Semana 07 - Lectura - Artículo 03

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 48

Accepted Manuscript

Combining operational research and Life Cycle Assessment to optimize municipal


solid waste collection in a district in Lima (Peru)

Alessandro Gilardino, Jonatan Rojas, Héctor Mattos, Gustavo Larrea-Gallegos, Ian


Vázquez-Rowe

PII: S0959-6526(17)30700-X
DOI: 10.1016/j.jclepro.2017.04.005
Reference: JCLP 9356

To appear in: Journal of Cleaner Production

Received Date: 1 December 2016


Revised Date: 28 March 2017
Accepted Date: 1 April 2017

Please cite this article as: Gilardino A, Rojas J, Mattos Hé, Larrea-Gallegos G, Vázquez-Rowe I,
Combining operational research and Life Cycle Assessment to optimize municipal solid waste collection
in a district in Lima (Peru), Journal of Cleaner Production (2017), doi: 10.1016/j.jclepro.2017.04.005.

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

1 Combining operational research and Life Cycle Assessment to optimize


2 municipal solid waste collection in a district in Lima (Peru)
3 Alessandro Gilardinoa,*, Jonatan Rojasa, Héctor Mattosa, Gustavo Larrea-Gallegosb, Ian
4 Vázquez-Roweb
a

PT
5 Department of Engineering, Pontificia Universidad Católica del Perú, 1801 Avenida
6 Universitaria, San Miguel, Lima 32, Peru
b
7 Peruvian LCA Network, Department of Engineering, Pontificia Universidad Católica del

RI
8 Perú, 1801 Avenida Universitaria, San Miguel, Lima 32, Peru

9 *Corresponding author: Alessandro Gilardino; e-mail: agilardino@pucp.pe

SC
10 Abstract

11 Waste management in cities has proved to be a complex task for policy-makers who seek

12
U
cost reductions when taking decisions. However, on many occasions substantial
AN
13 environmental gain can be attained through waste management optimization. In this
M

14 particular study, the main objective was to combine operational research techniques with

15 Life Cycle Assessment to locate collection sites for general and recyclable residues and
D

16 create an effective collection-route system for compactor trucks to attain a reduction in


TE

17 environmental impacts. The case study selected was the district of Comas, one of the most

18 populated districts in the city of Lima, a city that has experienced several episodes of poor
EP

19 waste collection management in recent years. For the location of the collection sites a

20 mathematical optimization model was proposed to decide where to locate them, as well as
C

21 the number of containers required at each site, considering that general and recyclable
AC

22 residues would not be collected at the same site due to the restricted space available. For

23 the creation of collection routes, a heuristic approach based on the Vehicle Routing

24 Problem was constructed, taking into account the available working time of each collection

25 truck to assign them collection routes and minimize the number of compactor trucks. These

1
ACCEPTED MANUSCRIPT

26 results, when combined with LCA, focused mainly on greenhouse gas (GHG) emissions,

27 air pollution impact categories and the depletion of abiotic resources. In addition, a

28 sensitivity analysis was developed by varying three different parameters with respect to the

29 proposed baseline scenario, comparing the several impact categories. Computed results

PT
30 show that implementing a container collection system reduces the number of compactor

RI
31 trucks required by up to 50% as compared to the current door-to-door collection system.

32 Moreover, GHG emissions where reduced by almost 14% with the proposal, mostly due to

SC
33 a 41% reduction in distance traveled. Reductions in terms of particulate matter formation

34 and photochemical oxidant formation were even more significant, especially considering

35
U
that air pollution is an important health hazard in the district analyzed, whereas most
AN
36 scenarios showed a slight increase in resource depletion due to the inclusion of container

37 infrastructure in the system.


M

38 Keywords: allocation optimization; heuristic; industrial ecology; LCA; Peru; waste


D

39 management.
TE

40 1. Introduction

41 One of the major problems of the ever-growing world population is that year after
EP

42 year higher amounts of residues are generated due to increasing consumption rates

43 (Hoornweg et al., 2013). These rates directly impact the depletion of raw materials and
C

44 environmental pollution due to lack of correct recirculation flows (Koppelaar and Wiekard,
AC

45 2013). As a counter measure, several countries have increased their spending on

46 environmental protection programs, including waste management programs (DEFRA,

47 2016; Institute for Protection and Environmental Investigation, 2016). For instance, the

48 European Union’s environmental protection programs have increased their budget by

2
ACCEPTED MANUSCRIPT

49 almost 40% in the period 2003-20131. Interestingly, by 2013, 41% of this budget

50 corresponded to waste management activities (European Union, 2016). As a result, the

51 combination of several measures within these waste management programs has allowed the

52 European Union to decrease its waste production per capita almost 10% in the period 1995-

PT
53 2014 (European Union, 2016a, 2016b).

RI
54 Governments and private waste management institutions are continuously looking

for innovative techniques to optimize resources such as fossil fuels, materials and costs

SC
55

56 (Laforest, 2014). For instance, in the collection stage of waste management certain studies

U
57 have developed methods to optimize the scheduling of trucks (Jing and Borestein, 2008). In
AN
58 this context, the best known algorithm is the Traveling Salesman Problem (TSP) which

59 seeks the cheapest route to visit all the clients in a group and return to the starting point at
M

60 the end (Applegate et al., 2007; Matai et al., 2010). This algorithm, which was developed in

61 the XVII century, derived in the creation of many other routing algorithms, such as: i) the
D

62 Vehicle Routing Problem (VRP), which is a variation that considers the maximum capacity
TE

63 of the vehicles (Toth et al., 2002); ii) the Vehicle Routing Problem with Time Windows,

64 which takes into account a specific time interval to visit each customer (Kallehauge, 2008);
EP

65 and, iii) the Period Vehicle Routing Problem (PVRP) (Francis et al., 2006). The latter is
C

66 very similar to the VRP; however, the main difference is that it creates a visiting schedule
AC

67 for several days or other time dimensions in a specific period of time (Christofides and

68 Beasley, 1984).

69 These examples represent only a small sample of a large pool of routing algorithms

70 that can be solved using exact methods (Pillac et al., 2013). However, in most cases it is

1
The increase rate refers to nominal monetary values, without considering inflation-adjusted values.

3
ACCEPTED MANUSCRIPT

71 necessary to develop heuristic methods based on these exact algorithms to solve routing

72 problems, although there are also other techniques that can be used such as stochastics

73 methods (Gendreau et al., 1996) and simulation (Li et al., 2008). The need to use these

74 methods is linked to the fact that the larger the number of clients to be served, the more

PT
75 complicated it will be to solve the problem (Tore and Irnich, 2005). In this sense, several

RI
76 heuristic methods have been developed in the literature (Braekers et al., 2016). For

77 instance, the one proposed by Ho et al. (2008) developed a genetic algorithm for the multi-

SC
78 depot VRP. In Tas et al. (2014) a tabu search algorithm for the VRP with flexible Time

79 Windows (VRPTW) was proposed. Another research article published by Solomon (1987)

80
U
presented several heuristics to solve the VRPTW. Finally, Li et al. (2003) proposed a
AN
81 metaheuristic algorithm to solve the VRPTW. Heuristics algorithms can also be found

82 applied to waste collection, such as in Texeira et al. (2004) and Huang et al. (2015), who
M

83 proposed a heuristic algorithm, to create a collection route for recyclable residues and to
D

84 collect waste from collection points, respectively.


TE

85 Similarly, several mathematical algorithms have been proposed as decision criteria

86 to locate containers in residential areas when shifting from a door-to-door waste collection
EP

87 system to the use of containers (Tralhao et al., 2010). This change has been shown to save
C

88 at least 30% of costs in the collection stage, according to interviews with municipal waste
AC

89 managers (technical staff from the Municipalities of Comas, Rimac and San Borja, personal

90 communication, March 2016; technical staff from Contenur, personal communication, May

91 2016).

92 Currently, it is developing and emerging countries which present the most critical

93 problems in terms of waste management due to the lack of buoyant economic resources and

4
ACCEPTED MANUSCRIPT

94 skilled labor (Burnley, 2007). In the specific case of Peru, for instance, only 42% of the

95 territory benefits from waste management plans. However, most of these existing plans lack

96 efficient management systems (INEI, 2014), despite the large increase in the amount of

97 waste generated. For instance, in Lima, the capital city, alone, the total amount of waste

PT
98 generated has increased from 1412 t in 2000 to 2828 t in 2014, a 100% increase (INEI,

RI
99 2015). In addition, in those districts where waste management plans do exist, only a

100 minority collect residues in a segregated way, complying with the national technical

SC
101 specification that considers five different types of non-hazardous residues: organic, plastics,

102 paper and cardboard, glass and metals (NTP, 2005). Another important issue is the lack of

103
U
certified sanitary landfills. According to OEFA (2014), there are only 10 certified sanitary
AN
104 landfills in the country, five of which are located in Lima. Hence, many urban areas of the

105 nation have to use illegal dumpsters to dispose of their residues (OEFA, 2014).
M

106 Regardless of the high costs linked to erratic waste management and policy, these
D

107 tend to be related to increased environmental impacts. For instance, Kinobe et al. (2015)
TE

108 demonstrated that optimizing waste collection and disposal in Kampala (Uganda) could

109 reduce GHG emissions significantly. In fact, the use of environmental management
EP

110 methodologies, such as Life Cycle Assessment – LCA (ISO 2006a), which has been
C

