- Email: [email protected]
Analysis of collection systems for sorted household waste in Spain
Waste Management 32 (2012) 1623–1633
Contents lists available at SciVerse ScienceDirect
Waste Management journal homepage: www.elsevier.com/locate/wasman
Country report
Analysis of collection systems for sorted household waste in Spain Antonio Gallardo ⇑, María D. Bovea, Francisco J. Colomer, Míriam Prades Department of Mechanical Engineering and Construction, Universidad Jaume I, Av. Sos Baynat s/n, E-12071, Castellón, Spain
a r t i c l e
i n f o
Article history: Received 17 October 2011 Accepted 5 April 2012 Available online 18 May 2012 Keywords: Beta regression Collection system Efficiency indicator
a b s t r a c t This work analyses the separate collection systems used in Spanish towns with between 5000 and 50,000 inhabitants. The study looks at the systems and their efficiency by means of the indicators fractioning rate, quality in container rate and separation rate. The results obtained are compared with those from a similar study conducted earlier that was applied to towns and cities with populations over 50,000. It can be concluded that the most widely implemented system in Spain involves the collection of mixed waste from kerbside bins and picking up paper/cardboard, glass and lightweight packaging from dropoff points. Findings show that the best system is the one that collects mixed waste, organic material and multiproduct waste door-to-door, and glass from drop-off points. The indicator separation rate made it possible to establish beta regression models to analyse the influence of the following logistic variables: inhabitants per point (people/pt), time (years) and frequency of collection (freq). From these models it can be seen that people/pt has a negative effect on all the fractions, while freq and years have a positive effect in the case of paper. Ó 2012 Elsevier Ltd. All rights reserved.
1. Introduction In order to fulfil both the European targets regarding waste (Directive 2004/12/EC) and the Spanish regulations, which make it mandatory for towns with over 5000 inhabitants to implement separate collection systems, town councils have had to design new models of refuse collection. This explains why a wide range of separate collection systems can be found both nationwide throughout Spain and across Europe. This situation has given rise to a number of studies and comparisons in the different countries in which they are in use. For example, in Sweden, Dahlén et al. (2007) conducted a study in which three separate collection systems were compared: kerbside collection of recyclable and biodegradable waste, kerbside collection of recyclable material and collection of recyclable waste at drop-off points. This study analyses the efficiency of collection systems from the point of view of the amounts of waste recovered, as a means of determining how close they come to the figures established by law. Nevertheless, there are also other approaches to assess the efficiency of collection systems, for instance, environmental and economic approaches. Many authors have focused their work on this kind of analysis. Environmental assessments were carried out in areas of Spain by Muñoz et al. (2003), Güereca et al. (2006), Iriarte et al. (2009), Rives et al. (2010) and Bovea et al. (2010) in order to quantify and compare the environmental impacts of different se⇑ Corresponding author. Tel.: +34 964 728 187; fax: +34 964 728 106. E-mail address: [email protected] (A. Gallardo). 0956-053X/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.wasman.2012.04.006
lective collection systems. Economic assessments were performed in Portugal by Gomes et al. (2008) to compare different alternative methods of collecting organic waste. Woodard et al. (2001) examined the waste that was recovered as a means of analysing the amounts of waste that were sent to landfills in two collection programmes in England. Wilson and Williams (2007) also determined the recycling rates after a new collection system had been implemented in Darwen (England). In Spain, Ayerbe and Pérez (2005) carried out an analysis to determine which system is best for collecting lightweight packaging. Berbel et al. (2001) conducted a similar study in Córdoba. In this case the authors compared two types of lightweight packaging collection: with containers for only this type of waste and with containers that also included inert waste. Gallardo et al. (2010) studied the separate collection systems in Spanish towns and cities with over 50,000 inhabitants by means of a set of efficiency indicators. Later, Ibáñez et al. (2011) used beta regression models to analyse the influence of a series of variables on efficiency indicators. This work analyses the collection systems for sorted household waste used in Spanish towns with between 5000 and 50,000 inhabitants from the point of view of the amounts of waste that are recovered. For this purpose a sample of the population to be studied was taken and a survey was sent to each member in order to obtain the information that was needed. The efficiency of each of the systems found was analysed by means of the indicators defined in Gallardo et al. (2010). Likewise, the influence of a series of logistic variables on the amount of waste collected separately was examined using beta regression models.
1624
A. Gallardo et al. / Waste Management 32 (2012) 1623–1633
Together, this work and the study by Gallardo et al. (2010) provide us with the first complete overview of the current state of selective waste collection in Spain.
2. Materials and methods 2.1. Determination of the sample The aim of this work is to determine which separate collection systems are implemented in Spanish towns with between 5000 and 50,000 inhabitants and to analyse their efficiency. Altogether, in 2008, 1145 municipalities met this condition (INE, 2008). Studying the whole population would be a complicated task that would require a great deal of time and effort. A sample was therefore defined to represent the whole population in terms of the characteristics to be studied. In this case, the characteristic that is to be studied in each individual (i.e. in each town) is the collection system for sorted household waste that is used. In a previous study (Gallardo et al., 2010) it was concluded that there are four main collection systems in Spanish towns and cities with over 50,000 inhabitants. These four systems were used to draw up a pilot survey and the data that we are going to use to define the sample is the proportion of towns and cities that employ each system. The variable to be studied in each individual is categorical, that is, it indicates a category, which in this case means belonging to a particular type of collection system. The proportions will therefore be the percentages of towns and cities that there are in each system with respect to the total number of towns and cities studied. The formulation for calculating a sample size, n0, depends on the aim of the study and on the characteristics to be analysed. In the case of categorical variables, according to Bartlett et al. (2001), the formula to be used is the following:
n0 ¼
z2a pq 2
d
ð1Þ
;
where za is the chosen level of confidence, which is determined by the value of a. Normally a confidence level of 95% is employed (a = 0.05), with which za = 1.96. p is the expected prevalence of the parameter that is to be estimated, which in our case is the proportion of towns and cities in each system, and q = 1 p. d is the precision, that is to say, the acceptable margin of error for the proportion that is estimated. A value of 5% is usually taken for such purposes. If the resulting sample size exceeds the population size by 5%, the corrected sample size formula, n1, must be used:
n1 ¼
n0 ; 1 þ n0 =N
From this calculation it was found that the sample size to be taken into account should be the one corresponding to system 1 (369) and thus the other sample sizes will also be included. If 5% of the size of the population is calculated, that is to say, 0.051145 = 57, the sample size obtained exceeds the size of the population. As mentioned above, in these cases corrected formula (2) should be used, the result being a sample size of 279. In Spain there are 17 Autonomous Communities (AC), each with different numbers of towns with between 5000 and 50,000 inhabitants. Hence, the number of municipalities considered in each AC is proportional to the sample size. Once the sample size has been obtained, the next step is to determine which towns need to be included in the sample. They must be chosen at random and to do so the algorithm shown below was created (Fig. 1) and run in the software application R (R Development Core Team, 2008). There is a database for each AC with municipalities between 5000 and 50,000 inhabitants. The input data of the algorithm are the different databases and the sample size that was obtained in the previous section. The algorithm returns towns at random. 2.2. Designing the survey Once the towns that are going to be included in the study have been determined, the final version of the survey is designed and sent out by post. In each survey, respondents are asked for the following information: General information about the municipality: number of inhabitants, surface area, existence of information campaigns and collection system in use. With regard to each of the waste fractions that are collected separately: tons collected per year; year the fraction was implemented in the selective collection system; number of containers; location of containers (door-to-door, kerbside, drop-off points) and frequency of collection. 3. Results and discussion 3.1. Participation Six months after sending out the 279 surveys, all the information received was pooled for its analysis. Documentation was also obtained about other municipalities from the official websites of the AC that publish this kind of data. All the data obtained were
ð2Þ
where N is the size of the population (N = 1145). The data gathered in the pilot survey made it possible to know the number of municipalities in each system, thus providing the first approximation to the calculation of the sample size using Eq. (1). The results obtained are shown in Table 1.
