Model for the separate collection of packaging waste in Portuguese low-performing recycling regions

Journal of Environmental Management xxx (2017) 1e12

Contents lists available at ScienceDirect

Journal of Environmental Management journal homepage: www.elsevier.com/locate/jenvman

Research article

Model for the separate collection of packaging waste in Portuguese low-performing recycling regions V. Oliveira a, V. Sousa b, J.M. Vaz c, C. Dias-Ferreira a, d, * a Research Centre for Natural Resources, Environment and Society (CERNAS), College of Agriculture (ESAC), Polytechnic Institute of Coimbra, Bencanta, 3045-601, Coimbra, Portugal b Department of Civil Engineering, Architecture and GeoResources, Tecnico Lisboa - IST, Av. Rovisco Pais, Lisbon, Portugal c ECOGESTUS Lda, Waste Management Consulting, Figueira da Foz, Portugal d
rio de Santiago, 3810-193 Aveiro, Portugal Materials and Ceramic Engineering Department, CICECO, University of Aveiro, Campus Universita

a r t i c l e i n f o

a b s t r a c t

Article history: Received 19 February 2017 Received in revised form 3 April 2017 Accepted 22 April 2017 Available online xxx

Separate collection of packaging waste (glass; plastic/metals; paper/cardboard), is currently a widespread practice throughout Europe. It enables the recovery of good quality recyclable materials. However, separate collection performance are quite heterogeneous, with some countries reaching higher levels than others. In the present work, separate collection of packaging waste has been evaluated in a lowperformance recycling region in Portugal in order to investigate which factors are most affecting the performance in bring-bank collection system. The variability of separate collection yields (kg per inhabitant per year) among 42 municipalities was scrutinized for the year 2015 against possible explanatory factors. A total of 14 possible explanatory factors were analysed, falling into two groups: socio-economic/demographic and waste collection service related. Regression models were built in an attempt to evaluate the individual effect of each factor on separate collection yields and predict changes on the collection yields by acting on those factors. The best model obtained is capable to explain 73% of the variation found in the separate collection yields. The model includes the following statistically significant indicators affecting the success of separate collection yields: i) inhabitants per bring-bank; ii) relative accessibility to bring-banks; iii) degree of urbanization; iv) number of school years attended; and v) area. The model presented in this work was developed specifically for the bring-bank system, has an explanatory power and quantifies the impact of each factor on separate collection yields. It can therefore be used as a support tool by local and regional waste management authorities in the definition of future strategies to increase collection of recyclables of good quality and to achieve national and regional targets. © 2017 Elsevier Ltd. All rights reserved.

Keywords: Bring-bank Household waste Modelling Packaging waste Separate collection Waste management

1. Introduction Separate collection of packaging waste (glass, paper/cardboard and plastic/metal) is currently in place throughout Europe. This practice enables the recovery of recyclables of higher quality than those recovered from unsorted waste at mechanical and biological treatment facilities (MBT). The quality is important in the perspective of further re-integration of the recovered materials in

* Corresponding author. Research Centre for Natural Resources, Environment and Society (CERNAS), College of Agriculture (ESAC), Polytechnic Institute of Coimbra, Bencanta, 3045-601, Coimbra, Portugal. E-mail address: [email protected] (C. Dias-Ferreira).

production cycles as secondary raw materials. Average separate collection performance of glass, paper/cardboard, plastic/metal and bio-waste from residual waste/mixed municipal waste in the capital cities of the 28 European Union member states was 18.6% in 2015, but performances were quite heterogeneous across the 28 cities (Fig. 1) (European Commission, 2015). A coastal region in the centre of Portugal where bring-bank separate collection schemes are being used and monitored was selected for this study. The separate collection of packaging waste from households in this region in 2015 was 7.6%, standing well below the EU-28 capital cities average of 18.6% (Fig. 1). Consequently, in this region, separate collection still has a significant potential for improvement and needs to do so. Separate collection

http://dx.doi.org/10.1016/j.jenvman.2017.04.065 0301-4797/© 2017 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Oliveira, V., et al., Model for the separate collection of packaging waste in Portuguese low-performing recycling regions, Journal of Environmental Management (2017), http://dx.doi.org/10.1016/j.jenvman.2017.04.065

2

V. Oliveira et al. / Journal of Environmental Management xxx (2017) 1e12

Fig. 1. Separate collection performances in EU-28 capital cities (European Commission, 2015).

is one of the objectives of the Portuguese Strategic Plan for Municipal Solid Waste Management (MAOTE, 2014), approved in 2014 following European Union regulations (Directive 2008/98/EC, on Waste, currently under revision (European Parliament and of the Council, 2008); Directive 99/31/EC on Landfill Waste (European Council, 1999); and Directive 2004/12/EC on Packaging and Packaging Waste (European Parliament and of the Council, 2004)). In 2015, the European Commission proposed a new Circular Economy Package where new targets were established for 2030: i) reduction of landfilling to a maximum of 10% of municipal waste; ii) increase re-use and recycling of municipal waste to 65%, iii) increase recycling of packaging waste to 75% and iv) ban on landfilling of separately collected waste. Given the low separate collection performance of coastal region in the centre of Portugal, the pressures to attain National and European targets related to waste recycling are even more critical, justifying furthermore the context for the present research. Previous studies on separate waste collection have addressed the environmental, economic and social costs and benefits in the ~o et al., 2014; recycling system (Da Cruz et al., 2014, 2012; Ferra Ferreira et al., 2014a), the identification and analysis of engineering indexes for municipal solid waste (MSW) management systems (Gamberini et al., 2013), the economies of density and size in municipal solid waste recycling (Carvalho and Marques, 2014), the economic viability of packaging recycling (Marques et al., 2014) and the viability of implementing the separate collection of bio-waste in restaurant and canteens (Rodrigues et al., 2015). Technical aspects related to the characterization of waste collection systems and operation performance of separate collection were also analysed by Rodrigues et al. (2016) in the Greater Lisbon area, in Portugal, by Gallardo et al. (2012, 2010), in Spain. In addition, a robust non-parametric method based on conditional order-m efficiency was used by Guerrini et al. (2017) to assess the performance of drivers in MSW services in 40 municipalities in Verona province, Italy. The Waste and Resources Action Programme (WRAP) evaluated the performance of kerbside dry recycling in the UK during 2008/ 09 (Wrap, 2010). In this work, a model was put forward in which three factors accounted for 42% of the variation in recycling yields, namely the levels of deprivation (higher levels of deprivation leading to lower performance), the range of materials targeted (more materials leading to higher collection yields) and frequency

