Journal of Cleaner Production 256 (2020) 120387

Contents lists available at ScienceDirect

Journal of Cleaner Production

journal homepage: [Link]/locate/jclepro

Landfill area estimation based on solid waste collection prediction

using ANN model and final waste disposal options
Md. Maruful Hoque a, *, M. Tauhid Ur Rahman b
a
Climate Change Lab, Department of Civil Engineering, Military Institute of Science & Technology, Mirpur, Dhaka, 1216, Bangladesh
b
Department of Civil Engineering, Military Institute of Science & Technology, Mirpur, Dhaka, 1216, Bangladesh

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

Article history: Public health of inhabitants has been affected by the increase of unsound waste management in the cities
Received 14 March 2019 of developing countries. Solid waste management has received extensive attention from the munici-
Received in revised form palities. A sustainable plan and design of solid waste management system of a city necessitate accurate
13 January 2020
forecasting of the solid waste generated and collected by the city authorities in the landfill for final
Accepted 1 February 2020
Available online 4 February 2020
disposal and other renewable energy options. In the study, an Artificial Neural Network (ANN) surrogate
model was used to predict solid waste collected from the year 2012e2016 at Matuail landfill site of Dhaka
Handling Editor: CT Lee South City Corporation (DSCC). 120 monthly solid waste quantity and vehicle trip number getting from
weighbridge were used as inputs data into the model. 70% of the data used for the model training, 15% of
Keywords: the data used for validation and 15% of that used for testing. The remaining 60 monthly waste quantities
Municipal solid waste are used as output to develop the model. Feed-forward back propagation neural network was used with
Forecasting model the hyperbolic tangent sigmoid activation function and the LevenbergeMarquardt optimization method.
Artificial neural network The model with 2-5-1-1 topology is selected as the best topology based on the performance metrics i.e.,
Disposal
the minimum value of MSE and high value of regression. The ANN based solid waste forecasting model
Landfill
performed to be promising using the available weighbridge data with the coefficient of determination
Levenbergemarquardt
(R2) for training and testing 0.85 and 0.86. The developed model can be alternatively used successfully
with weight bridge software in the landfill to efficiently forecast solid waste collection, particularly that
in the countries with a similar demographic and social environment. Considering other proposed
alternative disposal options and waste characterization the required landfill area are estimated and
found that the landfill authority can save the valuable urban landfill area requirement up to 28.6%. The
result shows that the innovative proposed method of landfill area estimation by ANN and final disposal
methods can be used alternatively that helps for better planning and management of the landfill site.
© 2020 Elsevier Ltd. All rights reserved.

1. Introduction (Chen, 2018). MSW management is still remaining one of the most
challenges for the local authorities of developing countries to
1.1. Solid waste management at the final disposal site protect the environment, reduce the risk of public health (Younes
et al., 2015) and to preserve natural resources (Abbasi and
Economic development, urbanization, improved lifestyle and Hanandeh, 2016). The quantification and prediction of solid waste
life expectancy of city inhabitants are contributing to the increasing play a vital role in effective solid waste management system
quantity and complex nature of the solid waste (Shams et al., 2017). planning (Sun and Chungpaibulpatana, 2017), design and operation
Almost every day a new type of product generate new waste (Singh and Satija, 2016). Solid waste management, from its gener-
component that added in the city waste stream. Municipal Solid ation to final treatment, is a very complex which is usually pro-
Waste (MSW) which is comprised of mainly urban household cessed in one of three ways: landfilling, biological treatment and
wastes with a small portion of commercial or industrial wastes thermal treatment (Chen and Wang, 2017). Landfilling method is
commonly used over other two methods for approximately 95% of
the MSW collected worldwide (Foo and Hameed, 2009) considering
the environmental, financial, social and other urban development
* Corresponding author.
factors (Chen and Wang, 2017). The city authority of developing
E-mail address: marufpwd@[Link] (Md.M. Hoque).

[Link]
0959-6526/© 2020 Elsevier Ltd. All rights reserved.
2 Md.M. Hoque, M.T.U. Rahman / Journal of Cleaner Production 256 (2020) 120387

countries has a lot of problems to handle solid waste in sustainable approach. ANNs are defined as computing systems which can be
manners (Liikanena et al., 2018). Among the different factors few trained to learn a complex relationship between two or more in-
selected factors that are responsible for the poor or unsound solid dependent variables or data sets. Basically, interconnecting simple
waste management at final disposal site are less governmental processing nodes forming these parallel computing systems. The
capacity, institutional preparedness, lack of community awareness networks that are mathematically challenging, neural networks
and involvement, staffs mindset and commitment, lack of (NNs) utilize a matrix programming environment for those. In
improved collection equipment, lack of human resources, land neuro-computing or artificial intelligence among the available
scarcity for landfill, inexperienced operation and maintenance of artificial neural networks, the feed-forward neural network (FFNN)
landfill, financial constraint, training of staffs, lack of baseline study is one of the most significant historical developments. The feed-
and insufficient or no solid waste data (Santiban ~ ez-Aguilar et al., forward neural network is a nonlinear function of its inputs
2017). which composed of the functions of its neurons. Artificial Neural
Networks (ANNs) have been provided a great platform for different
1.2. Dhaka solid waste final disposal site management problems fields of study such as Photocatalytic degradation prediction in at-
and prospects mospheric research (Bararpour et al., 2017), compressive strength
forecasting of silica fume concrete in construction materials research
Solid waste has been a major concern in Dhaka city (Islam and (Behnood and Golafshani, 2018), end-of-life vehicles prediction in
Shafi, 2004) for recent years and the city has been suffering from economic and management research (Hao et al., 2018), predicting
unmanaged solid waste (Hasan, 1998; JICA, 2005 & BIGD, 2015). The the amount of separately-collected packaging waste in municipal
MSW of Dhaka city is collected by authorized primary collection solid waste research (Oliveira et al., 2018). ANN has gained supe-
service providers (door to door collection) to the nearby newly riority to conventional models in forecasting modeling study such
constructed mini transfer stations or large container carrier loca- as solid waste forecasting due to its surge of innovation and so-
tions or compactor locations in case of the narrow street. The phistication for solving complex problems with high speed (Abdoli
collected wastes then transferred to the landfill site. The MSW et al., 2011).
transferred to the landfill in Dhaka City contains food and vegeta- The development of forecasting municipal solid waste generation
bles (62.72%), paper product (9.83%), Plastic, leather, rubber (7.75%), is a growing body of literature (Abbasi and Hanandeh, 2016) that
metals (0.28), glass and ceramics (0.65%), wood (5.88%), green includes simple academic or research-oriented model to highly
waste (9.88%), moisture (5%), rock, dirt, debris (6.91%), rags, textile, technical heterogeneous complex applied model for both short
jute (3.18%), medical/chemical 0.37% and other solids (19.63%) as term like weekly solid waste prediction (Noori et al., 2009b) and long
shown in Table 3. In the year 2016 over about 2000 tones domestic, term time frame for the year 2011-2032 (Abdoli et al., 2011). Based on
commercial and industrial mixed municipal wastes were sent to the methodological approaches those model can be categorized
the Matuail landfill site in the Dhaka city mention in Fig. 2. The city into time series models that are to use the past data and their dis-
authorities have taken a lot of initiatives for better SWM and reduce tribution to determine future waste trends and factor models that is
the impacts of SWM (Mahmud, 2018). Transforming open dumping to use socioeconomic and other explanatory variables to explain and
site into a so-called sanitary landfill site (Yousuf and Rahman, 2009) predict waste quantities along with the inherent relationship of
is one of them. The installation of weighbridge in the year 2007 at factors responsible for that (Chung, 2010). Whereas Abbasi and
the landfill site has (Yousuf and Rahman, 2009) a milestone step. Hanandeh (2016) are classified into five major categories such as
That is capable to provide data on collected solid waste including descriptive statistical methods, regression analysis methods, ma-
trips numbers. A land scarce Bangladesh like other developing terial flow model, time series analysis and artificial intelligence
countries disposal of solid waste is a persistent problem including models. Most of the surrogate models are developed for solid waste
sufficient data on solid waste (Mmereki et al., 2016). It is unfortu- generation and the contribution of influential factors for the gener-
nate that the data available from weighbridge, are not utilized by ation (Antanasijevi c et al., 2013).
the Waste Management Department (WMD) of DSCC for the
improvement of SWM at the landfill site. There are a lot of oppor- 1.4. Gap analysis of solid waste forecasting models based on ANN
tunities to use these historical solid waste data for accurate fore-
casting of solid waste (Goel et al., 2017) for the planning of SWM Application of ANN for forecasting solid waste is summarized in
management. The predicted solid waste data can be used for the Table 2. Masebinu et al. (2017) found that historical length of the
better and effective operation of the landfill site, estimating the data, its quality and preciseness are critical for a reliable neural
future life span of the landfill site, and optimization of other network model performance after comparing an African city solid
disposal options like composting, waste to energy, incineration, etc. waste data with that of a European city. The model of the study
to save urban land area. testing was not carried out due to insufficient data. Collection fre-
quency is an important factor for determining the disposal rate at
1.3. Application of artificial neural network in solid waste the landfill, which is ignored in the study. In a city-scale study, Sun
prediction and Chungpaibulpatana (2017) showed, how ANN gives highly
accurate results of waste generation prediction and how the
In comparison with the traditional statistics, these can capable influencing factors such as total number of residents, native people
with the sophisticated nonlinear functions in multi-dimensional aged 15e59 year, total people aged 15e59 years, number of
spaces (Kannangara et al., 2017). The concept of ANNs is encour- households, income per household, and number of tourists
aged and stirred by the structure and functional characteristics of contribute to the waste generation. The prediction study was
human neurons and biological neural networks (Abbasi and conducted using waste data collected by the municipality which is
Hanandeh, 2016). Millions of interconnected cells which is called a major portion of actual waste generation (the collection fre-
neuron are formed both animal and human nervous systems quency is not mentioned) in the Bankok. In another study, Azadi
(Agatonovic-Kustrin and Beresford, 2000). Each neuron deals with and karimi-Jashni (2016) verified the performance of artificial
incoming signals in several ways that make a complex arrange- neural network (‘4-9-1’ structure, ‘Transig’ activation function, ‘LM’
ment. Similarly, a correlated set of artificial neurons built ANNs, and training function, R2 ¼ 0.86) in predicting the mean seasonal
for the computation, it processes information using a connectionist municipal solid waste generation rate for 20 cities of Fars province
 Md.M. Hoque, M.T.U. Rahman / Journal of Cleaner Production 256 (2020) 120387 3