Table 1 Summary of the calculation of n0 for each collection system.
p q d za n0
System 1
System 2
System 3
System 4
0.6 0.4 0.05 1.96 369
0.09 0.91 0.05 1.96 126
0.01 0.99 0.05 1.96 15
0.3 0.7 0.05 1.96 323
Fig. 1. Algorithm for determining which towns belong to the sample.
A. Gallardo et al. / Waste Management 32 (2012) 1623–1633
1625
from years 2008 and 2009. Adding the 95 answered surveys to the 20 towns whose details were obtained from other sources makes a total of 115 towns for which information is available, i.e. 41% of all the towns in the sample. The participation rate was not higher due to the fact that the survey was long and most City Councils subcontract the collection of the different fractions out to different enterprises. This fact makes it more difficult to obtain the data.
over 50,000 (5%). This means that in small towns it is easier to implement selected kerbside collection of lightweight packaging and mixed waste. Systems 4 to 8 are only used in towns in Catalonia and the Balearic Islands, because under the Communities’ own legislation organic waste must be separated out from the rest. A new fraction, multiproduct, is also introduced with the aim of optimising collection (systems 6 and 7).
3.2. Collection systems implemented
3.3. Composition
One of the aims of these surveys was to determine which separate collection systems are currently in use in Spanish towns in terms of the location of the containers. In addition to the four systems pointed out in the works by Gallardo et al. (2010), four others were also obtained, some of which can be distinguished from the others only by small differences. Their characteristics are as follows:
It was difficult to obtain data about the composition of the urban waste because many towns do not know the overall composition. This is due to the high degree of fractioning of the collection systems and to the fact that different enterprises are responsible for handling each of the fractions. The best way to find the average composition of the urban waste in the towns that were studied was to use the data about the composition of the mixed waste fraction, since these data were known in some towns. These data provided the average composition of the mixed waste fraction for each of the collection systems and this was extended to the other towns with the same system for which no information was available. The composition of the separately collected waste fractions were added to the different fractions found in the mixed waste fraction. In the case of the lightweight packaging fraction, data were available for 31 municipalities, in which all the different systems were represented. The main results per system are shown in Table 3. Finally, the average composition of the urban waste was obtained from the average composition of each collection system. A graph with the average composition can be seen in Fig. 3. It can be observed that the main component of the waste is organic waste (37.29%), followed by paper/cardboard (18.43%) and plastic (10.86%).
System 1: separation into four fractions (mixed waste, paper/ cardboard, glass and lightweight packaging). The mixed waste is picked up from kerbside bins, whereas the paper/cardboard, glass and packaging are collected from drop-off points. System 2: separation into four fractions (mixed waste, paper/ cardboard, glass and lightweight packaging). The mixed waste and lightweight packaging are picked up from kerbside bins, whereas the paper/cardboard and glass are collected from drop-off points. System 3: separation into four fractions (mixed waste, organic waste, paper/cardboard and glass). The mixed waste and organic waste are picked up from kerbside bins, whereas the paper/cardboard and glass are collected from drop-off points. System 4: separation into five fractions (mixed waste, organic waste, paper/cardboard, glass and lightweight packaging). The mixed waste and organic waste are picked up from kerbside bins, whereas the paper/cardboard, glass and lightweight packaging are collected from drop-off points. System 5: separation into five fractions (mixed waste, organic waste, paper/cardboard, glass and lightweight packaging). The mixed waste is picked up from kerbside bins and the paper/ cardboard, glass and lightweight packaging are collected from drop-off points. Organic waste is collected door-to-door. System 6: separation into four fractions (mixed waste, organic waste, glass and multiproduct).1 The mixed waste and organic waste are picked up from kerbside bins, whereas the multiproduct and glass are collected from drop-off points. System 7: separation into four fractions (mixed waste, organic waste, glass and multiproduct). The mixed waste, organic waste and multiproduct are picked up door-to-door, while glass is collected from drop-off points. System 8: separation into five fractions (mixed waste, organic waste, paper/cardboard, glass and lightweight packaging). All fractions are collected at kerbside. The eight types of collection system can be seen in the diagram in Fig. 2. Table 2 shows the number of towns that have implemented each of the systems described above. On comparing the information from Table 2, where the findings of the study of towns and cities with more than 50,000 inhabitants (Gallardo et al., 2010) are given in the third column, it can be seen that in both cases system 1 is still the most widely used throughout the municipalities taken into account in the study (54% of the total number of towns and cities). It can also be concluded that in this case more municipalities use system 2 (21%) than in the towns and cities with populations 1
Multiproduct: lightweight packaging and paper/cardboard.
3.4. Efficiency indices As in the work by Gallardo et al. (2010), the following efficiency indicators, defined in that work – fractioning rate (FRi), quality in container rate (QCRi) and separation rate (SRi) – were calculated for each of the separately collected fractions.
FRi ¼
Gross amount of waste collected in container for i Total amount of urban waste 100ð%Þ
QCRi ¼
SRi ¼
Amount of waste collected correctly in container for i Gross amount of waste collected in container for i 100ð%Þ
Gross amount of waste collected in container for i Total amount of i waste generated 100ð%Þ
These indicators are presented below according to the collection system used. 3.4.1. Degrees of fractioning and quality Fig. 4a–d shows the different FRi obtained per system and Table 4 shows the QCRi for organic waste (QCRo), lightweight packaging (QCRlp) and multiproduct (QCRmult). In the case of glass and paper, personal communications were received from the disposal companies and the values are QCRp = 99% and QCRg = 97%. These are the values for all the systems. From the FRi and QCRi that were calculated it can be seen which system works best. The best results for FRo are found in system 7,
1626
A. Gallardo et al. / Waste Management 32 (2012) 1623–1633
Fig. 2. Diagram showing the selective collection systems.
Table 2 Number of towns using each system. System
No. towns 5000–50,000 inhab.
No. towns >50,000 inhab.
1 2 3 4 5 6 7 8
52 21 7 16 2 2 2 1
20 2 2 15 – – – –
Table 3 Composition of lightweight packaging.