of refuse collections (fortnightly refuse collection leading to higher performance in dry recycling collection in comparison to weekly refuse collections) (Wrap, 2010). The contribution of kerbside collection to the increase of separation rates of paper and plastic has more recently been assessed also in some municipalities of the Czech Republic (Struk, 2017), together with other factors (higher density of drop-off sites and existence of incentive programs). In Portugal, the main collection system for separate recyclable is the bring-bank system, as opposed to the kerbside collection in use in other countries. The difference between them is that while in kerbside collection (also called “door-to-door”) each household has its own waste container that is placed outside the household, on the kerbside, for collection, in the bring-bank system the waste is placed in larger collective containers at the road side (Fig. 2). Since no study was found evaluating the factors influencing separate collection for the case of bring-bank collection system, the present research is a contribution to fill this gap. The objectives are: i) to identify and quantify the relevant factors affecting the success of separate collection services; and ii) develop a model specific for simulating the performance of the bring-bank system. The remainder of this work is organised as follows: Section 2 introduces the study area; section 3 presents the methodological approach used; Section 4 presents and discusses the results; and Section 5 lays out the conclusions and relevant implications of this work. 2. Study area The study area (Fig. 3) is located in the coast of the “Centro” Region of Portugal, bordered by Oporto in the North and by Lisbon in the South. It comprises 42 municipalities, a population of 1,245,241 inhabitants (12% of the total Portuguese population) and covers an area of 8848 km2 (9.6% of national territory). This region has economic and socio-cultural similarities with various areas away from the main urban centers in the Mediterranean countries (e.g., Spain, Italy, Greece). The main flows for municipal solid waste in the study area are presented in Fig. 4, showing that almost 500,000 tonnes were discarded in 2015 (data compiled from ERSUC (2016) and VALORLIS (2016)), representing roughly 10.4% of all municipal solid waste in Portugal (APA, 2016). Separate collection of packaging waste in this region is carried out by two multimunicipal waste management companies, “ERSUC, S.A.” (hereafter referred as ERSUC) and “VALORLIS, S.A.” (hereafter referred as VALORLIS). The main separate collection process for household packaging waste in the study area is bring-banks (95.7%), so called “ecopoints” depicted in Fig. 2a). Civic amenity recycling sites are enclosed sites or buildings where citizens can drop-off different recyclable fraction, such as packaging waste, wood, household hazardous waste, electric and electronic waste, garden waste and others. These civic amenity recycling sites are responsible for the collection of the remaining 4.3% of separately collected packaging waste in the study area (ERSUC, 2016; VALORLIS, 2016). The bring-bank usually comprises 3 units: i) a blue container for paper/cardboard packaging; ii) a yellow container for plastic/metal packaging; and iii) a green container for glass packaging. Collected material is delivered at three integrated centers for treatment and recovery of municipal solid waste. These centers are managed by the multimunicipal waste management companies of the study area, ERSUC and VALORLIS. Packaging materials sorted at these integrated centers are then sent for recycling through SPV e “Sociedade Ponto Verde”, Portugal's nonprofit organisation responsible for recycling packaging waste established under the framework of extended producer's responsibility. The SPV finances the collection and separation of

Please cite this article in press as: Oliveira, V., et al., Model for the separate collection of packaging waste in Portuguese low-performing recycling regions, Journal of Environmental Management (2017), http://dx.doi.org/10.1016/j.jenvman.2017.04.065

V. Oliveira et al. / Journal of Environmental Management xxx (2017) 1e12

3

Fig. 2. Collection of dry recyclables using: (a) Bring-banks; (b) kerbside collection.

Fig. 3. Geographic settings of the study area located in the coast of the “Centro” region in Portugal and comprising the administrative regions of Aveiro, Coimbra and Leiria.

packaging waste discarded by households by giving a financial retribution to collected packaging waste and guarantees the recycling of such waste by reselling it to recycling companies. In 2015, the separate collection rate for packaging waste (percentage of municipal packaging waste separately collected, compared to total municipal waste) was 7.6%, corresponding to 30.7 kg/inh/yr (kg per inhabitant per year) in ERSUC's service area and 7.7% (29.2 kg/inh/yr) in VALORLIS0 service area, being very close to the national average of 7.9%. The 2020 target (MAOTE, 2014) for separate collection is 46 kg/inh/yr for ERSUC and 42 kg/inh/yr for VALORLIS. These targets were established taking into account the national strategy for the municipal waste sector and considering the specific characteristics of ERSUC and VALORLIS, such as amount of MSW managed, MSW composition, infra-structures, etc. The municipalities in the study area are therefore lagging behind, and the pressures to attain targets related to separate collection are even more critical in this region. Nonetheless this limitation, between 2010 and 2015, tonnages of separately collected packaging waste decreased in the study area, instead of increasing, both in absolute value as well as in percentage of total waste (Fig. 5). Therefore, separate collection growth rate for packaging waste was negative in this period (9.2% for ERSUC and 19.1% for VALORLIS). Separate collection yields for each collected material followed the overall pattern, and the growth rate was mostly negative during

the period 2010e2015, except for plastic/metal packaging by ERSUC, where an increase of 31.5% took place (Table 1). In this period, the growth rates for the collection of glass packaging waste were 13.6% for ERSUC and 12.2% for VALORLIS, and for paper/ cardboard packaging were 23.1% for ERSUC and 29.9% for VALORLIS. Considering the differences on the specific weight of the 3 types of wastes collected separately, in particular the glass with 2500 kg/m3 being significantly heavier, glass packaging (based on weight) is almost 50% of all packaging waste collected in bringbanks in the study area. 3. Methodology The approach used to assess the factors influencing the performance of separate collection of packaging materials was: i) identification of indicators deemed likely to affect separate collection yields; ii) determination of the value of each indicator for all the municipalities in the study area; iii) assess the potential relation between separate collection yields and the indicators using regression analysis. The regression analysis also to develop a model for predicting the separate waste collection yields including a combination of statistically significant indicators, from which the relative importance of each of the indicators included could be evaluated.

Please cite this article in press as: Oliveira, V., et al., Model for the separate collection of packaging waste in Portuguese low-performing recycling regions, Journal of Environmental Management (2017), http://dx.doi.org/10.1016/j.jenvman.2017.04.065

4

V. Oliveira et al. / Journal of Environmental Management xxx (2017) 1e12

Fig. 4. Waste management and main waste flow (in tonnes) of municipal solid waste in the coast of “Centro” region of Portugal in 2015 (administrative regions of Aveiro, Coimbra and Leiria).