111 applied on several occasions with the objetive of identifying environmental impacts linked
AC

112 to waste management activities, has been shown to provide detailed insights in terms of

113 identifying environmental hotspots and sustainable alternatives in waste management. For

114 instance, Muñoz et al. (2004) conducted an LCA to assess seven possible waste

115 management scenarios for the province of Gipuzkoa (Basque Country – Spain) for the 2016

116 time horizon. Rives et al. (2010) performed an LCA to quantify and compare the

5
ACCEPTED MANUSCRIPT

117 environmental impact of 14 types of waste containers commonly used in waste

118 management, whereas Iriarte et al. (2008) assessed the environmental impacts of three

119 selective waste collection systems commonly used in densely populated areas: mobile

120 pneumatic, multi-container and door-to-door collection. Finally, Pukkinen et al. (2012)

PT
121 have used LCA to compare the feasibility of shifting from a door-to-door collection to a

RI
122 pneumatic alternative, demonstrating that although emissions would be outsourced form

123 residential areas with the pneumatic alternative, the overall impact would increase under

SC
124 this management system. However, as far as we were able to ascertain, there is currently no

125 literature available in which operational research techniques have been combined with

126
U
robust environmental management methods (e.g., LCA) to compute environmental gains in
AN
127 the waste management sector.
M

128 Based on this discussion, the main aim of this paper was to consider the problem of

129 locating waste collection sites in a district in northern Lima proposing a mathematical
D

130 programming model and the creation of adequate collection routes for municipal waste
TE

131 compactor trucks. In addition, LCA was used to analyze the environmental benefits of these

132 changes in the waste management plan in the period 2016-2022. This study is addressed to
EP

133 municipal waste management decision-makers to help them in policy making and allow
C

134 them to prove that container collection systems are environmentally more efficient than
AC

135 door-to-door collection.

136 2. Materials and Methods

137 2.1 Studied area information and current waste management system

6
ACCEPTED MANUSCRIPT

138 The studied area was the district of Comas, a lower middle class area located in

139 northern Lima, with a total extension of 48.7 km2 (APEIM, 2015). In 2015, the forecasted

140 population of the district was 524,894, making it one of the most populated in Peru (INEI,

141 2009). According to the latest characterization study developed in the district of Comas in

PT
142 2014, waste generation per capita was 0.54 kg/per-day, with a density of 133.9 kg/m3

RI
143 (Municipality of Comas, 2014a). This generation was based on a study performed in June

144 2014 by the Municipality of Comas in which a total of 100 households were randomly

SC
145 sampled to analyze waste generation and composition in the district. Table 1 presents this

146 composition, segregating the recyclable residues from general residues. It should be noted

147
U
that per capita generation represents the amount of waste produced in households,
AN
148 excluding that of industries, markets and street cleaning (Municipality of Comas, 2014b).

149 Conventional collection routes analyzed in the current study do not target waste engendered
M

150 in these excluded sectors.


D

151 Table 1. Household waste composition in the district of Comas in 2014 (Municipality of
TE

152 Comas, 2014a).

Total waste generation in metric


EP

Residue Percentage (%) tons for 2014


(t/year)
General 91.61 92,753
C

Tetrapak / Cans 1.75 1772


Paper 3.50 3544
AC

Glass 1.89 1914


Plastic 1.25 1266

153 Population growth is also an important parameter to be taken into account within the

154 framework of the study, since it will make it possible to predict the total amount of waste

155 generated in the district in the time horizon of the study (i.e., 2016-2022). Hence, based on

7
ACCEPTED MANUSCRIPT

156 the data provided by the National Statistics Institute (INEI, 2009), who forecasted the

157 population of the district up to 2015 based on the 2009 census, a linear regression technique

158 was applied to forecast the annual increase up to 2022, as described by Gujarati and Porter

159 (2010). In addition, despite the fact that the per capita generation of waste for the district is

PT
160 known, it was necessary to forecast the per capita generation for the time horizon of the

RI
161 study. A limitation observed in this respect was the fact that the Municipality lacked a

162 forecasting method to calculate future increases in waste generation per capita, as well as

SC
163 changes in waste composition. However, a recent publication from INEI regarding

164 increases in waste generation in districts of Lima in the period 2012-2015 suggests that the

165
U
annual augmentation can be fixed at approximately 1% (INEI, 2016). Hence, it was
AN
166 assumed that this value could be used as a proxy throughout the projected time period (see

167 Table 2).


M

168 Table 2. Population, per capita generation and total waste generation forecast for the
D

169 district of Comas in the period 2016-2022 (INEI, 2009).


TE

Waste generation per


Total waste generation
Year Population capita
(metric ton/day)
EP

(kg/day)
2016 531,406 0.543 288.76
2017 534,675 0.549 293.44
2018 537,944 0.554 298.18
C

2019 541,213 0.56 303.00


AC

2020 544,482 0.565 307.87


2021 547,751 0.571 312.57
2022 551,020 0.576 317.58
170

171 Currently, waste in Comas is collected unsorted on a daily basis with a door-to-door

172 collection system. A total of 18 compactor trucks are in charge of collecting the waste

173 throughout the district. The Municipality provided spatial information regarding the 58

8
ACCEPTED MANUSCRIPT

174 routes that are currently performed by the trucks (Leoncio Sicha Punil, Public Cleaning

175 Supervisor, Municipality of Comas, personal communication, April 2015). The total

176 distance travelled was estimated as being 431 km per day and truck, which was calculated

177 by measuring every collection route. The collection service is provided by a private

PT
178 company who charges approximately 33 USD to the city council per metric ton including

RI
179 the cost of the landfill.

2.2 Model formulation for collection sites

SC
180

181 2.2.1 Clustering and maximum distance considered

182
U
The location of waste disposal units, known as containers, constitutes a nonconvex
AN
183 problem due to the large number of dwellings to serve (Feldman et al., 1966). One way to
M

184 solve this complexity is through the use of demand aggregation (Francis et al., 2004; Jang

185 and Lee, 2015). There are several ways to aggregate dwellings. In Tralhao et al. (2010),
D

186 dwellings were aggregated into 20 meter long linear sectors and 10 meter radius node
TE

187 sectors. Both in Ghiani et al. (2012) and Ghiani et al. (2014) dwellings were grouped

188 according to their position; a cluster can group all the citizens of a street, a portion of it, or
EP

189 a single home. In any of these three cases, the waste generation of a cluster or sector was

190 the sum of the waste generation of each dwelling assigned to the cluster. Thereafter, the
C

191 distance from a cluster to a candidate collection site was calculated from the centroid of the
AC

192 cluster.

193 In the current case study, all dwellings have been clustered considering a linear

194 distance range from 40 to 70 meters, mainly due to the fact that almost all dwellings are one

195 or two-storey houses. In other words, the lower distance range corresponds to a high

9
ACCEPTED MANUSCRIPT

196 density of dwellings, whereas areas with lower density are assigned a higher distance range.

197 Based on this rationle, a total of 97,659 dwellings in the studied area were grouped into

198 5,473 clusters. Tralhão et al (2010) and Zhou et al (2005) used 200 meters as the threshold

199 distance from the centroid of each cluster to a candidate collection site. This value was the

PT
200 one which appeared to be adequate for our case of study and was therefore assumed as the

RI
201 threshold distance.

2.2.2 Types of containers considered

SC
202

203 Based on economic factors, only surface rear-loading containers were considered

U
204 for the assessment. In addition, it should be noted that general and recyclable residue
AN
205 collection sites should be located separately. The main reason for this assumption is based

206 on the outstanding difference between the per capita generation of general waste and the
M

207 per capita generation of recyclable waste. Nevertheless, four different types of recyclable
D

208 residues were taken into account: glass, paper, tetrapak/cans and plastic. A daily collection
TE

209 was assumed for general residues, whereas two different collection frequencies were

210 considered for the recyclable residues: paper collection was modeled on a three times per
EP

211 week basis, and the remaining recyclable residues were modelled for twice per week. These

212 frequencies were considered taking into account the higher per capita generation of paper
C

213 as compared to other recyclable residues (Municipality of Comas, 2014; Municipality of


AC

214 Ate, 2014; Municipality of Breña, 2016).

215 For general residues, 1100 L containers were considered. Three different types of

216 container arrangements were considered at each collection site: two, three or four

217 containers. Regarding recyclable residues, 800 L containers were considered; two or four

10
ACCEPTED MANUSCRIPT

218 containers were considered at each site. In the first case, half a container was assigned to

219 each of the four types of recyclable residues. For the second case, one container was

220 assigned to each one.

PT
221 2.2.3 Methodological assumptions

222 For the sake of clarity and considering the extension of the territory analyzed, the

RI
223 area of the district was divided into 9 different plots. Table 3 shows the number of clusters,

SC
224 candidate collection sites, dwellings and citizens in each of these zones.

225 Table 3. Total number of clusters, candidate collection sites, dwellings and citizens per

226 zone.
U
AN
Number of candidate Number of dwellings
Zone Number of clusters Population per zone
collection sites per zone
M

1 283 71 5251 29,628


2 476 97 8850 49,934
D

3 603 177 12,684 71,567


4 299 37 6903 38,949
TE

5 386 91 6539 36,895


6 526 180 9437 53,246
7 1392 364 21,643 122,116
EP

8 1275 308 22,017 124,226


9 233 50 4335 24,459
TOTAL VALUES 97,659 551,020
C

227
AC

228 2.2.4 Model presentation

229 The mathematical model was built to assure that a cluster of citizens is always

230 allocated to the closest collection site, which was not necessarily the case in the previous

231 version of the model. Table 4 presents the sets, parameters and variables relevant for the

232 mathematical model.