System System System System System System System a
1 2 4 5 6 7 8
% Plastic
% Metal
% LPBa
% Inappropriate
50.83 49.14 56.19 56.34 37.08 37.08 56.19
12.95 13.40 12.98 7.84 7.52 7.52 12.98
13.54 10.16 8.56 25.34 4.51 4.51 8.56
22.69 27.30 22.28 10.49 50.90 50.90 22.28
LPB: liquid packaging board.
which also has the best QCRo. This system employs door-to-door collection, which is very practical for citizens because they do not have to go anywhere to deposit their waste. The system is suitable for towns and cities in which containers or bins can be placed inside buildings. The worst results for FRo and QCRo are for systems 5 and 3, respectively. The low value of FRo is due to the fact that the level of participation by citizens is very poor and they prefer to leave their waste in kerbside bins. Although the FRo is low, it can be seen how the QCRo is high (93.12%), which means that the few people who do use this system of collection do so correctly. The low value of QCRo in system 3 can be explained by the
Fig. 3. Average composition of urban waste.
proximity of the mixed waste container, because if the mixed waste container is full or if users are unsure about which bin to use, they can leave their mixed waste in the organic waste container. In turn, the mixed waste bin in system 3 contained approximately 40% of organic waste and hence campaigns should be carried out in order to provide citizens with further information about this kind of waste collection. With regard to the FR of lightweight packaging and multiproduct, it can be seen how system 2 is the best in the case of FRlp and the best QCRlp is found in system 5. The best FRmult and QCRmult values for multiproduct, on the other hand, are those offered by system 7. As in the case of organic waste, these values can be explained by the fact that it is a door-to-door collection system, which is the most practical for citizens. System 2 presents a lower QCRlp value, due to the proximity to the mixed waste bin. Overall, the rest of the QCRlp and QCRmult are similar to those of the other systems, which means that citizens are already familiar with this type of collection and many of them sort their waste correctly.
A. Gallardo et al. / Waste Management 32 (2012) 1623–1633
Fig. 4. Degrees of fractioning of all systems (FRi).
1627
1628
A. Gallardo et al. / Waste Management 32 (2012) 1623–1633
Fig. 4 (continued)
A. Gallardo et al. / Waste Management 32 (2012) 1623–1633
Fig. 4 (continued)
1629
1630
A. Gallardo et al. / Waste Management 32 (2012) 1623–1633
Fig. 4 (continued)
A. Gallardo et al. / Waste Management 32 (2012) 1623–1633 Table 4 Degrees of quality of each system. System
1
2
3
4
5
6
7
8
QCRo (%) QCRlp (%) QCRmult (%)
– 79.22 –
– 72.67 –
68.51 – –
83.82 78.09 –
93.12 88.39 –
90.80 – 74.33
97.67 – 76.81
92.96 77.64 –
3.4.2. Separation rate Table 5 shows the mean SRi for each collection system. It can be seen how the values of system 7 are still the highest of all the fractions that are collected selectively, which leads us to conclude that it is the best system. The efficiency and the best system were obtained from the point of view of the amounts recovered. Nevertheless, in order to have a well-dimensioned system, environmental and economic analyses should be conducted to reach a balance between the best options in each approach.
The analysis will examine how these variables affect the systems used to collect the paper/cardboard, glass and packaging fractions at the drop-off points. The multiproduct fraction will not be analysed due to the fact that it is implemented in very few towns, which makes it impossible to carry out a statistical analysis. As far as the organic waste fraction is concerned, some cases involve door-to-door collection and this makes it difficult to define variables such as Radius (Ri). The variable SR is defined as a proportion, that is, it is found within the interval (0, 1) and is asymmetrical, as can be seen in the histograms in Fig. 5. For this kind of variables, linear regression models are not always suitable because the variable of interest will exceed the upper and lower limits of the interval. One alternative for these situations is that proposed by Ferrari and Cribari-Neto (2004) and Ibáñez et al. (2011). Their proposal consists in using a regression model based on the beta distribution, which is very flexible for proportions. The density function of the beta distribution is:
f ðy; l; /Þ ¼ 3.4.3. Beta regression models In order to conduct the analysis of the indicator SR, several logistics variables were defined to study the effect they had on each of the SR. The variables are: people/pt.i: inhabitants per collection point of the fraction i = p, g, lp. R.i: radius of action of each point (metres). freq.i: frequency of collection (days/week). years.i: number of years since the system was first implemented. Another interesting variable to consider in the analysis was the existence of information campaigns. However, although the survey did ask about this subject, all the cities answered that they have campaigns, so it is not a variable. Table 5 Separation rates of each system. System
1
2
3
4
5
6
7
8
SRo (%) SRp (%) SRg (%) SRlp (%) SRmult (%)
– 27.07 45.97 15.44 –
– 14.37 39.46 22.66 –
71.51 27.67 56.46 – –
24.50 50.06 44.95 21.54 –
12.92 26.30 35.52 12.79 –
33.44 – 39.14 – 40.96
76.22 – 91.89 – 97.98
37.85 69.52 59.54 33.77 –
1631
Cð/Þ l/1 ð1 yÞð1lÞ/1 ; y Cðl/Þ
ð3Þ
where 0 < y < 1, 0 < l < 1, u > 0. l is the mean of the variable response and u is the dispersion parameter. Although these models are centred on the interval (0, 1), they can be used for any interval (a, b), where a and b are finite numbers and a < b. In this case the ya variable to be modelled instead of y would be ba . This analysis was performed using the statistical software application R (R Development Core Team, 2008), and more specifically the package betareg (Zeileis, 2009). In all three models, all the possible combinations were carried out with the logistic variables using the Akaike information criterion (AIC) (Akaike, 1974):
^ þ 2k; AIC ¼ 2lðbÞ ^ is the log-likelihood of the model, b ^ is the value that where lðbÞ maximises the log-likelihood and k is the number of estimated parameters. There are several procedures for selecting variables based on AIC and in this case the Stepwise method was used. This method consists in removing and/or adding at each step the variable that keeps the AIC value as low as possible and this process is repeated until it can no longer be reduced. Results are shown in Table 6. The table presents the estimated coefficients and their p-value, as well as the AIC obtained in the first and last steps of the Stepwise analysis. It can be seen how the variables people/pt.i and freq.i are present in all three models, while the variable R.i does not exert any
Fig. 5. Histograms of the separation rate of paper/cardboard, glass and lightweight packaging.