Fig. 5. Tonnages of municipal waste and of separately collected packaging waste (103 tons) between 2010 and 2015 at the area of influence of (a) ERSUC; (b) VALORLIS.

Table 1 Amount of glass, plastic/metal and paper/cardboard (kg/inh/yr) separately collected by ERSUC and VALORLIS between the years 2010 and 2015.

ERSUC

VALORLIS

Glass Plastic/metal Paper/cardboard Total Glass Plastic/metal Paper/cardboard Total

2010

2011

2012

2013

2014

2015

17.6 5.4 10.8 33.8 14.7 7.7 13.7 36.1

17.5 5.7 11.2 34.4 14.4 7.8 12.4 34.6

15.8 5.5 9.5 30.8 13.1 7.4 10.5 31.0

15.3 5.2 8.3 28.8 12.8 7.0 9.8 29.6

14.8 7.4 8.2 30.4 12.3 6.9 9.6 28.8

15.2 7.1 8.3 30.7 12.9 6.7 9.6 29.2

3.1. Selection of relevant indicators Separate collection yields are linked to many different characteristics of the inhabitants, households, collection service, municipality, structure of housing, and so on (Wrap, 2010). For the purpose of this work we have considered in the initial stage a wide range of

descriptors which fell into 2 groups: socio-economic/demographic and waste collection service (Tables 2 and 3). This does not preclude the existence of other factors that might also be relevant and which were not included in this investigation. Of these 20 possible indicators, five were promptly discarded either due to lack of available data for their calculation at the required desegregation level

Please cite this article in press as: Oliveira, V., et al., Model for the separate collection of packaging waste in Portuguese low-performing recycling regions, Journal of Environmental Management (2017), http://dx.doi.org/10.1016/j.jenvman.2017.04.065

V. Oliveira et al. / Journal of Environmental Management xxx (2017) 1e12

5

Table 2 Socio-economic/demographic indicators. Socio-economic/demographic

Rationale

Population

Higher number of inhabitants are associated with urban areas, and these are usually related to higher tonnages of waste and of packaging waste. In municipalities with large territorial surface the specific effort associated with waste collection is comparatively larger High population density is usually associated with urbanization. However, population density might not be a good descriptor for municipalities with one big, highly-populated city and a vast surrounding area of low population density: the decreased value for population density obtained in this case (due to large lowpopulated areas) might not reflect the actual urbanization level of the vast majority of the population. So this indicator was dropped and Degree of urbanization was used instead. Expresses the percentage of the population living in urban areas. The definition of urban area encompasses two parameters: population density and the presence of agglomerates above a certain number of inhabitants. This is believed to better reflect the urbanization level than the indicator Population density. The higher the purchase power the more affordable goods become and the higher the consumption. Higher consumption leads to more packaging waste being generated. Refers to the purchase power compared to the national average. It is similar to Purchase power per capita, but it is a relative value instead of absolute. Deprived areas are associated to lower recycling performance, often because deprived households are less able to prioritise participating in recycling activities (2010). The implementation of separate collection in Portugal started less than 20 years ago. Population over 65 years old had relatively less contact during their lifetime with education, training and awareness-raising campaigns related to separate collection. Therefore, waste disposal practices that are no longer the ones recommended in present days might be in place, just because “old habits die hard”. In addition, in case of reduced mobility it is more difficult to travel/walk to the bring-bank to dispose of packaging waste than to the unsorted waste container (which is closer) This indicator reflects the level of education of the population. In principle higher educated people are more aware of environmental concerns. Nowadays separate recycling is already taught in primary school (first cycle), for children 6e10 years old. This is reinforced in the following years at 2nd and 3rd cycles (10e15 years old). So as the number of school years attended increases so does the contact with awareness and environmental education campaigns. This indicator reflects the percentage of the adult population which have not been involved in educational campaigns carried out at school, and are therefore less aware of the importance of separate recycling. This indicator might be co-related to Population over 65 years old, because compulsory education level is currently the 9th year.

Area (km2) Population density (inhabitants/km2)

Degree of urbanization (%)

Purchase power per capita (%) Purchase power index Deprivation index Population over 65 years old (%)

Number of school years attended

Population over 15 years old without education or with only the first cycle of education (%)

(municipality) or else because other more suitable indicators could be used instead (rationale presented in Tables 2 and 3). 3.2. Calculation of indicators The indicators were determined with information from 3 main sources: i) waste management companies annual reports; ii) the Portuguese water, sanitation and waste regulatory entity database; and iii) national statistical data. Collection yields and performances are expressed in kg of waste collected per inhabitant per year (kg/ inh/yr) and can refer to different types of waste: total municipal solid waste, unsorted/residual waste; packaging waste, etc. Tonnages of unsorted waste, separately collected packaging and the number of containers for recyclables for each of the 42 municipalities in the study area were compiled from the annual reports published by the two waste management companies operating in this area, ERSUC and VALORLIS. The data refers to 2015 (most recent year for which there is available information) (ERSUC, 2016; VALORLIS, 2016). The term “packaging waste” in this work refers to glass, paper/cardboard and plastic/metal packaging wastes that are source-segregated at the household level, deposited in bring-banks or in civic amenity drop-off sites, and consequently separately collected. In addition to packaging waste, other materials might also be separately collected at the municipalities, such as electrical appliances, batteries, textiles, large furniture items, etc. So packaging waste does not equal overall separate collection. To estimate the relative accessibility to bring-banks, which compares waste deposition sites for bring-banks and for unsorted containers, it was assumed that at each bring-bank site there are 3 containers for recyclables and that at each unsorted waste site there is one container for unsorted waste. Even though this is not always the case, this assumption verifies most of times.

The Portuguese entity regulating water, sanitation and waste,
ERSAR (Entidade Reguladora dos Serviços de Aguas e Resíduos), maintains a database compiles data reported by the municipalities every year. Most information related to the waste collection system was retrieved directly from the ERSAR database, for the years 2014 and 2015 (ERSAR, 2015), namely the parameters accessibility to separate collection services, number of unsorted containers and number of civic amenity drop-off sites. Socio and demographic estimates were based on the national census for 2011 (Statistics Portugal, 2011a). This includes population, population over 65 years old and education-related parameters (population over 15 years old without education or with only the first cycle of education and number of school years attended). Population estimate was further used to calculate yields per capita and inhabitants per bring-bank. The area indicator was collected from the national statistical database for 2015 (Statistics Portugal, 2016) and was used to estimate the bring-banks per surface area and civic amenity drop-off sites per surface area. The economic indicators of purchase power per capita and purchase power index were based on a purchase power municipality report for 2013 (Statistics Portugal, 2013a). The degree of urbanization was based on the classification attributed to the electoral wards/neighbourhoods in each municipality. Each electoral wards in Portugal is currently classified as either “rural”, “averagely urban”, or “mostly urban” (Statistics Portugal, 2014a). The calculation of this indicator considered the percentage of inhabitants in neighbourhoods classified as “averagely urban” and “mostly urban”, compared to the total population of the municipality (Table 4), with the inhabitants at each electoral ward compiled from Statistics Portugal (2014b). Deprivation can take numerous forms (income, employment, education and training, health and disability, housing, services, crime, living environment, and so on (Smith et al., 2015). In the

Please cite this article in press as: Oliveira, V., et al., Model for the separate collection of packaging waste in Portuguese low-performing recycling regions, Journal of Environmental Management (2017), http://dx.doi.org/10.1016/j.jenvman.2017.04.065

6

V. Oliveira et al. / Journal of Environmental Management xxx (2017) 1e12

Table 3 Waste collection service indicators. Waste collection

Rationale

Bring-banks per surface area (number/km2)

In the case where there is no bring-bank available close by then there is no place where packaging waste can be deposited, so the practice of separating packaging waste is not encouraged. Accessibility to separate collection services (%) This indicator measures the percentage of the population within 200 m of a bring-bank for packaging waste or served with door to door (/kerbside) collection for packaging waste. People who wish to separate waste will only do so if afterwards there is a sufficiently close-by place where they can place this waste. This indicator is expected to be related with the indicator bring-banks per surface area. Relative accessibility to bring-banks This indicator measures the ratio between bring-banks and unsorted waste deposition sites. Ideally, places to deposit unsorted waste should also include bring-banks, so that citizens wishing to separate waste would not need to walk extra. Ratio between installed capacity for unsorted When there is a limited volume for unsorted waste then more attention is given to reducing as much as possible that and for separate collection fraction of waste, by either adopting prevention measures on waste generation or by transferring part of the waste to other fluxes (e.g. separate collection). However, information to calculate this indicator is not available, so it will not be considered further in this study. Frequency of waste collection Less collection frequency for unsorted waste and more frequent collection for recyclables fosters separate collection because, together with installed capacity, it limits the physical volume available for each fraction of waste. However, this indicator is more relevant in case of door-to-door/kerbside collection. In the present situation the collection frequency for unsorted waste is the same throughout the study area (daily, except on Sunday) and separate collection is irregular, so this indicator was not used. Civic amenity drop-off sites per surface area Related to access to civic amenities sites for dropping off recyclable waste. Civic amenity drop-off sites are responsible for 2 (number/km ) 4.7% of separately collected waste in the study area. The tonnage of packaging waste deposited in civic amenity drop-off sites is of relatively low magnitude when compared to bring-banks, so this indicator is not expected to be significant Inhabitants per bring-bank Might influence separate collection rates in case the collection is not frequent enough and bring-banks get full. Facing this situation, most people will likely place the initially source-segregated packaging waste into the unsorted waste container, instead of in the bring-bank. This indicator is expected to be related with Number of complaints related to separate collection Number of complains related to separate Complains are often related to the container being full or otherwise non-operational. When facing this situation most collection people will likely place the packaging waste initially source-segregated in the unsorted waste container, instead of in the bring-bank. However, only the overall number of complaints are available on the database, and there is not specific information on separate collection-related complaints. This indicator is not to be considered any further in this work Number of packaging materials targeted for Usually, the more materials targeted, the higher the yields. Throughout the study area bring-banks systematically collect 3 bring-bank collection fluxes of materials: paper/cardboard, plastic/metal/composites and glass. In some cases, batteries are also collected but the magnitude in tonnages is much lower than for the other materials. Given the constant number of materials targeted throughout the study area this indicator will not vary among municipalities, and as so, it is not expected to influence separate collection rates. Therefore it will not be used further in this work. Waste management company Waste management companies responsible for separate waste collection might have intrinsic characteristics that affect performance in their specific area (for instance educational campaigns, frequency of collection services, internal policies related to collection services, cleaning of bring-banks, and so on).

Table 4 Methods of calculation of indicators. Notes: UR ¼ unemployment rate (in %), MND ¼ rate of incidence of mandatory notification diseases (per 1000 inhabitants), CR ¼ crime rate (in %). Indicator

Calculation formula

Degree of urbanization

population of “averagely urban”þpopulation of “mostly urban” total population ðUR5ÞþðMND2ÞþðCR1Þ 8 ðnumber of recyclable containers=3Þ number ofunsorted waste containers number of civic amenity sites area population number of bring banks number of bring banks area

Deprivation index Relative accessibility to bring-banks Civic amenity sites per surface area Inhabitants per bring-bank Bring-banks per surface area

present work, lack of employment, lack of health and exposure to crime were considered to build the indicator. These parameters were selected because of their availability at the level of disaggregation required (municipality). The deprivation index was constructed considering the unemployment rate (UR, in %) (Statistics Portugal, 2011b), the rate of incidence of mandatory notification diseases (MND, per 1000 inhabitants) (Statistics Portugal, 2007) and the crime rate (CR, in %) (Statistics Portugal, 2013b). The calculation formulas used to estimate the degree of urbanization, deprivation index, relative accessibility to bring-banks, civic amenity sites per surface area, inhabitants per bring-bank and bringbanks per surface area are presented in Table 4. 3.3. Statistical analysis on the indicators influencing separate collection yields Regression analysis is a statistical tool for developing models

that best fit a set of data observation. In the process, a relation between the independent variables and the dependent variable is established. In the present work it was used to evaluate the relation between the different socio-economic/demographic and waste collection service indicators (independent variables) and separate collection yields (dependent variable). Based on the rationale presented in Tables 2 and 3 above, the following 14 independent variables were evaluated: Socio-economic/demographic indicators: i) Degree of urbanization (%); ii) Purchase power per capita (%); iii) Purchase power index; iv) Deprivation index; v) Population over 65 years old (%); vi) Number of school years attended; vii) Population over 15 years old without education or with only the first cycle of education (%); viii) Population; and ix) Area (km2); Waste collection service: x) Bring-banks per surface area (number/km2); xi) Accessibility to separate collection services (%); xii) Relative accessibility to bring-banks; xiii) Civic amenity drop-off sites per surface area (number/km2); and xiv)

Please cite this article in press as: Oliveira, V., et al., Model for the separate collection of packaging waste in Portuguese low-performing recycling regions, Journal of Environmental Management (2017), http://dx.doi.org/10.1016/j.jenvman.2017.04.065

V. Oliveira et al. / Journal of Environmental Management xxx (2017) 1e12

inhabitants per bring-bank. The regression analysis carried out entailed: i) simple linear regression; ii) multiple linear regression; and iii) multiple nonlinear regression. Prior to the regression, descriptive statistics were computed and the existence of outliers in data was analysed using the Modified Thompson t test (t ¼ 1.926; the cases considered outliers arejz-scorej t), considering a significance level of 0.05. Also, the underlying regression models assumptions were tested: i) the existence of outliers and influential cases in the linear regression analyses was evaluated based on the Cook's distance (D) criterion of D > 4/n, where n is the number of cases in the sample; ii) the relation between the independent variables was evaluated diagnosing multicollinearity based on variance inflation factor (VIF) criterion of VIF > 10; and iii) normality and variances equality were tested using the Shapiro-Wilk and Levene's tests, respectively. In the multiple linear and non-linear regression the outliers were removed from regression analysis, but not for the simple linear regression since they were developed as a basis for comparison only. Simple linear regression models were built using each of the 14 indicators as independent variables individually to assess their explanatory power. This was done to build a base for assessing the increase on explanatory power due to the combination of various indicators and to check if the variable with more explanatory power individually also were the ones included in the multiple linear regression models. In the multiple linear regression the independent variable to include in the model were evaluated testing the best subsets and using the Akaike Information Criterion (AICC), based on the likelihood of the training set given the model and adjusted to penalize overly complex models. Other options were also tested for selecting the independent variables (e.g., forward stepwise) and compare the models (e.g., adjusted coefficient of determination - R2 adjusted, F statistic), but the results were either equivalent or worse. The comparison of the separate collection yields achieved by ERSUC and VALORLIS was also done for the year 2015, using an independent samples t-test to analyse the values of the municipalities covered by each company. Microsoft Office EXCEL 2013™ and IBM SPSS Statistics™ version 24 software were used for the statistical analysis of the data. 4. Results and discussion 4.1. Comparison of separate collection yields between municipal waste management companies The separate collection of packaging waste in each municipality of the study area in 2015 is shown in Fig. 6, with values ranging from 43.2 kg/inh/yr (municipality of “Cantanhede”, pop. 36,234) down to 16.5 kg/inh/yr (municipality of “Pampilhosa da Serra”, pop. 4257). The first factor investigated was the existence of differences between the administrative regions of “Aveiro”, “Coimbra” and “Leiria”. These might be due to intrinsic regional characteristics or to different policies of waste management companies operating in the study area, such as educational campaigns, frequency of collection services, cleaning of bring-banks, and so on. ERSUC operates in two administrative regions (“Coimbra” and “Aveiro”) and VALORLIS in one (“Leiria”). A comparison of the means of the municipalities served by ERSUC and VALORLIS through an independent samples t-test concluded that there is no statistically significant difference (t-test ¼ 0.101, p-value ¼ 0.922) in the separate collection yields between the two waste management companies. Pre-requisites for the statistical analysis were verified prior to the statistical test, ensuring its validity (the

7

Fig. 6. Separate collection yields (kg/inh/yr) in each municipality of the study area.

Levene's test indicates equality of variances (F-test ¼ 1.232, pvalue ¼ 0.274) and the Shapiro-Wilk test indicates normality of both samples (W ¼ 0.966, p-value ¼ 0.331 e ERSUC; W ¼ 0.915, pvalue ¼ 0.470 e VALORLIS). Since there is no statistically significant difference between the means of separate collection yield in the municipalities served by ERSUC and VALORLIS, the waste management company responsible for the separate collection was not considered further in this work as an explanatory factor for separate collection yields and the data from both companies were analysed together. 4.2. Influential factors in separate collection yields Prior to the regression, the municipalities of “Coimbra”, “Leiria”, ~o da Madeira” were identified as “Pampilhosa da Serra” and “S~ ao Joa main outliers of data. The municipalities of “Coimbra” and “Leiria” have the highest number of inhabitants (“Coimbra”: 143,396 inhabitants and “Leiria”: 126,879 inhabitants). “Coimbra” is also the municipality with the highest purchase power index, which influences positively the number of school years attended. “Pampilhosa da Serra” is a rural municipality with a small population (4481 inhabitants), the smallest education level (4.4 years) and the highest aging population (42%). With only 7.9 km2, the municipality of “S~ ao Jo~ ao da Madeira” is the smallest in area in Portugal. These 4 municipalities are very distinct from the average of the 42 municipalities, explaining the reason why they are outliers. However, these municipalities were not excluded because the data are real and do not result from any errors. The relations between separate collection yields and each of the 14 independent variables, assessed individually by means of simple linear regression, are shown in Fig. 7. The high dispersion of data and consequently the low R2 (<40%) achieved in every case indicates that, individually, none of the variables is enough to explain the variability of separate collection yields in the different municipalities, even though the analysis of variance (Table 5) shows that most independent variables are statistically significant (at a significance level of 0.050). That means that even though these variables were found to statistically affect yields, their explanatory power is very weak. Therefore, a combination of variables was tested next, using multiple linear regression. The multiple linear regression analysis was carried out testing in an informed way all possible combinations of the independent variables. The model with the highest explanatory power included 5 variables: inhabitants per bring-bank, relative accessibility to bring-

Please cite this article in press as: Oliveira, V., et al., Model for the separate collection of packaging waste in Portuguese low-performing recycling regions, Journal of Environmental Management (2017), http://dx.doi.org/10.1016/j.jenvman.2017.04.065

8

V. Oliveira et al. / Journal of Environmental Management xxx (2017) 1e12

Fig. 7. Scatter plots of the separate collection yields with the various variables characterizing socio-economic/demographic characteristics of the municipalities and the collection systems.

Please cite this article in press as: Oliveira, V., et al., Model for the separate collection of packaging waste in Portuguese low-performing recycling regions, Journal of Environmental Management (2017), http://dx.doi.org/10.1016/j.jenvman.2017.04.065

V. Oliveira et al. / Journal of Environmental Management xxx (2017) 1e12

banks, area, number of schools years attended and degree of urbanization. Since according to the Cook distance (D) criterion 3 munici
m” D ¼ 0.126 and palities (“Sever do Vouga” D ¼ 0.118), “Oure “Pombal” D ¼ 0.139) were identified as outliers or influential cases that could influence the model precision, data from these municipalities were not used for model development. The best multiple linear regression model achieved an adjusted R2 of 0.664 (R2 of 0.709). This represented a significant improvement on the relation between independent and dependent variables than the previous simple linear regression models. The ANOVA results are shown in Table 6, confirming that the model is statistically significant (p-value < 0.050). However, since there is no reason why the relation should be linear, a multiple non-linear regression model was additionally built using the same independent variables. The non-linear multiple regression model with constant achieved an adjusted R2 of 0.692, which is higher than that of the linear multiple regression model (adjusted R2 of 0.664). The corresponding ANOVA results are also shown in Table 6, confirming that the model is statistically significant (p-value < 0.050). In addition to the improvement on explanatory power, the multiple non-linear regression model allowed the removal of the constant from without decreasing the R2, contrarily to multiple linear regression where the removal of the constant decreased the R2 from 0.709 to 0.695. The ANOVA table of the multiple non-linear regression model without constant is presented in Table 6.

4.3. Model for the estimation of separate collection yields for packaging waste Based on the several statistical tests described in the previous section (section 4.3), the best-fit model achieved is the multiple non-linear regression model with no constant 5 independent variables: area, number of school years attended, degree of urbanization, relative accessibility to bring-banks and inhabitants per bring-bank. The equation for this model is:

separate collection yieldðkg=inh=yrÞ ¼ 2:55E4 *area1:635 þ 3:954*number of school years attended0:994 þ 4:911E5 *degree of urbanization2:614 þ 163:813*relative accessibility to bring  banks1:859  5:759E9 *inhabi tan ts per bring  bank3:786 in which the independent variables are described in Tables 2e4 The option for the model with no constant in detriment of the model with constant was based on the fact that if all variables tend to zero so should the separate collection yield. The model presented has an adjusted R2 of 0.693 (R2 of 0.732), which is quite high for a model trying to predict human behaviour related to waste and separate collection. The adjusted R2 is also higher than that for the kerbside separate collection model developed for the U.K. settings (adjusted R2 of 0.501), which was considered as having an “extremely impressive degree of explanatory power” by the authors (Wrap, 2010). A study carried out by Ferreira et al. (2014b) which forecast the number of collections required per year for a paper/cardboard required bring-bang using 10 explanatory factors achieved an R2 of 0.636, which is also lower than the one obtained in this investigation. To some degree there is always some level of relation between the supposedly independent variables, particularly when

9

discussing waste collection systems. For instance, the degree of urbanization might be correlated with area or with relative accessibility to bring-banks (the later since in more urbanised areas the relative accessibility to bring-banks is expected to increase). The inclusion of independent variables strongly correlated between each other (multicollinearity) in a regression model constitutes a problem, because it artificially increases the R square. The diagnostic of multicollinearity (based on criterion of VIF > 10) showed that multicollinearity is not a problem for the model presented in this work (Table 7). This means that the independent variables used in the model are not over-correlated. Therefore, R2 obtained is not artificially high, and the model presented in this work is statistically robust. All the variables identified as statistically relevant in the regression model increase separate collection yields, except inhabitants per bring-bank, which decreases the yields. These behaviour is expected and reasonable, as discussed previously in Tables 2 and 3. The variable with the highest relative weight (that is, the one which has got the highest influence in separate collection yields) is inhabitants per bring-bank. Following, in descending order are the number of school years attended, relative accessibility to bring-banks and degree of urbanization (with similar explanatory power), and finally, the area, which is the variable with the smallest contribution to separate collection yields (Fig. 8). It is noted the simple linear regression models with area and inhabitants per bring-bank showed a lower explanatory power for separate collection yields and were not statistically significant. However in the multiple regression models these variables were statistically significant and consequently included in the model. Thus, statistically, the variables of area, number of school years attended, degree of urbanization, relative accessibility to bring-banks and inhabitants per bring-bank partially explain the separate collection yields, and these 5 variables all together explain 73% of the variation found in the separate collection yields in the municipalities. It is important to highlight that the model presented in this work is applicable to estimate total packaging waste from households for the type of waste collection system that is the subject of this study, namely bring-banks for separate collection and collective kerbside containers for unsorted waste. The actual separate collection yields against the yields which are predicted by the regression model for each municipality of the study area is shown in Fig. 9. The straight line shows the predicted performance values, and each point the actual value. The distance between the points and the expected value highlights the difference between actual performances of the municipalities in separate collection against predicted performances. Differences between predicted performances and real performances can be due to one (or more) relevant factors not being included in the model. Differences can also be explained by the random degree of variation affecting people's behaviour in relation to separate collection, and these random variations are not easy to account for, systematically. Other explanations can be errors in data acquisition and the co-weighing of material separately collected from different municipalities. This is even more plausible considering that waste management companies cover several municipalities, and sometimes circuits are not independent, being therefore difficult to properly allocate tonnages. 5. Conclusions In this work, a powerful explanatory model describing separate collection of packaging waste using bring-banks is presented. This model is based on multiple, non-linear regression analysis, and

Please cite this article in press as: Oliveira, V., et al., Model for the separate collection of packaging waste in Portuguese low-performing recycling regions, Journal of Environmental Management (2017), http://dx.doi.org/10.1016/j.jenvman.2017.04.065

10

V. Oliveira et al. / Journal of Environmental Management xxx (2017) 1e12

Table 5 ANOVA results of the simple linear regression models for each of the variables characterizing the municipalities and separate collection systems (Values in bold indicate that data is not statistically significant). Variables Population

Population over 15 years old without education or with only the first cycle of education Population over 65 years old

Area

Purchase power index

Purchase power per capita

Number of school years attended

Degree of urbanization

Deprivation index

Accessibility to separate collection services

Civic amenity drop-off sites per surface area

Relative accessibility to bring-banks

Bring-banks per surface area

Inhabitants per bring-bank

Regression Residual Total Regression Residual Total Regression Residual Total Regression Residual Total Regression Residual Total Regression Residual Total Regression Residual Total Regression Residual Total Regression Residual Total Regression Residual Total Regression Residual Total Regression Residual Total Regression Residual Total Regression Residual Total

Sum of squares

df

Mean square

F

Sig.

R2

290.923 1330.141 1621.063 335.570 1285.493 1621.063 171.902 1449.161 1621.063 1.776 1619.287 1621.063 312.862 1308.201 1621.063 425.011 1196.053 1621.063 478.489 1142.574 1621.063 189.928 1431.135 1621.063 101.415 1519.649 1621.063 549.902 1071.161 1621.063 18.402 1602.661 1621.063 636.457 984.606 1621.063 123.722 1497.341 1621.063 7.281 1613.782 1621.063

1 40 41 1 40 41 1 40 41 1 40 41 1 40 41 1 40 41 1 40 41 1 40 41 1 40 41 1 40 41 1 40 41 1 40 41 1 40 41 1 40 41

290.923 33.254

8.749

0.005

0.179

335.570 32.137

10.442

0.002

0.207

171.902 36.229

4.745

0.035

0.106

1.776 40.482

0.044

0.835

0.001

312.862 32.705

9.566

0.004

0.193

425.011 29.901

14.214

0.001

0.262

478.489 28.564

16.751

0.000

0.295

189.928 35.778

5.308

0.026

0.117

101.415 37.991

2.670

0.110

0.063

549.902 26.779

20.535

5E-05

0.340

18.402 40.067

0.459

0.502

0.011

636.457 24.615

25.856

9E-06

0.392

123.723 37.434

3.305

0.077

0.076

7.281 40.345

0.180

0.673

0.004

Table 6 ANOVA results of multiple linear regression model with constant and multiple non-linear regression models with and without constant. Model Multiple linear regression

Multiple non-linear regression with constant

Multiple non-linear regression without constant

Regression Residuals Total Regression Residuals Total Regression Residuals Total

Sum of squares

df

Mean squares

F

Sig.

R2

1046.474 430.420 1476.893 1081.557 395.336 1476.893 30,272.003 395.378 30,667.381

5 33 38 5 33 38 5 34 39

209.295 13.043

16.046

0.000

0.709

216.311 11.980

18.056

0.000

0.732

6054.401 11.629

520.640

0.000

0.732

comparison of R2 against earlier models highlight the high degree of explanatory power of the model proposed. It is possible to identify the separate effects of each of the 5 factors found as most influential for separate collections yields for packaging waste, which, in descending order of influence are inhabitants per bringbank, number of school years attended, relative accessibility to bring-banks and degree of urbanization and area of the municipality. To achieve national targets, it is necessary to increase separate collection yields, and this means changing people's behaviour regarding separate collection. Of the 5 factors found in this work as most influent to separate collection yields, two are related to the

waste collection system: inhabitants per bring-bank (which currently varies across municipalities from 109 to 327 inhabitants per bring-bank), and relative accessibility to bring-banks (ranging from 0.02 to 0.26), while the remaining are socio-economic/ demographic factors. Since factors related to the waste collection service can be more easily changed in a short time (as opposed to the socio-economic/demographic factors), taking measures to decrease the number of inhabitants per bring-bank and increase the relative accessibility to bring-banks would allow achieving higher yields. In both cases the indicators could be achieved by increasing either the total or the relative number of bring-banks, compared to

Please cite this article in press as: Oliveira, V., et al., Model for the separate collection of packaging waste in Portuguese low-performing recycling regions, Journal of Environmental Management (2017), http://dx.doi.org/10.1016/j.jenvman.2017.04.065

V. Oliveira et al. / Journal of Environmental Management xxx (2017) 1e12

11

Table 7 Coefficients for separate collection yield non-linear multiple regression model. Model

1

Coefficients

Inhabitants per bring-bank Relative accessibility to bring-banks Area Number of school years attended Degree of urbanization

Unstandardized Coefficients

Standardized Coefficients

B

Std Error

Beta

5.759E9 163.813 0.000 3.954 4.911E5

0.000 38.646 0.000 0.244 0.000

-0.230 0.114 0.083 0.894 0.183

t

Sig.

Collinearity Statistics Tolerance

VIF

4.535 4.239 2.495 16.197 3.787

0.000 0.000 0.018 0.000 0.001

0.147 0.525 0.346 0.124 0.162

6.792 1.904 2.894 8.035 6.182

definition of the best and most effective strategies to increase separate collection yields for packaging waste. Acknowledgements ~o para a C. Dias-Ferreira gratefully acknowledges FCT e Fundaça ^ncia e para a Tecnologia for financial support (SFRH/BPD/ Cie 100717/2014). References Fig. 8. Relative weight of the variables considered in multiple non-linear regression model.

Fig. 9. Scatter plot of the actual separation collection yields against predicted yields for the municipalities in the study area in the year 2015.

unsorted waste. The model allows not only the identification of the most influential factors but provides also a quantitative assessment, meaning that theoretically it would be possible to estimate the number of bring-banks that need to be installed in the field to achieve the targets for separate collection of packaging waste. It should be noted that the model presented in this work was built using data from the study area and considering bring-banks collection system for packaging waste discarded by households. As such before its application to other regions (or different time spans) the model should be first calibrated with data from the new settings/conditions. Nevertheless this limitation, given the statistical robustness of the model and the high degree of explanatory power achieved, the model proposed here can be used by local authorities and waste management companies to help in the


rio Anual 2015. Departamento de Resíduos, APA, 2016. RES
IDUOS URBANOS - Relato ^ncia Portuguesa do Ambiente I.P., Amadora, Portugal, p. 41 (In Portuguese). Age Carvalho, Pedro, Marques, Rui Cunha, 2014. Economies of size and density in municipal solid waste recycling in Portugal. Waste Manag. 34 (1), 12e20. http:// dx.doi.org/10.1016/j.wasman.2013.10.004. ~es, Pedro, Marques, Rui Cunha, 2012. Economic cost da Cruz, Nuno Ferreira, Simo recovery in the recycling of packaging waste: the case of Portugal. J. Clean. Prod. 37, 8e18. http://dx.doi.org/10.1016/j.jclepro.2012.05.043. ~es, P., Marques, R.C., 2014. Packaging waste da Cruz, N.F., Ferreira, S., Cabral, M., Simo recycling in Europe: is the industry paying for it? Waste Manag. (New York, N.Y.) 34 (2), 298e308. http://dx.doi.org/10.1016/j.wasman.2013.10.035.

rio Anual dos Serviços de Aguas ERSAR, 2015. Relato e Resíduos em Portugal (2015): ~o do setor de a
guas e resíduos - relato
rio dados indiVolume 1-Caracterizaça cadores (In Portuguese).
lidos do Centro, S.A. Relato
rio e Contas 2015, ERSUC, 2016. ERSUC - Resíduos So p. 117 (In Portuguese). European Commission, 2015. Assessment of Separate Collection Schemes in the 28 Capitals of the EU (070201/ENV/2014/691401/SFRA/A2) - Final Report. European Council, 1999. Council Directive 1999/31/EC of 26 April 1999 on the Landfill of Waste. Official Journal of the European Communities L 182/1-19. European Parliament and of the Council, 2004. Directive 2004/12/EC of the European Parliament and of the Council of 11 February 2004 Amending Directive 94/62/EC on Packaging and Packaging Waste. Official Journal of the European Communities L47/26-31. European Parliament and of the Council, 2008. Directive 2008/98/EC of the European Parliament and the Council of 19 November 2008 on Waste and Repealing certain Directives (Text With EEA Relevance). Official Journal of the European Communities L 312/3-30. Ferr~ ao, Paulo, Ribeiro, Paulo, Rodrigues, Jo~ ao, Marques, Alexandra, Preto, Miguel, ^s, 2014. EnvironAmaral, Miguel, Domingos, Tiago, Lopes, Ana, Costa, e Ine mental, economic and social costs and benefits of a packaging waste management system: a Portuguese case study. Resour. Conserv. Recycl. 85, 67e78. http://dx.doi.org/10.1016/j.resconrec.2013.10.020. Ferreira, S., Cabral, M., da Cruz, N.F., Marques, R.C., 2014a. Economic and environmental impacts of the recycling system in Portugal. J. Clean. Prod. 79, 219e230. http://dx.doi.org/10.1016/j.jclepro.2014.05.026. Ferreira, J.A., Figueiredo, M.C., Oliveira, J.A., 2014b. Forecasting household packaging waste Generation : a case study. In: Lect. Notes Comput. Sci. 14th Int. Conf. Comput. Sci. Its Appl. 8581. Gallardo, A., Bovea, M.D., Colomer, F.J., Prades, M., Carlos, M., 2010. Comparison of different collection systems for sorted household waste in Spain. Waste Manag. (New York, N.Y.) 30 (12), 2430e2439. http://dx.doi.org/10.1016/ j.wasman.2010.05.026. Gallardo, A., Bovea, M.D., Colomer, F.J., Prades, M., 2012. Analysis of collection systems for sorted household waste in Spain. Waste Manag. (New York, N.Y.) 32 (9), 1623e1633. http://dx.doi.org/10.1016/j.wasman.2012.04.006. Gamberini, R., Del Buono, D., Lolli, F., Rimini, B., 2013. Municipal solid waste management: identification and analysis of engineering indexes representing demand and costs generated in virtuous Italian communities. Waste Manag. 33 (11), 2532e2540. http://dx.doi.org/10.1016/j.wasman.2013.06.003. Guerrini, Andrea, Carvalho, Pedro, Romano, Giulia, Cunha Marques, Rui, Leardini, Chiara, 2017. Assessing efficiency drivers in municipal solid waste collection services through a non-parametric method. J. Clean. Prod. 147,

Please cite this article in press as: Oliveira, V., et al., Model for the separate collection of packaging waste in Portuguese low-performing recycling regions, Journal of Environmental Management (2017), http://dx.doi.org/10.1016/j.jenvman.2017.04.065

12

V. Oliveira et al. / Journal of Environmental Management xxx (2017) 1e12

431e441. http://dx.doi.org/10.1016/j.jclepro.2017.01.079.
gico para os Resíduos Urbanos MAOTE, 2014. Portaria no187-A/2014: Plano Estrate
rio do Ambiente, Ordena(PERSU 2020), para Portugal Continental. Ministe
rio e Energia, Portugal, p. 87 (In Portuguese). mento do Territo ~ es, Pedro, Faria Ferreira, Sandra, Marques, Rui Cunha, da Cruz, Nuno Ferreira, Simo Pereira, Marta Cabral, De Jaeger, Simon, 2014. Economic viability of packaging waste recycling systems: a comparison between Belgium and Portugal. Resour. Conserv. Recycl. 85, 22e33. http://dx.doi.org/10.1016/j.resconrec.2013.12.015. Rodrigues, J., Oliveira, V., Lopes, P., Dias-Ferreira, C., 2015. Door-to-Door collection of food and kitchen waste in city centers under the framework of multimunicipal waste management systems in Portugal: the case study of Aveiro. Waste Biomass Valoriz. 6 (5), 647e656. http://dx.doi.org/10.1007/s12649-015-9366-3. Rodrigues, Susana, Martinho, Graça, Pires, Ana, 2016. Waste collection systems. Part B: benchmarking indicators. Benchmarking of the great Lisbon area, Portugal. J. Clean. Prod. 139, 230e241. http://dx.doi.org/10.1016/j.jclepro.2016.07.146. Smith, T., Noble, M., Noble, S., Wright, G., McLennan, D., Plunkett, E., 2015. The English Indices of Deprivation 2015-Technical Report. Department for Communities and Local Government, London, UK, p. 126. Statistics Portugal, 2007. Saúde [WWW Document]. https://ine.pt (Accessed 23 January 2017). Statistics Portugal, 2011a. Censos 2011 Resultados Definitivos - Portugal. In:

Portuguese. Portugal, Lisboa, p. 560. Statistics Portugal, 2011b. Mercado de Trabalho [WWW Document]. https://ine.pt/ (Accessed 23 January 2017). ~ es de vida e cidadania [WWW Document]. Statistics Portugal, 2013a. Condiço https://ine.pt (Accessed 23 March 2016). Statistics Portugal, 2013b. Crime e justiça criminal [WWW Document]. https://ine. pt/ (Accessed 23 January 2017).

rio Te
cnico, p. 20 (In Statistics Portugal, 2014a. Tipologia de Areas Urbanas - Relato Portuguese). ~o [WWW Document]. https://ine.pt (accessed 23 Statistics Portugal, 2014b. Populaça January 2017).
rio [WWW Document]. https://ine.pt/ (Accessed 23 Statistics Portugal, 2016. Territo January 2017). Struk, Michal, 2017. Distance and incentives matter: the separation of recyclable municipal waste. Resour. Conserv. Recycl. 122, 155e162. http://dx.doi.org/ 10.1016/j.resconrec.2017.01.023. ~o e Tratamento de Resíduos So
lidos, S.A. VALORLIS, 2016. Valorlis - Valorizaça
rio & Contas 2015, p. 126 (In Portuguese). Relato Wrap, 2010. Analysis of Kerbside Dry Recycling Performance in the UK 2008/09Summary Report, p. 60.

Please cite this article in press as: Oliveira, V., et al., Model for the separate collection of packaging waste in Portuguese low-performing recycling regions, Journal of Environmental Management (2017), http://dx.doi.org/10.1016/j.jenvman.2017.04.065