11
ACCEPTED MANUSCRIPT

233 Table 4. Description of sets, parameters and variables used in the mathematical model.

Symbol Description
Sets
C Set of clusters
S Set of candidate collection sites

PT
M Set of container arrangements for general residues
L Set of type of recyclable residues

RI
P Set of container arrangement for recyclable residues
Parameters
cm Capacity of a general residues container arrangement m  

SC
P_max Maximum occupation percentage for a general residues container
cp Capacity of a recyclable residues container arrangement   
P_maxrec Maximum occupation percentage for a recyclable residues container

U
qi Daily general residues generation for cluster   
q_recil Daily recyclable residues generation of type   for cluster   
AN
dij Distance form a cluster    to a candidate collection site

D_max Maximum threshold distance (in our case D_max = 200 meters)
Const A constant, equals to 0.25
M

Variables
δi Maximum distance variation from a cluster    to a general residues collection site
δ_reci Maximum distance variation from a cluster    to a recyclable residues collection site
D

Binary variable that takes value of 1 if cluster    will be allocated to general collection site
xij

 , 0 otherwise.
Binary variable that takes value of 1 if cluster    will be allocated to recyclable collection site
TE

x_recij

 , 0 otherwise.
Binary variable that takes value of 1 if candidate collection site
 is activated with an
zjm
arrangement  , 0 otherwise.
Binary variable that takes value of 1 if candidate collection site
 is activated for recyclable
EP

yjlp
residue type   with an arrangement   , 0 otherwise.
234
C

235    ∑ ∑  + ∑ ∑ ∑  (1)
AC

236 ∑  = 1,    (2)

237 ∑  ≤ 1 ,


 (3)

238 ∑# "  ≤ $∑  % &_ (,


 (4)

239 ∑):,-. /,-0 ) ≤ 1 − ∑    ,


 (5)

12
ACCEPTED MANUSCRIPT

240 1  ≤ 2_ ( + 3 ,   ,


 (6)

241 0 ≤ 3 ≤ 30 (7)

242 ∑ _6% = 1,    (8)

PT
243 ∑  = 1,   ,
 (9)

244  − 7 = 0,   ,   ,  ,


 ,  ≠ (10)

RI
245 ∑# "6 _6% ≤ $∑  % &_ (6%,   ,
 (11)

SC
246 ∑):,-. /,-0 _6%) ≤ 1 − (∑ ∑  ); <=   ,
 (12)

1 _6% ≤ 2_ ( + 3_6% ,   ,




U
247 AN (13)

248 0 ≤ 3_6% ≤ 30 (14)

249 ∑ ∑  ; <= + ∑  ≤ 1,


 (15)
M

250 The objective function minimizes the total number of general and recyclable
D

251 collection sites. Constraints (2) to (7) are aimed to fix the location of the general
TE

252 residuescollection sites. Constraint (2) ensures that every cluster    is allocated to one

253 and only one collection site within the threshold distance (D_max). Constraint (3)
EP

254 determines that if the candidate collection site


 is activated, then only one type of

255 container arrangement   can be placed. Constraint (4) limits the capacity, determining
C

256 the maximum occupation percentage of containers. Constraint (5) makes it necessary to
AC

257 allocate each cluster to the closest collection sites, ensuring that citizens will walk as little

258 as possible to dispose of their waste. When variable  takes a value of 1, all variables )

259 which fulfill the equation 1 + 3 < 1) will take values of 0 in a way that every cluster

260    will be allocated to the closest collection site


 . Constraint (6) ensures that

261 distances between clusters and their respective collection sites are equal or lower than a
13
ACCEPTED MANUSCRIPT

262 maximum distance D_max. More specifically, this constraint is considered a value of 3 ,

263 which is the maximum distance variation allowed between a cluster and a collection site.

264 This means that if a cluster cannot be allocated to a collection site within the threshold

265 distance D_max, it can still be allocated to a collection site within a maximum threshold

PT
266 distance equal D_max + 3 . Constraint (7) determines the maximum value of the variable

RI
267 3 , which in the case study is 30 meters.

SC
268 Constraints (8) to (14) are aimed at fixing the location of recyclable residue

269 collection sites. Firstly, constraint (8) ensures that every cluster    is allocated to one and

U
270 only one collection site within the threshold distance (D_max). Secondly, constraint (9)
AN
271 implies that if the potential collection site
 is activated, for every type of residue  

272 one and only one type of container arrangement    can be placed at that site. Thereafter,
M

273 constraint (10) means that if the candidate collection site


 is activated, then all types of

274 recyclable residues must have the same container arrangement, that is, if for residue  = 1
D

275 container arrangement  = 1 is used, then for residues  = 2, 3, 4 p = 1 must be used in the
TE

276 same type of container arrangement. Constraint (11) is a capacity constraint which
EP

277 determines the maximum occupation percentage of containers. Constraint (12) makes it

278 necessary to allocate each cluster to the closest collection site ensuring that citizens will
C

279 walk as little as possible to leave their waste. When variable  takes value of 1, all
AC

280 variables _6%) which fulfill the equation 1 + 3_6% < 1) will take values of 0 in a

281 way that every cluster    is allocated to the closest collection site
 . Constraints (13)

282 and (14) work the same way as constraints (6) and (7), respectively.

14
ACCEPTED MANUSCRIPT

283 Finally, constraint (15) ensures that if at a candidate collection site


 a general

284 residue collection site is placed, that same spot cannot host a recyclable residue collection

285 site or vice versa.

PT
286 2.3 Model formulation for collection routes

287 The collection stage also includes the creation of routes for compactor trucks.

RI
288 Therefore, an algorithm is proposed to create those routes and minimize the total distance.

SC
289 2.3.1 Methodological assumptions

U
290 The estimated amount of residues in every collection site was obtained based on the
AN
291 citizens allocation developed in the optimization model. Moreover, the distances between

292 every collection site were obtained using the Floyd – Warshall algorithm (Aini et al, 2012).
M

293 Thereafter, an average speed of 20 km/h was considered to estimate the traveling time

294 between collection sites for compactor trucks in charge of collecting general residues (i.e.,
D

295 organics). In addition, the service time for each collection site was estimated considering
TE

296 the number of containers in every collection site. For the compactor trucks a capacity of 20

297 m3 (Geesinknorba, 2016), a compaction ratio of 3:1 and a maximum occupation percentage
EP

298 of 95% were considered. Finally, two different shifts (8 and 12 h) per day were considered.
C

299 Regarding recyclable residues, non-motorized vehicles represent the current way in which
AC

300 they are collected in other districts of Lima that do recycle their residues. Therefore, the

301 collection of these residues was assumed to be done in this manner.

302 2.3.2 The heuristic approach

15
ACCEPTED MANUSCRIPT

303 For the case study presented in this article, a heuristic approach was developed to

304 solve the VRP by adding one constraint: every vehicle has a maximum time available to

305 visit customers. This approach allows the optimization of the number of vehicles needed to

306 serve all customers, because a vehicle can return more than once to the depot. In addition, it

PT
307 allows the identification of the number of routes that a vehicle will have to serve regarding

RI
308 the available time. Therefore, the following pseudocode was coded in Matlab (Mathworks,

309 2016):

SC
310 1: procedure CreateCollectionRoutes (number_of_clients, Time, Residues, Set_clients,
311 Service_Time)
312 2: VisitedClients ← zeros
U
AN
313 3: FindRoute ← null
314 4: " ← Residues
M

315 5: i ← warehouse
316 6: = ← Time
D

317 7: InsertedClients ← null


318 8: < ← Service_Time
TE

319 9: insertionCost ← null


320 10: minimumCost ← +∞
EP

321 11: traveltime ← null


322 12: repeat
13: for all j ϵ Set_clients do
C

323
324 14: if B<=1 =<C = 0 and insertionCost < minimumCost then
AC

325 15: insertionCost ← insertionCost + = + < + =,DEFG)HIJG - =,DEFG)HIJG


326 16: minimumCost ← insertionCost
327 17: end if
328 18: end for
329 19: traveltime = traveltime + = + < + =,DEFG)HIJG - =,DEFG)HIJG
330 20: if traveltime < Max_Time and " < residualCapacity

16
ACCEPTED MANUSCRIPT

331 21: residualCapacity ← residualCapacity - "


332 22: p←j
333 23: FindRoute ← FindRoute ⋃ { j }
334 24: InsertedClients ← InsertedClients + 1
335 25: else if traveltime < Max_time and " > residualCapacity

PT
336 26: i ← warehouse
337 27: residualCapacity ← Max_capacity

RI
338 28: traveltime ← traveltime - = - < - =,DEFG)HIJG + =,DEFG)HIJG
339 29: FindRoute ← FindRoute ⋃ {warehouse}

SC
340 30: else if traveltime > Max_time
341 31: i ← warehouse

U
342 32: residualCapacity ← Max_capacity
traveltime ← traveltime - = - < - =,DEFG)HIJG + =,DEFG)HIJG
AN
343 33:
344 34: FindRoute ← FindRoute ⋃ {warehouse}
345 35: // a new truck will be required or a new schedule will start, a new route will be
M

346 created
347 36: end if
D

348 37: until InsertedClients = number_of_clients


TE

349 38: return FindRoute


350 39: end procedure
EP

351 2.4 Environmental assessment of the proposed waste management system

352 An internationally standardized environmental assessment methodology, LCA, was


C

353 used to compare the environmental impacts of the current and proposed waste management
AC

354 collection systems. The framework used to conduct the study was the ISO 14040/14044

355 standards (ISO, 2006a, 2006b).

