (MSW) in urban centers throughout the globe (Ravindra et al., 2015).
Development of infrastructural services for MSW management (e.g., waste col-
C

lection systems, sanitary landfills, waste processing facility etc.) in the urban
areas are not able to keep a pace with the rapid urbanization rate (JnNURM,
AC

2013). Though, governments have been trying to tackle this problem in a safe,
hygienic and economically viable manner using environmentally sound technolo-
gies (ESTs) (Asnani, 2006).
MSW (Management and Handling) rules, 2000 entrust the Indian municipal-
ities with the responsibilities for collection, storage, transportation, processing

2
 ACCEPTED MANUSCRIPT

and disposal of MSW. In India, a significant amount of the total expenditure

i.e. 85% (Ghose et al., 2006) is spent on collection, which is higher than typical

PT
value between 50-70% (Sonesson, 2000). Improvement in the design of the col-
lection system can result in a significant saving, thereby a large proportion of
the fund (Greco et al., 2014). Therefore, the cost effective collection system and

RI
finding the appropriate number and location of transfer stations (TSs), similar
to other location problems (e.g. fire stations, health care system), should be

SC
well designed (Owen and Daskin, 1998). TSs play an important role in collec-
tion system such as: (i) reduction in transportation cost; (ii) reduction in the
volumes of solid waste because of the compaction process; (iii) reduction in the

U
traffic of the city (Cui et al., 2011). Nevertheless, choosing best locations for
the TSs is the major problem (Ramachandra, 2006). Therefore, this paper pro-
AN
poses a facility location model to choose best locations and numbers of TSs in
an economically optimal manner. Facility location models are extensively used
M
to tackle this problem and found to be very useful. Proposed model for loca-
tion of the TSs will help to improve the efficiency while to decrease the costs
and pollution level. Furthermore, this model will contribute to the existing
D

literature on mathematical models for optimum locations of TSs and will also
be helpful for researchers and municipal officials for designing an economically
TE

optimal collection system.

1.1. Transfer Stations (TSs)

EP

A TS is an intermediate station between final disposal options and generation

sources to increase the efficiency of the MSW management system (Ramachan-
C

dra and Bachamanda, 2007). TSs are constructed close to residential areas to
receive and hold waste from primary collection vehicles (PCVs) until it can be
AC

moved to secondary collection vehicles (SCVs) for transportation to the final

waste disposal or processing options (Kirca and Erkip, 1988). TSs are vital
infrastructural unit of waste collection system for any municipality (O’Malley,
2003). On the basis of SCVs loading, TSs may be classified as: (i) direct load;
(ii) storage load; and (iii) combined direct load and discharge load (USEPA,

3
 ACCEPTED MANUSCRIPT

Primary Collection
Capacity 3 ton

Vehicles Shutter
PCVs

PT
Tipping Floor Elevated Platform
Ramp System

MSW

RI
Compactor

Capacity 20 ton
Secondary Collection

SC
SCVs
Vehicles

Ground Surface

U
AN
Figure 1: Conceptual diagram showing various components of a mechanized direct load trans-
fer station; where PCVs and SCVs are primary collection vehicles and secondary collection
vehicles respectively.
M
2002). In this study, feasibility of having mechanized direct load TS with com-
pactor has been demonstrated for Indian urban centers. Functioning of this
D

type of TS through a conceptual diagram is shown in the Figure 1. This type of

TSs are being used by Coimbatore City Municipal Corporation (CCMC), Tamil
TE

Nadu; Surat Municipal Corporation (SMC), Gujarat and Municipal Corpora-

tion of Greater Mumbai (MCGM), Maharashtra, India currently.
EP

1.2. Why TSs in Urban Centers?

USEPA (2002) primarily explains the need, uses and justifications of having
TSs in any community. This section conveys some of the justifications for having
C

TSs, exclusively in Indian urban centers.

AC

Environmentally Sound Technology (ESTs). ESTs are cost effective and energy
efficient technologies which do not pollute air, land or water (Ramachandra and
Bachamanda, 2007). The factors for TSs associated with environmentally sound
technologies are (i) proposed TSs are able to accommodate all the PCVs already
in use by the municipal corporation; and (ii) these TSs are constructed by

4
 ACCEPTED MANUSCRIPT

indigenous manufacturer; and hence gained local expertise. These indigenously

manufactured TSs are cost effective and does not depend on foreign experts for

PT
maintenance.

Flexibility of Choosing New Locations for Disposal. TSs in a collection system

RI
provide the flexibility of choosing new locations for final disposal or process-
ing by reducing the dependency on PCVs to access these locations. Therefore,
new facilities (sanitary landfills, waste to energy plants, composting plants etc.)

SC
can be located by giving preferences to public health and environmental issues.
Moreover, to minimize the health hazards due to flies, mosquitoes, rodents and
stray animals final disposal or processing facilities may be established signifi-

U
cantly away from the urbanized areas.

AN
Economic Viability. Traveling time for the PCVs can be reduced by including
TSs in the collection system and hence the primary collection cost (USEPA,
M
2002). Also, inclusion of TSs will reduce number of PCVs. However, break even
point determination is considered as a classic approach to examine economic
viability of TSs (Phelps et al., 1996). This approach provides a comparison
D

of TS cost, hauling cost of SCVs collectively and hauling cost of PCVs. The
TE

hauling costs of PCVs and SCVs have two components: (i) variable cost; and
(ii) fixed cost. Variable cost is further divided into fuel consumption, tire cost,
maintenance cost, depreciation, engine oil cost, grease cost and wages of the
EP

crew etc. (World Bank, 2005; Dhingra et al., 2011). Fixed cost is the sum of
vehicles capital cost (only for SCVs), taxes, fines and toll charges etc. (Ahmad
et al., 2013).
C

The TS cost includes establishment cost, operation and maintenance cost,

AC

calculated per unit capacity (Ramachandra, 2006). The distances between fi-
nal disposal or processing sites and generation source are the only key factors
to decide the need of TSs. Phelps et al. (1996) found this dependency in the
following derivation:

5
 ACCEPTED MANUSCRIPT

Hauling Cost with SCVs

PT
Hauling Cost with PCVs
150 Cost of TS
Hauling Cost (Rs./ton)

RI
100

SC
50

U
0
AN “Break Even” Point

0 5 10 15 20
M
Distance from Source to Integrated MSW Facility (km)

Figure 2: Break even analysis to determine economic viability of transfer stations (TS) by
following USEPA (2002) manual and based on the information provided by Surat Municipal
D

Corporation (SMC), where PCVs and SCVs are primary collection vehicles and secondary
collection vehicles respectively.
TE

Let hauling cost with PCVs be a Rs./ton/km1 , TS cost be b Rs./ton and haul-
ing cost with SCVs be c Rs./ton/km. If the distance between generation source
EP

and TS is x km and the distance between TS and disposal site is y km. Then,
C

Hauling cost with TSs (Rs./ton): T C1 = ax + b + cy (1)

Hauling cost without TSs (Rs./ton): T C2 = ax + ay (2)

AC

TSs are economically viable iff T C1 ≤ T C2

11 Rs. = 0.015 USD as on 16th December 2015

6
 ACCEPTED MANUSCRIPT

i.e. ax + b + cy ≤ ax + ay (3)
b
y≥ (4)
(a − c)

PT
Break even distance is site specific and subject to vary with different input
parameters (USEPA, 2002). Figure 2 shows break even distance of 10.44 km in

RI
single trip for the city of Surat, India i.e. TSs are economically viable if single
trip distances of collection vehicles are more than 10.44 km. In this calculation,

SC
hauling costs for PVCs and SCVs are derived using Dhingra et al. (2011)’s
methodology and TS cost is taken from SMC’s divisional office.