in Iran using the variables population, solid waste collection fre- this study is to estimate landfill area by building an ANN-based
quency, maximum seasonal temperature and altitude. In this study prediction model for MSW collection at a landfill site that will
model, the mean seasonal waste collection rate prediction is forecast the MSW collected by the year 2030 at the Matuail landfill
possible instead of mean seasonal municipal generation rate. In a site of Dhaka city. That study outcome will reduce the uncertainty
tourist and tropical Langkawi Island in Malaysia, Shamshiry et al. of predicting model due to the use of less influencing factors and
(2014) found ANN (R2 ¼ 0.98) is a better modeling tool with high reliable data support for the landfill managers, planners, and con-
accuracy for prediction of weekly solid waste generation using solid tractors in the planning phase for calculation of landfill area. The
waste collection service contributing variables like type of trucks study results then can also use by the decision-makers as a basis for
and their trips, number of personnel in per trips and fuel cost. This future landfill management planning and evaluation. This knowl-
is a short term study where the seasonal variation is not reflected. edge can also provide a rationale for landfill area determination and
At a national scale, Antanasijevic et al. (2013) applied two different its different phases of operation.
ANN structures (R2 ¼ 0.930, R2 ¼ 0.981) in Bulgaria and Serbia to So it is evident that the forecasting of solid waste by ANN also
model and forecast municipal waste generation using sustainability depends on a lot of influential factors. Those ANN prediction
indicators such as gross domestic product, domestic material con- models are for any particular locality. Any model is area-specific
sumption, and resource productivity. The role of direct factors on and may not apply to the other city, region or country. The objec-
the waste generation and location-specific influence are not tive of the research paper, to estimate landfill area based on the
considered and discussed in the study. In another study, Shahabi solid waste prediction at a landfill in Dhaka Bangladesh using ANN
et al., (2012) developed a feed-forward artificial neural network and the final disposal options, will be a novel approach for future
model (R2 ¼ 0.648) to predict the weight of waste generation in study that will also save costly urban land area for future landfill
Saqqez city of Iran. Abdoli et al. (2011) were ANN modeled development.
considering population, household income and maximum tem-
perature factors for the long-term prediction of solid waste gen- 2. Materials and methods
eration in the city of Mashhad, Iran. The specific practical
application of the ANN model is not evaluated. Jalili and Noori Solid waste data at landfill site are time series data with cross-
(2008) found ANN (R2 ¼ 0.746) model has more advantages in sectional data so we may consider these as pooled data character-
comparison with traditional methods in predicting the weekly istics (Statistics Solutions, 2018) used for time series analysis to
municipal solid waste generation in the same city of Iran. They determine the patterns of time series or simply describe the un-
mentioned that there are few effective factors for waste generation derlying forces and structure that produced the time series and to
predictions. The study did not quantify or consider the factors in the fit the model for forecasting (STAT 510 | Applied Time Series
waste generation model prediction. For example, collection fre- Analysis, 2018). The major uses of such model are summarized by
quency is most crucial because the waste management authority is depicting the important features of the time series pattern, inter-
managing the collected waste only. In Tehran, Iran, waste genera- connection of past and future, predict the future values of the time
tion prediction by ANN was also successfully applied by Noori et al. series and serve as a process monitoring and control standard
(2009a and 2009b) using a short term like weekly data and without (Engineering Statistics Handbook, 2012). The different stages of the
recommending the practical use of the model for the waste research study are shown in Fig. 1.
manager.
A large number of researchers (few are mentioned in the 2.1. Study area
Table A-1) are carried out ANN modeling for predicting solid waste
generation, collection with varied data type, demographic area and According to 2011 census data of Bangladesh, the population
ANN network topology. They conclude that the ANN is a promising under Dhaka South City Corporation (DSCC) are about more than 30
approach for solid waste forecasting study. million and the forecasted populations in the year 2030 will be

1.5. Significance of the current research

DSCC Solid waste data Literature review: Application of
It is evident from the literature study that in developing coun- collection from weighbridge ANN Model for forecasting solid
at landfill site waste generation at landfill
tries majority of solid waste generated remaining unmanaged
(Ogwueleka, 2009; Ahsan et al., 2014) and rest are collected for
crude dumping or control dumping (Diaz, 2017; Duan et al., 2017).
The forecasting model of solid waste generation results are uncer-
ANN Model developments
tain due to the dynamic and unpredictable nature of social, economic using MATLAB 2016 Selecting Effective inputs
and demographic factors (Chhay et al., 2018) which are different
Sensitivity study
according to the scale (e.g., municipality, region or country) and
Model performance metrics
scope (e.g., weekly, seasonal, annual) of the application of the Forecasting monthly MSW
model (Oliveira et al., 2018). On the contrary, the weighbridge at the collection by ANN
landfill site produces solid waste data that are more reliable, fewer
biases due to less socio-economic factors and readily available.
Those data might be useful for predicting solid waste at the landfill Required land area Landfill area calculation based
calculation based on for waste density and waste
site. This future quantity of waste is crucial for plan, design, and Forecasted waste composition analysis considering
operation of landfill sites and other disposal option choices. Only a quantity (volume) proposed solid waste disposal
few research has discussed the application of ANN for predicting ti
solid waste at the landfill site using the weighbridge data. There is
urgent need to able to develop an ANN model for the solid waste
forecasting in a real case scenario at a landfill site particularly, as Landfill area estimation
most of the recent studies have focused on ANN models developed
for forecasting of MSW generation at source. The main objective of Fig. 1. Flowchart depicting the stages of the research study.
4 Md.M. Hoque, M.T.U. Rahman / Journal of Cleaner Production 256 (2020) 120387

Table 1
Summary of descriptive statistics.