356 The main goal of the LCA was to determine the environmental trade-offs of

357 implementing a container collection system for waste in the district of Comas in the period

17
ACCEPTED MANUSCRIPT

358 2016-2022. Hence, the function of the system was defined as the need to optimize the

359 collection of municipal solid waste during the time horizon (2016-2022) in the district of

360 Comas, Lima. Based on this functionality, the selected functional unit (FU) was the

361 collection of one (1) metric ton of waste (i.e., including general and recyclable wastes) in

PT
362 the analyzed area.

RI
363 The system boundary makes it possible to set an interface between the product

system and the environment and define which unit processes will be included in the

SC
364

365 assessment (Rivela et al., 2006). The system boundary for current and proposed waste

U
366 collection systems is shown in Figure 1. Production of working clothes, tools and security
AN
367 features such as shovels, helmets, gloves, boots, etc. were excluded from the system due to

368 lack of information, as well as the expected low relevance of these items in the overall
M

369 system (Muñoz et al., 2014). Moreover, human labor was excluded from the boundary

370 based on common practice in LCA (Rugani et al., 2012).


D
TE

371 Figure 1. Graphical representation of the system boundary of the production system under

372 assessement.
C EP
AC

18
ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE
C EP
AC

373

374 It should be noted that in comparative LCA studies like the one presented, there is a

375 technical requirement in the ISO 14044 standard to perform a critical review, for which an

376 academic peer review does not qualify (ISO, 2006b). However, this step was excluded from

377 the study due to budgetary constraints.

19
ACCEPTED MANUSCRIPT

378 2.4.1 Data acquisition

379 Data were obtained from interviews and consultations with city council waste

380 managers, waste container and compactor trucks distributors (Gustavo Fuentes, Public

PT
381 Cleaning Supervisor, Municipality of Rímac, personal communication; Urban Equipment,

382 2016). In addition, available data were adapted from the ecoinvent v3.1 database

RI
383 (ecoinvent, 2016). Table 5 presents the main data sources for each data type.

SC
384 Table 5. List and description of the main data sources, as well as the dataset modifications

385 that were performed for the computation of the results.

Item Data type


U Sources
AN
Interviews with technical staff of Contenur, a
Useful lifetime
company specialized in container management.
Data for production and maintenance of the containers
Production process and were obtained from Rives et al. (2010), a study that
M

Containers maintenance included the inventory for 14 different types of


containers.
The distance from the port of Valencia (Spain) to the
Transportation distance port of Callao (Peru) was assumed since Peru imports
D

most containers from Spain (SeaRates, 2016).


Data were obtained from interviews with municipal
Useful lifetime
TE

waste managers.
Production process and
Data were retrieved from Spielman et al. (2007).
maintenance
Compactor trucks
The distance from the port of Rotterdam (The
Netherlands) to the port of Callao (Peru) was assumed
EP

Transportation distance
since the company Geesinknorba has its production
facility in The Netherlands (SeaRates, 2016).
Data were obtained from interviews with technical
Useful lifetime staff from Hi-Riser, a representative company in the
C

sector.
Non-motorized Data were obtained from interviews with technical
AC

load vehicles Weight of vehicle and material staff from Hi-Riser, a representative company in the
sector.
Production process Data were retrieved from Spielman et al. (2007).
Ntziachristos et al. (2013). Diesel emissions were
calculated based on the EMEP-Corinair, assuming
Diesel Diesel emissions
that the compactor trucks follow Euro III emission
standards.
Vázquez-Rowe et al. (2015a) adapted the electricity
mix inventory available in ecoinvent for Peru to the
Electricity Electricity mix
electricity mix conditions in 2013, which were used in
the current study.

20
ACCEPTED MANUSCRIPT

386

387 For diesel emissions, kilometers traveled were used as the reference unit and total

388 distances were calculated as a sum of the distance traveled of the entire compactor truck

389 fleet. For the current scenario, a total traveled distance of 431.6 km per day was modeled.

PT
390 Calculations for this total distance are shown in Section S1 of the Supporting Material

RI
391 (SM).

2.4.2 Life Cycle Inventory (LCI)

SC
392

393 The Life Cycle Inventory (LCI) enables the provision of computational data to link

U
394 the operational inputs and outputs to the assessment methods in order to calculate the
AN
395 environmental impacts (ISO, 2006). Inventory data were divided into five main

396 subsystems, as detailed in Table 6.


M

397 Table 6. Reference flows constructed to represent current waste management conditions
D

398 and the proposed future scenario (Data reported per functional unit: 1 t of collected waste).
TE

Current Proposed
Item Material Unit ecoinvent inventories
scenario scenario
HDPE granulate - 299.24 g
EP

Injection molding {RER},


Production process - 299.24 g
Containers processing
Transport, freight sea,
Transportation - 3.39 tkm
C

transoceanic ship

Tap water, tap water production


AC

Containers Water - 31.03 kg with electricity modeled from


maintenance Peru
Electricity - 1.52 MJ
Iron - 16.49 g Cast iron , production
Bicycle - 4.12E-5 p Bicycle , production
Non-
Transport, freight sea,
motorized load Transportation - 128.16 kgkm
transoceanic ship
vehicles
Metal working, average for steel
Production process - 16.49 g
product manufacturing

21
ACCEPTED MANUSCRIPT

Diesel, petroleum refinery


Diesel 298.66 176.21 g
operation
Municipal waste collection
Waste collection 2.36E-05 9.17E-06 units
service by 21 metric ton lorry

Road 0.033 0.013 my Road infrastructure maintenance


Compactor

PT
trucks Transport, freight sea,
Transportation 8.29 3.22 tkm
transoceanic ship

Maintenance, lorry 28 metric

RI
Maintenance 2.03E-05 1.20E-05 p ton, with peruvian electricity
mix

SC
p= piece.

399

400 2.4.3 Life Cycle Impact Assessment


U
AN
401 Considering the fact that consequences on other systems engendered from changes
M

402 in the production system analyzed were not taken into consideration, a retrospetive (i.e.,

403 attributional) LCA perspective was chosen. The assessment method selected to calculate
D

404 the results was ReCiPe midpoint-hierarchist (Goedkoop et al., 2009). Midpoint impact
TE

405 categories were used in order to report environmental impacts based on the links in the

406 cause-effect chain of an impact category (Bare et al., 2000). Moreover, The hierarchist
EP

407 perspective was used since it considers the main consensus policy approaches linked to
C

408 time horizons, such as the 100-year horizon for climate change-linked emissions (Vázquez-
AC

409 Rowe et al., 2015b). The impact categories considered were photochemical oxidant

410 formation (POF), acidification potential (AP), particular matter formation (PMF), water

411 depletion (WD), metal depletion (MD) and fossil depletion (FD). Global warming potential

412 (GWP), was computed using the IPCC method, in order to consider the climate change

413 characterization factors with a a timeframe of 100 years, updated to 2013 (IPCC, 2013).

22
ACCEPTED MANUSCRIPT

414 The rationale behind this selection was based on the importance of the use of different

415 types of resources in the system under analysis, as well as on the impact that the production

416 system produces in the urban environment where it is located. Therefore, air pollution

417 impact categories, such as PMF and POF, were included. Other impact categories,

PT
418 especially those linked to damages to the ecosystem, were excluded from the scope of the

RI
419 study. The software used to conduct the analysis was SimaPro v8.2 (PRè Consultants,

420 2016).

SC
421 2.4.4 Sensitivity analysis

U
422 A series of alternate scenarios to the proposed baseline scenario were modelled in
AN
423 order to assess the effect that alternative assumptions could have on the LCA results.

424 Therefore, three additional options to replace the proposed scenario were considered. In the
M

425 first place, scenarios A1-A3 considered different collection rates for the trucks. Hence,
D

426 while the proposed scenario assumed a 95% filling percentage for the collection trucks, A1-
TE

427 A3 considered filling percentages of 80%, 90% and 100%, respectively. Secondly, scenario

428 B1 considered that the maintenance of the containers was less intensive in terms of water
EP

429 consumption. Therefore, scenario B1 considers that containers will not receive

430 maintenance, reducing their lifespan from 5 to 2.5 years. Finally, scenario C1 considered
C

431 different emission standard regulations for the compactor trucks. Hence, while EURO 3
AC

432 emission standards were modelled in the proposed baseline scenario, EURO 4 standards

433 were modelled for scenario C1 (Ntziachristos et al., 2013).

434 3. Results

435 3.1 Location of the collection sites

23
ACCEPTED MANUSCRIPT

436 The optimization model was solved using AMPL language and IBM ILOG CPLEX

437 12.6 solver (IBM, 2016). A total of 9 data files were created, one for each zone. Thereafter,

438 the optimization model was solved with each of the data files. Figure 2 presents the

439 activated collection sites for general (Figure 2a) and recyclable residues (Figure 2b).

PT
440 Similarly, Table 7 shows the number of general and recyclable collection sites and the

RI
441 number of bins located in each zone.