Frequency of Waste Collection. “Just-in-time” collection system is currently

U
practiced in many municipalities, for example, Agartala, Ahmedabad, Coimbat-

AN
ore, Dehradun, Madurai, Nashik, Panaji and Surat. In this collection system,
PCVs visit some certain places and citizens have to take their solid waste out
at these particular places on or before the arrival of PCVs. Collection vehicles
M
also use some specific signals (e.g. bells, rings etc.) to mark their presence.
Inclusion of TSs will allow PCVs to travel less distance, offload and come back
D

to the collection area promptly. According to Sharholy et al. (2008), fraction of

compost-able materials is 40% to 60% in urban areas. Compost-able materials
TE

need to be collected frequently, else, bad odor, flies, insect vectors and stray
animals may cause nuisance. Conclusively, there is a need of “Just-in-time”
collection system with higher collection frequency, which can be achieved using
EP

TSs.

Efficiency of Collection System. Cui et al. (2011) stated that the MSW man-
C

agement system with TSs ensues reduction in the traffic of the city by bringing
down the number of vehicles to-and-from processing or disposal facilities. As far
AC

as Indian urban centers are concerned, traffic due to waste collection vehicles
coincide with the city’s routine traffic because of 9 a.m. - 6 p.m. work culture,
and make this problem more acute. Therefore, TS installation will reduce traffic
congestion, crashes, noise level and air emissions due to less quantity of fuel used
(Ross, 1999). Also, final disposal and processing sites in a MSW management

7
 ACCEPTED MANUSCRIPT

system with out TSs are often found to be crowded with the collection vehicles.
Inclusion of TSs in MSW management system diverts this fleet into PCVs and

PT
SCVs. Consequently, reduction in the waiting times at TSs or final disposal and
processing sites. This time can be utilized in making more efficient collection
system.

RI
The primary characteristics of given justifications are particularly pertinent
to economical and environmental nature. Break even point determination and

SC
ESTs (cost effectiveness criteria) justifications are economical in nature, whereas
frequency, efficiency of waste collection and flexibility in choosing new locations
for processing/disposal are environmental in nature. Apparently, justification

U
pertaining to environmental nature are inherent within the establishment of
TSs. Therefore, this study emphasizes on economical aspects for the locations
AN
of TSs. Next section elucidates the literature survey on mathematical models
developed to find best locations for TSs in an economically optimal way.
M
2. Facility Location Models for the Locations of TSs: A Review
D

A careful planning is required in order to execute MSW management sys-

tem in an economically optimal way. Over the past decades, several studies
TE

have used different mathematical programming techniques (MPTs), such as lin-

ear programming, nonlinear programming, integer programming (IP),dynamic
programming (DP) and multi-objective and goal programming to evaluate the
EP

optimum cost for MSW management systems. All results indicate that MPTs
are useful tools for evaluating least cost strategies for MSW management prob-
lems. Table 1 lists several facility location models which have been used to find
C

specifically the best locations for TSs. Facility location problems are a special
AC

branch of operation research and should not be integrated with vehicle routing
problem due to their different planning periods (Nagy and Salhi, 2007). How-
ever, a vehicle routing problem may be solved after locating the TSs on best
locations. Classification of past studies has been made as per their study areas
and applied approaches. Literature review reports that no such model has been

8
 ACCEPTED MANUSCRIPT

Table 1: Literature survey on models developed specifically to find the best locations for
transfer stations, where MILP is mixed integer linear program and MINLP is mixed integer
non-linear program.

PT
Author Study Area Approach
Marks and Liebman (1970) Baltimore (USA) Heuristic
Jenkins (1982) Toronto (Canada) Heuristic
Kirca and Erkip (1988) Istanbul (Turkey) Heuristic

RI
Or and Curi (1993) Izmir (Turkey) MILP
Kulcar (1996) Brussels (Belgium) MILP
Chang and Lin (1997) Taiwan (ROC) MILP
Badran and El-Haggar (2006) Port Said (Egypt) MILP

SC
Komilis (2008) Athens (Greece) MILP
Lyeme (2011) Dar es Salaam (Tanzania) MILP
Chatzouridis and Komilis (2012) Hellenic (Greece) MINLP
Eiselt and Marianov (2014) Biobío (Chile) MINLP

U
AN
developed for Indian urban centers. Moreover, it can be seen in the Table 1,
mixed integer linear programming (MILP) and heuristic are most commonly
used approaches. However, Chatzouridis and Komilis (2012) used mixed inte-
M
ger nonlinear programming (MINLP) to develop a methodology to design an
economically optimal collection system for Hellenic region (Greece). MINLP
models are difficult to solve (Coello, 2011), but paves the way for much pre-
D

cise research. In these days, computational advancements have made possible

TE

to solve nonlinear models and hence precise research. In current study rooted
on the literature review, a MINLP model has been proposed for urban centers.
Next section provides the explanation of this mathematical model.
EP

3. Methodology
C

The methodology proposed here provides a logical framework to help mu-

nicipal officials for development of a MSW management system with TSs in an
AC

economically optimal manner. This methodology also emphasizes on the inter-

connection of geographical information system (GIS) and mathematical models
in optimizing the overall cost by choosing best locations for TSs. GIS analysis
is an integral part of this methodology and has been used for data compilation
for mathematical model. This methodology consists of two basic elements (see

9
 ACCEPTED MANUSCRIPT

Figure 3): (i) a mathematical model to optimize overall cost for MSW manage-
ment by choosing best locations for TSs (Mathematical analysis); and (ii) GIS

PT
tools to create a data inventory for the mathematical model (GIS analysis).
Following subsections provide the detailed discussion of these analyses.

Methodology

RI
Mathematical Analysis GIS Analysis
Vectors

SC
Potential locations for TSs

• Creation of network dataset

• Road network

§ Network Analyst

§ Development of
Spatial Database
• Potential locations

• Finding shortest routes

Development of a facility • Ward boundaries
location model (MINLP)
• Waste generation point sources

U
Optimal solution to the Attribute Tables
mathematical model • Ward-wise population

Best location for TSs

AN • Latitude and longitude of
waste generation point sources
and potential locations
M
Figure 3: Scheme describing the methodology for choosing best locations for transfer station.
D

3.1. Mathematical Analyses

TE

The locations for TSs are perhaps limited to certain sites, known as potential
locations. The proposed mathematical model is to find the best locations out of
potential locations for TSs. Consequently, it requires a binary variable to select
EP

the best locations. These problems (0 − 1 integer programming problems) are

listed in Karp’s 21 N P-complete problems (Karp, 1972). N P-completeness of a
problem means that no polynomial time algorithm is able to solve these problems
C

(Murray and Shanbhag, 2006). Apparently, these problems are difficult to solve,
AC

but results more accurate solutions. Constraints used for the strategic allocation
of TSs are the sources of nonlinearity in this model. Following are the steps
followed to model:

10
 ACCEPTED MANUSCRIPT

3.1.1. Selection of Potential Locations

According to USEPA (1995) choosing a suitable location for TS is a major

PT
and challenging issue. Potential locations should be chosen by considering health
hazards due to insects, nuisance because of bad odor, stray animals and noise due
to PCVs, SCVs and machines (Ramachandra and Bachamanda, 2007). MCGM

RI
officials emphasizes on road accessibility and proximity to the sources with high
generation rate to the potential locations. Nevertheless, it is recommended to

SC
have a discussion with the local governing bodies while choosing the potential
locations. In the present study, potential locations have been chosen as per the
suggestions of NMC officials and recommendation of MCGM officials.

U
3.1.2. Selection of Best Locations
AN
After choosing the potential locations, the next step is to find the best loca-
tions. Best locations are the locations which have to be chosen in an economi-
cally optimal manner among the potential locations. In order to determine the
M
best locations, facility location model has been adopted. Facility location mod-
els are the broad classes of optimization models that have been widely studied
D

in the operations research community (Mirchandani and Francis, 1990). Also,

these models have been adopted by many researchers (see Table 1) to find the
TE

best locations for TSs.

3.1.3. Nonlinear Facility Location Model

EP

This subsection shows the mathematical representation of a facility location

model for TSs location. This model, named as TSS model, has been concep-
tualized from the pioneer work of Or and Curi (1993), Badran and El-Haggar
C

(2006), Komilis (2008) and Chatzouridis and Komilis (2012).