Variables Month(year) n Min Max Median Average S

trip no. 2012 12 310 483 390 397 62.4

2013 12 405 480 445 442 22
2014 12 438 553 502 492 35.3
2015 12 460 603 497 497 43
2016 12 475 539 498 498 20.6

waste (ton) 2012 12 1076.2 1763 1267.4 1353.4 260.97

2013 12 1064.9 1701.3 1531.8 1501.4 162.96
2014 12 1181.4 2098.8 1750.2 1685.2 256.6
2015 12 1306.1 2295.5 1754.7 1764.8 245.4
2016 12 1659.4 2539.7 2018.2 1997.1 245.7

waste generation at sources such as total population, GDP, elec-

tricity demand, employment and unemployment numbers
(Kollikkathara et al., 2010), etc. The major reported factors of DSCC
Fig. 2. Matuail landfill site in Dhaka South city. responsible for waste collection at landfills are seasonal variation,
vehicle performance, drivers intention, budget for repair and
maintenance, a workforce of WMD, political commitment, con-
4894416. The Matuail landfill which was the largest and only one struction of mini secondary transfer station (STS), increase primary
recognized sanitary landfill in Bangladesh (Yousuf and Rahman, collection service provider (PCSP) numbers. All the above mention
2009) maintained by DSCC (BIGD, 2015). It is located about 8 km factors are reflected in the trip numbers of the waste collection
from the zero point, Gulistan in the south of Dhaka city. It is starting vehicles. So, in the present study, the trip number is considered
off with an open dump of 20 ha area in 1995, a further 20 ha has rather than the time variation of waste collection for ease of pre-
been extended to develop so-called Matuail sanitary landfill site in diction analysis and model evaluated based on the performance
2005. In the initial landfill area, the waste height is up to 70 feet metrics as mentioned earlier section.
(about 21 m) high. There is an almost 60 feet (about 18 m) waste
height in the adjacent 40 ha, which is used as a controlled landfill
from the year 2008. The basic requirements of a sanitary landfill 2.3.2. Data preprocessing
such as daily soil cover, leachate treatment, gas collection facility To predict solid waste at the landfill site, in this study trip
are barely met at Matuail. The present landfill is predicted to reach numbers and monthly data are used as input variables. Trip
its capacity within a year. Now the city authority is planning to numbers for each monthly collected solid waste and twelve months
acquire a further 81 ha adjacent land for the expansion of the in each year from the year 2012 to the year 2016 are used. Table 1
landfill site. If the current landfilling practice is prevailing in the described the statistical summary of the data. The maximum
future, it takes only five to seven years to reach the capacity. So, quantity shows the summer season and Muslim festival whereas
recycling, incineration, refuse-derived fuel, etc. have to be taken to the minimum quantity represents winter season waste generation.
extend the landfill life. Location of the study area shown in Fig. 2. From the data, it is evident that every the waste collection
increased gradually that shows the rapid industrialization and ur-
banization in the DSCC. The measured solid waste in each month
2.2. Solid waste data
was considered as an output variable in the model. The data set for
the entire five years consisting of 180 records (as input 60 no. trip
For the current forecasting study, total 60 months or 5 years
numbers and 60 no. months, as output 60 no. waste quantity) for all
(2012e2016) solid waste landfill data are collected from the waste
variables that are used in this study. The initial 120 observations
management department (WMD) of DSCC. In Table 1 the descrip-
were used to develop prediction models. The remaining observa-
tive statistics are shown. Monthly solid waste data was monitored
tions were used as test and validation data to verify the prediction
at the Matuail Landfill site in Dhaka, Bangladesh under DSCC using
models. The ANN models (see Fig. 3) are developed to predict the
weigh-bridge. The vehicle trips number per month also obtained
solid waste collected at the Matuail landfill Site. The material bal-
from the WMD of the DSCC.
ance is essential in chemistry, where the sum of reactants and
The number of trips and waste amount maintained by weigh-
products should converge to one. Here, the ANN model was not
bridge data except few manual data during maintenance of
learned from these boundary conditions, that is the sum of inputs
weigh-ridge and other unavoidable reasons. The DSCC landfill ex-
and sum of output must be same (100%). It is also difficult to
perts who had monitored the same site for several years suggested
analyze completely a complex matter such as daily trip number of
that the filtering of monthly solid waste data are not required for
vehicles (which ultimately depends on different factors as already
noise data created for the daily weighbridge data. So the unpro-
described) and their waste carrying time (month) and then esti-
cessed raw data were used for the study.
mate the average monthly (including the seasonal effect) trip
numbers with waste volume. Logically, we converge them to 100%
2.3. Model development by adding ‘others’. Here, in terms of the collections of waste by
average trip numbers, the monthly waste quantity is defined
2.3.1. Selecting effective inputs approximately 80%. The rest is probably other factors that affect the
The collected monthly solid waste data at landfill Weigh Bridge trip numbers. But it did not define it in the ANN model setup. The
are the percentage of total solid waste generation at source level in average trip numbers and the months in the year (whether the
the city. According to the WMD of DSCC, about 53%e65% solid summer or rainy season) exactly makes 100%. And the solid waste
waste collected by the DSCC. The fluctuation of the solid waste quantity makes up 100%. The real inputs and output variables are
collection depends on several factors including the factors for solid less than the variables trained and tested in the ANN model.
 Md.M. Hoque, M.T.U. Rahman / Journal of Cleaner Production 256 (2020) 120387 5

Table 2
ANN model training, testing and validation results.

ANN model structure MSE Optimum epoch No. of neuron Regression (R) Epoch equation

Training Data Testing Data Validation All

2-1-1-1 35063 5 1 0.805 0.81 0.68 0.79 11 0.66*targetþ4.8eþ03

2-2-1-1 50963 7 2 0.9086 0.85 0.82 0.87 13 0.77*targetþ3.5eþ02
2-3-1-1 2Eþ05 4 3 0.6872 0.68 0.87 0.68 10 0.77*targetþ3.5eþ03
2-4-1-1 7631 0 4 0.9101 0.99 0.96 0.93 6 0.85*targetþ2.5eþ02
2-5-1-1 9228 27 5 0.9932 0.96 0.95 0.98 33 0.96*targetþ70
2-6-1-1 32184 3 6 0.6744 0.66 0.84 0.69 9 0.49*targetþ7.3eþ02
2-7-1-1 3Eþ05 1 7 0.1248 0.13 0.43 0.12 2 0.0.18*targetþ9.2eþ02
2-8-1-1 21599 6 8 0.7757 0.80 0.68 0.75 12 0.88*targetþ2.6eþ02
2-9-1-1 11278 5 9 0.9358 0.95 0.90 0.93 11 0.87*targetþ1.7eþ02
2-10-1-1 36543 3 10 0.8981 0.93 0.88 0.90 9 0.86*targetþ3eþ02
2-11-1-1 9116 9 11 0.9321 0.80 0.96 0.91 15 0.96*targetþ94
2-12-1-1 6Eþ05 1 12 0.0577 0.83 0.53 0.22 1 0.43*targetþ2.1eþ03
2-13-1-1 2Eþ05 69 13 0.404 0.001 0.59 0.38 70 0.79*targetþ4.5eþ02
2-14-1-1 50614 4 14 0.9662 0.46 0.89 0.90 4 0.76*targetþ4.1eþ02
2-15-1-1 20320 4 15 0.8962 0.83 0.77 0.86 10 0.78*targetþ3.6eþ02
2-16-1-1 18087 8 16 0.9449 0.97 0.91 0.94 14 0.87*targetþ2eþ02
2-17-1-1 27394 1 17 0.7336 0.51 0.74 0.70 7 0.55*targetþ6.4eþ02
2-18-1-1 3Eþ05 1 18 0.1534 8.6e-17 0.46 0.19 8 0.32*targetþ7.7eþ02
2-19-1-1 32043 11 19 0.9978 0.66 0.77 0.94 17 0.88*targetþ2.1eþ02
2-20-1-1 46213 4 20 0.9618 0.58 0.50 0.82 10 0.77*targetþ3.9eþ02
2-21-1-1 18241 6 21 0.9447 0.45 0.90 0.69 12 0.85*targetþ2.5eþ19
2-22-1-1 17902 3 22 0.9547 0.62 0.92 0.89 9 0.91*targetþ2.1eþ02
2-23-1-1 3Eþ05 1 23 0.0909 0.50 0.50 0.19 7 0.34*targetþ7.8eþ02
2-24-1-1 1Eþ05 30 24 0.8023 0.61 0.74 0.72 36 1.1*target þ -3.8eþ02