1632
A. Gallardo et al. / Waste Management 32 (2012) 1623–1633
Table 6 Beta regression models for each separation rate. GSp
Constant people/pt.i R.i years.i freq.i AIC (initial) AIC (final)
GSg
GSlp
Coefficient
p-Value
Coefficient
p-Value
Coefficient
p-Value
1.516 0.003 – 0.095 0.276
0.0001 0.0003 – 0.0020 0.0000
0.431 0.003 – – 0.062
0.016 0.000 – – 0.243
1.232 0.002 – – 0.052
0.000 0.002 – – 0.216
40.06 42.03
41.47 45.14
influence whatsoever in any of the cases. The variables people/pt.i have a negative influence, that is to say, the more inhabitants there are per point, the lower the SRi values will be. The fact that there are many inhabitants for a single collection point means that there are citizens who must walk a long way to deposit their waste. As a result these citizens tend to collaborate less in the collection. The variable freq.i, on the other hand, has a positive influence on the SRi although in the models of the SRg and SRlp it is not significant (p-value >0.05). This influence means that the higher the frequency of collection is, the higher the values of SRi are. This can be explained by the fact that if the frequency is low, overflows are produced and this may mean that citizens end up depositing the waste they sorted in their homes in the mixed waste bin. In the model of the SRp the variable years.p has a positive influence, which is obvious because the longer a system has been implemented, the more familiar citizens are with that method of collection. Once the influences have been determined, waste managers have to reach a balance between high efficiency indicators and economic and environmental factors in order to have a well-dimensioned system. 4. Conclusions This study and the one presented by Gallardo et al. (2010), together, make up a complete analysis of the current status of collection systems for sorted urban waste in Spanish towns and cities with over 5000 inhabitants, which are the ones that, by law, must have this kind of collection system. The analysis also states which factors are the ones that have the greatest influence on the functioning of each system. In this study eight different separate collection systems were found through the surveys sent to the towns. Four of the systems are the same as those found in the earlier project. The most widely implemented system is still the one that collects paper/cardboard, glass and lightweight packaging from drop-off points and mixed waste from kerbside bins. One new finding was that in some towns the multiproduct fraction (paper/cardboard and lightweight packaging) is collected with the aim of optimising waste collection. In Spain it is difficult to obtain the general composition of urban waste because of the high degree of fractioning. Administrations and entities responsible for municipal solid waste management should collect suitable data about their systems in order to be able to analyse the systems they use and introduce improvements into their management. The composition was found using data about the composition of the mixed waste fraction, and organic waste is still the main component (37%). The efficiency indicators quality in container rate (QCR) and fractioning rate (FR) were used to determine that the best system is number 7 (door-to-door collection of mixed waste, organic waste and multiproduct, and collection of glass at drop-off points). System 7 has the best QCR and FR for organic waste and for the multiproduct fraction.
62.15 66.14
The SRp, SRg and SRlp were analysed in terms of logistic variables. Beta regression models were used because they are the best suited for this kind of data. In the three beta regression models that were obtained, the variable people/pt.i had a negative influence on the SR: the more inhabitants there are per point, the lower the SR values will be. This means that if there are many inhabitants for a single collection point, there will be citizens who live further away and must travel a greater distance to deposit their waste. As a result these citizens collaborate less in the collection. In the beta regression model for paper, the variables freq.p and years.p exert a positive influence on the SRp. This freq.p can be explained by the fact that if the frequency is low, overflows are produced and this can mean that citizens end up depositing the waste that they sorted in their homes in the mixed waste bin. The variable years.p can be explained by the fact that the longer a system has been implemented, the more familiar citizens will be with the method of collection. This study analyses the efficiency of the different collection systems from the point of view of the amount of waste that is recovered. Another study on the environmental and economic efficiency should be conducted in order to find a balance between all the points of view. Acknowledgements We are grateful to the Spanish Ministry of the Environment and Rural and Marine Affairs for co-funding the project ‘‘Diseño de un modelo para la gestión de la recogida de residuos urbanos en las poblaciones españolas’’ (No. 150/pc08/3-02.4). Our gratitude also goes to the Conselleria de Educación de la Generalitat Valenciana for funding for the project ‘‘Diseño de un modelo para la gestión de la recogida de residuos urbanos en las poblaciones españolas’’ (No. ACOMP/2010/232). Finally, special thanks go to all the Town Councils that so kindly provided us with data in the surveys, since without their collaboration this work would not have been possible. References Akaike, H., 1974. A new look at the statistical model identification. IEEE Trans. Automat. Control 19 (6), 716–723. Ayerbe, S., Pérez, S., 2005. Alternativas para la recogida de envases ligeros. Ing. Quím. 423, 203–207. Bartlett, J., Kotrlik, J., Higgins, C., 2001. Organizational research: determining appropriate sample size in survey researches. Inf. Technol. Learn. Perform. J. 19 (1), 43–50. Berbel, J., Peñuelas, J., Ortiz, J., Gómez, M., 2001. Análisis comparado de modelos de recogida selectiva de envases/orgánico. Residuos 59, 52–57. Bovea, M.D., Ibáñez-Forés, V., Gallardo, A., Colomer-Mendoza, F.J., 2010. Environmental assessment of alternative municipal solid waste management strategies. A Spanish case study. Waste Manag. 30, 2383–2395. Dahlén, L., Vukicevic, S., Meijer, J., Lagerkvist, A., 2007. Comparison of different collection systems for sorted household waste in Sweden. Waste Manag. 27 (10), 1298–1305. Ferrari, S., Cribari-Neto, F., 2004. Beta regression for modelling rates and proportions. J. Appl. Stat. 31 (7), 799–815.
A. Gallardo et al. / Waste Management 32 (2012) 1623–1633 Gallardo, A., Bovea, M., Colomer, F., Prades, M., Carlos, M., 2010. Comparison of different collection systems for sorted household waste in Spain. Waste Manag. 31 (7), 379–406. Gomes, A., Matos, M., Carvalho, I., 2008. Separate collection of the biodegradable fraction of MSW: an economic assessment. Waste Manag. 28 (10), 1711–1719. Güereca, L.P., Gassó, S., Baldasano, J.M., Jiménez-Guerrero, P., 2006. Life cycle assessment of two biowaste management systems for Barcelona, Spain. Resour. Cons. Recycl. 49 (1), 32–48. Ibáñez, M.V., Prades, M., Simó, A., 2011. Modelling municipal waste separation rates using generalized linear models and beta regression. Resour. Cons. Recycl. 55 (12), 1129–1138. INE, 2008. Demografía y población. Cifras de población y censos demográficos. Instituto Nacional de Estadística. Iriarte, A., Gabarrell, X., Rieradevall, J., 2009. LCA of selective waste collection systems in dense urban areas. Waste Manage. 29, 903–914.
1633
Muñoz, I., Rieradevall, J., Doménech, X., Milà, L.L., 2003. LCA application to integrated waste management planning in Guipuzkoa (Spain). Int. J. Life Cycle Ass. 9 (4), 272–280. R Development Core Team, 2008. In: Computing, R.F. (Ed.), R: A Language and Environment for Statistical Computing. Vienna, Austria. Rives, J., Rieradevall, J., Gabarrell, X., 2010. LCA comparison of container systems in municipal waste management. Waste Manage. 30, 949–957. Wilson, C.D.H., Williams, I.D., 2007. Kerbside collection: a case study from the north-west of England. Resour. Cons. Recycl. 52, 381–394. Woodard, R., Harder, M., Bench, M., Philip, M., 2001. Evaluating the performance of a fortnightly collection of household waste separated into compostables, recyclates and refuse in south England. Resour. Conserv. Recycl. 31 (3), 265– 284. Zeileis, A., 2009. betareg: Beta Regression. Retrieved from:.