AC

This model is consists of a linear objective function and some nonlinear

constraints. The novelties of TSS model are (i) inclusion of on-road distances,
(ii) consideration of heterogeneity in data; and (iii) strategic allocation of TSs
by introducing constraints which restrict each and every source i for not trans-
porting their waste to two different TSs. Figure 4 represents the conceptual

11
 ACCEPTED MANUSCRIPT

Sources(i) TSs(j) Presorting Units(k) Processing(l) Final Disposal(m)

i=1 Eij = 1 [D1 = 1]

PT
Eij =0 j=1
i=2 l=1
[D2 = 0]
i=3 j=2 k=1 m=1

RI
• •
i=4 [D3 = 0] l=2
• •
j=3
• •
•
i=5
• •

SC
•
• •
k=P m=R
•
•
•
l=Q
•
•
j=O

U
i=N [Dj = 1]
[N >> O >> Q >> R ≈ P ]

AN
Figure 4: Conceptual representation for the proposed TSS model.

structure of TSS model. In the figure, Di s and Eij s are the binary variables,
M
Dj s take 1 if a TS is to be set up at candidate location (j = 1, 2, . . . , O),
whereas Eij s take 1 if source (i = 1, 2, . . . , N ) chooses TS (j = 1, 2, . . . , O).
D

Solid lines in Figure 4 represent the actual flow of waste from sources to TSs
(j = 1, 2, . . . , O) units optimally, whereas dashed lines represent the possible
TE

flow of waste from sources to TSs. Possible flow of waste from TSs to presorting
units (k = 1, 2, . . . , P ) and then to processing facilities (l = 1, 2, . . . , Q) and
finally to disposal sites (m = 1, 2, . . . , R) has been shown using dotted lines.
EP

Some municipalities have also adopted integrated solid waste management sys-
tems in which all presorting units, processing facilities and final disposal sites
are at the same place, which is represented through a dashed rectangular in the
C

Figure 4.
AC

Assumptions. Segregation at source is a major issue and there is no organized

system for segregation formally in many developing countries like India (Ku-
mar and Pandit, 2013; Bharti et al., 2015). Although, a number of NGOs are
making their efforts in the direction of making people aware of the benefits they
can achieve from this practice. Also, after enactment of MSW management

12
 ACCEPTED MANUSCRIPT

and handling rules 2000 in India, many municipalities constructed final dis-
posal/processing facilities (e.g. landfills, composting plants etc.) but, very few

PT
among them have TSs. Considering these situations, the following assumptions
have been imposed:

1. There is no waste separation at the source.

RI
2. Transportation cost is proportional to the distance traveled.

SC
3. Traffic jams are not considered in the movement of PCVs and SCVs.

4. MSW handling is assumed daily.

5. The running costs of TSs and existing facilities are described as daily cost.

U
6. Municipalities have final disposal and processing facilities.
AN
7. Without loss of generality, final disposal and processing facilities include
composting plants, presorting units, Refuse-derived fuel (RDF) plants,
M
inert processing facility and sanitary landfills.

8. The establishment costs of the existing facilities are not considered in this
D

formulation.

The TSS model for MSW management system is developed as follows:

TE

TSS Model Indexes, Parameters and Variables. Following indexes are used for
TSS model:
EP

N Total number of waste generation point sources

i Index for waste generation point source (i = 1, 2, . . . , N )
O Total number of potential locations for TSs
C

j Index for potential locations for TSs (j = 1, 2, . . . , O)

P Total number of presorting units
AC

k Index for presorting units (k = 1, 2, . . . , P )

Q Total number of processing facilities
l Index for processing facilities (l = 1, 2, . . . , Q)
R Total number of final disposal sites
m Index for final disposal sites (m = 1, 2, . . . , R)

13
 ACCEPTED MANUSCRIPT

Input parameters used for TSS model are as follows:

C The set of l where composting is not possible

PT
{
C The complement set of C
F The set of l where RDF plants are not available

RI
{
F The complement set of F
I The set of l where inert processing plants are not available
I{

SC
The complement set of I
Fj The establishment cost of the TS ‘j’ represented as daily cost for the
given time horizon

U
RTj Running cost of the TS ‘j’ in Rs./ton
RSk Running cost of presorting unit ‘k’ in Rs./ton
RPl
RFm
AN
Running cost of processing facility ‘l’ in Rs./ton
Running cost of final disposal site ‘m’ in Rs./ton
TSKij Transportation cost of one ton of MSW from source (i = 1, 2, . . . , N )
M
to TS (j = 1, 2, . . . , O) in Rs.
TSIjk Transportation cost of one ton of MSW from TS (j = 1, 2, . . . , O) to presorting
D

unit (k = 1, 2, . . . , P ) in Rs.
TSMkl Transportation cost of one ton of MSW from presorting unit (k = 1, 2, . . . , P )
TE

to processing facility (l = 1, 2, . . . , Q) in Rs.

TSJlm Transportation cost of one ton of MSW from processing facility (l = 1, 2, . . . , Q)
EP

to final disposal facility (m = 1, 2, . . . , R) in Rs.

αk Fraction of incoming MSW to presorting unit ‘k’ which has been recycled
REP Revenue from recycling in Rs./ton
C

βl Fraction of incoming MSW to composting plant l ∈ C { which has been converted

to the manure
AC

REC Revenue from composting plant in Rs./ton

γl Fraction of incoming MSW to RDF plant l ∈ F{ which has been converted to
the fuel pallets

14
 ACCEPTED MANUSCRIPT

RER Revenue from RDF plant in Rs./ton

δl Fraction of incoming MSW to inert processing plant l ∈ I{ which has been

PT
converted to the construction materials
REI Revenue from inert processing plant in Rs./ton
Gi The amount of MSW generated at source (i = 1, 2, . . . , N ) in TPD

RI
CTSj The capacity of TS ‘j’ in TPD
CPUk The capacity of presorting unit ‘k’ in TPD

SC
ζl Fraction of MSW which needs to be sent to the composting plant
CCPl The capacity of composting plant l ∈ C{ in TPD
ηl Fraction of MSW which needs to be sent to the RDF plant

U
CRDl The capacity of RDF plant l ∈ F{ in TPD
θl
CIPl
AN
The fraction of MSW which needs to be sent to the inert processing plant
The capacity of inert processing plant l ∈ I{ in TPD
PTS The number of potential locations for TSs
M
RCl Fraction of rejects from composting plant l ∈ C{
RRl Fraction of rejects from RDF plant l ∈ F{
RIl Fraction of rejects from inert processing plant l ∈ I{
D

Variables used for TSS model are as follows:

TE

Dj Binary variable (= 0 or 1), which is to take one if a TS is to be set up at

candidate location (j = 1, 2, . . . , O)
XKij Amount (in ton) of daily MSW to be transferred from source (i = 1, 2, . . . , N )
EP

to TS (j = 1, 2, . . . , O)
XIjk Amount (in ton) of daily MSW to be transferred from TS (j = 1, 2, . . . , O)
to presorting unit (k = 1, 2, . . . , P )
C

XMkl Amount (in ton) of daily MSW to be transferred from presorting unit
AC

(k = 1, 2, . . . , P ) to processing facility (l = 1, 2, . . . , Q)
XJlm Amount (in ton) of daily MSW to be transferred from processing facility
(l = 1, 2, . . . , Q) to final disposal sites (m = 1, 2, . . . , R)
Eij Binary variable (= 0 or 1), which is to take one if TS (j = 1, 2, . . . , O)
has been chosen for the source (i = 1, 2, . . . , N )

15
 ACCEPTED MANUSCRIPT

Objective Function. Let Z be the objective function calculated after subtracting

revenues of all processing facilities from the summation of functioning cost and

PT
transportation cost.

min Z = [Eq.(6) + Eq.(7) − Eq.(8) − Eq.(9) − Eq.(10) − Eq.(11)] (5)

RI
• The functioning cost (FC) is obtained by taking the sum of daily running
cost of TSs, other facilities and the fixed cost of the TSs.