Table 3
Municipal solid waste characterization/composition analysis in Dhaka City.

Waste Category Hossain et al., 2000 Huda (2002) Enayetullah and Hashmi (2006) JICA (2005) Average HHEECL (2017) This study Average

Food & Vegetable 70 59.91 67.65 56.5 68 54.28 62.72

Paper product 4 11.21 9.73 9.25 10 14.81 9.83
Plastic, lather, Rubber 5 17.67 5.1 2.33 6 10.42 7.75
Metals 0.13 0.15 0.26 NA NA 0.61 0.28
Glass and ceramics 0.25 NA 1.13 NA NA 0.58 0.65
Wood 0.16 NA 4.2 9.92 7 8.12 5.88
Green Waste 11 8.76 NA NA NA NA 9.88
Moisture 5 NA NA NA NA NA 5
Rock, dirt, debris NA 2.3 8.79 15.33 1 7.13 6.91
Rags, Textile, Jute NA NA 2.5 NA NA 3.87 3.18
Medicine/chemical NA NA 0.64 NA NA 0.11 0.37
Other 65 NA NA 6.25 7 0.27 19.63

Table 4
Average solid waste density at Matuail landfill site, Dhaka south city, Bangladesh.

waste category waste density (kg/m3) compaction ratio compacted density (kg/m3) Average waste composition (kg/m3) Category density (Kg/m3)

Food & Vegetable 223 3 669 62.72 419.62

Paper product 50 6.7 335 9.83 32.94
Plastic, lather, Rubber 90 10 900 7.75 69.78
Metals 306 6.7 2050.2 0.29 5.89
Glass and ceramics 350 2.5 875 0.65 5.72
Wood 96 3.3 316.8 5.88 18.63
Green Waste 210 1.5 315 9.88 31.12
Moisture 1000 1 1000 5.00 50.00
Rock, dirt, debris 313 4 1252 6.91 86.51
Rags, Textile, Jute 94 6.7 629.8 3.19 20.06
Medicine/chemical 160 2.6 416 0.38 1.56
Other 450 1 450 19.63 88.34
Total average waste density (kg/m3) 830.17

According to the degree of freedom, the input variables are two and model must be evaluated accordingly. In future studies, the
the output variable is just one. Since the ANN model was not boundary conditions may be informed into the ANN model and
learned with the above boundary conditions, the ANN model freely more precise results can be obtained.
predicts the waste quantity. And successfully the sum of the fore- To forecast one step ahead of the input that is solid waste
casted results fluctuates around 100%. The deviation is at most 5% counted data at time instance t is Wt is used in the model. The
theoretically and concentrating mainly at 3%. The results of ANN simplest form of the time series forecasting can be represented in
6 Md.M. Hoque, M.T.U. Rahman / Journal of Cleaner Production 256 (2020) 120387

1
A¼ (3)
1 þ expð  xÞ
The basic concept is enhancing the accuracy in the forecasting or
prediction by attaining at a function that minimizes the error (E)
between the input (actual) and output (predicted) variables.

hX
n i2
Eminðx;bÞ ¼ f X; W; b  y (4)
j¼1

In each training, mode uses random numbers of the weights and

proceeds iteratively. The most critical issue in the ANN model is
how to determine the numbers of hidden layers. It is essential to
estimate the appropriate number of neurons to avoid over- or
under-fitting. The numbers of input and output variables are fixed
the numbers of neuron in the input and output layers. The trial and
error method determines the number of hidden neurons. For any
given case of input and output historical data, the major governing
factors for an optimal network topology are the number of training
cases, percentages of testing and validation data, the amount of
noise in data and the complexity of the function.
The neural fitting tool of Matlab (2016) use a three-layer feed-
forward neural network with sigmoid hidden neurons and linear
output neurons to solve an input-output fitting problem. The ANN
network is trained with popular Levenberg-Marquardt back-prop-
agation algorithm (Shahabi et al., 2012), and it connects between a
data set of numeric inputs and a set of numeric targets. The
‘trainlm’ which is the fastest backpropagation algorithm in the
MATLAB toolbox that updates the weight and bias values according
Fig. 3. ANN surrogate model development schematic. to the Levenberg-Marquardt optimization. The neural fitting app of
MATLAB selects data, create and train a network, and validate its
performance using mean square error and other goodness of fit
Eq. (1):
criteria.
Wtþ1 ¼ f ðWt ; Wt1 ; ………:Wtn1 Þ (1)
In the above equation, the output or target waste amount at a
2.4. Goodness of fit
time (tþ1) is a function of the solid waste at a previous time. So the
one step ahead forecast of solid waste Wtþ1 is computed using n
In this paper, the prediction performance of the proposed
number of the past observations like Wt, Wt-1, … …. Wt-n-1. Where
forecasting models or the goodness of fit is evaluated through the
n was 120 in this study. Real weighbridge data on collected solid
following statistical tools: coefficient of determination (R2) and
waste at Matuail landfill obtained from the DSCC, Dhaka,
Mean Square Error (MSE). R2 and MSE are used to measure the
Bangladesh to compare the forecasting ability of artificial
discrepancy between simulated data and model data. The value of
intelligence.
R2 varies from 0 to 1. R2 values close to unity and small MSE values
are an indication of a good fit, more explicitly the simulated values
2.3.3. Theory and development of ANN surrogate model and model values have a good match.
The input layer consist of nodes (trip numbers and the monthly The coefficient of determination, R2, is described in the
solid waste amount at the landfill site) are required to forecast following Eq. (5).
monthly solid waste. Each node or neuron in a specific layer is
Pn
collected with all nodes of the following layer, called hidden or ðYo; i  Ym; iÞ2
target layer. A signal in every neuron or node of the hidden or target R2 ¼ 1  Pi¼1
n 2
(5)
i¼1 ðYo  Ym; iÞ
layer is then formed. The outcome is the function of the linear
integration of all incoming inputs. For example, the output value of
Y is determined with the sum of the individual weighted variables X
n
ðYo; i  Ym; iÞ2 ¼ The residual sum of square
(XW) and bias (b) that impacts the activation function A:
i¼1

X1 W1 þ X1 W1 þ X1 W1 þ ……… Xn Wn þ b ¼ Y ¼ A
X
n
X
n
ðYo  Ym; iÞ2 ¼ The total sum of square
¼ ðX1 W1 þ bÞ (2)
i¼1
j¼1

Where, A is called the transfer or activation function through which 1 X

n

the input signals (X1, X2, X3————Xn) and weighted values (W1,
Y¼ Yo; i; The mean of observed values
n
i¼1
W2, W3 …. Wn) are passed. The sigmoid transfer used in the study is
shown in Eq. (3): Using the Eq. (5) the MSE is calculated.
 Md.M. Hoque, M.T.U. Rahman / Journal of Cleaner Production 256 (2020) 120387 7

Table 5
Landfill height (waste height at landfills) in the literature.

Landfill Heights (m) Reference

1 <5 m to >30 m CPHEEO (2000)

2 3 me6 m (manual landfill) Jaramillo, (2003)
3 10 m- 25m Cointreau, (2004)
4 20 me25 m Pichtel, (2005)
5 <30m, > 30 m (huge landfill) Munawar and Fellner, (2013)
6 25 m Paul et al., (2014)
7 15 me30 m Younes et al., (2015)

Pn 2
i¼1 ðYo; i  Ym; iÞ
MSE ¼ (6)
n

Yo ¼ Simulated values
Ym ¼ model values
n ¼ Number of data sets or observations.

All the above computations were run on Windows 10 operating

system based computer and MATLAB Version R2016 of Mathworks
Inc.

2.5. Landfill area estimation in this study Fig. 4. Scatter plot of an ANN for both training and test set of a monthly solid waste for
sensitivity analysis.

To determine the required landfill area for waste disposal, the

developed ANN model has been implemented to forecast the solid 3.2. Model performance metrics
waste collection at the landfill site from the weighbridge data. In
this case, the general formula has been used for the landfill area The ANN model training, testing and validation results with the
estimation: goodness of fit for the monthly solid waste data that are shown in
Table 2. The best ANN network is selected based on MSE and the
1:5
Landfill Area ¼ Cr  Ls  P  (7) values of R for training, validation, and testing and overall.
ðrbulk  HÞ
Considering the criteria the 2-5-1-1 ANN topology gives the best
results. The best ANN structure achieved the learning ability (MSE)
Where Cr ¼ Waste collection rate (kg/capita year)
value 9228 for overall regression value 0.98 (Fig. 4).
Ls ¼ Life span of the landfill (years)
P¼ Population (number of persons) 3.3. Forecasting monthly MSW collection by ANN
rbulk ¼ waste bulk density (kg/m3),
H ¼ desired height of the landfill (m). After selecting the input variables different types of neuron
numbers were tested to determine the best function performance.
The landfill capacity is much depended on the height of the The input layer contained the trip number and month of the year,
landfill. As mention in Table 5, the height varies from 3 m (in case of and the output layer contained the solid waste quantity collection
small landfill) to more than 30 m (huge landfill). The existing at the landfill site. The number of neurons remained constant (five)
landfill height of 20 m is considered in this case. The predicted and the employed training algorithm was a hybrid. Fig. 5 shows the
waste collection at the landfill and the waste disposal options of final ANN structure.
draft DSCC waste management master plan 2030 will be utilized to The best model structures were used to get an effective evalu-
estimate the required landfill area in this study. To estimate the ation of ANN model performance. 70% of the data was used for
required waste disposal area the average solid waste composition training, 15% for testing and 15% of the dataset for validation.
from the literature will be included. Monthly MSW collection at Matuail landfill in Dhaka city was
predicted by ANN as shown in Fig. 7. The accuracy of the model
3. Results during training and testing stages are shown in Fig. 6.

3.1. Sensitivity study 3.4. Landfill area estimation

A sensitivity study on different networks with different hidden In this study implement aims of proposed draft 2030 Dhaka
neurons was conducted for solid waste collected at landfill site South City Master Plan for waste disposal options in Dhaka,
from the year 2012e2016 to find out the optimum number of Bangladesh is considered. Fig. 8 shows the projected solid waste
hidden nodes (neurons), as shown in Table 3. Both training and disposal options based on the forecasted model and the proposed
testing cases the coefficient of determination (R2) is close to unity draft 2030 DSCC solid waste master plan. The demand for landfill
near the same number of neurons (Fig. 4). A maximum correlation will remain although the alternative disposal options will be initi-
coefficient value can be clearly recognized at five neurons for the ated in Dhaka as a part of the future sustainable solid waste
case of all training, testing, and validation. management.
8 Md.M. Hoque, M.T.U. Rahman / Journal of Cleaner Production 256 (2020) 120387

Fig. 5. Schematic diagram of the ANN model for forecasting of solid waste collection.

waste density is required. The average waste composition reported

from literature during the period 2000e2018, which is shown in
Table 3 to estimate the average density of the municipal solid waste
in Dhaka City.
Based on the Dhaka solid waste composition Table 4 lists the
average solid waste density. The commonly used solid waste den-
sity at sanitary landfills, that is, 800 kg/m3 (Chong et al., 2005). In
this case, the estimated average waste density at the Matuail
landfill after compaction is approximately 830.17 kg/m3.
The forecasted amount of solid waste disposed at Matuail
Landfill up to the year 2030 is about 7845981.986 m3. The detail
estimated volume for two options are shown in Table 6. This
amount including 10% commingled solid waste incinerated. The
waste height in landfills varies and increases as the required
disposal area decreases. In the literature, the waste height at the
landfill is summarized in the Table 5. But in case of Matuail pro-
posed landfill site, the previous two adjacent landfill height limit
the maximum waste height that was 20 m. In the present study, a
waste disposal height of 20 m is considered.

4. Discussion

4.1. Parameters and correlation matrix

Only two parameters (trip number and month) are used as input
Fig. 6. Scatter plot of an ANN for both training and test set of a monthly solid waste
collection at Matuail landfill site, Dhaka South City, Bangladesh. in the model. Between these independents variable, it is evident
that the dependent variable (MSW) is most correlated with the trip
numbers which ultimately presents a number of factors as
According to the proposed draft master plan approximately mentioned in the introduction section. The bias and weights are
more than two third of the generated solid waste will be landfilled, randomly generated the MATLAB NN tool and ready the model for
as shown in Fig. 8. To estimate the required landfill area the solid forecasting. The accuracy of the forecasting model is measured by
 Md.M. Hoque, M.T.U. Rahman / Journal of Cleaner Production 256 (2020) 120387 9

mean square error (MSE) which ultimately depends on the preci- R2, more than 0.85. During the test stage, the differences between
sion achieved during the training stage. Fig. 6 schematically depicts the observed data and predicted value were not substantial
the ANN architecture used in this solid waste collection forecasting (R2 ¼ 0.86). The ANN model showed good agreement with the
in the study. It is evident from Fig. 5 plots that most of the data observed data, and close predictions of the observed peaks as
points lie on the straight line which is the proof of good perfor- clearly seen from Fig. 8. The high peaks are the months when the
mance of the developed model for sensitivity analysis. The trend of solid waste collection in the landfill site is maximum during the
training and testing set of a monthly solid waste data are the same. summer fruit season and Muslim religious festival (Eid Ul Azha). The
Initially, the coefficient of determination value for both train and generalization and performance of ANN-based forecasting model
test data matches very well compare to the increasing of neuron are quite satisfactory but more data is required to calibrate the ANN
numbers. model as also mentioned in the study of Pandey et al. (2016).