Contents lists available at SciVerse ScienceDirect
Waste Management journal homepage: www.elsevier.com/locate/wasman
Country report
Analysis of collection systems for sorted household waste in Spain Antonio Gallardo ⇑, María D. Bovea, Francisco J. Colomer, Míriam Prades Department of Mechanical Engineering and Construction, Universidad Jaume I, Av. Sos Baynat s/n, E-12071, Castellón, Spain
a r t i c l e
i n f o
Article history: Received 17 October 2011 Accepted 5 April 2012 Available online 18 May 2012 Keywords: Beta regression Collection system Efficiency indicator
a b s t r a c t This work analyses the separate collection systems used in Spanish towns with between 5000 and 50,000 inhabitants. The study looks at the systems and their efficiency by means of the indicators fractioning rate, quality in container rate and separation rate. The results obtained are compared with those from a similar study conducted earlier that was applied to towns and cities with populations over 50,000. It can be concluded that the most widely implemented system in Spain involves the collection of mixed waste from kerbside bins and picking up paper/cardboard, glass and lightweight packaging from dropoff points. Findings show that the best system is the one that collects mixed waste, organic material and multiproduct waste door-to-door, and glass from drop-off points. The indicator separation rate made it possible to establish beta regression models to analyse the influence of the following logistic variables: inhabitants per point (people/pt), time (years) and frequency of collection (freq). From these models it can be seen that people/pt has a negative effect on all the fractions, while freq and years have a positive effect in the case of paper. Ó 2012 Elsevier Ltd. All rights reserved.
1. Introduction In order to fulfil both the European targets regarding waste (Directive 2004/12/EC) and the Spanish regulations, which make it mandatory for towns with over 5000 inhabitants to implement separate collection systems, town councils have had to design new models of refuse collection. This explains why a wide range of separate collection systems can be found both nationwide throughout Spain and across Europe. This situation has given rise to a number of studies and comparisons in the different countries in which they are in use. For example, in Sweden, Dahlén et al. (2007) conducted a study in which three separate collection systems were compared: kerbside collection of recyclable and biodegradable waste, kerbside collection of recyclable material and collection of recyclable waste at drop-off points. This study analyses the efficiency of collection systems from the point of view of the amounts of waste recovered, as a means of determining how close they come to the figures established by law. Nevertheless, there are also other approaches to assess the efficiency of collection systems, for instance, environmental and economic approaches. Many authors have focused their work on this kind of analysis. Environmental assessments were carried out in areas of Spain by Muñoz et al. (2003), Güereca et al. (2006), Iriarte et al. (2009), Rives et al. (2010) and Bovea et al. (2010) in order to quantify and compare the environmental impacts of different se⇑ Corresponding author. Tel.: +34 964 728 187; fax: +34 964 728 106. E-mail address: [email protected] (A. Gallardo). 0956-053X/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.wasman.2012.04.006
lective collection systems. Economic assessments were performed in Portugal by Gomes et al. (2008) to compare different alternative methods of collecting organic waste. Woodard et al. (2001) examined the waste that was recovered as a means of analysing the amounts of waste that were sent to landfills in two collection programmes in England. Wilson and Williams (2007) also determined the recycling rates after a new collection system had been implemented in Darwen (England). In Spain, Ayerbe and Pérez (2005) carried out an analysis to determine which system is best for collecting lightweight packaging. Berbel et al. (2001) conducted a similar study in Córdoba. In this case the authors compared two types of lightweight packaging collection: with containers for only this type of waste and with containers that also included inert waste. Gallardo et al. (2010) studied the separate collection systems in Spanish towns and cities with over 50,000 inhabitants by means of a set of efficiency indicators. Later, Ibáñez et al. (2011) used beta regression models to analyse the influence of a series of variables on efficiency indicators. This work analyses the collection systems for sorted household waste used in Spanish towns with between 5000 and 50,000 inhabitants from the point of view of the amounts of waste that are recovered. For this purpose a sample of the population to be studied was taken and a survey was sent to each member in order to obtain the information that was needed. The efficiency of each of the systems found was analysed by means of the indicators defined in Gallardo et al. (2010). Likewise, the influence of a series of logistic variables on the amount of waste collected separately was examined using beta regression models.
1624
A. Gallardo et al. / Waste Management 32 (2012) 1623–1633
Together, this work and the study by Gallardo et al. (2010) provide us with the first complete overview of the current state of selective waste collection in Spain.
2. Materials and methods 2.1. Determination of the sample The aim of this work is to determine which separate collection systems are implemented in Spanish towns with between 5000 and 50,000 inhabitants and to analyse their efficiency. Altogether, in 2008, 1145 municipalities met this condition (INE, 2008). Studying the whole population would be a complicated task that would require a great deal of time and effort. A sample was therefore defined to represent the whole population in terms of the characteristics to be studied. In this case, the characteristic that is to be studied in each individual (i.e. in each town) is the collection system for sorted household waste that is used. In a previous study (Gallardo et al., 2010) it was concluded that there are four main collection systems in Spanish towns and cities with over 50,000 inhabitants. These four systems were used to draw up a pilot survey and the data that we are going to use to define the sample is the proportion of towns and cities that employ each system. The variable to be studied in each individual is categorical, that is, it indicates a category, which in this case means belonging to a particular type of collection system. The proportions will therefore be the percentages of towns and cities that there are in each system with respect to the total number of towns and cities studied. The formulation for calculating a sample size, n0, depends on the aim of the study and on the characteristics to be analysed. In the case of categorical variables, according to Bartlett et al. (2001), the formula to be used is the following:
n0 ¼
z2a pq 2
d
ð1Þ
;
where za is the chosen level of confidence, which is determined by the value of a. Normally a confidence level of 95% is employed (a = 0.05), with which za = 1.96. p is the expected prevalence of the parameter that is to be estimated, which in our case is the proportion of towns and cities in each system, and q = 1 p. d is the precision, that is to say, the acceptable margin of error for the proportion that is estimated. A value of 5% is usually taken for such purposes. If the resulting sample size exceeds the population size by 5%, the corrected sample size formula, n1, must be used:
n1 ¼
n0 ; 1 þ n0 =N
From this calculation it was found that the sample size to be taken into account should be the one corresponding to system 1 (369) and thus the other sample sizes will also be included. If 5% of the size of the population is calculated, that is to say, 0.051145 = 57, the sample size obtained exceeds the size of the population. As mentioned above, in these cases corrected formula (2) should be used, the result being a sample size of 279. In Spain there are 17 Autonomous Communities (AC), each with different numbers of towns with between 5000 and 50,000 inhabitants. Hence, the number of municipalities considered in each AC is proportional to the sample size. Once the sample size has been obtained, the next step is to determine which towns need to be included in the sample. They must be chosen at random and to do so the algorithm shown below was created (Fig. 1) and run in the software application R (R Development Core Team, 2008). There is a database for each AC with municipalities between 5000 and 50,000 inhabitants. The input data of the algorithm are the different databases and the sample size that was obtained in the previous section. The algorithm returns towns at random. 2.2. Designing the survey Once the towns that are going to be included in the study have been determined, the final version of the survey is designed and sent out by post. In each survey, respondents are asked for the following information: General information about the municipality: number of inhabitants, surface area, existence of information campaigns and collection system in use. With regard to each of the waste fractions that are collected separately: tons collected per year; year the fraction was implemented in the selective collection system; number of containers; location of containers (door-to-door, kerbside, drop-off points) and frequency of collection. 3. Results and discussion 3.1. Participation Six months after sending out the 279 surveys, all the information received was pooled for its analysis. Documentation was also obtained about other municipalities from the official websites of the AC that publish this kind of data. All the data obtained were
ð2Þ
where N is the size of the population (N = 1145). The data gathered in the pilot survey made it possible to know the number of municipalities in each system, thus providing the first approximation to the calculation of the sample size using Eq. (1). The results obtained are shown in Table 1.