SC
O
X N X
X O O X
X P
FC = Fj ∗ Dj + RTj ∗ XKij + RSk ∗ XIjk
j=1 i=1 j=1 j=1 k=1

U
X Q
P X Q X
X R
+ RPl ∗ XMkl + RFm ∗ XJlm (6)
k=1 l=1
AN l=1 m=1

• Transportation cost (TC) is due to the transportation of MSW (i) from

generation sources to TSs; (ii) from TSs to presorting units; (iii) from
M
presorting units to processing facilities; and (iv) from processing facilities
to final disposal sites. TC has been estimated using GIS based shortest
D

routing tool.
TE

X O
N X O X
X P
TC = TSKij ∗ XKij + TSIjk ∗ XIjk
i=1 j=1 j=1 k=1

X Q
P X Q X
X R
EP

+ TSMkl ∗ XMkl + TSJlm ∗ XJlm (7)

k=1 l=1 l=1 m=1

• Revenue from recycling is obtained by selling the recyclables.

C

O X
X P
AC

REP ∗ αk ∗ XIjk (8)

j=1 k=1

• Revenue from composting plant is obtained by selling the compost.

P X
X
REC ∗ βl ∗ ζl ∗ XMkl (9)
k=1 l∈C{

16
 ACCEPTED MANUSCRIPT

• Revenue from RDF plant is obtained by selling the fuel pellets.

P X
X

PT
RER ∗ γl ∗ ηl ∗ XMkl (10)
k=1 l∈F{

• Revenue from inert processing plant is obtained by selling the construction

RI
materials.

P X

SC
X
REI ∗ δl ∗ θl ∗ XMkl (11)
k=1 l∈I{

Constraints. The constraints in this model are mainly classified into four cat-

U
egories: (i) mass balance; (ii) capacity; (iii) binary; and (iv) non-negativity

AN
constraints. Explanation of each constraint is given below:

(1) All generated waste at source ‘i’ is sent to the TS ‘j’.

M
O
X
XKij = Gi ∀ i = 1, 2, . . . , N (12)
j=1
D

(2) All waste coming to TS ‘j’ shall be transported to presorting unit ‘k’.
TE

N
X P
X
XKij = XIjk ∀ j = 1, 2, . . . , O (13)
i=1 k=1
EP

(3) All the remnants of presorting units ‘k’ shall be transported to respective
processing facility ‘l’.
C

O
X Q
X
(1 − αk ) ∗ XIjk = XMkl ∀ k = 1, 2, . . . , P (14)
AC

j=1 l=1

17
 ACCEPTED MANUSCRIPT

(4) TS ‘j’ will be set up if and only if it is selected by model.

N
X

PT
XKij ≤ CTSj ∗ Dj ∀ j = 1, 2, . . . , O (15)
i=1
P
X
XIjk ≤ CTSj ∗ Dj ∀ j = 1, 2, . . . , O (16)

RI
k=1

(5) The quantity of waste transported to the presorting unit ‘k’ must not

SC
exceed its capacity.

O
X
XIjk ≤ CPUk ∀ k = 1, 2, . . . , P (17)

U
j=1

AN
(6) The quantity of waste transported to composting plant must not exceed
its capacity.
M
P
X
ζl ∗ XMkl ≤ CCPl ∀ l ∈ C{ (18)
k=1
ζl = 0 ∀l ∈ C (19)
D

(7) The quantity of waste transported to the RDF plant must not exceed its
TE

capacity.

P
X
EP

ηl ∗ XMkl ≤ CRDl ∀ l ∈ F{ (20)

k=1
ηl = 0 ∀l ∈ F (21)
C

(8) The quantity of waste transported to the inert processing plant must not
AC

exceed its capacity.

P
X
θl ∗ XMkl ≤ CIPl ∀ l ∈ I{ (22)
k=1
θl = 0 ∀l ∈ I (23)

18
 ACCEPTED MANUSCRIPT

(9) The allowable number of TSs should not exceed a specific limit PTS.

O
X

PT
Dj ≤ PTS (24)
j=1

(10) If source ‘i’ chooses ‘j’ as its TS, then all the waste generated at source

RI
‘i’ will be transferred to TS ‘j’ only. These constraints strategize the
allocation of TSs for waste generation sources.

SC
Gi ∗ Eij ≥ XKij ∀ i = 1, 2, . . . , N, j = 1, 2, . . . , O (25)
O
X
Eij = 1 ∀ i = 1, 2, . . . , N (26)

U
j=1

AN
(11) Rejects of processing plants shall be equal to the amount transported to
the disposal sites.
M
P
XX P
XX
RCl ∗ ζl ∗ XMkl + RRl ∗ ηl ∗ XMkl
l∈C{ k=1 l∈F{ k=1
Q X
D

P
XX X R
+ RIl ∗ θl ∗ XMkl = XJlm (27)
l∈I{ k=1 l=1 m=1
TE

(12) Non-negativity constraints.

EP

XKij , XIjk , XMkl , XJlm ≥ 0 ∀ i, j, k, l, m (28)

To provide input data for the TSS model, GIS tools are used. The steps
C

followed to create data inventory using GIS are explained in the next section.
AC

3.2. GIS Analysis

GIS is a prominent channel for importing, managing and analyzing spatial

data (Longley, 2005). GIS also produces an advanced modeling framework
to analyze spatial problems associated with MSW management (Chalkias and
Lasaridi, 2009). In this study, GIS tools have been used to find (i) shortest

19
 ACCEPTED MANUSCRIPT

routes for waste collection vehicles to estimate transportation cost (see Equation
7); and (ii) geographical positions of waste generation sources and potential

PT
locations for TSs. Steps followed to create data inventory are as follows:

Step 1 Development of spatial database for the study area covering road net-
works (routes), ward boundaries, census data and planned locations for

RI
final disposal facilities.

Step 2 Preparation of a list of waste generation sources and potential locations

SC
for TSs as per the requirement.

Step 3 Shortest route analysis for each pair of distances between sources and

U
facilities using Network Analyst (NA) tool of ArcGIS 10.2.

AN
Next section provides the implementation of developed methodology on the
city of Nashik, India.
M
4. A Case Study on the City of Nashik

This section deals with the application of developed methodology to the city
D

of Nashik, India.
TE
C EP
AC

Figure 5: Primary collection vehicle (Ghanta Gadis) at the time of MSW collection in Panch-
vati division (Nashik).

4.1. Overview of Study Area and Existing MSW Management System

Nashik is located in the northwest of Maharashtra state in India (see Figure
8(d)). The total area of the city is 259 km2 . Presently, NMC collects approxi-

20
 ACCEPTED MANUSCRIPT

mately 350 tons MSW per day with waste generation rate of approximately 235
g/c/day. It contains compost-able materials (40%); recyclables (25%); ash, fine

PT
earth and others (18.80%); and textiles, plastic, rubber (16.20%) approximately.
Existing collection system in the Nashik city involves door to door collection of
MSW through PCVs known as Ghanta Gadis (see Figure 5) and transportation

RI
to integrated MSW management facility. The integrated MSW management
facility is situated in the south-west extreme of the city and known as Khat

SC
Prakalp site. Facilities available at Khat Prakalp site include a presorting unit,
an aerobic composting plant, an inert processing plant, a leachate treatment
plant, a RDF plant, animal carcasses incinerator and a sanitary landfill.

U
1,800 260 AN 2020
Waste Generation
Rate (g/c/day)

255
d
1,600 250 cte
je 2011
ro
Population (Thousands)

245
M
1,400 P
240
1,200 235

Projected
230 2001
D

1,000 2012 2014 2017 2020

Year
TE

800
1991
600
1981
400
EP

1971
200
1968 1982 1995 2009 2023
Year
C

Figure 6: Projected population and waste generation rate (shown as inset) for Nashik by the
AC

end of the year 2020.