Figure 2. Graphical representation of the suggested location of activated collection sites for

SC
442

443 general (Figure 2a) and recyclable (Figure 2b) residues in the district of Comas following

U
444 the proposed baseline scenario. AN
M
D
TE
C EP
AC

445

24
ACCEPTED MANUSCRIPT

446 Table 7. Number of activated collection sites and containers for general and recyclable

447 residues.

Number of containers Number of containers


Activated general Activated recyclable
Zone for the collection of for the collection of
collection sites collection sites

PT
general residues recyclable residues
1 32 18 125 70
2 53 28 210 110

RI
3 76 41 296 160
4 42 24 164 96
5 39 21 153 84

SC
6 57 32 222 118
7 134 69 518 264
8 142 80 543 302

U
9 27 15 103 58
TOTAL VALUES 2334 1262
AN
448

3.2 Collection routes


M

449

450 3.2.1 Collection routes for general residues


D
TE

451 After solving the heuristic algorithm, 13 collection routes were obtained (see Figure

452 3). A total of 7 compactor trucks were identified to serve the routes obtained in each of the
EP

453 two daily shifts. The total distance traveled for the 7 trucks was 254 km, a 41.1% reduction

454 as compared to the current scenario. As shown in Figure 3, the blue colored route
C

455 represents the routes in the diurnal shifts, whereas the red colored route represents the route
AC

456 taken by collector truck 1 in the night shift. Although Figure 3 only shows the routes for

457 collector truck 1, the routes for the remaining trucks are represented in Figures S1 – S6 in

458 Section S2 of the SM.

459

25
ACCEPTED MANUSCRIPT

460 Figure 3. Graphical representation of the different collection routes performed by

461 compactor truck 1 in the district of Comas following the proposed baseline scenario. The

462 blue line represents the diurnal shift route; the red line represents the nocturnal shift route.

PT
RI
U SC
AN
M
D
TE

463
EP

464
C

465 3.2.2 Collection routes for recyclable residues


AC

466 The estimated amount of residues in every collection site, distances between every

467 collection site, and service time for each collection site were obtained using the same

468 assumptions as for the general residues collection routes. In addition, an average speed of 8

469 km/h was considered to estimate the traveling time between collection sites. The collection

26
ACCEPTED MANUSCRIPT

470 vehicle that was modelled was a 6 m3 capacity, non-motorized freight. Finally, two shifts of

471 8 h per day were considered. The algorithm was coded in Matlab (Mathworks, 2016).

472 Figure 4 depicts the collection routes for plastic (Figure 4a), tetrapak/cans (Figure

PT
473 4b) and glass residues (Figure 4c) performed by the non-motorized vehicles. The algorithm

474 computed a total of 6 collection routes for plastic, 7 for tetrapak/cans and 7 collection

RI
475 routes for glass that would be performed twice per week in any of the two available shifts.

As an example, for plastic collection routes, Figure 4a shows that every Monday and

SC
476

477 Thursday one vehicle will cover the red colored route in the diurnal shift and the blue

U
478 colored route in the night shift. In the case of Figure 4c, an additonal green colored route
AN
479 will be covered by an extra vehicle that will use that route once per day in either the diurnal

480 or nocturnal shift. Tuesday and Friday routes, as well as the Wednesday and Saturday
M

481 itineraries, are available in Figures S7 and S8 in the SM.


D

482
TE

483

484
EP

485
C
AC

27
ACCEPTED MANUSCRIPT

486 Figure 4. Graphical representation of the different routes performed by non-motorized

487 vehicles on Mondays and Thursdays following the proposed baseline scenario to collect

488 plastics (Figure 4a), tetrapak/cans (Figure 4b) and glass (Figure 4c). The blue line

489 represents the diurnal shift route; the red line represents the nocturnal shift route; the green

PT
490 line represents an additional route for collecting glass (see Figure 4c), that can be either

RI
491 diurnal or nocturnal.

U SC
AN
M
D

492
TE

493
EP

494 Figure 5 presents the collection routes computed for paper residues. The algorithm

495 calculated 7 collection routes that are performed three times per week, in any of the two
C

496 available shifts. For instance, Figure 5a shows the route performed on Mondays,
AC

497 Wednesdays and Fridays. In this particular case, two vehicles are supposed to fulfill the

498 entire route to cover the selected collection sites. The first one would cover the red colored

499 route in the day shift and the blue colored route in the night shift; the second vehicle, in

500 contrast, would cover the purple colored route in the shift day and the green colored route

501 in the night shift. Hence, a total of 6 municipal waste collection vehicles would be

28
ACCEPTED MANUSCRIPT

502 necessary to cover the collection of tetrapak-cans and glass residues (4) and paper residues

503 (2). Figure 5b presents the same results for Tuesdays, Thursdays and Saturdays.

504 Figure 5. Graphical representation of the different collection routes performed by non-

PT
505 motorized vehicles in the district of Comas to collect paper residues following the proposed

506 baseline scenario. The blue line represents the diurnal shift route; the red line represents the

RI
507 nocturnal shift route; the green and purple lines represent additional routes for collecting

paper, which can be either diurnal or nocturnal.

SC
508

U
AN
M
D
TE
EP

509
C

510
AC

511 3.3 Life Cycle Assessment results

512 When computing the environmental impact results in the LCA software (i.e.,

513 SimaPro v8.2), results show that there would be substantial improvements for most impact

514 categories (Figure 6). On the one hand, in terms of MD, reduction of impacts in the

29
ACCEPTED MANUSCRIPT

515 proposed scenario would be of 50%. Important reductions are also attained for the air

516 pollution categories, such as PMF (35%) and POF (31%), as well as for AP (28%). GHG

517 emissions are also reduced by 13% from 3.19 kg CO2eq per t of collected residue to

518 approximately 2.78 kg CO2eq. Although these reductions are less relevant than for the

PT
519 abovementioned impact categories, the use of the proposed collection system would imply

RI
520 the savings of approximately 317.5 t of CO2eq in the period 2016-2022. In both scenarios

521 the main activity contributing to GHG emissions involves the use of the municipal waste

SC
522 collection trucks. However, in the current scenario 100% of the contribution is linked to the

523 trucks, since there are no containers or other infrastructure used in waste management.

524
U
However, in the proposed scenario this contribution is reduced to 56%. Interestingly, the
AN
525 use of the containers would represent 41% of total emissions, while only 3% would be

526 attributable to the production and maintenance of non-motorized vehicles. For air pollution
M

527 categories, as mentioned above, the potential reductions were substantially higher than for
D

528 GHG emissions. In fact, this if of the utmost importance considering that urban emissions
TE

529 of POF ad PMF would be reduced by approximately 41%.

530 On the other hand, two impact categories linked to the depletion of natural abiotic
EP

531 resources, FD and WD, showed increases in environmental impact by shifting to the

532 proposed optimized collection systems. For FD this increase is of 6.9%, due mainly to the
C

533 production of the containers (53%). In the case of WD, however, the increase is of 167%,
AC

534 which is almost entirely due to the maintenance needed to elongate the lifespan of the

535 containers. Hence, in the current situation approximately 21 L of water are used to collect 1

536 metric ton of waste. Of this amount, it should be noted that approximately 99% is linked to

537 outsourced water use through energy, production of materials for the trucks, and production

30
ACCEPTED MANUSCRIPT

538 of diesel. In contrast, for the proposed scenario, 82% of the 56.1 L per FU would be used

539 directly in the Municipality of Comas to clean the containers.

540 Figure 6. Environmental impacts of the current and proposed scenarios for selected impact
541 categories. Results reported per functional unit (1 t of collected waste).

PT
Global warming potential Acidification potential
3.5 21.0

3.0 18.0

RI
2.5 15.0
kg CO2 eq

g SO2 eq
2.0 12.0

1.5 9.0

SC
1.0 6.0

0.5 3.0

0.0 0.0
CS PBS CS PBS

U
Photochemical oxidant formation Particulate matter formation
10
24.0
AN
8.5
20.0

7
g PM-10 eq

16.0
g C2H4 eq

12.0 5.5
M

8.0 4

4.0 2.5

0.0 1
CS PBS CS PBS
D

Water depletion Metal depletion


TE

0.6
56.0
0.5
48.0

40.0 0.4
kg Fe eq

32.0
L eq

0.3
EP

24.0
0.2
16.0
0.1
8.0

0.0 0.0
C

CS PBS CS PBS

Fossil depletion
AC

1.4

1.2

1.0
CS - Current Scenario
kg oil eq

0.8

0.6
PBS - Proposed Baseline Scenario
0.4

0.2

0.0
CS PBS
542

543

31
ACCEPTED MANUSCRIPT

544 3.3.5 Sensitivity analysis

545 The results from the sensitivity analysis for scenarios A1-A3 show, for all impact

546 categories, similar results as those in the proposed baseline scenario (see Table 8).