Table 1 Summary of the calculation of n0 for each collection system.
p q d za n0
System 1
System 2
System 3
System 4
0.6 0.4 0.05 1.96 369
0.09 0.91 0.05 1.96 126
0.01 0.99 0.05 1.96 15
0.3 0.7 0.05 1.96 323
Fig. 1. Algorithm for determining which towns belong to the sample.
A. Gallardo et al. / Waste Management 32 (2012) 1623–1633
1625
from years 2008 and 2009. Adding the 95 answered surveys to the 20 towns whose details were obtained from other sources makes a total of 115 towns for which information is available, i.e. 41% of all the towns in the sample. The participation rate was not higher due to the fact that the survey was long and most City Councils subcontract the collection of the different fractions out to different enterprises. This fact makes it more difficult to obtain the data.
over 50,000 (5%). This means that in small towns it is easier to implement selected kerbside collection of lightweight packaging and mixed waste. Systems 4 to 8 are only used in towns in Catalonia and the Balearic Islands, because under the Communities’ own legislation organic waste must be separated out from the rest. A new fraction, multiproduct, is also introduced with the aim of optimising collection (systems 6 and 7).
3.2. Collection systems implemented
3.3. Composition
One of the aims of these surveys was to determine which separate collection systems are currently in use in Spanish towns in terms of the location of the containers. In addition to the four systems pointed out in the works by Gallardo et al. (2010), four others were also obtained, some of which can be distinguished from the others only by small differences. Their characteristics are as follows:
It was difficult to obtain data about the composition of the urban waste because many towns do not know the overall composition. This is due to the high degree of fractioning of the collection systems and to the fact that different enterprises are responsible for handling each of the fractions. The best way to find the average composition of the urban waste in the towns that were studied was to use the data about the composition of the mixed waste fraction, since these data were known in some towns. These data provided the average composition of the mixed waste fraction for each of the collection systems and this was extended to the other towns with the same system for which no information was available. The composition of the separately collected waste fractions were added to the different fractions found in the mixed waste fraction. In the case of the lightweight packaging fraction, data were available for 31 municipalities, in which all the different systems were represented. The main results per system are shown in Table 3. Finally, the average composition of the urban waste was obtained from the average composition of each collection system. A graph with the average composition can be seen in Fig. 3. It can be observed that the main component of the waste is organic waste (37.29%), followed by paper/cardboard (18.43%) and plastic (10.86%).
System 1: separation into four fractions (mixed waste, paper/ cardboard, glass and lightweight packaging). The mixed waste is picked up from kerbside bins, whereas the paper/cardboard, glass and packaging are collected from drop-off points. System 2: separation into four fractions (mixed waste, paper/ cardboard, glass and lightweight packaging). The mixed waste and lightweight packaging are picked up from kerbside bins, whereas the paper/cardboard and glass are collected from drop-off points. System 3: separation into four fractions (mixed waste, organic waste, paper/cardboard and glass). The mixed waste and organic waste are picked up from kerbside bins, whereas the paper/cardboard and glass are collected from drop-off points. System 4: separation into five fractions (mixed waste, organic waste, paper/cardboard, glass and lightweight packaging). The mixed waste and organic waste are picked up from kerbside bins, whereas the paper/cardboard, glass and lightweight packaging are collected from drop-off points. System 5: separation into five fractions (mixed waste, organic waste, paper/cardboard, glass and lightweight packaging). The mixed waste is picked up from kerbside bins and the paper/ cardboard, glass and lightweight packaging are collected from drop-off points. Organic waste is collected door-to-door. System 6: separation into four fractions (mixed waste, organic waste, glass and multiproduct).1 The mixed waste and organic waste are picked up from kerbside bins, whereas the multiproduct and glass are collected from drop-off points. System 7: separation into four fractions (mixed waste, organic waste, glass and multiproduct). The mixed waste, organic waste and multiproduct are picked up door-to-door, while glass is collected from drop-off points. System 8: separation into five fractions (mixed waste, organic waste, paper/cardboard, glass and lightweight packaging). All fractions are collected at kerbside. The eight types of collection system can be seen in the diagram in Fig. 2. Table 2 shows the number of towns that have implemented each of the systems described above. On comparing the information from Table 2, where the findings of the study of towns and cities with more than 50,000 inhabitants (Gallardo et al., 2010) are given in the third column, it can be seen that in both cases system 1 is still the most widely used throughout the municipalities taken into account in the study (54% of the total number of towns and cities). It can also be concluded that in this case more municipalities use system 2 (21%) than in the towns and cities with populations 1
Multiproduct: lightweight packaging and paper/cardboard.
3.4. Efficiency indices As in the work by Gallardo et al. (2010), the following efficiency indicators, defined in that work – fractioning rate (FRi), quality in container rate (QCRi) and separation rate (SRi) – were calculated for each of the separately collected fractions.
FRi ¼
Gross amount of waste collected in container for i Total amount of urban waste 100ð%Þ
QCRi ¼
SRi ¼
Amount of waste collected correctly in container for i Gross amount of waste collected in container for i 100ð%Þ
Gross amount of waste collected in container for i Total amount of i waste generated 100ð%Þ
These indicators are presented below according to the collection system used. 3.4.1. Degrees of fractioning and quality Fig. 4a–d shows the different FRi obtained per system and Table 4 shows the QCRi for organic waste (QCRo), lightweight packaging (QCRlp) and multiproduct (QCRmult). In the case of glass and paper, personal communications were received from the disposal companies and the values are QCRp = 99% and QCRg = 97%. These are the values for all the systems. From the FRi and QCRi that were calculated it can be seen which system works best. The best results for FRo are found in system 7,
1626
A. Gallardo et al. / Waste Management 32 (2012) 1623–1633
Fig. 2. Diagram showing the selective collection systems.
Table 2 Number of towns using each system. System
No. towns 5000–50,000 inhab.
No. towns >50,000 inhab.
1 2 3 4 5 6 7 8
52 21 7 16 2 2 2 1
20 2 2 15 – – – –
Table 3 Composition of lightweight packaging.