The time horizon for this study is five years (2015-2020). According to
the Census of India (2011), NMC had a population of 1486973 and present
population is estimated to be 1759783 (projected by the end of the year 2020 by
arithmetical increase method). Arithmetical increase method is suitable for old

21
 ACCEPTED MANUSCRIPT

Municipal Solid Waste

Market Household: Door Commercial

PT
(including Vegetable Market) to Door Collection

Ghanta Gadi (PCVs) Street Sweeping

RI
Transfer Stations

SC
Large Trucks (SCVs) Recyclables
Khat Prakalp
Site
Presorting Unit

U
Aerobic Composting Inert Processing RDF Plant

Compost
AN
Rejects, Non-
Recyclable Materials Fuel Pallets
M
Sanitary Landfill
D

Figure 7: Schematic diagram of MSW flow in the proposed system with TSs for Nashik.
TE

and large cities with considerable development (JnNURM, 2011) (see Figure 6).
Figure 6 also shows the annual increment of 1.33% (Pattnaik and Reddy, 2010)
in per capita waste generation as an inset.
EP

4.2. Proposed MSW Management System with TSs

Proposed system for MSW management of NMC is described through a

C

schematic diagram (given in Figure 7). Indian MSW generally subsumes resi-
dential or household waste, commercial waste, market waste 2 , street sweeping,
AC

construction and demolition (C&D) waste3 (Ramachandra, 2011). MSW from

2 Some fraction of market waste i.e. vegetable waste, is directly transported to the com-

posting plants using some special vehicles (see Figure 7).

3 NMC has identified some specific sites for the dumping of C&D waste, hence no C&D

waste enters into MSW stream.

22
 ACCEPTED MANUSCRIPT

residential sources is collected in door-to-door collection manner and is sent to

the TSs using Ghanta Gadis. After that, SCVs will be used to transport MSW

PT
from TSs to the Khat Prakalp site. This proposed system in the Nashik city
further examined under two different scenarios (scenarios (I) and (II)). Scenario
(I) depicts the current situation of the city and assumes existing facilities to

RI
be persistent by the end of the year 2020. Scenario (II) depicts a situation in
which the existing Khat Prakalp site is assumed to be closed before the end of

SC
the year 2020 due to the limited remaining capacity. In scenario (II), a new site
is proposed for integrate MSW management plant in the west of Nashik city
(approximately 20 km away from the city boundary).

U
In door-to-door collection system, MSW is collected from the doorsteps of
residents. However, the whole city is not homogeneous in terms of population
AN
densities of wards and economic activities, which reflects in the waste generation
data as well. Next section addresses this heterogeneity in the data compilation.
M
4.3. Need of Heterogeneity in Data Compilation

Indian urban areas have been spreading spatially and are indicated by their
D

urban sprawl (Tiwari, 2002). These urban sprawls are heterogeneous in nature.
This heterogeneity propagates to the population density. Furthermore, Gallardo
TE

et al. (2015) observed that number of residents and their economic activities vary
throughout the urban area, and hence the waste generation data. In order to
analyze the effect of heterogeneity of data throughout the urban areas in terms
EP

of waste load and measurement methods for distances, three distinct cases have
been considered.
C

Case (a) In this case, centroid of the wards of Nashik city are assumed to be the
AC

waste generation point sources (see Figure 8(a)). Each and every centroid
has the waste load equal to the waste generated in the respective ward
per day. Also, the distances between generation sources to final disposal
and processing sites are euclidean distances. This case is analogous to the
work done by Kulcar (1996); Badran and El-Haggar (2006).

23
 AC
C
Maharashtra
EP
INDIA
TE
D

24
Study Area
M (d)

AN
ACCEPTED MANUSCRIPT

(a)
(a)
U (b) SC (c)

RI
PT
Figure 8: (a), (b) and (c) describes cases explained in Section 4.3; (d) is the layout of the study region.
 ACCEPTED MANUSCRIPT

However, these studies have not addressed the factors such as: (i) on-
road measurements for distances between sources and processing/disposal

PT
facilities since roads are actual mode of waste transportations in most of
the Indian cities; and (ii) on-road locations of waste generation sources
since PCVs use road networks to collect the waste. Cases (b) and (c) have

RI
4
been introduced below by taking these factors into consideration.

Case (b) In this case, waste generation sources are assumed to be at the cen-

SC
ter of each and every spell of 500 meters on road network (see Figure
8(b)) with uniform generation of waste. Also, the distances between these
sources to final processing and disposal sites are on road distances.

U
Case (c) In this case, waste generation sources are assumed to be at the center
AN
of each and every spell of 500 meters on road network with waste load
distribution according to the population densities of the wards of NMC
(see Figure 8(c)). However, the measurement of distances is similar to
M
case (b).
D

The next section presents and discuss the results obtained by implementing
the developed methodology (see Section 3) on the city of Nashik.
TE

5. Results and Discussion

EP

The optimization model (see Section 3.1.3) has been written in AMPL (a
mathematical programming language) (Fourer et al., 1990) and the solutions
have been obtained using KNITRO 5.2 (nonlinear interior-point trust region
C

optimizer) (Byrd et al., 2006), a nonlinear programming solver at IIT Bombay,

Mumbai, India. The solutions to the model provide the optimum number and
AC

capacity of TSs required, best locations and details for the amount of MSW
need to be transferred using PCVs and SCVs.

4 Constraints Eq.(25) and Eq.(26) are not applicable in case (a).

25
 ACCEPTED MANUSCRIPT

As a part of results, following subsections explain (i) how to choose the opti-
mum capacity and number of TSs, (ii) selection of best locations for all different

PT
scenarios and cases described in Section 4.2 and 4.3 respectively, (iii) econom-
ical viability of having TSs in Nashik city; and (iv) sensitivity analysis to find
the most sensitive parameter in MSW management system.

RI
5.1. Optimum Capacity for TSs

SC
300
Case (a)
Case (b)

U
2
Capacity of a TS Unit (TPD)

250 Case (c)

AN

Number of TSs
200
3
M
150
4
D

100 5
6
TE

7
8
50
172 174 176 178 180 182 184 186 188
EP

Overall/Total Cost for MSW Management (Thousands Rs./day)

Figure 9: Scheme describing approach to find optimum number and capacity of TSs.
C

Finding out the optimum capacity for the TSs is one of the major goals for an
ideal MSW management system (Rafiee et al., 2011). Capacity of a TS should
AC

be neither very small (increases the overall cost; see the pattern in the Figure 9),
nor very big (makes TSs crowded with vehicles, and hence reduction in collection
efficiency). Consequently, the capacity for a TS has been kept in between 56
(least possible capacity if all potential locations are selected) and 300 TPD
(maximum possible capacity which can be built on available land with NMC).

26
 ACCEPTED MANUSCRIPT

In this study, an approach has been developed to find out optimum capacity
and number of TSs. Developed approach needs multiple runs for TSS model

PT
(see Section 3.1.3) for all three cases (see Section 4.3). The steps implemented
in this approach can be described as follows:

1. To particularize the TS cost (see the solid horizontal lines in the Figure 9

RI
and Appendix A).

SC
2. To find the dividend for a specific capacity of TSs (see the area between
any two dashed horizontal lines in the Figure 9).

3. To observe the behavior of optimum values for all cases and scenarios

U
using multiple runs.

AN
Figure 9 shows that if the number of TSs are more (or the capacity of TSs is
less), then the difference between the overall cost (X-axis) is comparatively low.
Moreover, vertical slabs provides the budget constraints and horizontal slabs
M
shows that the cost is minimum if TSs are used fully and not partially. Also,
Figure 9 concludes that TS capacity of 200 TPD is optimum for Nashik city.
D

5.2. Best Locations for TSs in Nashik

TE

After finding the optimum capacity and number of TSs, next step is to find
optimal overall cost for MSW management for NMC with best locations. For
this purpose, optimization model has been run for both of the scenarios. The
EP

major difference in the scenarios is: scenario (I) considers current situation to be
persistent by the end of year 2020; whereas scenario (II) considers existing Khat
Prakalp site to be closed before 2020. Best locations for scenario (I) in all cases
C

(see Section 4.3) have shown in this section. However, for scenario (II), only
AC

case (c) has been considered, since case (c) is most realistic case. Changes in
the best locations with respective cases and scenarios are discussed in following
paragraphs.