PT
547 However, slight increases in environmental impact were identified for scenarios A1 (80%

548 filling percentage) and A2 (90% filling percentage), due to the increase in the distance

RI
549 traveled by trucks (305 and 275 km per day, respectively) and the amount of maintenance

required. For the 100% filling percentage, lower impacts than the proposed baseline

SC
550

551 scenario are observed, due to the reduction in the distance travelled per truck (248

U
552 kilometers per day) and the amount of maintenance required. Results for scenario B1 show
AN
553 an increase of 23% in WD impact category, which is a lower increase than in the proposed

554 baseline scenario or scenarios A1-A3. This is attributable to the lack of maintenance of the
M

555 containers, with a subsequent reduction in the use of water. For GWP and FD, increases of

556 11% and 55%, respectively, are observed with respect to the current scenario. In both cases,
D

557 these rises that are not observed in any other of the modelled projected scenarios are due to
TE

558 the increased production and transportation of containers given their reduced lifespan, as a

559 result of the lack of maintenance. Results for scenario C1 show an important reduction in
EP

560 POF (34%), PF (41%) and AP (41%) if EURO 4 emission standards are assumed rather
C

561 than EURO 3.


AC

562

32
ACCEPTED MANUSCRIPT

563 Table 8. Environmental impacts per impact category for the different scenarios. Results

564 reported per functional unit (1 metric ton of collected waste).

PT
RI
U SC
AN
M
D
TE
C EP
AC

33
ACCEPTED MANUSCRIPT

565

PT
Proposed Scenario A Scenario B Scenario C
Current
Impact Unit baseline a
scenario Variation
category
(CS)
scenario A1 Variationb A2 Variationb A3 Variationb B1 Variationb C1 Variationb
(PBS)

RI
GWP kg CO2eq 3.190 2.780 -13% 2.903 -9% 2.800 -12% 2.706 -15% 3.550 11% 2.780 -13%
kg
POF 0.020 0.014 -28% 0.015 -23% 0.015 -26% 0.014 -29% 0.017 -12% 0.013 -34%

SC
C2H4eq
AP kg SO2eq 0.023 0.016 -31% 0.017 -24% 0.016 -28% 0.016 -31% 0.020 -14% 0.014 -41%
kg
PMF 0.009 0.006 -35% 0.006 -30% 0.006 -33% 0.006 -35% 0.007 -23% 0.005 -41%

U
PM10eq
WD m3 eq 0.021 0.056 167% 0.057 171% 0.056 169% 0.056 167% 0.026 23% 0.056 167%

AN
MD kg Fe eq 0.507 0.254 -50% 0.261 -48% 0.257 -49% 0.253 -50% 0.265 -48% 0.254 -50%
FD kg oil eq 1.176 1.258 7% 1.323 12% 1.285 9% 1.249 6% 1.819 55% 1.258 7%
a

M
566 This value refers to the percentage variation between the current scenario and the proposed baseline scenario.
b
567 This value refers to the percentage variation between the current scenario and the alternative proposed scenarios.
568 GWP= global warming potential; POF= photochemical oxidant formation; AP= acidification potential; PMF= particulate matter formation; WD= water depletion;

D
569 MD= metal depletion; and, FD= fossil depletion.

TE
570

571
EP
572
C
AC

573

574

34
ACCEPTED MANUSCRIPT

575 4. Discussion

576 The mathematical optimization model used in the case study to improve the

577 collection of waste in the Municipality of Comas has proved to be an interesting basis for

578 shifting to a container system. Furthermore, whenever the environmental impacts of the

PT
579 proposed container system scenario are compared to those from the current state of waste

RI
580 collection in the district, substantial environmental gains are observed in most impact

581 categories.

SC
582 For instance, the comparison of both waste management systems pointed out that

U
583 container systems can reduce GHG emissions by almost 14% (Figure 6). Although these
AN
584 gains are not as relevant as for other impact categories assessed, implementing this type of

585 management system on a broader level throughout most of Lima’s districts, considering
M

586 that they represent one third of the nation’s population, could entail substantial GHG

587 emission reductions for a country that has recently ratified the Treaty of Paris to combat
D

588 climate change. In fact, according to the “Intended Nationally Determined Contributions”
TE

589 (iNDCs) presented by the Peruvian Ministry of the Environment to reduce GHG emissions

590 by 2030 in the framework of this Treaty, a total of 2585 kt of CO2eq are planned to be
EP

591 reduced thanks to the implementation of a set of five specific mitigation plans in landfills,
C

592 as shown in Table 9 (MINAM, 2015). Through the optimization of collection routes in the
AC

593 Municipality of Comas alone, 317 t CO2eq would be reduced in the period 2016-2022.

594 Consequently, it seems plausible to assume that the optimization of waste collection on a

595 greater scale nationally could be an attractive option to enforce a more ambitious mitigation

596 strategy in the residues sector.

35
ACCEPTED MANUSCRIPT

597 Table 9. GHG emissions mitigation plans presented by the Peruvian government linked to

598 residues in order to comply with the Treaty of Paris. Comparison with GHG emission

599 reductions computed in the current study.

PT
GHG emissions mitigation by 2030
Mitigation options in the residues sector
(Mt CO2eq)

RI
Capture and combustion of methane in landfills 1.587
Capture and combustion of methane in other landfills 0.291
Implementation of semi aerobic technology in landfills 0.473

SC
Composting in landfills 0.217
Recycling in landfills 0.017
TOTAL MITIGATION 2.585

U
Optimization of waste collection routes (current
3.17E-4
study)1
1
AN
GHG emissions reductions for the current study represent the period 2016-2022.

600 Despite the increase in GHG emissions linked to the production of containers with
M

601 the new proposed scenario, these gains are achieved primarily through the reduction in the

602 number of collector trucks, as well as in their optimized routes to reduce the use of fossil
D

603 fuels. Moreover, considering the scenarios conducted for the sensitivity analysis, only
TE

604 scenario B1 provides differences in terms of GHG emissions as compared to the proposed

605 baseline scenario.


EP

606 The reduction in the amount of diesel used to power the waste collector trucks
C

607 thanks to a reduction in distance travelled, however, has more significant benefits in terms
AC

608 of air pollution impact categories, such as POF and PMF. In both cases the reductions

609 attained in the proposed scenario are above 30%. If locally emitted impacts are considered

610 exclusively, the reduction increases to approximately 41%, a very relevant improvement in

611 an urban area in the north of Lima where air quality tends to be one of the worst among

612 metropolitan areas in Latin America. For instance, according to the World Health

36
ACCEPTED MANUSCRIPT

613 Organization (WHO) two out of the three districts with PM-10 measurements in the city of

614 Lima presented values between 47% and 140% higher than those recommended by the

615 WHO (INEI, 2015; WHO, 2016). Although Comas is not among the districts assessed, a

616 study conducted by Tashiro and Taniyama (2002) suggested that the wind in Lima tends to

PT
617 carry pollutants from the downtown area to the residential areas in the northeastern area of

RI
618 the city, where Comas is located. Consequently, the optimized waste collection would be a

619 small step toward reducing the high levels of air pollution in many districts and could be

SC
620 used as part of the strategies to improve traffic flow and quality of life in the city (Harrison

621 et al., 2010; Restrepo et al., 2016; Progiou and Ziomas, 2015; WHO, 2013).

U
AN
622 The higher use of water in the proposed baseline scenario is inevitable to maintain a

623 high lifespan for the containers. However, considering that water is a scarce resource in
M

624 Lima and in most of the Peruvian coast, with water scarcity indexes close to 1 (Pfister et al.,

625 2009; Pfister and Bayer, 2014), reducing the maintenance of containers appears to be a
D

626 feasible strategy. However, as shown in the sensitivity analysis results, these improvements
TE

627 in terms of WD would have negative consequences regarding most remaining categories,

628 namely GWP, which increase from 2.78 kg of CO2eq with the proposal to 3.55 kg of CO2eq
EP

629 in this scenario (24% of increase) and FD, which increase from 1.258 kg oil eq with the
C

630 proposal to 1.891 kg oil eq in this scenario.


AC

631 5. Conclusions

632 The use of operational research in this particular case study enabled the

633 minimization of the number of collection sites by installing and using the available capacity

634 of containers, while satisfying the maximum walking distance for citizens in the district of

37
ACCEPTED MANUSCRIPT

635 Comas without mixing recyclable and general residues. In addition, the minimization of the

636 distance traveled by compactor trucks was achieved. This enabled the number of

637 compactor trucks required to be reduced by more than half.

PT
638 When computing the LCA results, these demonstrate that the optimization of

639 collection residues may provide important environmental impact reductions. For instance,

RI
640 in the case study proposed, these reductions are mainly linked to air pollution, which is

critical in the area assessed, although environmental savings are also observed in terms of

SC
641

642 GHG emissions and depletion of resources. Hence, the assessment of multiple

U
643 environmental areas of protection, as performed in the study, allows the identification of
AN
644 trade-offs between different categories.

645 Current research is focusing on developing inventories for landfills in Peru and
M

646 recycling companies to allow LCA practitioners to develop a cradle-to-grave LCA


D

647 including the final disposal of residues or their recycling or valorization. This assessment
TE

648 will allow municipal waste managers to take into account the importance of recycling or

649 other valorization routes to reduce the environmental impacts of a waste management
EP

650 system. Finally, from a methodological perspective, an heuristic approach for the location

651 of collection sites could develop alternative scenarios to the one presented in this study
C

652 considering the variation of different parameters, such as varying rates for containers, in
AC

653 waste per capita generation or regarding the density of the residues.

654 Acknowledgements

38
ACCEPTED MANUSCRIPT

655 The authors thank the Peruvian LCA Network of the Pontificia Universidad Católica del

656 Perú for operational support. Gustavo Larrea-Gallegos thanks the Consejo Nacional de

657 Ciencia, Tecnología e Innovación Tecnológica (CONCYTEC) for financial support.

PT
658 References

659 Aini, A., Salehipour, A., 2012. Speeding up the Floyd-Warshal algorithm for the cycled

RI
660 shortest path problem. App. Math. Lett. 25 (1), 1-5.

SC
661 APEIM, 2015. Niveles socioeconómicos 2015. Asociación Peruana de Empresas de
662 Investigación de Mercados. Retrieved from: www.apeim.com.pe/wp-
663 context/themes/apeim/docs/nse/APEIM-NSE-2015.pdf. Last accessed: February 2nd 2017
664 [in Spanish].
U
AN
665 Applegate, D., Bixby, R., Chvátal, V., Cook, W., 2007. The Traveling Salesman Problem.
666 Princeton: Princeton University Press.
M

667 ArcGIS, 2016. ArcMap. Retrieved from: desktop.arcgis.com/es/desktop/. Last accessed:


668 September 6th 2016.
D

Badran, M., El-Haggar, S., 2006. Optimization of municipal solid waste management.
TE

669
670 Waste Manage. 26(5), 534-545.
EP

671 Bare, J.C., Hofstetter, P., Pennington, D.W., De Haes, H.A.U., 2000. Midpoints versus
672 endpoints: the sacrifices and benefits. Int. J. Life Cycle Assess. 5 (6), 319-326.
C

673 Braekers, K., Ramaekers, K., Van Nieuwenhuyse, I., 2016. The vehicle routing problem:
674 State of the art classification and review. Comput. Ind. Eng. 99, 300-313.
AC

675 Burnley, S., 2007. A review of municipal solid waste composition in the United Kingdom.
676 Waste Manage. 27 (10), 1274-1285.

677 Christofides, N., Beasley, J., 1984. The Period Routing Problem. Networks, 14 (2), 237-
678 256.

39
ACCEPTED MANUSCRIPT

679 DEFRA, 2016. Digest of Waste and Resource Statistics - 2015 Edition. Department ofr
680 Environment Food & Rural Affairs. Retrieved from:
681 www.gov.uk/government/collections/waste-and-recycling-statistics. Last accessed: August
682 8th 2016.

PT
683 European Environmental Agency, 2013. Managing municipal solid waste - a review of
684 achievements in 32 European countries. Retrieved from:
www.eea.europa.eu/publications/managing-municipal-solid-waste. Last accessed: March

RI
685
686 14th 2016.

SC
687 European Union, 2013. Glossary: Classification of functions of government (COFOG).
688 Retrieved from: ec.europa.eu/eurostat/statistics-
explained/index.php/Glossary:Classification_of_the_functions_of_government_(COFOG).

U
689
690 Last accessed: March 7th 2016.
AN
691 European Union, 2016a. Eurostat Statistics Explained. Retrieved from:
692 ec.europa.eu/eurostat/statistics-
M

693 explained/index.php/Environmental_protection_expenditure#Environmental_protection_ex
694 penditure_by_the_public_sector. Last accessed: May 31st 2016.
D

695 European Union, 2016b. Eurostat Statistics Explained. Retrieved from:


TE

696 ec.europa.eu/eurostat/statistics-
697 explained/index.php/Population_and_population_change_statistics. Last accessed:
EP

th
698 February 17 2017.

699 Feldman, E., Lehrer, F., Ray, T., 1966. Warehouse location under continuous economies of
C

700 scale. Manage. Sci. 12 (9), 670-684.


AC

701 Francis, P., Smilowits, K., Tzur, M., 2006. The Period Vehicle Routing Problem with
702 Service Choice. Transportation Sciences, 40 (4), 439-454.

703

40
ACCEPTED MANUSCRIPT

704 Francis, R., Lowe, T., Tamir, A., Emir-Farinas, H., 2004. A framework for demand point
705 and solution space aggregation analysis for location models. Eur. J. Oper. Res. 159, 574-
706 585.

707 Geesinknorba, 2016.. Productos. Retrieved from:

PT
708 www.geesinknorba.com/es/productos/residuos-domesticos/serie-n3/. Last accessed: May
709 11th 2016.

RI
710 Gendreau, M., Laporte, G., Séguin, R., 1996. Stochastic vehicle ruting. Eur. J. Oper. Res.
711 88 (1), 3-12.

SC
712 Ghiani, G., Demetrio, L., Manni, E., Triki, C., 2012. Capacitated location of collection sites
713 in an urban waste management system. Waste Manage. 32 (7), 1291-1296.

714
U
Ghiani, G., Manni, A. M., Toraldo, M., 2014. The impact of an efficient collection sites
AN
715 location on the zoning phase in municipal solid waste management. Waste Manage. 34
716 (11), 1949-1956.
M

717 Gilardino, A., Rojas, J., 2015. Improvement proposal in the Urban Solid Waste
718 Management system in a district in northern Lima, using Operation Research tools. CILCA
D

719 Conference 2015, (pp. 398-405). Lima. ISBN= 978-9972-674-11-2.


TE

720 Goedkoop, M., Heijungs, R., Huijbregts, M., de Schryver, A., Struijs, J., van Zelm, R.,
721 2009. ReCiPe 2008. A Life Cycle Impact Assessment Method Which Comprises
722 Harmonised Category Indicators at the Midpoint and the Endpoint Level. Report I:
EP

723 Characterisation. Ministry of Housing, Spatial Planning and Environment (VROM).


724 Retrieved from: www.lcia-recipe.info (last accessed: 19.05.15.).
C

725 Gujarati, D., Porter, D., 2010. Econometrics (Fifth ed.). McGraw-Hill International: New
726 York.
AC

727 Ho, W., Ho, G., Ji, P., Lau, H., 2008. A hybrid genetic algorithm for the multi-depot
728 vehicle routing problem. Artif. Intell. 21 (4), 548-557.

729 Hoornweg, D., Bhada-Tata, P., Kennedy, C., 2013. Waste production must peak this
730 century. Nature, 502 (7473), 615-617.

41
ACCEPTED MANUSCRIPT

731 Huang, S.-H., Lin, P.C., 2015. Vehicle routing-scheduling for municipal waste collection
732 system under the "Keep Trash off the Ground" policy. Omega, 55, 24-37.

733 IBM, 2016. IBM ILOG CPLEX Optimization Studio. Retrieved from: www-
734 03.ibm.com/software/products/es/ibmilogcpleoptistud. Last accessed: September 5th 2016.

PT
735 INEI, 2009. Peru: Population estimates per gender for Departments, Provinces and Districts
736 (2000-2015). Peruvian Institute of Statistics. Lima [in Spanish].

RI
737 INEI, 2014. Peru: Enviromental statistics yearbook 2014. Peruvian Institute of Statistics.
738 Retrieved from:

SC
739 www.inei.gob.pe/media/MenuRecursivo/publicaciones_digitales/Est/Lib1197/libro.pdf.
740 Last accessed: February 5th 2017.

741
U
INEI, 2015. Peru: Environmental statistics yearbook 2015. Peruvian Institute of Statistics.
AN
742 Retrieved from:
743 www.inei.gob.pe/media/MenuRecursivo/publicaciones_digitales/Est/Lib1342/libro.pdf.
M

744 Last accessed: February 10th 2017.

745 INEI, 2016. Peru: Environmental statistics yearbooks. Retrieved from:


D

746 www.inei.gob.pe/estadisticas/indice-tematico/medio-ambiente/. Last accessed: January 17th


TE

747 2017.

748 Institute for Protection and Environmental Investigation, 2016. ISPRA Ambiente. Report
EP

749 on Urban Residues 2015. Retrieved from:


750 www.isprambiente.gov.it/files/pubblicazioni/rapporti/rifiuti-urbani-
751 2015/RapportoRifiutiUrbani_Ed.2015n.230_Vers.Integrale_agg22_12_2015.pdf. Last
C

th
752 accessed: August 19 2016 [in Italian].
AC

753 ISO, 2006. International Standarisation Organization. ISO 14040 - Enviornmental


754 management - Life cycle assessment - Principles and framework.

755 IPCC, 2013. Climate Change 2013. The Physical Science Basis. Working Group I
756 contribution to the 5th Assessment Report of the IPCC. Intergovernamental Panel on

42
ACCEPTED MANUSCRIPT

757 Climate Change. Retrieved from: www.climatechange2013.org. Last accessed: June 30th
758 2016.

759 Iriate, A., Gabarrell, X., Rieradevall, J., 2008. LCA of selective waste collection systems in
760 dense urban areas. Waste Manage. 29 (2), 909-914.

PT
761 Jang, H., Lee, T., 2015. Demand point aggregation method for covering problems with
762 gradual coverage. Comput. Oper. Res. 60, 1-13.

RI
763 Jing-Quan, L., Borenstein, D.M., 2008. Truck scheduling for solid waste collection in the
764 City of Porto Alegre, Brazil. Omega, 36(6), 1133-1149.

SC
765 Kallehauge, B., 2008. Formulations and exact algorithms for the vehicle routing problem
with time windows. Comput. Oper. Res. 35 (7), 2307-2330.

U
766 AN
767 Kinobe, J.R., Bosona, T., Gebresenbet, G., Niwagaba, C.B., Vinnerås, B., 2015.
768 Optimization of waste collection and disposal in Kampala city. Habitat Int. 49, 126-137.
M

769 Koppelaar, R., Wiekard, H., 2013. Asessing phosphate rock depletion and phosphorus
770 recycling options. Global Environ. Chang. 23 (6), 1454-1466.
D

771 Laforest, V., 2014. Assessment of emerging and innovative techniques considering best
TE

772 available techniques performances. Resour. Conserv. Rec. 92, 11-24.

773 Li, H., Lim, A., 2003. Local search with annealing-like restarts to solve the VRPTW. Eur.
EP

774 J. Oper. Res. 150 (1), 115-127.

775 Li, J.-Q., Borenstein, D., Mirchandani, P., 2008. Truck scheduling for solid waste
C

776 collection in the City of Porto. Omega, 36 (6), 1133-1149.


AC

777 Martínez-Blanco, J., Colón, J., Gabarrell, X., Font, X., Sánchez, A., Artola, A., Rieradevall,
778 J., 2010. The use of life cycle assessment for the comparison of biowaste composting at
779 home and full scale. Waste Manage. 30 (6), 983-994.

780 Mesa, J., Boffey, B., 1996. A review of extensive faclity location in networks. Eur. J. Oper.
781 Res. 95, 592-603.

43
ACCEPTED MANUSCRIPT

782 MINAM, 2015. Construyendo participativamente la contribución nacional: propuesta del


783 Perú (iNDC) para consulta pública. Responsabilidad climática para aumentar la
784 competitividad y mejorar comportamientos socio-ambientales. Peruvian propisal for
785 intended nationally determined contributions (iNDCs). Peruvian Ministry of the
786 Environment. Ministerio del Ambiente, Perú. September 2015 [in Spanish].

PT
787 Municipality of Ate, 2014. Solid Waste Characterization Study. Retrieved from:
www.muniate.gob.pe/ate/files/documentosPlaneamientoOrganizacion/GESTION_RESIDU

RI
788
789 OS_SOLIDOS/2011/Estudio%20de%20Caracterizaci%C3%B3n%20de%20Residuos%20S
%C3%B3lidos%20Municipales%202014.pdf. Last accessed: February 14th 2017 [in

SC
790
791 Spanish].

Municipality of Breña, 2016 Solid Waste Characterization Study. Retrieved from:

U
792
793 www.munibrena.gob.pe/servicios/estudio-de-residuos-solidos-de-brena.pdf. Last accessed:
AN
794 February 14th 2017 [in Spanish].

795 Municipality of Comas, 2014a. Solid Waste Characterization Study. Retrieved from:
M

796 www.municomas.gob.pe/anuncios/Estudio_de_Caracterizacion_de_Residuos_domiciliarios
797 .pdf. Last accessed: February 14th 2017 [in Spanish].
D

798 Municipality of Comas, 2014b. Municipal Solid Waste management plan for the
TE

799 Municipailty of Comas (2014-2016) Retrieved from:


800 www.municomas.gob.pe/anuncios/Plan%20Manejo%20RRSS%20Comas%20Julio%20201
EP

801 4%20echo%20por%20alternativa.pdf. Last accessed: February 14th 2017 [in Spanish].

802 Muñoz, I., Rieradevall, J., Doménech, X., Milà i Canals, L., 2004. LCA Aplication to
C

803 Integrated Waste Management Planning in Gipuzkoa (Spain). Int. J. Life Cycle Assess. 9
AC

804 (4), 1-9.

805 NTP, 2005. Gestión ambiental. Gestión de residuos. Código de colores para los dispositivos
806 de almacenamiento de residuos. Norma Técnica Peruana: NTP 900.058 – 2005 [in
807 Spanish].

44
ACCEPTED MANUSCRIPT

808 Ntziachristos, L., Samaras, Z., 2013. EMEP/EEA emission inventory guidebook 2013
809 update Sept 2014. Retrieved from: www.eea.europa.eu/publications/emep-eea-guidebook-
810 2013. Last accessed: february 14th 2017.

811 OEFA, 2014. Organismo de Información y Fiscalización Ambiental. Municipal Waste

PT
812 management environmental supervision. Lima.

813 Pfister, S., Bayer, P., 2014. Monthly water stress: spatially and temporally explicit

RI
814 consumptive water footprint of global crop production. J. Clean. Prod. 73, 52-62.

815 Pfister, S., Koehler, A., Hellweg, S., 2009. Assessing the Environmental Impacts of

SC
816 Freshwater Consumption in LCA. Environ. Sci. Technol. 43, 4098-4104.

817 Pillac, V., Gendreau, M., Guéret, C., Medaglia, A.L., 2013. A review of dynamic vehicle
818 routing problems. Eur. J. Oper. Res. 225 (1), 1-11.

819
U
PRé Consultants, 2016. PRé Consultants. Retrieved from: www.pre-sustainability.com/.
AN
820 Last accessed: February 17th 2017.

821 Rivela, B., Moreira, M.T., Muñoz, I., Rieradevall, J., Feijoo, G., 2006. Life cycle asessment
M

822 of wood wastes: A case study of ephemeral architecture. Sci. Total Environ. 357 (1), 1-11.
D

823 Rives, J., Rieradevall, J., Gabarrell, X., 2010. LCA comparison of container systems in
824 municipal solid waste management. Waste Manage. 30 (6), 949-957.
TE

825 Ropke, S., 2005. Heuristic and exact algorithms for vehicle routing problems. PhD thesis,
826 Computer Science Department, University of Copenhagen, Copenhagen, Denmark.
EP

827 Rugani, B., Panasiuk, D., Benetto, E., 2012. An input–output based framework to evaluate
828 human labour in life cycle assessment. Int. J. Life Cycle Assess. 17, 795–812.
C
AC

829 SeaRates, 2016. SeaRates. Retrieved from: www.searates.com/es/reference/portdistance/.


830 Last accessed: February 17th 2017.

831 Solomon, M., 1987. Algorithms for the Vehicle Routing and Scheduling Problems with
832 Time Window Constraints. Oper. Res. 35 (2), 254-265.

833 Spielman, M., Bauer, C., Dones, R., Tuchschmid, M., 2007. Transport Services. ecoinvent
834 report No. 14. Swiss Centre for Life Cycle Inventories, Dübendorf, Switzerland.

45
ACCEPTED MANUSCRIPT

835 Tas, D., Jabali, O., Van Woensel, T., 2014. A Vehicle Routing Problem with Flexible Time
836 Windows. Comp. Oper. Res. 52, 39-54.

837 Tashiro, Y., Tasiyama, T., 2002. Atmospheric NO2 and CO concentration in Lima, Peru.
838 Environ. Int. 28, 227-233.

PT
839 Texeira, J., Pais Antunes, A., Pinho de Sousa, J., 2004. Recyclable waste collection
840 planning - a casu study. Eur. J. Oper. Res. 158 (3), 543-554.

RI
841 Tore, B., Irnich, S., 2005. Local Search for Vehicle Routing and Scheduling Problems:
842 Review and Conceptual Integration. J. Heuristics, 11 (4), 267-306.

SC
843 Toth, P., Vigo, D., 2002. The vehicle routing problem, Society for industrial and applied
mathematics. 1st Edition. SIAM Monographs on Discrete Mathematics and Applications.

U
844
845 Philadelphia, US. ISBN: 978-0898714982.
AN
846 Tralhão, L., Coutinho-Rodrigues, J.A., 2010. A multiobjective modeling approach to locate
847 multi-compartment containers for urban-sorted waste. Waste Manage. 30 (12), 2418-2429.
M

848 Urban Equipment, 2016. Urban Equipment. Retrieved from:


www.equipamientourbano.com.pe/soluciones2.html. Last accessed: February 14th 2017 [in
D

849
850 Spanish].
TE

851 Vázquez-Rowe, I., Reyna, J., García-Torres, S., Kahhat, R., 2015. Is climate change
852 centrism an optimal policy making strategy to set national electricity mixes? Appl. Energ.
EP

853 159, 108-116.

854 Vázquez-Rowe, I., Golkowska, K., Lebuf, V., Vaneeckhaute, C., Michels, E., Meers, E.,
C

855 Benetto, E., Koster, D., 2015. Environmental assessment of digestate treatment
AC

856 technologies using LCA methodology. Waste Manage. 43, 442-459.

857 Zhou, K., Kondo, A., Wada, T., Watanabe, K., 2005. Allocation Problem of Garbage
858 Containers Based on Residential Awareness. Proceedings of Ninth International
859 Conference on Computers in Urban Planing and Urban Management, (p. 124). London, UK

46
ACCEPTED MANUSCRIPT

Combining operational research and Life Cycle Assessment to optimize


municipal solid waste collection in a district in Lima (Peru)
Alessandro Gilardinoa,*, Jonatan Rojasa, Héctor Mattosa, Gustavo Larrea-Gallegosb, Ian
Vázquez-Roweb
a

PT
Department of Engineering, Pontificia Universidad Católica del Perú, 1801 Avenida
Universitaria, San Miguel, Lima 32, Peru
b
Peruvian LCA Network, Department of Engineering, Pontificia Universidad Católica del

RI
Perú, 1801 Avenida Universitaria, San Miguel, Lima 32, Peru

*Corresponding author: Alessandro Gilardino; e-mail: agilardino@pucp.pe

SC
Highlights

U
AN
An exact optimization algorithm was used to locate waste collection sites
• A heuristic approach was modelled to create collection routes for compactor trucks
• The proposal allows reducing the number of compactor trucks and travelled
M

distance
• LCA was computed to evaluate the environmental impacts of the proposal
• Global warming and air pollution were reduced significantly with the proposed
D

model
TE
C EP
AC

You might also like