System System System System System System System a
1 2 4 5 6 7 8
% Plastic
% Metal
% LPBa
% Inappropriate
50.83 49.14 56.19 56.34 37.08 37.08 56.19
12.95 13.40 12.98 7.84 7.52 7.52 12.98
13.54 10.16 8.56 25.34 4.51 4.51 8.56
22.69 27.30 22.28 10.49 50.90 50.90 22.28
LPB: liquid packaging board.
which also has the best QCRo. This system employs door-to-door collection, which is very practical for citizens because they do not have to go anywhere to deposit their waste. The system is suitable for towns and cities in which containers or bins can be placed inside buildings. The worst results for FRo and QCRo are for systems 5 and 3, respectively. The low value of FRo is due to the fact that the level of participation by citizens is very poor and they prefer to leave their waste in kerbside bins. Although the FRo is low, it can be seen how the QCRo is high (93.12%), which means that the few people who do use this system of collection do so correctly. The low value of QCRo in system 3 can be explained by the
Fig. 3. Average composition of urban waste.
proximity of the mixed waste container, because if the mixed waste container is full or if users are unsure about which bin to use, they can leave their mixed waste in the organic waste container. In turn, the mixed waste bin in system 3 contained approximately 40% of organic waste and hence campaigns should be carried out in order to provide citizens with further information about this kind of waste collection. With regard to the FR of lightweight packaging and multiproduct, it can be seen how system 2 is the best in the case of FRlp and the best QCRlp is found in system 5. The best FRmult and QCRmult values for multiproduct, on the other hand, are those offered by system 7. As in the case of organic waste, these values can be explained by the fact that it is a door-to-door collection system, which is the most practical for citizens. System 2 presents a lower QCRlp value, due to the proximity to the mixed waste bin. Overall, the rest of the QCRlp and QCRmult are similar to those of the other systems, which means that citizens are already familiar with this type of collection and many of them sort their waste correctly.
A. Gallardo et al. / Waste Management 32 (2012) 1623–1633
Fig. 4. Degrees of fractioning of all systems (FRi).
1627
1628
A. Gallardo et al. / Waste Management 32 (2012) 1623–1633
Fig. 4 (continued)
A. Gallardo et al. / Waste Management 32 (2012) 1623–1633
Fig. 4 (continued)
1629
1630
A. Gallardo et al. / Waste Management 32 (2012) 1623–1633
Fig. 4 (continued)
A. Gallardo et al. / Waste Management 32 (2012) 1623–1633 Table 4 Degrees of quality of each system. System
1
2
3
4
5
6
7
8
QCRo (%) QCRlp (%) QCRmult (%)
– 79.22 –
– 72.67 –
68.51 – –
83.82 78.09 –
93.12 88.39 –
90.80 – 74.33
97.67 – 76.81
92.96 77.64 –
3.4.2. Separation rate Table 5 shows the mean SRi for each collection system. It can be seen how the values of system 7 are still the highest of all the fractions that are collected selectively, which leads us to conclude that it is the best system. The efficiency and the best system were obtained from the point of view of the amounts recovered. Nevertheless, in order to have a well-dimensioned system, environmental and economic analyses should be conducted to reach a balance between the best options in each approach.
The analysis will examine how these variables affect the systems used to collect the paper/cardboard, glass and packaging fractions at the drop-off points. The multiproduct fraction will not be analysed due to the fact that it is implemented in very few towns, which makes it impossible to carry out a statistical analysis. As far as the organic waste fraction is concerned, some cases involve door-to-door collection and this makes it difficult to define variables such as Radius (Ri). The variable SR is defined as a proportion, that is, it is found within the interval (0, 1) and is asymmetrical, as can be seen in the histograms in Fig. 5. For this kind of variables, linear regression models are not always suitable because the variable of interest will exceed the upper and lower limits of the interval. One alternative for these situations is that proposed by Ferrari and Cribari-Neto (2004) and Ibáñez et al. (2011). Their proposal consists in using a regression model based on the beta distribution, which is very flexible for proportions. The density function of the beta distribution is:
f ðy; l; /Þ ¼ 3.4.3. Beta regression models In order to conduct the analysis of the indicator SR, several logistics variables were defined to study the effect they had on each of the SR. The variables are: people/pt.i: inhabitants per collection point of the fraction i = p, g, lp. R.i: radius of action of each point (metres). freq.i: frequency of collection (days/week). years.i: number of years since the system was first implemented. Another interesting variable to consider in the analysis was the existence of information campaigns. However, although the survey did ask about this subject, all the cities answered that they have campaigns, so it is not a variable. Table 5 Separation rates of each system. System
1
2
3
4
5
6
7
8
SRo (%) SRp (%) SRg (%) SRlp (%) SRmult (%)
– 27.07 45.97 15.44 –
– 14.37 39.46 22.66 –
71.51 27.67 56.46 – –
24.50 50.06 44.95 21.54 –
12.92 26.30 35.52 12.79 –
33.44 – 39.14 – 40.96
76.22 – 91.89 – 97.98
37.85 69.52 59.54 33.77 –
1631
Cð/Þ l/1 ð1 yÞð1lÞ/1 ; y Cðl/Þ
ð3Þ
where 0 < y < 1, 0 < l < 1, u > 0. l is the mean of the variable response and u is the dispersion parameter. Although these models are centred on the interval (0, 1), they can be used for any interval (a, b), where a and b are finite numbers and a < b. In this case the ya variable to be modelled instead of y would be ba . This analysis was performed using the statistical software application R (R Development Core Team, 2008), and more specifically the package betareg (Zeileis, 2009). In all three models, all the possible combinations were carried out with the logistic variables using the Akaike information criterion (AIC) (Akaike, 1974):
^ þ 2k; AIC ¼ 2lðbÞ ^ is the log-likelihood of the model, b ^ is the value that where lðbÞ maximises the log-likelihood and k is the number of estimated parameters. There are several procedures for selecting variables based on AIC and in this case the Stepwise method was used. This method consists in removing and/or adding at each step the variable that keeps the AIC value as low as possible and this process is repeated until it can no longer be reduced. Results are shown in Table 6. The table presents the estimated coefficients and their p-value, as well as the AIC obtained in the first and last steps of the Stepwise analysis. It can be seen how the variables people/pt.i and freq.i are present in all three models, while the variable R.i does not exert any
Fig. 5. Histograms of the separation rate of paper/cardboard, glass and lightweight packaging.