Case (a). In case (a) with scenario (I), three best locations are chosen and
shown as black stars in the Figure 10. Red stars in the Figure 10 are potential

27
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
Figure 10: Best locations for scenario (I) with case (a).

locations which are not the best locations for TSs. Optimization model for case
D

(a) has 1005 constraints and 880 variables. Optimized overall cost for this case
is Rs. 172178 (USD 2609) approximately.
TE

Cases (b) and (c). In cases (b) and (c) with scenario (I), three best locations
are chosen and shown as black stars in the Figure 11. Red stars in the Figure
EP

11 are potential locations which are not the best locations for TSs. It should be
noted that the best locations for cases (b) and (c) are the same, but different
from case (a). Optimization model for cases (b) and (c) have 9186 constraints
C

and 8152 variables. Optimized overall cost for cases (b) and (c) are Rs. 175396
AC

(USD 2658) and Rs. 175853 (USD 2665 ) approximately.

Scenario (II) with case (c) chooses three best locations (shown as black stars
in the Figure 12). Red stars in the Figure 12 are potential locations which are
not the best locations for TSs. Optimized overall cost in this case is Rs. 203337
(USD 3081) approximately.

28
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
Figure 11: Best locations for scenario (I) with cases (b) and (c).
D
TE
C EP
AC

Figure 12: Best locations for scenario (II) with case (c), where ISWM is integrated solid waste
management.

29
 ACCEPTED MANUSCRIPT

After finding the best locations for TSs with optimum capacity, next step is
to find whether these combinations are economically viable for the city or not.

PT
5.3. Economic Viability of TSs in Nashik

The economic viability of TSs in the city of Nashik with optimum condi-

RI
tions has been analyzed in this section. Break even point calculation, a classic
approach to test economic viability (as described in Section 1.2), has been done

SC
for Nashik. In this calculation hauling costs are found using Dhingra et al.
(2011)’s methodology for SCVs. Data collected from SMC and MCGM is used
to calculate TS cost (see Appendix A for detailed calculations).

U
500 400
Hauling Cost with PCVs
Hauling Cost with SCVs

AN

Frequency of Waste Generation

400 TS Cost
Scenario 1 300
Hauling Cost (Rs/day)

Scenario 2
300

Point Source
200

200
M
100
100

0
0
D

Break Even Point

- 100 - 100
0 10 20 30 40
TE

Distance from Source to Integrated MSW Facility (km)

Figure 13: Comparison between break even distance and frequency distribution of waste
generation sources of Nashik city in both of the scenarios.
EP

Figure 13 represents a comparison between break even distance and fre-

quency distribution of distances from waste generation sources to Khat Prakalp
C

site and proposed integrated MSW management site in both of the scenar-
ios. Class intervals for frequency distribution of distances are calculated using
AC

Sturge’s law (Sturges, 1926). Frequency polygon for the distances between waste
generation sources and Khat Prakalp site in scenario (I) shows a peak in 8 − 10
class interval (shown red in the Figure 13); whereas scenario (II) shows a peak
in 32 − 34 class interval i.e. maximum number of waste generation sources lies
in this interval. Break even distance for Nashik is found to be 16.13 km. It may

30
 ACCEPTED MANUSCRIPT

also be interpreted from the Figure 13 that TSs are not economically viable for
the city of Nashik as far as scenario (I) is concerned, since most of the genera-

PT
tion sources lie in a zone which has distances less than the break even distance.
However, in scenario (II) TSs are economically viable in the city of Nashik.

RI
5.4. Sensitivity Analysis

The assessment of TSS model’s behavior under uncertainty has been done
for further improvement using univariate sensitivity analysis (SA). Univariate

SC
SA is the study of sensitivity of the optimal solution against discrete changes
in its parameters (Karnavas et al., 1993). SA is performed only for case (c) in

U
scenario (I) because this case is more realistic than others.

AN
Overll Cost for MSW Management (Rs./day)

350,000
Z[Gi (α)]
Z[TSKij (α)]
300,000 Z[TSIjk (α)]
M
250,000
D

200,000
TE

150,000

100,000
EP

50,000

0
C

- 0.8 - 0.6 - 0.4 - 0.2 0 0.2 0.4 0.6 0.8

α
AC

Figure 14: Sensitivity analysis for Nashik, where Gi is the amount of MSW generated at
source (i = 1, 2, ..., N ) in TPD and TSKij , TSIjk are transportation cost of one ton of MSW
from source (i = 1, 2, ..., N ) to TS (j = 1, 2, ..., O), from TS (j = 1, 2, ..., O) to the MSW
management facility (k = 1, 2, ..., P ) respectively in Rs.

Waste generation at source i per day and transportation costs are chosen as
parameters for this analysis. As city of Nashik has integrated MSW management

31
 ACCEPTED MANUSCRIPT

Table 2: Fractional increment in the overall cost for MSW management (Z) with respect to
the fractional increment (α) in parameter’s values.

PT
XX
XXX α
- 0.75 -0.5 -0.25 0 0.25 0.5 0.75
Paramet XXXX
Zα [Gi (α)] - 0.73 - 0.48 - 0.24 0 0.23 0.48 0.72
Zα [TSij (α)] - 0.06 - 0.04 - 0.02 0 0.02 0.04 0.06

RI
Zα [TSjk (α)] -0.04 - 0.03 - 0.016 0 0.015 0.031 0.047

SC
plant, hence TSKij and TSIjk are the only non-zero transportation cost. Also,
index k represents the Khat Prakalp site here and not the presorting unit. Figure
14 and Table 2 include the results of this analysis. ‘α’ is defined as: Let X be

U
any parameter of TSS model discussed in the Section 3.1.3 and α ∈] − 1, 1[⊆ R,
then at α = a
AN
X(α) = X + a ∗ X (29)
M
where, X(α) is the numeric value of parameter X at α = a. Additionally,
Z[X(α)] represents the value of the objective function with the respective values
D

of X(α) and Zα [X(α)] represents fractional increment in the objective function

value with respect to the X(α) values.
TE

SA plays a crucial role in decision making under uncertainty (Woodward

et al., 2002). Decision makers know about the time window in which variation
in the specific parameters take place, and by knowing the sensitivity of that
EP

parameter MSW management planning would be much easier.

As can be seen in Figure 14 and Table 2 that waste generation is the most
C

sensitive parameter in a MSW management system. This seems intuitively

correct because all other cost parameters depend on the waste generation i.e.
AC

higher the generation implies higher the quantity of waste and higher total trans-
portation and processing costs. The results also indicate that the sensitivity of
the transportation of waste from sources to TSs is in between the sensitivity of
waste generation and transportation cost from TS to the facility parameters.

32
 ACCEPTED MANUSCRIPT

6. Conclusions

Many municipalities have included transfer stations (TSs) in their waste col-

PT
lection systems, however, there is a dearth of literature exploration and models
based on mathematical analysis and GIS for TSs locations. In this paper, a

RI
methodology, comprises of mathematical analysis and GIS analysis, has been
presented to provide a logical framework for helping municipal officials and
researchers for development of a MSW management system with TSs in an

SC
economically optimal manner. Also, economical and environmental justifica-
tions have provided for TSs installation in urban centers through experiences
of municipal officials and literature survey. This methodology has produced an

U
optimum collection system design including best locations, capacity and number

AN
of TSs and details for the amount of MSW need to be transferred using primary
collection vehicles (PCVs) and secondary collection vehicles (SCVs) as a result.
In this study, city of Nashik (India) has chosen as a case study for demon-
M
strating the developed methodology in two distinct scenarios. Further, based
on distance measurements and estimation of waste load on generation point
D

sources, three distinct cases have been considered (see Section 4.3). Among
these three cases, case (c) has more realistic quotients such as (i) PCVs and
TE

SCVs use roads as mode of transportation for MSW; and (ii) estimation of
precise waste load on generation point sources by taking ward population into
consideration. This model (see Section 3.1.3) is written in AMPL (a mathemat-
EP

ical programming language) with KNITRO 5.2 solver. For a time horizon of five
years, this model has revealed that daily overall costs for MSW management are
Rs. 172178 (USD 2609), Rs. 175396 (USD 2658) and Rs. 175853 (USD 2665)
C

for cases (a), (b) and (c) in scenario (I) and Rs. 203337 (USD 3081) for case (c)
AC

in scenario (II) respectively. In order to find the optimum capacity of TSs, an

approach (see Section 5.1) has been introduced, which examine all possible com-
binations by running optimization model multiple times. Also, this approach
indicates that cost is optimized if the facilities are used fully and not partially
as per their capacities. After finding the optimum capacity of TSs, next step

33
 ACCEPTED MANUSCRIPT

was to find best locations for TSs. As can be seen in the Section 5.2, best loca-
tions are different from case (a), in the cases (b) and (c) for scenario (I). These

PT
differences enunciate the need of heterogeneous data compilation (see Section
4.3). Due to the change in the direction and distances from waste generation
sources to the proposed integrated MSW management facility, best locations for

RI
scenario (II) are different from scenario (I). To examine the economic viability
of TSs, a comparison between break even distance and frequency distribution of

SC
waste generation point sources has been shown (see Section 5.3). As far as eco-
nomic viability is concerned, Nashik should not have TSs if the persist existing
conditions by the end of 2020 i.e. scenario (I). However, in scenario (II) TSs

U
are economically feasible in the city of Nashik. Further, the amount of waste
generated at each source per day is found to be the most sensitive parameter
AN
in MSW management system. A slight change in the numeric value of waste
generation makes a significant change in the overall cost for MSW management.
M
Sensitivity analysis adduces the presence of uncertainty in the MSW manage-
ment system parameters. Therefore, this study may be extended by including
uncertain parameter within the same framework. Further, the algorithms for
D

uncertainty analysis may be used to solve uncertain TSS model.

TE

Acknowledgements

The authors would like to thank the Editor-in-Chief, the Subject Editor and
EP

the anonymous Reviewers for providing extremely insightful comments, which

has improved the quality of manuscript. The authors would also like to thank
Nashik Municipal Corporation (NMC), Surat Municipal Corporation (SMC)
C

and Municipal Corporation of Greater Mumbai (MCGM) for sharing valuable

AC

information. Moreover, authors wish to acknowledge central library, IIT Bom-

bay; Industrial Engineering and Operation Research (IEOR), IIT Bombay for
providing computational support.

34
 ACCEPTED MANUSCRIPT

Appendix A. TS Cost

TS cost calculations are provided in this section. The TS cost includes

PT
establishment cost, operation and maintenance cost. The establishment cost
for a TS in the city of Nashik is shown in Table A.3.

RI
Table A.3: Establishment cost for TSs with their respective capacities for Nashik.

S.N. Capacity (TPD) Cost( Rs./day)

SC
1 ≤ 100 2739
2 100 - 200 3424
3 200 - 300 4100

U
Operation and maintenance costs include electricity consumptions, water

AN
consumptions, pest control charges and salaries of cleaning staffs, supervisors,
operators. Electricity consumption is a major portion of operation and main-
tenance cost. Tariffs for electricity supplies vary from one state to another and
M
are controlled by state electricity distributors. The calculations for electricity
consumptions for a TS unit (same as SMC) in the city of Nashik has described
D

as follows:
Electricity charges for one TS unit = 15000 Rs./day (227 USD/day) in SMC,
TE

∵ running time for TS = 10 min/cycle and number of cycles per day = 25.
∴ total running time = 250 min = 4.166 hr.
EP

Table A.4: Electricity supply tariff for SMC (derived from Torrent (2015)).

S.N. Consumption Slab (kWh or Unit) Rs./unit

C

1 0 - 50 3.50
2 51 - 100 3.90
AC

3 101 - 200 4.25

4 201 - 500 4.60
5 ≥ 501 4.95

Tariff for Electricity supply for three-phase in SMC is given in the Table
A.4. In addition to this, Rs. 75 (USD 1.14) per month are charged as demand

35
 ACCEPTED MANUSCRIPT

charges for commercial uses (Torrent, 2015).

Table A.5: Electricity supply tariff for NMC (derived from Mahavitran (2015)).

PT
S.N. Consumption Slab (kWh or Unit) Rs./unit
1 0 - 200 5.85
2 ≥ 201 8.38

RI
Now, let x be the units consumed by a TS unit above 501 units.

SC
i.e. 50 × 3.5 + 50 × 3.9 + 100 × 4.25 + 300 × 4.6 + x × 4.95 + 2.5 = 15000
i.e. x ≈ 2590 units.
Total unit consumption = 50 + 50 + 100 + 300 + 2590 = 3090 kW h

U
Power of a TS unit = 3090 ÷ 4.166 ≈ 742 kW

AN
For NMC, electricity supply tariffs for commercial uses are given in Table
A.3. In addition to this, Rs. 190 (USD 2.88) per month are charged as demand
charges (Mahavitran, 2015).
M
Hence, total electricity charges for a TS unit in NMC per day = 200 × 5.85 +
2890 × 8.38 + 6.34 ≈ 25396 Rs. (USD 384.88).
D

References
TE

Ahmad, A., Jain, S.S., Parida, M., 2013. Updating road user cost and its im-
plication, in: 2 nd International Conference on Emerging Trends in Engineer-
ing & Technology, College of Engineering, Teerthanker Mahaveer University,
EP

Moradabad.

Asnani, P., 2006. Solid waste management. India Infrastructure Report

C

570. URL: http://www.iitk.ac.in/3inetwork/html/reports/IIR2006/

AC

Solid_Waste.pdf. accessed on 12 May 2015.

Badran, M., El-Haggar, S., 2006. Optimization of municipal solid waste man-
agement in Port Said - Egypt. Waste Management 26, 534–545.

Bharti, O.K., Bharti, B., Sood, V., 2015. Segregation at household level and
local composting not only helps protect environment but is also a cost effective

36
 ACCEPTED MANUSCRIPT

proposition for ULBs: A pilot study. Indian Journal of Applied Research 5,

436–441.

PT
Byrd, R.H., Nocedal, J., Waltz, R.A., 2006. Knitro: An integrated package for
nonlinear optimization, in: Large-scale nonlinear optimization. Springer, pp.
35–59.

RI
Census of India, 2011. Provisional population totals - India data sheet. Office

SC
of the Registrar General Census Commissioner, Government of India. Indian
Census Bureau. URL: http://censusindia.gov.in/. accessed on 12 May
2015.

U
Chalkias, C., Lasaridi, K., 2009. A GIS based model for the optimisation of

AN
municipal solid waste collection: The case study of Nikea, Athens, Greece.
WSEAS Transactions on Environment and Development 10, 11–15.

Chang, N.B., Lin, Y., 1997. Optimal siting of transfer station locations in
M
a metropolitan solid waste management system. Journal of Environmental
Science & Health Part A 32, 2379–2401.
D

Chatzouridis, C., Komilis, D., 2012. A methodology to optimally site and design
TE

municipal solid waste transfer stations using binary programming. Resources,

Conservation and Recycling 60, 89–98.

Coello, C.A.C., 2011. Learning and Intelligent Optimization. Springer.

EP

Cui, L., Chen, L., Li, Y., Huang, G., Li, W., Xie, Y., 2011. An interval-based
regret-analysis method for identifying long-term municipal solid waste man-
C

agement policy under uncertainty. Journal of Environmental Management 92,

AC

1484–1494.

Datta, Pranati, 2006. Urbanisation in india, in: European Population Confer-

ence, Bratislava, Slovak Republic, June, pp. 21–24.

Dhingra, S.L., Patil, G.R., Sasane, G.S., 2011. Transport Infrastructure As-
sessment and Planning for Integrated Municipal Solid Waste Management

37
 ACCEPTED MANUSCRIPT

System for Mumbai Metropolitan Region (MMR). Technical Report. Mum-

bai Metropolitan Region Development Authority and Department of Civil

PT
Engineering Indian Institute of Technology Bombay, Mumbai.

Eiselt, H.A., Marianov, V., 2014. A bi-objective model for the location of
landfills for municipal solid waste. European Journal of Operational Research

RI
235, 187–194.

Fourer, R., Gay, D.M., Kernighan, B.W., 1990. A modeling language for math-

SC
ematical programming. Management Science 36, 519–554.

Gallardo, A., Carlos, M., Peris, M., Colomer, F., 2015. Methodology to design a

U
municipal solid waste generation and composition map: A case study. Waste
Management 36, 1–11.
AN
Ghose, M., Dikshit, A.K., Sharma, S., 2006. A GIS based transportation model
for solid waste disposal: A case study on Asansol municipality. Waste Man-
M
agement 26, 1287–1293.

Greco, G., Allegrini, M., Del Lungo, C., Savellini, P.G., Gabellini, L., 2014.
D

Drivers of solid waste collection costs. Empirical evidence from Italy. Journal
of Cleaner Production , In Press.
TE

Jenkins, L., 1982. Developing a solid waste management model for Toranto.
Information Systems and Operational Research 20, 237–247.
EP

JnNURM, 2011. Rapid Training Programme, Preparation of DPRs, Module 2.2:

Demand Assessment URL: http://jnnurm.nic.in/. accessed on 12 August
2015.
C

JnNURM, 2013. City Report: Surat Solid Waste Management Project under
AC

Jawaharlal Nehru National Urban Renewal Mission URL: http://jnnurm.

nic.in/. accessed on 11 June 2015.

Karnavas, W.J., Sanchez, P.J., Bahill, A.T., 1993. Sensitivity analyses of con-
tinuous and discrete systems in the time and frequency domains. Systems,
Man and Cybernetics, IEEE Transactions on 23, 488–501.

38
 ACCEPTED MANUSCRIPT

Karp, R.M., 1972. Reducibility among combinatorial problems. Springer.

Kirca, Ö., Erkip, N., 1988. Selecting transfer station locations for large solid

PT
waste systems. European Journal of Operational Research 35, 339–349.

Komilis, D., 2008. Conceptual modeling to optimize the haul and transfer of

RI
municipal solid waste. Waste Management 28, 2355–2365.

Kulcar, T., 1996. Optimizing solid waste collection in Brussels. European

SC
Journal of Operational Research 90, 71–77.

Kumar, V., Pandit, R., 2013. Problems of solid waste management in Indian

U
cities. International Journal of Scientific and Research Publications 3, 1–9.

AN
Longley, P., 2005. Geographic information systems and science. John Wiley &
Sons.

Lyeme, A., 2011. Optimization of municipal solid waste management system.

M
Master’s thesis. Department of Mathematics, University of Dar es salaam,
Tanzania.
D

Mahavitran, 2015. Maharashtra state electricity distribution co. ltd. URL:

TE

http://www.mahadiscom.in/tariff/Tariff-Booklet-aug-2012.pdf. ac-
cessed on 01 March 2015.

Marks, D.H., Liebman, J., 1970. Mathematical analysis of solid waste collection,
EP

in: Public Health Service Publication. Departement of Health, Education and

Welfare, Washington, D.C. volume 2104.
C

Mirchandani, P.B., Francis, R.L., 1990. Discrete location theory. Wiley-

Interscience.
AC

Murray, W., Shanbhag, U.V., 2006. A local relaxation method for nonlinear
facility location problems, in: Multiscale optimization methods and applica-
tions. Springer, pp. 173–204.

39
 ACCEPTED MANUSCRIPT

Nagy, G., Salhi, S., 2007. Location-routing: Issues, models and methods. Eu-
ropean Journal of Operational Research 177, 649–672.

PT
O’Malley, P., 2003. Controlling odor at transfer stations and MRFs: There is
more than one way to reduce complaints. MSW Management 13, 44–52.

RI
Or, I., Curi, K., 1993. Improving the efficiency of the solid waste collection
system in Izmir, Turkey, through mathematical programming. Waste Man-

SC
agement & Research 11, 297–311.

Owen, S.H., Daskin, M.S., 1998. Strategic facility location: A review. European
Journal of Operational Research 111, 423–447.

U
Pattnaik, S., Reddy, M.V., 2010. Assessment of municipal solid waste man-
AN
agement in Puducherry (Pondicherry), India. Resources, Conservation and
Recycling 54, 512–520.
M
Phelps, H., Heinke, G., Jonker, J., Ouano, E., Vandecasteele, C., 1996. Man-
agement of solid wastes. Technical Report. UNESCO: Series of Learning
Materials in Engineering Sciences. Paris.
D

Rafiee, R., Khorasani, N., Mahiny, A.S., Darvishsefat, A.A., Danekar, A.,
TE

Hasan, S.E., 2011. Siting transfer stations for municipal solid waste using
a spatial multi-criteria analysis. Environmental & Engineering Geoscience
17, 143–154.
EP

Ramachandra, T., 2006. Management of municipal solid waste. TERI press.

Ramachandra, T., 2011. NPTEL, IISc Bangalore, Municipal Solid Waste Man-
C

agement, Lecture 1: Introducing Municipal Solid Waste Management .

AC

Ramachandra, T., Bachamanda, S., 2007. Environmental audit of municipal

solid waste management. International Journal of Environmental Technology
and Management 7, 369–391.

40
 ACCEPTED MANUSCRIPT

Ravindra, K., Kaur, K., Mor, S., 2015. System analysis of municipal solid waste
management in Chandigarh and minimization practices for cleaner emissions.

PT
Journal of Cleaner Production 89, 251–256.

Ross, A., 1999. Road accidents: A global problem requiring urgent action.
Roads & Highways Topic Note RH-2, World Bank .

RI
Sharholy, M., Ahmad, K., Mahmood, G., Trivedi, R., 2008. Municipal solid

SC
waste management in Indian cities: A review. Waste Management 28, 459–
467.

Sonesson, U., 2000. Modelling of waste collection: A general approach to cal-

U
culate fuel consumption and time. Waste Management and Research 18,
115–123.
AN
Sturges, H.A., 1926. The choice of a class interval. Journal of the American
Statistical Association 21, 65–66.
M
Tiwari, G., 2002. Urban transport priorities: Meeting the challenge of socio-
economic diversity in cities, a case study of Delhi, India. Cities 19, 95–103.
D

Torrent, 2015. Tariff schedule for Surat license area of torrent power lim-
TE

ited - Surat. URL: http://www.torrentpower.com/customers/customers_

surat_tariff.php. accessed on 27 February 2015.
EP

United Nations, 2014. World urbanization prospects: The 2014 revision. De-
partment of Economic and Social Affairs, United Nations Publications. URL:
http://esa.un.org/unpd/wup/highlights/wup2014-highlights.pdf. ac-
C

cessed on 19 April 2015.

AC

USEPA, 1995. Decision-makers guide to solid waste management, volume ii.

United States Environmental Protection Agency (EPA) .

USEPA, 2002. Waste transfer stations: A manual for decision-making. United

States Environmental Protection Agency (EPA) .

41
 ACCEPTED MANUSCRIPT

Woodward, C., Grant, K., Maxwell, R., 2002. Applications of sensitivity analysis
to uncertainty quantification in variably saturated flow. Developments in

PT
Water Science 47, 73–80.

World Bank, 2005. India: Road transport service efficiency study. Document
of the World Bank, South Asia Regional Office, Energy & Infrastructure

RI
Operations Division.

U SC
AN
M
D
TE
C EP
AC

42
 ACCEPTED MANUSCRIPT

Highlights:

• An approach to study the feasibility of MSW transfer station siting is proposed

• A nonlinear facility location framework is developed to identify best sites
• A GIS based shortest routing network is proposed to represent the data inventory
•

PT
Three scenarios are considered to address heterogeneity in waste generation data
• A comprehensive parametric sensitivity analysis is performed
• Proposed generic framework is demonstrated on the Nashik City in India

RI
U SC
AN
M
D
TE
C EP
AC