4.2. Optimum neuron numbers

4.5. Landfill area estimation based on the ANN model and final
A network of five hidden neurons in a hidden layer is used to disposal methods
develop the prediction model in this study. Our findings (row five of
Table 2) that the network containing the five neurons gives the From Fig. 8 we get the required landfill area considering the
lowest prediction errors and maximum correlation coefficient alternative disposal options will be proposed in the master plan.
(0.98) replicates the findings of optimum neuron numbers five in a The results based on the monthly solid waste data used as input. All
hidden layer for the ANN prediction model developed by Mokhtari the solid wastes are now coming from the area under the DSCC.
et al. (2014) to simulate the compression ratio of municipal solid During our study, DSCC has already expanded its jurisdiction,
waste. almost one and half times the present jurisdiction. That will not
In this study, five hidden neurons are enough for a single hidden affect the result too much. Because most the new area is rural, low
layer. As the input data variables are only two, so one hidden layer agricultural lands, wetlands accept a few small urban centers. The
with five neurons give satisfactory results. solid wastes of those urban centers are coming into main solid
waste stream unofficially. The chance for substantial urbanization
of the new area within the year 2030 that affect the solid waste
4.3. ANN based sensitivity study model performance collection quantity in the Matuail landfill is very low.
From Table 3 it is seen that there is a huge gap of solid waste
From Table 2 it can be observed that 1-9-1 is the best ANN composition study in the study area. The most reliable and rigorous
model due to minimum mean square error (9228) and maximum composition study was conducted by JICA during the formulation of
correlation (R ¼ 0.98) compared to other ANN models as already the first solid waste master plan of Dhaka city (JICA, 2005). Our
mentioned in section 4.2. The plots in Fig. 5 show that the accuracy present composition study findings are almost similar to that study.
of the model network on training data (R ¼ 0.99) is better than the There are two exceptions, one is the paper product waste category
testing data (R ¼ 0.96). Because in the training phase, the network (10 in JICA study and 14.81 in this study) and another is the plastic,
alters the weight values of inputs and output to obtain the best lather, rubber category (5.1 in JICA study and 10.42 in this study).
fitness whereas in the testing phase the output represents the This waste categories also represent the rapid urbanization and
performance of the trained model without adjusting the weights. industrialization of Dhaka City (Collier et al., 2017). Table 4 repre-
sents the calculated average solid waste density at Matuail landfill
4.4. ANN based MSW forecasting model performance site based on the average waste. It should be noted that only this
study waste composition analysis is conducted on collected waste
As evident from Fig. 7, the results indicated that the model at the landfill.
performed well during the training stage with a high goodness of fit The forecasted amount of solid waste disposed at Matuail

Fig. 7. Monthly solid waste collection forecasting at Matuail landfill site using the ANN model.
10 Md.M. Hoque, M.T.U. Rahman / Journal of Cleaner Production 256 (2020) 120387

Landfill Recycling Incineration Composting

5000
4500
4000

Solid Waste (Ton)

3500
3000
2500
2000
1500
1000
500
0
2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030
Projected Year

Fig. 8. Projected solid waste alternatives in Dhaka south city.

landfill up to the year 2030 is about 7845981.986 m3. This amount The implication of each coefficient and variable in the ANN
including 10% commingled solid waste incinerated. The final model is not easy to deduce, unlike in regression analysis. Although
landfill area should be extended by approximately 10e15% to slope the impact of these variables can be determined from the correla-
stability, dressing of slope, enable the use of daily and final soil tion matrix, for example, it is difficult to clearly identify the impact,
covers (Cointreau, 2004). An additional expansion of 40e50% is also as in regression analysis. In summary, rather than increase pre-
required to facilitate a receiving area and other treatment utilities dictability, ANN based model can be seen as comprising a “black
and administration buildings (CPHEEO, 2000). Pariatamby and box” model that has lower explanatory power similar finding from
Tanaka (2014) recommended that the landfill area should be at Pandey et al. (2016). For policy implications, ANN based model
least 60 ha to facilitate of leachate and gas collection and treatment. needs to be analyzed in combination with regression analysis
The large landfill areas more than 127 ha, should be avoided to because no neural network can clearly show what each model
reduce the operational impact and transportation cost (Seok Lim coefficient means.
and Missios, 2007). The ideal proposed number of future landfills
for DSCC is one with an area of 90 ha without considering the
5. Conclusion
recycling, composting and incineration, whereas landfill area can
be saved about 28.6% (25.74 ha) if the disposal methods are
This study reviews solid waste generation forecasting method-
considered (Table 6 and Table 7). This finding also similar to the
ologies by ANN. It shows an ANN model with minimum variable
recent extension of the landfill area of 81 ha by the DSCC.
inputs to predict monthly solid waste collection with acceptable
degree of accuracy and to estimate the future landfill area in
4.6. Implications of this study Bangladesh. The ANN based solid waste forecasting model at the
landfill site is performed to be promising using the available
When forecasting the solid waste collection using the ANN weighbridge data with the coefficient of determination (R2) for
model, the predicted value of the ANN model, was closest to the training and testing 0.85 and 0.86. The proposed ANN modelling
actual value, Fig. 7. Then this forecasted value is used to estimate approach can be alternatively used successfully with weight bridge
landfill area, Table 7. The ANN model, which could be forecasted software in the landfill to efficiently forecast solid waste collection,
from time-series data from the landfill weighbridge, are expected particularly that in the countries with a similar demographic and
to be useful in the SWM study of DSCC. In specific, it seems to apply social environment. The required landfill area for waste disposal
to prediction research in the SWM studies to plan and management decreased up to 28.6% of its former value (90 ha) considering the
of final disposal options in a sustainable way. proposed disposal options. The result shows that the innovative

Table 6
Estimated volume of waste collected at Matuail landfill (2018e2030).

Waste Category Waste Composition (%) Compacted density (kg/m3) Final waste volume at Landfill (a) Final waste volume options (b)

Food & Vegetable 62.72 669 4491858.28 3206913.80

Paper product 9.83 335 1406303.27 1004015.06
Plastic, lather, Rubber 7.75 900 412732.79 294666.13
Metals 0.28 2050.2 6718.38 4796.515
Glass and ceramics 0.65 875 35772.54 25539.42
Wood 5.88 316.8 889232.21 634857.75
Green Waste 9.88 315 1502690.09 1072829.39
Moisture 5 1000 239548.26 171022.90
Rock, dirt, debris & Misc 6.91 1252 264421.49 188780.87
Rags, Textile, Jute 3.185 629.8 242286.83 172978.07
Medicine/chemical 0.375 416 43187.78 30833.45
Other 19.63 450 2089925.56 1492079.83
Total volume (m3) 11624677.54 8299313.23
 Md.M. Hoque, M.T.U. Rahman / Journal of Cleaner Production 256 (2020) 120387 11

Table 7
Estimated land areas and land saving by the two options.

Option Total waste volume (m3) Required disposal area (m2) ¼ waste volume*1.1*1.4/waste height Area (hectare) Savings of land (%)

(a) 11624677.54 895100.1705 89.51 28.60

b) 8299313.23 639047.1187 63.90

proposed method of landfill area estimation by ANN and final Appendix

disposal methods is suitable for better policy, planning and man-

Table A-1
Application of ANN for forecasting solid waste generation study.

Sl no. Study Area Solid Waste Data Type Network (H ¼ hidden no. N ¼ neuron) Remarks/reference

1 Johannesburg, South Africa Yearly ANN (1H5N) Masebinu et al., (2017)

2 Bangkok, Thailand Yearly ANN (1H1N) Sun and Chungpaibulpatana, (2017)
3 Fars, Iran Monthly ANN (1H9N) Azadi and karimi-Jashni (2016)
4 Langkawi Island, Malaysia weekly ANN (3H3N) Shamshiry et al., (2014)
5 Belgrade, Serbia Kg/capita ANN (1H10N) Antanasijevi c et al. (2013)
6 Gujarat, India Yearly ANN (1H3N) Patel and Meka, (2013)
7 Saqqez city, Iran weekly ANN(1H16N) Shahabi et al., (2012)
8 Eluru, A.P, India yearly ANN Kumar et al., (2011)
9 Tehran, Iran Monthly ANN(1H10N) Abdoli et al., (2011)
10 Tehran, Iran weekly ANN(1H5N) Noori et al. (2009a)
11 Tehran, Iran weekly ANN(1H4N) Noori et al. (2009b)
12 Mashhad, Iran weekly ANN(1H16N) Jalili and Noori, (2008)

agement of the landfill site. In the future study the other influential
factors like number of staff, number of waste collection vehicles
and heavy equipment in operation at the landfill site can be References
considered as inputs to the ANN model development. The quanti-
tative modeling approach ANN should be further introduced with Abbasi, M., Hanandeh, A. El, 2016. Forecasting municipal solid waste generation
using artificial intelligence modelling approaches. Waste Manag. 56, 13e22.
practical improvement of the algorithm and boundary conditions in
[Link]
next stage of this research study and also compared with another Abdoli, M.A., Nezhad, M.F., Sede, R.S., Behboudian, S., 2011. Long term forecasting of
modeling approach like ANFIS (adaptive neuro-fuzzy inference solid waste generation by the artificial neural networks. Environ. Prog. Sustain.
system) to get a more efficient forecasting model. 31, 628e636. [Link]
Agatonovic-Kustrin, S., Beresford, R., 2000. Basic concepts of artificial neural
network (ANN) modeling and its application in pharmaceutical research.
Declaration of competing interest J. Pharmaceut. Biomed. 22 (5), 717e727. [Link]
7085(99)00272-1.
Ahsan, A., Alamgir, M., Elsergany, M., Shams, S., Rowshon, M.K., Daud, N.N., 2014.
The authors declare that they have no known competing Assessment of municipal solid waste management system in a developing
financial interests or personal relationships that could have country. Chinese J. of Eng. 2014 [Link] Article ID
appeared to influence the work reported in this paper. 561935.
Antanasijevic, D., Pocajt, V., Popovi
c, I., Red
zi
c, N., Ristic, M., 2013. The forecasting of
municipal waste generation using artificial neural networks and sustainability
indicators. Sustain. Sci. 8 (1), 37e46. [Link]
CRediT authorship contribution statement 9.
Azadi, S., Karimi-Jashni, A., 2016. Verifying the performance of artificial neural
network and multiple linear regression in predicting the mean seasonal
Md. Maruful Hoque: Conceptualization, Data curation, Meth-
municipal solid waste generation rate: a case study of Fars province, Iran. Waste
odology, Software, Validation, Writing - original draft, Visualiza- Manag. 48, 14e23. [Link]
tion, Investigation, Writing - review & editing. M. Tauhid Ur Bararpour, S.T., Feylizadeh, M.R., Delparish, A., Qanbarzadeh, M., Raeiszadeh, M.,
Rahman: Supervision, Project administration, Writing - review & Feilizadeh, M., 2017. Investigation of 2-nitrophenol solar degradation in the
simultaneous presence of K S O and H O: using experimental design and 2 2 8 2
editing. 2 artificial neural network. J. Clean. Prod. 176, 1154e1162. [Link] 10.1016/j.
jclepro.2017.11.191.
Behnood, A., Golafshani, E.M., 2018. Predicting the compressive strength of silica
Acknowledgments fume concrete using hybrid artificial neural network with multi-objective grey
wolves. J. Clean. Prod. 202, 54e64. [Link]
BIGD, 2015. BRAC Institute of Governance and Development, the State of Cities:
The authors are thankful to Dhaka South City Corporation Solid Waste Management of Dhaka Citye towards Decentralized Governance.
(DSCC) for supporting field activities, including MSW weighbridge BRAC University, Dhaka. [Link]
20OF%20CITIES%20Report%202015%20SWM%20in%20Dhaka%20Towards%
data supply. Special thanks to the anonymous editors and re- 20Decentralised%[Link]. (Accessed 21 November 2018).
viewers for their valuable comments to improve the paper. The Chen, Y.C., 2018. Effects of urbanization on municipal solid waste composition.
assistance of Climate change Lab of Civil Engineering (CE), Military Waste Manag. 79, 828e836. [Link]
Chen, Y.C., Wang, C.T., 2017. Municipal solid waste (MSW) incineration’s potential
Institute of Science and Technology (MIST) for their laptop, PC,
contribution to electricity production and economic revenue in Taiwan.
internet and other soft-wares for MSW forecasting is appreciated. J. Taiwan Energy. 4, 93e106.
Only funding for this research is received from the Higher Educa- CPHEEO, 2000. Central Public Health and Environmental Engineering Organization,
tion Quality Enhancement Project (HEQEP) of University Grants Manual on Municipal Solid Waste Management, Ministry of Urban Develop-
ment. government of India, New Delhi. [Link]
Comission (UGC), Dhaka, Bangladesh (grant no. CP-3143, for [Link]/files/Manual%20on%20municipal%20solid%20waste%
climate change lab of CE, MIST) and is gratefully acknowledged. 20management_%20MoUD_GOI_2000.pdf. (Accessed 10 December 2018).
12 Md.M. Hoque, M.T.U. Rahman / Journal of Cleaner Production 256 (2020) 120387

Chhay, L., Reyad, M.A.H., Suy, R., Islam, M.R., Mian, M.M., 2018. Municipal solid management-projects-gone-waste. (Accessed 7 July 2018).
waste generation in China: influencing factor analysis and multi-model fore- Masebinu, S.O., Akinlabi, E.T., Muzenda, E., Aboyade, A.O., Mbohwa, C.,
casting. J. Mater. Cycles. Waste. 20 (3) [Link] Manyuchi, M.M., Naidoo, P., 2017. Municipal solid waste data quality on artificial
4. neural network performance. In: Paper Presented at 2nd International Engi-
Chong, T.L., Matsufuji, Y., Hassan, M.N., 2005. Implementation of the semi-aerobic neering Conference, Minna, Nigeria.
landfill system (Fukuoka method) in developing countries: a Malaysia cost Matlab, 2016. MathWorks. Natick, Massachusetts, United States.
analysis. Waste Manag. 25, 702e711. Mmereki, D., Baldwin, A., Li, B., 2016. A comparative analysis of solid waste man-
Chung, S.S., 2010. Projecting municipal solid waste: the case of Hong Kong SAR. agement in developed, developing and lesser developed countries. Environ.
Resour. Conserv. Recycl. 54, 759e768. [Link] Technol. Rev. 5 (1), 120e141. [Link]
[Link].2009.11.012. Mokhtari, M., Heshmati, R., Ali, Akbar, Shariatmadari, Nader, 2014. Compression
Cointreau, S., 2004. Sanitary Landfill Design and Siting Criteria. Sanitary Landfill ratio of municipal solid waste simulation using artificial neural network and
Siting and Design Guidance. [Link] adaptive neurofuzzy system. Earth Sci. Res. J. 18 (2), 165e171. [Link]
461871468139209227/Sanitary-landfill-design-and-siting-criteria. (Accessed [Link]/[Link]?script¼sci_
15 November 2018). arttext&pid¼S179461902014000200010&lng¼en&tlng¼en. (Accessed 17 June
Collier, P., Anthony, J., Venables, 2017. Autumn 2017. Urbanization in developing 2019).
economies: the assessment, oxford review of economic policy, 33, pp. 355e372. Munawar, E., Fellner, J., 2013. Guidelines for Design and Operation of Municipal
[Link] 3. Solid Waste Landfills in Tropical Climates. Technische Universita €t Wien, Vienna,
Diaz, L.F., 2017. Waste management in developing countries and the circular Austria, 2013.
economy. Waste Manag. Res. 35 (1), 1e2. [Link] Noori, R., Abdoli, M., Ghazizade, M.J., Samieifard, R., 2009a. Comparison of neural
0734242X16681406. network and principal component- regression analysis to predict the solid
Duan, H., Li, J., Liu, G., 2017. Developing countries: growing threat of urban waste waste generation in Tehran. Iran. J. Public Health 38 (1), 74e84.
dumps. Nature 546, 599. [Link] Noori, R., Karbassi, A., Sabahi, M.S., 2009b. Evaluation of PCA and gamma test
Engineering Statistics Handbook, 2012. Definitions, Applications and Techniques. techniques on ANN operation for weekly solid waste prediction. J. Environ.
Retrieved from. [Link] Manag. 91, 767e771.
[Link]. (Accessed 21 December 2018). Ogwueleka, T.C., 2009. Municipal solid waste characteristics and management in
Enayetullah, I., Hashmi, Q.S.I., 2006. Community based solid waste management Nigeria. Iran J. Environ. Health Sci. Eng. 6, 173e180.
through public private community partnerships: experience of waste concern Oliveira, V., Sousa, V., Dias-Ferreira, C., 2018. Artificial neural network modelling of
in Bangladesh. In: Paper Presented in 3R Asia Conference, Tokyo, Japan. the amount of separately-collected household packaging waste. J. Clean. Prod.
Foo, K.Y., Hameed, B.H., 2009. An overview of landfill leachate treatment via acti- 210, 401e409. [Link]
vated carbon adsorption process. J. Hazard Mater. 171, 54e60. Pandey, D.S., Das, S., Pan, I., Leahy, J.J., Kwapinski, W., 2016. Artificial neural network
Goel, S., Ranjan, V.P., Bardhan, B., Hazra, T., 2017. Forecasting Solid Waste Generation based modelling approach for municipal solid waste gasification in a fluidized
Rates. Modelling Trends in Solid and Hazardous Waste Management, bed reactor. Waste Manag. 58, 202e213.
pp. 35e64. [Link] Pariatamby, A., Tanaka, M., 2014. Municipal Solid Waste Management in Asia and
Hao, H., Zhang, Q., Wang, Z., Zhang, J., 2018. Forecasting the number of end-of-life the Pacific Islands. Challenges and Strategic Solutions. Springer Singapore.
vehicles using a hybrid model based on grey model and artificial neural [Link]
network. J. Clean. Prod. 202, 684e696. [Link] Patel, V., Meka, S., 2013. Forecasting of municipal solid waste generation for me-
[Link].2018.08.176. dium scale towns located in state of Gujarat, India. Int. J. Innovative Res. Sci. 2
Hasan, S., 1998. Problems of municipal waste management in Bangladesh: an in- (9), 4707e4716.
quiry into its nature. Habitat Int. 22 (2), 191e202. [Link] Paul, k., Dutta, A., Krishna, A.P., 2014. A comprehensive study on landfill site se-
S0197-3975(97)00039-8. lection for Kolkata City, India. J. Air Waste Ma. 64 (7), 846e861. [Link]
Henan Huatai Environmental Engineering Co, LTD (HHEECL), 2017. Survey of Waste 10.1080/10962247.2014.896834.
Management Industry in Bangladesh - Taking Dhaka City as an Example. http:// Pichtel, J., 2005. Waste Management Practices-Municipal, Hazardous, and Indus-
[Link]/News/Industry_News/2017/1213/[Link]. (Accessed 22 trial. Taylor and Francis Group, LLC, New York.
December 2018). STAT 510 | Applied time series analysis, 2018. Overview of Time Series Character-
Hossain, M.D., Badruzzaman, A.B.M., Ali, M.A., 2000. Characterization of Municipal istics. Retrieved from. [Link]
Solid Waste of Dhaka City, Research Report. Environmental Engineering Divi- 1. (Accessed 22 December 2018).
sion, Department of Civil Engineering, BUET, Dhaka. Santiban ~ ez-Aguilar, J.E., Flores-Tlacuahuac, A., Rivera-Toledo, M., Ponce-Ortega, J.M.,
Huda, K.M.N., 2002. Municipal solid waste management Dhaka city perspective. In: 2017. Dynamic optimization for the planning of a waste management system
Ahmed, M.F., Tanveer, S.A., Badruzzaman, A.B.M. (Eds.), pp. 732e733. involving multiple cities. J. Clean. Prod. 165, 190e203. [Link]
Bangladesh Environment, 2002(2). [Link].2017.07.063.
Islam, N., Shafi, S.A., 2004. Solid Waste Management and the Urban Poor in Dhaka. Seok Lim, J., Missios, P., 2007. Does size really matter? Landfill scale impacts on
Uurban Infrastructure and Public Service Delivery for the Urban Poor Regional property values. Appl. Econ. Lett. 14, 719e723.
Focus: Asia. New Delhi, India. [Link] Shahabi, H., Khezri, S., Ahmad, B.B., Zabihi, H., 2012. Application of artificial neural
Islam%20and%20Shafi.doc. (Accessed 21 June 2018). network in prediction of municipal solid waste generation (case study: Saqqez
JICA, 2005. Japan International Cooperation Agency, the Study on Solid Waste city in Kurdistan province). World Appl. Sci. J. 20 (2), 336e343. [Link]
Management in Dhaka City: Clean Dhaka Master Plan. Dhaka City Corporation. 10.5829/[Link].2012.20.02.3769.
The People’s Republic of Bangladesh. [Link] Shams, S., Sahu, J.N., Rahman, S.M.S., Ahsan, A., 2017. Sustainable waste manage-
[Link]. (Accessed 1 July 2018). ment policy in Bangladesh for reduction of greenhouse gas. Sustain. Cities Soc.
Jalili, G.M., Noori, R., 2008. Prediction of municipal solid waste generation by use of 33, 18e26. [Link]
artificial neural network: a case study of Mashhad. Int. J. Environ. Res. 2, 13e22. Shamshiry, E., Mokhtar, M.B., Abdulai, A., 2014. Comparison of artificial neural
Jaramillo, J., 2003. Guidelines for the Design, Construction and Operation of Manual network (ANN) and multiple regression analysis for predicting the amount of
Sanitary Landfills. Pan American Center for Sanitary Engineering and Environ- solid waste generation in a tourist and tropical area - Langkawi Island. In: In-
mental Sciences. [Link] (Accessed 10 ternational Conference on Biological, Civil and Environmental Enginnering,
November 2018). pp. 161e166. [Link]
Kannangara, M., Dua, R., Ahmadi, L., Bensebaa, F., 2017. Modeling and prediction of Singh, D., Satija, A., 2016. Prediction of municipal solid waste generation for opti-
regional municipal solid waste generation and diversion in Canada using ma- mum planning and management with artificial neural network d case study:
chine learning approaches. Waste Manag. 74, 3e15. [Link] faridabad City in Haryana State (India). Int. J. Sys. Assur. Eng. Manage. 1e7
[Link].2017.11.057. [Link]
Kollikkathara, N., Feng, H., Yu, D., 2010. A system dynamic modeling approach for Statistics Solutions, 2018. Time Series Analysis. [Link]
evaluating municipal solid waste generation, landfill capacity and related cost time-series-analysis/. (Accessed 21 December 2018).
management issues. Waste Manag. 30 (11), 2194e2203. [Link] Sun, N., Chungpaibulpatana, S., 2017. Development of an appropriate model for
[Link].2010.05.012. forecasting municipal solid waste generation in Bangkok. Energy Procedia 138,
Kumar, J.S., Subbaiah, K.V., Rao, P., 2011. Prediction of municipal solid waste with 907e912. [Link]
RBF network-a case study of Eluru, AP, India. Int. J. Innovative Manage. Technol. Younes, M.K., Nopiah, Z.M., Basri, N.E.A., Basri, H., Abushammala, M.F.M.,
2, 238e243. Younes, M.Y., 2015. Landfill area estimation based on integrated waste disposal
Liikanena, M., Havukainena, J., Vianab, E., Horttanainena, M., 2018. Steps towards options and solid waste forecasting using modified ANFIS model. Waste Manag.
more environmentally sustainable municipal solid waste management e a life 55, 3e11. [Link]
cycle assessment study of Sa ~o Paulo, Brazil. J. Clean. Prod. 196, 150e162. https:// Yousuf, T.B., Rahman, M.M., 2009. Transforming an open dump into a sanitary
[Link]/10.1016/[Link].2018.06.005. landfill: a development effort in waste management. J. Mater. Cycles Waste
Mahmud, A.H., 2018. Waste Management Projects Gone to Waste. Dhaka Tribune. Manag. 11 (3), 277e280. [Link]
[Link]