1632
A. Gallardo et al. / Waste Management 32 (2012) 1623–1633
Table 6 Beta regression models for each separation rate. GSp
Constant people/pt.i R.i years.i freq.i AIC (initial) AIC (final)
GSg
GSlp
Coefficient
p-Value
Coefficient
p-Value
Coefficient
p-Value
1.516 0.003 – 0.095 0.276
0.0001 0.0003 – 0.0020 0.0000
0.431 0.003 – – 0.062
0.016 0.000 – – 0.243
1.232 0.002 – – 0.052
0.000 0.002 – – 0.216
40.06 42.03
41.47 45.14
influence whatsoever in any of the cases. The variables people/pt.i have a negative influence, that is to say, the more inhabitants there are per point, the lower the SRi values will be. The fact that there are many inhabitants for a single collection point means that there are citizens who must walk a long way to deposit their waste. As a result these citizens tend to collaborate less in the collection. The variable freq.i, on the other hand, has a positive influence on the SRi although in the models of the SRg and SRlp it is not significant (p-value >0.05). This influence means that the higher the frequency of collection is, the higher the values of SRi are. This can be explained by the fact that if the frequency is low, overflows are produced and this may mean that citizens end up depositing the waste they sorted in their homes in the mixed waste bin. In the model of the SRp the variable years.p has a positive influence, which is obvious because the longer a system has been implemented, the more familiar citizens are with that method of collection. Once the influences have been determined, waste managers have to reach a balance between high efficiency indicators and economic and environmental factors in order to have a well-dimensioned system. 4. Conclusions This study and the one presented by Gallardo et al. (2010), together, make up a complete analysis of the current status of collection systems for sorted urban waste in Spanish towns and cities with over 5000 inhabitants, which are the ones that, by law, must have this kind of collection system. The analysis also states which factors are the ones that have the greatest influence on the functioning of each system. In this study eight different separate collection systems were found through the surveys sent to the towns. Four of the systems are the same as those found in the earlier project. The most widely implemented system is still the one that collects paper/cardboard, glass and lightweight packaging from drop-off points and mixed waste from kerbside bins. One new finding was that in some towns the multiproduct fraction (paper/cardboard and lightweight packaging) is collected with the aim of optimising waste collection. In Spain it is difficult to obtain the general composition of urban waste because of the high degree of fractioning. Administrations and entities responsible for municipal solid waste management should collect suitable data about their systems in order to be able to analyse the systems they use and introduce improvements into their management. The composition was found using data about the composition of the mixed waste fraction, and organic waste is still the main component (37%). The efficiency indicators quality in container rate (QCR) and fractioning rate (FR) were used to determine that the best system is number 7 (door-to-door collection of mixed waste, organic waste and multiproduct, and collection of glass at drop-off points). System 7 has the best QCR and FR for organic waste and for the multiproduct fraction.
62.15 66.14
The SRp, SRg and SRlp were analysed in terms of logistic variables. Beta regression models were used because they are the best suited for this kind of data. In the three beta regression models that were obtained, the variable people/pt.i had a negative influence on the SR: the more inhabitants there are per point, the lower the SR values will be. This means that if there are many inhabitants for a single collection point, there will be citizens who live further away and must travel a greater distance to deposit their waste. As a result these citizens collaborate less in the collection. In the beta regression model for paper, the variables freq.p and years.p exert a positive influence on the SRp. This freq.p can be explained by the fact that if the frequency is low, overflows are produced and this can mean that citizens end up depositing the waste that they sorted in their homes in the mixed waste bin. The variable years.p can be explained by the fact that the longer a system has been implemented, the more familiar citizens will be with the method of collection. This study analyses the efficiency of the different collection systems from the point of view of the amount of waste that is recovered. Another study on the environmental and economic efficiency should be conducted in order to find a balance between all the points of view. Acknowledgements We are grateful to the Spanish Ministry of the Environment and Rural and Marine Affairs for co-funding the project ‘‘Diseño de un modelo para la gestión de la recogida de residuos urbanos en las poblaciones españolas’’ (No. 150/pc08/3-02.4). Our gratitude also goes to the Conselleria de Educación de la Generalitat Valenciana for funding for the project ‘‘Diseño de un modelo para la gestión de la recogida de residuos urbanos en las poblaciones españolas’’ (No. ACOMP/2010/232). Finally, special thanks go to all the Town Councils that so kindly provided us with data in the surveys, since without their collaboration this work would not have been possible. References Akaike, H., 1974. A new look at the statistical model identification. IEEE Trans. Automat. Control 19 (6), 716–723. Ayerbe, S., Pérez, S., 2005. Alternativas para la recogida de envases ligeros. Ing. Quím. 423, 203–207. Bartlett, J., Kotrlik, J., Higgins, C., 2001. Organizational research: determining appropriate sample size in survey researches. Inf. Technol. Learn. Perform. J. 19 (1), 43–50. Berbel, J., Peñuelas, J., Ortiz, J., Gómez, M., 2001. Análisis comparado de modelos de recogida selectiva de envases/orgánico. Residuos 59, 52–57. Bovea, M.D., Ibáñez-Forés, V., Gallardo, A., Colomer-Mendoza, F.J., 2010. Environmental assessment of alternative municipal solid waste management strategies. A Spanish case study. Waste Manag. 30, 2383–2395. Dahlén, L., Vukicevic, S., Meijer, J., Lagerkvist, A., 2007. Comparison of different collection systems for sorted household waste in Sweden. Waste Manag. 27 (10), 1298–1305. Ferrari, S., Cribari-Neto, F., 2004. Beta regression for modelling rates and proportions. J. Appl. Stat. 31 (7), 799–815.
A. Gallardo et al. / Waste Management 32 (2012) 1623–1633 Gallardo, A., Bovea, M., Colomer, F., Prades, M., Carlos, M., 2010. Comparison of different collection systems for sorted household waste in Spain. Waste Manag. 31 (7), 379–406. Gomes, A., Matos, M., Carvalho, I., 2008. Separate collection of the biodegradable fraction of MSW: an economic assessment. Waste Manag. 28 (10), 1711–1719. Güereca, L.P., Gassó, S., Baldasano, J.M., Jiménez-Guerrero, P., 2006. Life cycle assessment of two biowaste management systems for Barcelona, Spain. Resour. Cons. Recycl. 49 (1), 32–48. Ibáñez, M.V., Prades, M., Simó, A., 2011. Modelling municipal waste separation rates using generalized linear models and beta regression. Resour. Cons. Recycl. 55 (12), 1129–1138. INE, 2008. Demografía y población. Cifras de población y censos demográficos. Instituto Nacional de Estadística. Iriarte, A., Gabarrell, X., Rieradevall, J., 2009. LCA of selective waste collection systems in dense urban areas. Waste Manage. 29, 903–914.
1633
Muñoz, I., Rieradevall, J., Doménech, X., Milà, L.L., 2003. LCA application to integrated waste management planning in Guipuzkoa (Spain). Int. J. Life Cycle Ass. 9 (4), 272–280. R Development Core Team, 2008. In: Computing, R.F. (Ed.), R: A Language and Environment for Statistical Computing. Vienna, Austria. Rives, J., Rieradevall, J., Gabarrell, X., 2010. LCA comparison of container systems in municipal waste management. Waste Manage. 30, 949–957. Wilson, C.D.H., Williams, I.D., 2007. Kerbside collection: a case study from the north-west of England. Resour. Cons. Recycl. 52, 381–394. Woodard, R., Harder, M., Bench, M., Philip, M., 2001. Evaluating the performance of a fortnightly collection of household waste separated into compostables, recyclates and refuse in south England. Resour. Conserv. Recycl. 31 (3), 265– 284. Zeileis, A., 2009. betareg: Beta Regression. Retrieved from: