- Email: [email protected]
Identifying key process parameters for uncertainty propagation in environmental life cycle assessment for sewage sludge and food waste treatment
Journal of Cleaner Production 174 (2018) 966e976
Contents lists available at ScienceDirect
Journal of Cleaner Production journal homepage: www.elsevier.com/locate/jclepro
Identifying key process parameters for uncertainty propagation in environmental life cycle assessment for sewage sludge and food waste treatment Sam L.H. Chiu, Irene M.C. Lo* Department of Civil and Environmental Engineering, The Hong Kong University of Science and Technology, Hong Kong, China
a r t i c l e i n f o
a b s t r a c t
Article history: Received 28 June 2017 Received in revised form 30 September 2017 Accepted 15 October 2017 Available online 16 October 2017
Life cycle assessment (LCA) results are often subject to uncertainty, which may lead to erroneous conclusions. This paper aims to tackle the parameter uncertainty involved in LCA. Hong Kong is selected as the reference city for the treatment of sewage sludge and food waste. A selection approach with sensitivity analysis is proposed to identify the key process parameters and Monte Carlo simulation is further conducted to propagate the uncertainty of the key process parameters identified. The results show that climate change is the major impact category among various impact categories that are generally considered in waste LCA studies. Scenarios 5 and 6, which consider anaerobic co-digestion (coAD) treatment, achieve the best performance in regard to the climate change impact. Scenario 6 which includes a combined cycle gas turbine system for biogas utilization has 6.75 104 kg avoided CO2e emissions and it is equal to 44% more avoided emissions compared to scenario 5 which applies a combined heat and power system. For the key process parameters identified, it is found that the electricity generation efficiencies in different waste treatment facilities, such as the incineration plant and the anaerobic digestion plant, have the greatest sensitivity to the result. Uncertainty propagation is then conducted to obtain the probability distribution functions in environmental performance of different scenarios. Scenario 6 has a 95% probability of achieving at least 5.32 104 kg avoided CO2e emissions, while the probability of scenario 5, which is the second best scenario, achieving the same avoid emissions is below 5%. It indicates a significant advantage of using combined cycle gas turbine over combined heat and power unit for biogas utilization in Hong Kong. The methodologies and results of this study provide comprehensive material that can be adapted for other areas planning sustainable sewage sludge and food waste treatment, as well as in tackling parameter uncertainty in general LCA studies. © 2017 Elsevier Ltd. All rights reserved.
Keywords: Anaerobic co-digestion Combined cycle gas turbine Combined heat and power Monte Carlo Organic waste Sensitivity analysis
1. Introduction Organic waste management has long been a major challenge in urban areas. Improper management of organic waste could create negative environmental impacts such as generating unpleasant odors, spreading disease, and contributing to global warming (Singh et al., 2011). Since the early 21st century, the disposal of organic waste in landfills has been banned by legislation in some European countries, such as in Switzerland and Sweden (EEA, 2017). This legislation is aimed at reducing the negative environmental consequences from landfilling and in promoting resource
* Corresponding author. E-mail address: [email protected] (I.M.C. Lo). https://doi.org/10.1016/j.jclepro.2017.10.164 0959-6526/© 2017 Elsevier Ltd. All rights reserved.
recovery by using alternative treatment methods. For instance, the European Environment Agency (EEA, 2013) estimated that the energy produced from organic waste could reach more than 50% of the total renewable energy generated in Europe in 2020. Regarding the composition of organic waste, sewage sludge and food waste represent the greatest portion in urban areas (Righi et al., 2013). For example, in Hong Kong, such wastes accounted for 89% of the total organic waste generated in 2015 (HKEPD, 2017), so if these wastes can be properly treated, the environmental impacts created would be minimized and useful resources would also be recovered at the same time. Hence, the identification of a cleaner and sustainable treatment method for sewage sludge and food waste is vital. In order to identify a waste treatment strategy for environmental sustainability, life cycle assessment (LCA) is widely used as it can quantify the environmental impacts of different treatment
S.L.H. Chiu, I.M.C. Lo / Journal of Cleaner Production 174 (2018) 966e976
scenarios in a scientific manner (Humbert et al., 2009). However, LCA results are often subject to uncertainty, and may lead to incorrect conclusions for the decision makers (Huijbregts et al., 2003). The International Organization for Standardization (2006) recommends that the uncertainty should be considered in order to improve the reliability of the LCA results. Regarding the types of uncertainty in LCA, Huijbregts (1998) categorized uncertainty into parameter uncertainty, model uncertainty, and scenario uncertainty while Assefa and Frostell (2004) categorized uncertainty into parameter uncertainty and model uncertainty, with both studies suggesting that parameter uncertainty is of the utmost importance. Sensitivity analysis and uncertainty propagation can be undertaken when investigating the parameter uncertainty involved in an LCA study (EC, 2010). As a number of process parameters are generally included in an LCA study, they may cause a large variation in the results due to the presence of individual parameter uncertainty. It is important to consider the parameter uncertainty involved in an LCA study in order to improve the reliability of the results. Those process parameters leading to the highest environmental gain or loss by only a small deviation are considered as key process parameters (Heijungs, 1996). The identification of the key process parameters is important as they suggest room for possible improvement of current treatment systems and the determination of the parameter significance for future study. In LCA studies, researchers generally study the influence of selected key process parameters that are regarded as sensitive or important to the results. Since the selection of the key process parameters is usually based on the knowledge or experience of the researcher (i.e., expert judgment), there could be considerable variation of the selected parameters across LCA studies (Laurent et al., 2014). For example, Kirkeby et al. (2006) and Evangelisti et al. (2014) evaluated the environmental performance of different food waste treatment technologies. In these two studies, four key parameters with potentially large impacts on the results were selected based on expert judgment, and only one parameter, the fugitive emissions of methane in the AD process, was identical in the two studies. Zhao and Deng (2014) only investigated the influence of the energy mix parameter without considering other process parameter in food waste LCA. It is generally agreed that the expert judgment based selection approach for the key process parameters may lead to overestimation or underestimation of the parameter importance, leading to wrong conclusions. In order to identify the key process parameters in consideration of all the process parameters involved in an LCA study, sensitivity analysis can be used to evaluate the influence of each process parameter to the result. The key process parameter indicates the greatest contribution to the environmental impact and hence provides a better understanding of the variability in the LCA result (Ning et al., 2013). Wolf et al. (2016) suggested that less data collection effort should be afforded for those parameters of minor importance in future studies. This approach was applied in LCA studies for wind power generation in order to identify the key process parameters when assessing the greenhouse gas (GHG) emissions (Padey et al., 2012). Regarding waste LCA studies, Eriksson and Baky (2010) identified the key process parameters for municipal solid waste management by using sensitivity analysis. Specifically for sewage sludge and food waste treatment, the key process parameters have not yet been identified in any published study. Uncertainty propagation quantifies the uncertainty of model results due to various process parameter uncertainties and aims to improve the robustness and reliability of the results by providing ranges of possible outputs. To date, very few LCA studies have applied uncertainty propagation with only 6% of waste LCA studies considered uncertainty propagation in their assessments (Xu et al.,
967
2015). The reasons can likely be attributed to the large effort required for the collection of uncertainty information on the parameters and the difficulty in determining uncertainty information for all process parameters in an LCA study (Wolf et al., 2016). Clavreul et al. (2012) proposed a general framework for uncertainty propagation by considering the key process parameters in waste LCA. Nevertheless, the selection process of key process parameters was not described in the study (i.e., how to determine the key process parameters). Determination of the key process parameters is important for possible improvement of treatment systems, as well as being used for uncertainty propagation. In order to identify a sustainable treatment method for sewage sludge and food waste, as well as quantify the parameter uncertainty involved in the LCA study, Hong Kong is chosen as a reference city in this study. Hong Kong has long been solely relying on the three strategic landfills for sewage sludge and food waste disposal. The local government introduced a new waste management policy entitled “Hong Kong Blueprint for Sustainable Use of Resources 2013e2022” (HKEB, 2013). The directions of the policy are to reduce the waste at source, as well as utilize the waste for sustainable uses such as recovering energy from waste treatment. To tackle the latter objective, the government aims to commission a couple of waste-related infrastructures for turning waste to energy, such as building a sewage sludge incineration facility (T-PARK), and organic waste treatment facilities (OWTF) for food waste treatment by anaerobic digestion (AD). In order to further raise the waste treatment capacity, the government proposes to use sewage sludge and food waste anaerobic co-digestion (coAD) in the existing sewage treatment works (STWs) (HKCEO, 2016). In the meantime, the relocation of three existing STWs is suggested to be feasible, according to the cavern development strategy in Hong Kong (CEDD, 2011). However, the treatment methods for the sewage sludge generated from the proposed cavern STWs have not yet been established, therefore, the evaluation of a sustainable waste treatment strategy is of paramount importance. With the use of AD and coAD in the future, a large amount of biogas will be produced in Hong Kong. It is a common practice to apply a combined heat and power (CHP) system for the biogas produced to generate both heat and electricity. The former can satisfy the heat load demand by the digesters and the latter can be used as a fuel source (USEPA, 2011). Meanwhile, since the heat cannot be efficiently transported over a long distance (Cromie et al., 2014) and the demand for heat in Hong Kong is limited, the heat generated from the CHP in existing STWs can only be used internally. If the heat is not fully utilized, the overall efficiency for the CHP is consequently greatly reduced (Cromie et al., 2014). As an alternative for biogas utilization, the use of a combined cycle gas turbine (CCGT) for upgraded biogas has gained more attention recently, achieving around 55% efficiency for electricity generation (Gutierrez et al., 2016). To the best of the authors’ knowledge, studies on CCGT are still sparse in regard to biogas utilization. In particular, for countries or cities with low heat demand, evaluation of CCGT for biogas utilization in generating electricity is essential. Based on the literature review, researches on identifying key process parameters and tackling the uncertainty in waste LCA are scarce while these are necessary for process improvement and to strengthen the result reliability. To fill these gaps, the major objective of this study is to identify the key process parameters and tackling the uncertainty for an LCA study for sewage sludge and food waste treatment in Hong Kong. To achieve this major objective, four specific objectives are included: (i) to determine the key environmental impact and analyze the process contributions of the various waste treatment scenarios for sewage sludge and food waste using LCA; (ii) to evaluate whether CHP or CCGT is more environmentally friendly for biogas utilization; (iii) to identify the
968
S.L.H. Chiu, I.M.C. Lo / Journal of Cleaner Production 174 (2018) 966e976
key process parameters using a proposed selection approach with sensitivity analysis; and (iv) to propagate the uncertainty of the key process parameters identified. 2. Materials and methods 2.1. Framework of the study Fig. 1 illustrates the process flow of the current study. This study evaluates the environmental performance of different treatment scenarios for sewage sludge and food waste with reference to the cavern development strategy in Hong Kong using LCA. With the LCA results, the key impact category, indicating the highest
contribution to the total environmental performance, is determined. Sensitivity analysis is then conducted in order to identify the key process parameters. In order to examine whether the selection approach by sensitivity analysis is statistically acceptable, multiple linear regression is conducted to determine if the selected key process parameters are statistically significant so as to affect the result. After that, the uncertainty information such as the probability distribution function and uncertainty range is further collected and assigned to the key process parameters. Finally, Monte Carlo simulation is conducted to propagate the parameter uncertainty involved in the study in order to obtain the probability distribution functions of different scenarios based on the environmental performance.
Fig. 1. Process flow diagram of the current study.
S.L.H. Chiu, I.M.C. Lo / Journal of Cleaner Production 174 (2018) 966e976
2.2. System boundary and description The LCA is conducted with reference to ISO 14040 and 14044 (ISO, 2006a, 2006b). Six scenarios are proposed for sewage sludge and food waste treatment, representing all possible treatment scenarios for sewage sludge and food waste in Hong Kong. Scenario 1 is regarded as the baseline scenario as it represents the current treatment methods for these wastes in Hong Kong. A schematic
969
diagram of the various scenarios' system boundaries is shown in Fig. 2, and a detailed description of different waste treatment processes is given in Table 1. A 10:3 wet weight mixing ratio is adopted for coAD treatment, and for fair comparison, the same ratio for sewage sludge and food waste is applied in each treatment scenario. The functional unit in this study is defined as 350 t/d of sewage sludge and 105 t/d of food waste based on a 10:3 wet weight mixing ratio, taking into account for the estimated amount
Fig. 2. System boundary of the treatment scenarios for treating sewage sludge and food waste in accordance to the cavern development strategy in Hong Kong.
970
S.L.H. Chiu, I.M.C. Lo / Journal of Cleaner Production 174 (2018) 966e976
Table 1 Description of the feasible waste treatment methods in this study. Feasible treatment/disposal method
Waste
Description
West New Territories (WENT) Landfill
Food waste, fly ashes and bottom ashes from T-PARK
T-PARK (Sludge incinerator)
Sewage sludge
Organic waste treatment facility (OWTF)
Food waste
Anaerobic digestion
Sewage sludge
Anaerobic co-digestion
Sewage sludge and food waste
Waste is transported to the WENT landfill whenever landfill disposal is required. The landfill gas recovery rate is with reference to Woon and Lo (2014). Uncollected landfill gas is combusted in a flaring system and flared into the atmosphere. Complete combustion is assumed in this flaring process. Leachate is collected and pumped to a landfill leachate treatment plant. A generator fuelled by landfill gas is installed to provide heat and electricity for internal use. T-PARK has been operated since 2016 and is located at Tuen Mun, Hong Kong, which is next to the WENT Landfill. The maximum treatment capacity of T-PARK is 2000 t/d of sewage sludge (with 30% solid content). The heat energy generated from the incineration process is recovered and turned into electricity that can support the needs of the entire facility. Excess electricity is exported to the public grid (HKEPD, 2009a). OWTF adopts biological technologies, which consist of AD and composting to stabilize the food waste and turn it into biogas for energy recovery and useful compost products. In OWTF, the biogas produced is combusted to produce heat and electricity by the combined heat and power (CHP) system. The digestate is used to produce compost for land applications (HKEPD, 2009b). The compost produced avoids the application of artificial fertilizers. An anaerobic digester is built for treating the sewage sludge generated from the cavern STWs. The biogas generated during the AD process is combusted to produce heat and electricity by the CHP system. The digestate is transported and treated in T-PARK. The anaerobic digester, originally for sewage sludge AD in the cavern STWs, is used for coAD of sewage sludge and food. In this study, two biogas utilization methods for electricity generation are studied. The operational data of the anaerobic co-digestion process are collected from Yongyeon Wastewater Treatment Plant in South Korea (Ejlertsson and Magnusson, 2013). The same data are used in authors' previous study (Chiu et al., 2016) for the following reasons: (1) the plant is reconstructed for anaerobic co-digestion in 2010 in which the technology is compatible with current technology; and (2) Hong Kong and South Korea have a similar Asian diet and their food characteristics are similar. A mixing ratio of 10:3 by wet weight of the sewage sludge and food waste is used in this study, based on the South Korean plant. On the one hand, the biogas generated during the coAD process is combusted to produce heat and electricity by the CHP system in scenario 5. On the other hand, the biogas is utilized by the combined cycle gas turbine (CCGT) in scenario 6 to generate electricity. Biogas is first upgraded to 90% methane content (Woon et al., 2016) and then combusted by the CCGT system, with reference to Gutierrez et al. (2016). The remaining co-digestate is then transported to the T-PARK for combustion.
of sewage sludge from the cavern STWs (DSD, 2015). In addition, it is assumed that only the environmental impacts from the operational phase of the waste treatment facilities are considered, since they are regarded as the major environmental burdens (Gentil et al., 2010), while the environmental impacts of the construction and capital equipment are not included in this study. As mentioned in Cleary (2009), these items are considered as secondary environmental burdens and are relatively insignificant compared to the primary environmental emissions from the waste treatment process. Moreover, the environmental impact due to waste haulage is assumed to be negligible due to its low contribution, as observed in the authors’ previous waste LCA study in Hong Kong (Woon et al., 2016). The GHG and air pollutant emission factors for substituted electricity are collected from China Light & Power Hong Kong Company Limited (CLP, 2015) and those for substituted heat are from the Hong Kong and China Gas Company Limited (Towngas, 2015). CHP is applied for biogas utilization in scenarios 2, 3, 4, and 5. The heat generated is assumed to be used internally in the treatment plant as the demand for heat in Hong Kong is limited. However, it should be noted that successful applications of district heating or cooling systems in order to utilize the heat from CHP could be found in many countries. In order to compare the environmental performance of CHP and CCGT for biogas utilization, scenario 6 is proposed to apply CCGT to evaluate utilization of the biogas produced from the coAD treatment plant. It should be noted that all the key processes in scenario 6 are identical to scenario 5 except the use of CCGT replacing CHP. For scenario 5, the biogas produced is combusted to generate heat and electricity by the CHP system. The electricity generated is utilized by internal consumption and the excess electricity is exported to the public grid. Generally, 25% of the heat generated is utilized for the internal heat
demand by the anaerobic digesters (Woon et al., 2016). For scenario 6, the biogas is first upgraded to biomethane, with 90% methane content, for the application of CCGT to generate electricity (Cromie et al., 2014). The biogas is upgraded and purified by water scrubbing as it is the most suitable technology for biogas upgrading (Chiu and Lo, 2016). The life cycle inventory data on substrate composition and waste treatment processes are presented in Tables A.1 e A.7 in the supplementary information. The major assumptions and limitations of the scenario system boundary and data inventory include: (i) the data are mostly collected from local sources and they may not replicate global trends; (ii) only the environmental emissions from the operational phases are considered while the environmental emissions from construction and capital equipment are excluded as they are considered insignificant compared to the emissions from the operational phases (Cleary, 2009: Gentil et al., 2010); (iii) waste haulage is assumed to be negligible (Woon et al., 2016); (iv) heat produced from CHP is assumed to be used internally; and (v) the mixing ratio of sewage sludge and food waste in coAD is applied from a Korean plant data, and additional experiments are needed to determine the suitable mixing ratio of the substrates in Hong Kong. 2.3. Life cycle impact assessment SimaPro 7.2.4 software with ReCiPe version 1.04 (Goedkoop et al., 2009) is used for life cycle impact assessment. Four environmental impact categories are chosen as they are most commonly studied with reference to waste LCA studies, namely climate change, particulate matter formation, photochemical oxidant formation, and terrestrial acidification (Bernstad and la Cour Jansen, 2012). According to the ReCiPe method, climate change, photochemical oxidant formation and particulate matter
S.L.H. Chiu, I.M.C. Lo / Journal of Cleaner Production 174 (2018) 966e976
formation are categorized under the human health endpoint damage category by quantifying the midpoint impact to disabilityadjusted life years. Climate change and terrestrial acidification are categorized under the ecosystems endpoint damage category by quantifying the midpoint impact to the loss of species during a year. Among the midpoint impact categories studied, the midpoint impact with the highest contribution to the endpoint damage is identified as the key impact category. The key impact category is then used as a reference result for sensitivity analysis so as to identify the key process parameters. 2.4. Sensitivity analysis Sensitivity analysis is conducted for all process parameters in different scenarios based on the key impact category identified. All the process parameters involved in this study are presented in Table A.8 in the supplementary information. Sensitivity analysis can be classified into local sensitivity analysis (LSA) and global sensitivity analysis (GSA). LSA considers the output variability by changing the input parameter within a specific range. It is particularly useful in providing early approximation to identify the key parameters (Hemsath and Bandhosseini, 2015). Meanwhile, GSA differs from LSA in that all model parameters vary simultaneously when determining the parameter's sensitivity to overcoming the limitation of the linearity issue of LSA, and it requires specifying the input variability. Although GSA has its own merits in providing an overview of the influence of the inputs on the outputs, the conclusions from GSA should be drawn with care when the input variability is poorly known (Iooss and Lemaître, 2015). In addition, they are usually computationally expensive (Cariboni et al., 2007), and the choice of the method to use is driven by the computing cost in general (Saltelli et al., 2006). In the current study, LSA is chosen for sensitivity analysis so as to identify the key process parameters prior to uncertainty propagation. All the process parameters are assumed to be independent as their interactions are limited since they are mostly obtained from individual process. Sensitivity analysis is undertaken by varying one parameter at a time, while keeping the other parameters constant in order to determine the sensitivity of the parameters to variation in the input data variables. This is known as the one-at-a-time method (Morio, 2011). The minimum and maximum values of a parameter, which are considered as the amplitude of sensitivity analysis to reflect the degree of influence of the parameters on the assessment result (Boldrin et al., 2011), are collected from literature. By varying the process parameters with both minimum and maximum values, two sensitivity ratio (SR) values can be calculated, in which the higher SR value is chosen for the process parameter in order to reflect its sensitivity. To assess the influence of process parameters on the result, the SR is calculated for each process parameter using Eq. (1) (Clavreul et al., 2012). Since the SR can be a negative value for an avoided environmental impact result, the absolute value is applied for the SR with the aim of determining the sensitivity of the process parameter. A higher SR value indicates higher parameter sensitivity.
Final result Initial result Initial result SR ¼ Min or max process parameter Initial process parameter
971
solely determined by an exact value of SR, since there is no standard or reference for SR values. In addition, the SR value of individual process parameters can be different in each study due to the variation of the system boundaries. A more practical way to determine the key process parameters is by selecting parameters with high SR values. In this study, a selection approach based on the relative importance of the SR value is used to identify the key process parameters. The relative importance of the SR indicates the level of influence of the parameters to the result, representing the rank of the importance. It can be calculated by dividing an individual SR value with the total SR value of all the process parameters. The cumulative relative importance of SR can be obtained by summing up the relative importance of SR starting from the highest SR value. Once the cumulative relative importance of SR exceeds 50%, those process parameters used for the summation are identified as key process parameters. This selection process indicates that the selected parameters have contributed over half of the sensitivity among all the process parameters, while other parameters can be considered as trivial in regard to the result. In order to examine whether the proposed selection approach for the identification of key process parameters is reasonable, multiple linear regression is further conducted on the best scenario obtained based on the key impact category to investigate the significance of each process parameter to affect the result. Each process parameter in the best scenario is randomly varied within their possible ranges simultaneously with the degree of freedom of 30. It is noted that when the degree of freedom reaches 30, the t-distribution and the standard normal distribution are similar and the test result is regarded as reliable with sufficient number of trials (Anderson, 2013). The p-values obtained are then examined and those parameters with p-value lower than 0.05 are treated as significant according to the 95% confidence level, implying a statistically significant relationship with the result. The results from multiple linear regression can determine whether the key process parameters are statistically significant so as to affect the result, and therefore can support whether the proposed approach is statistically acceptable. Detailed calculation equations of the relative importance and cumulative relative importance of SR, and multiple linear regression are provided in Eq. A.1 and Eq. A.2 in the supplementary information. 2.6. Uncertainty propagation Monte Carlo simulation method is used to propagate the uncertainty of the selected key process parameters, which are represented by a stochastic variable with a probability distribution. By selecting a random sample from the probability distribution of the process parameters, Monte Carlo simulation calculates the impact results by repeating for 10,000 iterations. It has been built into commercial LCA software, and SimaPro 7.2.4 software is used to conduct the Monte Carlo simulation in the current study. The uncertainty information, which represents the input variability of each key process parameter, is then determined for the Monte Carlo simulation. The uncertainty information including probability distribution and uncertainty range are collected from the literature with detailed justifications provided in section 3.2.
(1)
Initial process parameter
3. Results and discussion 3.1. Process contributions to the key impact category
2.5. Identification of key process parameters Although the SR reflects the sensitivity of the process parameter to the result, the selection of key process parameters cannot be
As mentioned in Section 2.3, four midpoint impact categories are evaluated in this study and their relative contributions to the endpoint damage categories are presented in Table 2. Detailed results regarding the midpoint impact categories are presented in
972
S.L.H. Chiu, I.M.C. Lo / Journal of Cleaner Production 174 (2018) 966e976
Table 2 Midpoint impact category contribution to endpoint damage category in each scenario. Impact category
Unit
Scenario 1
Scenario 2
Scenario 3
Scenario 4
Scenario 5
Scenario 6
Climate change human health Photochemical oxidant formation Particulate matter formation Human healtha Climate change ecosystems Terrestrial acidification Ecosystemsa
DALYb DALY DALY DALY species-year species-year species-year
0.153 2.45 10-6 3.58 10-3 0.156 8.65 10-4 7.30 10-8 8.65 10e4
0.102 1.70 10-6 7.50 10-3 0.109 5.77 10-4 1.52 10-7 5.77 10e4
0.0176 1.23 10-7 4.06 10-4 0.0172 9.88 10-5 2.97 10-7 9.85 10e5
0.0334c 8.65 10-7 3.50 10-3 ¡0.0299 1.90 10-4 3.76 10-7 ¡1.90 10e4
0.0529 1.75 10-6 4.97 10-3 ¡0.0479 3.00 10-4 5.31 10-7 ¡3.01 10e4
0.0946 3.77 106 3.88 104 ¡0.0942 5.36 104 8.43 107 ¡5.37 104
a b c
Endpoint damage categories. DALY stands for Disability-adjusted life year. Negative value represents a positive impact to the environment.
Fig. 3. Process contributions by different scenarios to climate change impact category.
Table B.1 in the supplementary information. Regarding the endpoint damage categories for human health and ecosystems, scenarios 4, 5, and 6 have an overall positive environmental performance on human health and ecosystems in general, as a result of avoided emissions. It indicates the substitution effect of avoiding direct emissions from energy production using fossil fuels. Scenario 6 performs the best in both endpoint categories, with 52.3% more avoided impacts in human health, and 43.9% more avoided impacts in ecosystems, compared with scenario 5, which is the second best scenario. It indicates a better environmental performance is achieved with the use of CCGT over CHP for biogas utilization. Based on the contribution from the midpoint impact categories in Table 2, climate change has the highest impact in both endpoint damage categories and it is identified as the key impact category. Fig. 3 illustrates the environmental impacts related to climate change. A positive figure represents the emissions of waste treatment being greater than the avoided impacts resulting from resource utilization, and vice versa for a negative figure. Fig. 3 shows that scenarios 5 and 6 are the best two scenarios with 3.78 104 kg CO2e and 6.75 104 kg CO2e per functional unit, respectively. The avoided impacts are mainly contributed by the coAD process, bringing environmental benefits compared to those scenarios without this process. Apart from the coAD process in scenarios 5 and 6, the AD processes in scenarios 2, 3, and 4 also result in negative values, which indicates the avoided impacts are greater than the direct impacts of the process. In this study, a default value of 5% for fugitive CH4 emissions (IPCC, 2006) is applied for AD and coAD. The avoided impacts are contributed by the energy recovery system in generating heat and electricity, which considerably reduce the negative implications to the
environment. Meanwhile, it is interesting to note that the avoided impacts from scenario 4, which involve AD treatment for sewage sludge and food waste separately, exhibit similar reductions to scenario 5. This is because the compost produced from the food waste digestate is used as artificial fertilizer. For direct impacts, both the landfilling of food waste and incineration of sewage sludge or digestate at T-PARK create direct impacts on climate change. The landfill gas cannot be completely collected for the energy recovery system and flaring process. As a result, the uncollected landfill gas diffuses from the surface of the landfill and CH4 is released as a GHG to the atmosphere, resulting in climate change impacts. Regarding the sludge incineration in T-PARK, the most significant contribution during the incineration process is the N2O emission from the combustion of sewage sludge due to the high nitrogen content (Chiu et al., 2016). Since N2O is a potent GHG with a global warming potential of 298 (Forster et al., 2007), the incineration of sewage sludge makes a large contribution to climate change. Last but not least, the environmental impacts brought by sludge dewatering and food waste pre-treatment are relatively insignificant compared to other major processes.
3.2. Identification of key process parameters by the proposed selection approach Considering climate change impact as the key impact category, sensitivity analysis is then conducted for all the process parameters in order to calculate their respective SR values and relative importance. As mentioned in Section 2.5 in the methodology, when the cumulative relative importance of the SR value exceeds 50%, those process parameters that are used for the summation are
S.L.H. Chiu, I.M.C. Lo / Journal of Cleaner Production 174 (2018) 966e976 Table 3 Process parameters required to get the cumulative importance of SR value over 50%. Key process parameters
Scenario 1 Lower heating value of sewage sludge Energy recovery efficiency (sewage sludge incineration plant) Scenario 2 CHP efficiency (electricity) (AD/ coAD plant) Lower heating value of methane Percentage of methane in biogas Scenario 3 Lower heating value of sewage sludge Energy recovery efficiency (sewage sludge incineration plant) CHP efficiency (electricity) (AD/ coAD plant) Scenario 4 Lower heating value of methane CHP efficiency (electricity) (AD/ coAD plant) Percentage of methane in biogas Scenario 5 Lower heating value of methane CHP efficiency (electricity) (AD/ coAD plant) Energy recovery efficiency (sewage sludge incineration plant) Scenario 6 CCGT efficiency Lower heating value of methane Percentage of methane in biogas
Relative importance of SR value (%)
Cumulative relative importance of SR (%)
33.8
67.6
33.8
19.3
50.3
19.3 12.5 21.5
56.0
20.9 13.6
24.0 23.1
62.2
15.1 21.6 20.5
56.2
14.1
21.1 20.7 18.9
60.7
identified as key process parameters and are presented in Table 3. For detailed values of SR of all the process parameters, they can be found in in Table B.2 in the supplementary information. Multiple regression analysis is further conducted on scenario 6, which is the best scenario obtained based on the key impact category. The regression result is presented in Table B.3 and parameters of the lower heating value of digestate, energy recovery efficiency of sewage sludge incineration plant, biogas production rate, percentage of methane in biogas, lower heating value of methane, CCGT efficiency, and energy consumption for biogas upgrade attained p-values lower than 0.05, which implies that these parameters are significant to the result according to 95% confidence level. It is found that the key process parameters identified by the proposed selection approach are also regarded as significant parameters based on the multiple linear regression result. It indicates the proposed approach allows statistically acceptable selection of the key process parameters. Regarding the key process parameters identified, the percentage of methane in the biogas and the biogas production rate from the food waste depend on the characteristics of the waste. These parameters can be improved by applying appropriate pre-treatment technologies to the organic wastes, such as mechanical and thermal pre-treatment (Chiu and Lo, 2016). In order to maintain a good performance of the energy recovery efficiency of the sewage sludge incineration plant, CHP and CCGT efficiencies for electricity recovery, these energy generation facilities need to be frequently cleaned and maintained (DEFRA, 2013). The lower heating value of methane has relatively less fluctuation. The identification of the key process parameters is beneficial to future LCA studies of organic waste and extra efforts should be made to collect more reliable data for these key parameters in order to provide a more accurate result.
973
The uncertainty information including the probability distribution function and the uncertainty range are then determined for these key process parameters for uncertainty propagation as given in Table 4. The determination of the probability distribution function for process parameters can be classified into two groups. For the first group, the distribution function can be determined by statistical methods if extensive data are available (Huijbregts et al., 2003). However, it is generally impossible to differentiate distribution functions when sample sizes are very small (Penman, 2001). For the other group, an appropriate distribution can be selected by expert estimation when little information is available (Sonnemann et al., 2003). In the current study, the distribution function of the parameters is based on the expert judgment approach due to insufficient data. In order to determine the distribution function for the key process parameters selected, the relevant literature has been reviewed. For the energy recovery efficiency of the sewage sludge incineration plant, CHP and CCGT efficiencies for electricity of the AD or coAD plant, a lognormal distribution is suggested according to Galvez-Martos and Schoenberger (2014), as only positive values are possible. For the heating value of sewage sludge and methane, a lognormal distribution is used since it fits the dataset based on national inventory reports, IEA data and available national data for fuel (IPCC, 2006). A normal distribution is assigned to the percentage of methane in biogas and the biogas production rate of food waste with reference to Clavreul et al. (2012). The probability distribution function for empirical data is normal or lognormal in the current study, in line with Penman (2001). A 95% confidence interval is applied in the study and this means that there is a 95% probability that the confidence interval will contain the true population mean, indicating the range that the parameter would frequently fall into in most of the situations. For lognormal distribution, the 95% confidence interval is defined by dividing or multiplying the mean value with the squared geometric standard deviation. For normal distribution, the 95% confidence interval is approximately defined by subtracting twice the standard deviation and adding twice the standard deviation (Goedkoop et al., 2016). Three levels of the uncertainty range are proposed in order to reflect the parameter uncertainty: 50%, 25%, and 10% for high, medium, and low uncertainty of the key process parameters, respectively. The uncertainty of the lower heating value of sewage sludge is high since data on biomass as fuel are not well developed (IPCC, 2001). The percentage of methane in biogas and the biogas production rate from food waste are assumed to have medium uncertainty, as they are dependent on the characteristics of the waste. The energy recovery efficiency of the sewage sludge incineration plant, CHP and CCGT efficiencies for electricity of AD or coAD plant, and lower heating value of methane are subject to lower uncertainty as the technology is well developed and there is sufficient data to support using the lower heating value of methane. 3.3. Results with uncertainty propagation Fig. 4 illustrates the result of the probability distribution function and cumulative probability distributions of scenarios 1 to 5 under uncertainty. The probability distribution function graph aims to show the distribution of climate change values whereas the cumulative probability graph allows the identification of percentiles corresponding to the climate change impact presented on the horizontal axis. The statistical indices (i.e., mean, standard deviation, and possible range in the 95% confidence interval) of each scenario are presented in Table B.4. In Fig. 4a, the results show that scenarios 1, 2, and 3, the probability of having a higher climate change impact is notably greater than that of scenarios 4 and 5. Comparing scenarios 4 and 5, scenario 5 attains a better
974
S.L.H. Chiu, I.M.C. Lo / Journal of Cleaner Production 174 (2018) 966e976
Table 4 Probability distribution and uncertainty range of the key process parameters. a
Key process parameter
Unit
Probability distribution
Mean
Uncertainty range
GSD or SD
Energy recovery efficiency (Sewage sludge incineration plant) CHP efficiency for electricity (AD/coAD plant) Lower heating value of sewage sludge Lower heating value of methane Percentage of methane in biogas CCGT efficiency
%
Lognormal
20
±10%
1.05
Galvez-Martos and Schoenberger, 2014
%
Lognormal
36
±10%
1.05
Galvez-Martos and Schoenberger, 2014
MJ/kg dry basis
Lognormal
10
±50%
1.22
IPCC, 2006
kWh/m3
Lognormal
10
±10%
1.05
IPCC, 2006
%
Normal
60
±25%
7.5
Clavreul et al., 2012
%
Lognormal
55
±10%
1.05
Galvez-Martos and Schoenberger, 2014
a
References for probability distribution
GSD stands for geometric standard deviation, which is used for lognormal distribution. SD stands for standard deviation, which is used for normal distribution.
Fig. 4. (a) Probability distribution function in climate change impact; and (b) cumulative probability distribution in climate change impact of scenarios 1, 2, 3, 4, and 5.
performance in general. Yet, it might be possible that the climate change impact of scenario 4 is “more negative” than scenario 5 on some occasions. By referring to the cumulative probability graph in Fig. 4b, scenario 5 should at least avoid 2.56 104 kg CO2e with a 95% confidence interval, whilst the cumulative probability of scenario 4 in achieving this reduction is 31%. Based on the cumulative probability curve, the climate change performance of scenario 5 is always better than scenario 4 in most situations. The environmental performance of CHP and CCGT for biogas utilization is further evaluated and compared by examining scenarios 5 and 6. In both scenarios, the same amount of biogas is consumed, except that the biogas is required to be first upgraded to biomethane for the CCGT system in scenario 6. Sensitivity analysis is applied to scenario 6 after the assessment to determine its climate change impact. Fig. 5a shows the results of the probability distribution function and Fig. 5b presents the results of the cumulative probability distributions corresponding to the climate change impact of scenarios 5 and 6 under uncertainty. From the results, it is notable that scenario 6 attains better environmental
performance on climate change than that of scenario 5. The probability of scenario 6 achieving at least 5.32 104 kg CO2e avoided emissions is 95%, whilst it is below 5% for scenario 5, indicating a significant advantage of CCGT over CHP. Given that the CCGT system does not generate heat for internal consumption in the current study, and external heat is required for the anaerobic digesters, the avoided environmental impact is still greater in the CCGT scenario due to the higher efficiency in generating electricity. Particularly in the Hong Kong situation, the production of electricity is carbonintensive as coal contributes approximately 50% of the current fuel mix, whilst the heat generated by city gas is produced via naphtha and natural gas, serving as a cleaner fuel in terms of greenhouse gas emissions.
4. Conclusions This study applies LCA to evaluate the environmental performance of different treatment scenarios for sewage sludge and food waste with reference to the cavern development strategy in Hong
S.L.H. Chiu, I.M.C. Lo / Journal of Cleaner Production 174 (2018) 966e976
975
Fig. 5. (a) Probability distribution function in climate change impact; and (b) cumulative probability distribution in climate change impact of scenarios 5 and 6.
Kong. In addition, the key process parameters are identified and the uncertainty in this LCA is also tackled. Regarding the environmental performance of different treatment systems, the endpoint damage results indicate that scenarios 5 and 6, which involve coAD as the major treatment method, are more advantageous over other possible treatment methods for these wastes. Among the midpoint impact categories studied, climate change is the key impact category with the highest contribution to environmental performance. It is found that coAD treatment provides the greatest environmental benefits while landfilling results in the greatest environmental burdens in the waste treatment scenarios. The climate change impact category then serves as the reference result for sensitivity analysis in the proposed selection approach for the identification of the key process parameters. The parameters of electricity efficiency in different waste treatment facilities are the common key process parameters influencing the environmental performance, and therefore more attention should be paid on these process parameters in order to improve the environmental performance for cleaner production. Multiple linear regression is further conducted in order to justify whether the proposed methodology for key process parameters identification is statistically acceptable. The results show that the key process parameters identified from the proposed methodology are also regarded as significant parameters from the multiple linear regression result, at the 95% confidence level, indicating the parameter selection by the proposed approach is statistically acceptable. After uncertainty propagation, it is observed that scenarios 5 and 6 with coAD treatment perform better than other waste treatment scenarios in general. Regarding the use of CHP and CCGT for biogas utilization, scenario 6 has a 95% probability of achieving at least 5.32 104 kg CO2e avoided emissions while the probability of the second best scenario (i.e., scenario 5) achieving the same avoided emissions is below 5%, indicating a significant advantage of CCGT over CHP for biogas utilization in region with low heat demand, like Hong Kong. Based on the results from this study, it is concluded that coAD with CCGT application is the best treatment option for sewage sludge and food waste in accordance to the cavern development strategy
in Hong Kong. This study demonstrates an approach to identify the key process parameters and tackle the uncertainty in LCA. The identification of key process parameters is critical for environmental sustainability as these parameters are considered as hotspots with high influence to the environmental results. Besides, tackling the LCA uncertainty is also necessary as it provides stochastic result rather than deterministic result, hence the decision-makers can determine the likeliness to achieve particular environmental performance, especially when there are other selection criteria for consideration. Last but not least, this study only considers the environmental results as the performance indicator while modern waste treatment should also consider other important performance indicators such as life cycle costing and life cycle energy efficiency. They should be further studied with the consideration of their uncertainty involved by using the approach demonstrated in this study. Appendix A. Supplementary data Supplementary data related to this article can be found at https://doi.org/10.1016/j.jclepro.2017.10.164. References Anderson, A., 2013. Business Statistics for Dummies. John Wiley & Sons, United States. Assefa, G., Frostell, B., 2004. Social impact assessment within life cycle technology assessment. In: 4th SETAC World Congress. http://urn.kb.se/resolve?urn¼urn: nbn:se:kth:diva-88625 (assessed 07.08.17). Bernstad, A., la Cour Jansen, J., 2012. Review of comparative LCAs of food waste management systems e current status and potential improvements. Waste Manage 32 (12), 2439e2455. Boldrin, A., Neidel, T.L., Damgaard, A., Bhander, G.S., Møller, J., Christensen, T.H., 2011. Modelling of environmental impacts from biological treatment of organic municipal waste in EASEWASTE. Waste Manage 31 (4), 619e630. Cariboni, J., Gatelli, D., Liska, R., Saltelli, A., 2007. The role of sensitivity analysis in ecological modelling. Ecol. Model. 203, 167e182. CEDD, 2011. Enhancing Land Supply Strategy: Reclamation outside Victoria Harbour and Rock Cavern Development. Civil Engineering Development Department, Hong Kong. Chiu, S.L.H., Lo, I.M.C., 2016. Reviewing the anaerobic digestion and co-digestion
976
S.L.H. Chiu, I.M.C. Lo / Journal of Cleaner Production 174 (2018) 966e976
process of food waste from the perspectives on biogas production performance and environmental impacts. Environ. Sci. Technol. 23 (24), 24435e24450. Chiu, S.L.H., Lo, I.M.C., Woon, K.S., Yan, D.Y.S., 2016. Life cycle assessment of waste treatment strategy for sewage sludge and food waste in Macau: perspectives on environmental and energy production performance. Int. J. Life Cycle Assess. 21 (2), 176e189. Clavreul, J., Guyonnet, D., Christensen, T.H., 2012. Quantifying uncertainty in LCAmodelling of waste management systems. Waste Manage 32 (12), 2482e2495. Cleary, J., 2009. Life cycle assessments of municipal solid waste management systems: a comparative analysis of selected peer-reviewed literature. Environ. Int. 35, 1256e1266. CLP, 2015. Sustainability Report 2015. China Light & Power Company, Hong Kong. https://www.clpgroup.com/en/Sustainability-site/Report%20Archive%20% 20Year%20Document/SR_Full_2015_en.pdf (assessed 07.08.17). Cromie, T., Brown, J., Watters, P., Foley, A., Smyth, B., Rooney, D., Glover, S., Orozco, A., Groom, E., Irvine, C., 2014. Alternative Uses of Biogas Group Report. A Northern Ireland case study. Appendix 3.2, Biogas action plan for Northern Ireland. Northern Ireland. DEFRA, 2013. Incineration of Municipal Solid Waste. Department for Environment Food & Rural Affairs, United Kingdom. DSD, 2015. Sewage Treatment Works Information Sheet. Drainage Services Department, Hong Kong. http://www.dsd.gov.hk/EN/Publicity_and_ Publications/Publicity/Leaflets__Booklets___Fact_Sheets/index.html (Accessed 04.11.16). EC, 2010. International Reference Life Cycle Data System (ILCD) Handbook e General Guide for Life Cycle Assessment e Detailed Guidance, first ed. European Commission - Joint Research Centre - Institute for Environment and Sustainability, Luxemburg. EUR 24708. EN. EEA, 2013. EU Bioenergy Potential from a Resource-efficiency Perspective. Report no. 6/2013. European Environment Agency. EEA, 2017. Municipal Waste Management across European Countries. European Environment Agency. https://www.eea.europa.eu/themes/waste/municipalwaste/municipal-waste-management-across-european-countries (Accessed 03.08.17). Ejlertsson, J., Magnusson, B., 2013. Maximizing Biogas Production at a South Korean Biogas Plant, Scandinavian Biogas (South Korea). Eriksson, O., Baky, A., 2010. Identification and testing of potential key parameters in system analysis of municipal solid waste management. Resour. Conserv. Recy 54 (12), 1095e1099. Evangelisti, S., Lettieri, P., Borello, D., Clift, R., 2014. Life cycle assessment of energy from waste via anaerobic digestion: a UK case study. Waste Manage 34 (1), 226e237. Forster, P., Ramaswamy, V., Artaxo, P., Berntsen, T., Betts, R., Fahey, D.W., Haywood, J., Lean, J., Lowe, D.C., Myhre, G., Nganga, J., Prinn, R., Raga, G., Schulz, M., Van Dorland, R., 2007. Changes in atmospheric constituents and in radiative forcing. In: Climate Change 2007: the Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Gentil, E.C., Damgaard, A., Hauschild, M., Finnveden, G., Eriksson, O., Thorneloe, S., Kaplan, P.O., Barlaz, M., Muller, O., Matsui, Y., Ii, R., 2010. Models for waste life cycle assessment: review of technical assumptions. Waste Manage 30 (12), 2636e2648. Galvez-Martos, J.L., Schoenberger, H., 2014. An analysis of the use of life cycle assessment for waste co-incineration in cement kilns. Resour. Conserv. Recy 86, 118e131. Goedkoop, M.J., Heijungs, R., Huijbregts, M., De Schryver, A., Struijs, J., Van Zelm, R., 2009. ReCiPe 2008da Life Cycle Impact Assessment Method Which Comprises Harmonised Category Indicators at the Midpoint and the Endpoint Level, first ed. (Report I: Characterisation, First edition. Netherlands). Goedkoop, M.J., Oele, M., Leijting, J., Ponsioen, T., Meijer, E., 2016. Introduction to Consultant, Netherlands. LCA with SimaPro. PRe Gutierrez, E.C., Xia, A., Murphy, J.D., 2016. Can slurry biogas systems be cost effective without subsidy in Mexico? Renew. Energ 95, 22e30. Heijungs, R., 1996. Identification of key issues for further investigation in improving the reliability of life-cycle assessments. J. Clean. Prod. 4 (3e4), 159e166. Hemsath, T.L., Bandhosseini, K.A., 2015. Sensitivity analysis evaluating basic building geometry's effect on energy use. Renew. Energy 76, 526e538. HKCEO, 2016. Policy Address of Hong Kong. Hong Kong Office of the Chief Executive. http://www.policyaddress.gov.hk/2016/eng/index.html (Accessed 04.11.16). HKEB, 2013. Hong Kong Blueprint for Sustainable Use of Resources 2013-2022. Hong Kong Environmental Bureau, Hong Kong. HKEPD, 2009a. Sludge Treatment Facilities Environmental Impact Assessment Report. Hong Kong Environmental Protection Department, Hong Kong. HKEPD, 2009b. Organic Waste Treatment Facilities Phase 1 Environmental Impact
Assessment Report. Hong Kong Environmental Protection Department, Hong Kong. HKEPD, 2017. Monitoring of Solid Waste in Hong Kong Waste Statistics for 2015 (Hong Kong). Huijbregts, M.A., 1998. Application of uncertainty and variability in LCA. Int. J. Life Cycle Assess. 3 (5), 273e280. Huijbregts, M.A., Gilijamse, W., Ragas, A.M., Reijnders, L., 2003. Evaluating uncertainty in environmental life-cycle assessment. A case study comparing two insulation options for a Dutch one-family dwelling. Environ. Sci. Technol. 37 (11), 2600e2608. Humbert, S., Loerincik, Y., Rossi, V., Margni, M., Jolliet, O., 2009. Life cycle assessment of spray dried soluble coffee and comparison with alternatives (drip filter and capsule espresso). J. Clean. Prod. 17 (15), 1351e1358. Iooss, B., Lemaître, P., 2015. A review on global sensitivity analysis methods. In: Uncertainty Management in Simulation-optimization of Complex Systems (Pp. 101e122). Springer, US. IPCC, 2001. Good Practice Guidance and Uncertainty Management in National Greenhouse Gas Inventories. Intergovernmental Panel on Climate Change, Switzerland. IPCC, 2006. IPCC Guidelines for National Greenhouse Gases Inventories. Intergovernmental Panel on Climate Change, Switzerland. ISO 14040, 2006a. Environmental Management e Life Cycle Assessment e Life Cycle Assessment Principles and Framework. International Standards Organization, Geneva, Switzerland. ISO 14044, 2006b. Environmental Management e Life Cycle Assessment e Requirements and Guidelines. International Standards Organization, Geneva, Switzerland. Kirkeby, J.T., Birgisdottir, H., Hansen, T.L., Christensen, T.H., Bhander, G.S., Hauschild, M., 2006. Evaluation of environmental impacts from municipal solid waste management in the municipality of Aarhus, Denmark (EASEWASTE). Waste Manage. Res. 24 (1), 16e26. Laurent, A., Clavreul, J., Bernstad, A., Bakas, I., Niero, M., Gentil, E., Hauschild, M.Z., 2014. Review of LCA studies of solid waste management systemsePart II: methodological guidance for a better practice. Waste Manage 34 (3), 589e606. Morio, J., 2011. Global and local sensitivity analysis methods for a physical system. Eur. Phys. J. 32 (6), 1577. Ning, S.K., Chang, N.B., Hung, M.C., 2013. Comparative streamlined life cycle assessment for two types of municipal solid waste incinerator. J. Clean. Prod. 53, 56e66. Padey, P., Blanc, I., Le Boulch, D., Xiusheng, Z., 2012. A simplified life cycle approach for assessing greenhouse gas emissions of wind electricity. J. Ind. Ecol. 16 (s1), S28eS38. Penman, J., 2001. Good Practice Guidance and Uncertainty Management in National Greenhouse Gas Inventories. Intergovernmental Panel on Climate Change, Switzerland. Righi, S., Oliviero, L., Pedrini, M., Buscaroli, A., Della Casa, C., 2013. Life cycle assessment of management systems for sewage sludge and food waste: centralized and decentralized approaches. J. Clean. Prod. 44 (0), 8e17. Saltelli, A., Ratto, M., Tarantola, S., Campolongo, F., 2006. Sensitivity analysis practices: strategies for model-based inference. Reliab. Eng. Syst. Safe 91, 1109e1125. Singh, R.P., Singh, P., Araujo, A.S., Ibrahim, M.H., Sulaiman, O., 2011. Management of urban solid waste: vermicomposting a sustainable option. Resour. Conserv. Recy 55 (7), 719e729. Sonnemann, G.W., Schuhmacher, M., Castells, F., 2003. Uncertainty assessment by a Monte Carlo simulation in a life cycle inventory of electricity produced by a waste incinerator. J. Clean. Prod. 11 (3), 279e292. Towngas, 2015. The Hong Kong and China Gas Company Limited. Sustainability Report. Hong Kong. https://www.towngas.com/getmedia/73b36b28-8281476d-ac00-8d8ea6bd93cc/Sustainability-Summary-Report-2015.pdf.aspx? ext¼.pdf (assessed 07.08.17). USEPA, 2011. Protocol for Quantifying and Reporting the Performance of Anaerobic Digestion Systems for Livestock Manures. United States Environmental Protection Agency AgSTAR Program, USA. Wolf, P., Groen, E.A., Berg, W., Prochnow, A., Bokkers, E.A., Heijungs, R., de Boer, I.J., 2016. Assessing greenhouse gas emissions of milk production: which parameters are essential? Int. J. Life Cycle Assess. 22 (3), 441e455. Woon, K.S., Lo, I.M.C., Chiu, S.L.H., Yan, D.Y.S., 2016. Environmental assessment of food waste valorization in producing biogas for various types of energy use based on LCA approach. Waste Manage 50, 290e299. Xu, C., Shi, W., Hong, J., Zhang, F., Chen, W., 2015. Life cycle assessment of food waste-based biogas generation. Renew. Sust. Energ. Rev. 49, 169e177. Zhao, Y., Deng, W., 2014. Environmental impacts of different food waste resource technologies and the effects of energy mix. Resour. Conserv. Recy 92, 214e221.
Contents lists available at ScienceDirect
Journal of Cleaner Production journal homepage: www.elsevier.com/locate/jclepro
Identifying key process parameters for uncertainty propagation in environmental life cycle assessment for sewage sludge and food waste treatment Sam L.H. Chiu, Irene M.C. Lo* Department of Civil and Environmental Engineering, The Hong Kong University of Science and Technology, Hong Kong, China
a r t i c l e i n f o
a b s t r a c t
Article history: Received 28 June 2017 Received in revised form 30 September 2017 Accepted 15 October 2017 Available online 16 October 2017
Life cycle assessment (LCA) results are often subject to uncertainty, which may lead to erroneous conclusions. This paper aims to tackle the parameter uncertainty involved in LCA. Hong Kong is selected as the reference city for the treatment of sewage sludge and food waste. A selection approach with sensitivity analysis is proposed to identify the key process parameters and Monte Carlo simulation is further conducted to propagate the uncertainty of the key process parameters identified. The results show that climate change is the major impact category among various impact categories that are generally considered in waste LCA studies. Scenarios 5 and 6, which consider anaerobic co-digestion (coAD) treatment, achieve the best performance in regard to the climate change impact. Scenario 6 which includes a combined cycle gas turbine system for biogas utilization has 6.75 104 kg avoided CO2e emissions and it is equal to 44% more avoided emissions compared to scenario 5 which applies a combined heat and power system. For the key process parameters identified, it is found that the electricity generation efficiencies in different waste treatment facilities, such as the incineration plant and the anaerobic digestion plant, have the greatest sensitivity to the result. Uncertainty propagation is then conducted to obtain the probability distribution functions in environmental performance of different scenarios. Scenario 6 has a 95% probability of achieving at least 5.32 104 kg avoided CO2e emissions, while the probability of scenario 5, which is the second best scenario, achieving the same avoid emissions is below 5%. It indicates a significant advantage of using combined cycle gas turbine over combined heat and power unit for biogas utilization in Hong Kong. The methodologies and results of this study provide comprehensive material that can be adapted for other areas planning sustainable sewage sludge and food waste treatment, as well as in tackling parameter uncertainty in general LCA studies. © 2017 Elsevier Ltd. All rights reserved.
Keywords: Anaerobic co-digestion Combined cycle gas turbine Combined heat and power Monte Carlo Organic waste Sensitivity analysis
1. Introduction Organic waste management has long been a major challenge in urban areas. Improper management of organic waste could create negative environmental impacts such as generating unpleasant odors, spreading disease, and contributing to global warming (Singh et al., 2011). Since the early 21st century, the disposal of organic waste in landfills has been banned by legislation in some European countries, such as in Switzerland and Sweden (EEA, 2017). This legislation is aimed at reducing the negative environmental consequences from landfilling and in promoting resource
* Corresponding author. E-mail address: [email protected] (I.M.C. Lo). https://doi.org/10.1016/j.jclepro.2017.10.164 0959-6526/© 2017 Elsevier Ltd. All rights reserved.
recovery by using alternative treatment methods. For instance, the European Environment Agency (EEA, 2013) estimated that the energy produced from organic waste could reach more than 50% of the total renewable energy generated in Europe in 2020. Regarding the composition of organic waste, sewage sludge and food waste represent the greatest portion in urban areas (Righi et al., 2013). For example, in Hong Kong, such wastes accounted for 89% of the total organic waste generated in 2015 (HKEPD, 2017), so if these wastes can be properly treated, the environmental impacts created would be minimized and useful resources would also be recovered at the same time. Hence, the identification of a cleaner and sustainable treatment method for sewage sludge and food waste is vital. In order to identify a waste treatment strategy for environmental sustainability, life cycle assessment (LCA) is widely used as it can quantify the environmental impacts of different treatment
S.L.H. Chiu, I.M.C. Lo / Journal of Cleaner Production 174 (2018) 966e976
scenarios in a scientific manner (Humbert et al., 2009). However, LCA results are often subject to uncertainty, and may lead to incorrect conclusions for the decision makers (Huijbregts et al., 2003). The International Organization for Standardization (2006) recommends that the uncertainty should be considered in order to improve the reliability of the LCA results. Regarding the types of uncertainty in LCA, Huijbregts (1998) categorized uncertainty into parameter uncertainty, model uncertainty, and scenario uncertainty while Assefa and Frostell (2004) categorized uncertainty into parameter uncertainty and model uncertainty, with both studies suggesting that parameter uncertainty is of the utmost importance. Sensitivity analysis and uncertainty propagation can be undertaken when investigating the parameter uncertainty involved in an LCA study (EC, 2010). As a number of process parameters are generally included in an LCA study, they may cause a large variation in the results due to the presence of individual parameter uncertainty. It is important to consider the parameter uncertainty involved in an LCA study in order to improve the reliability of the results. Those process parameters leading to the highest environmental gain or loss by only a small deviation are considered as key process parameters (Heijungs, 1996). The identification of the key process parameters is important as they suggest room for possible improvement of current treatment systems and the determination of the parameter significance for future study. In LCA studies, researchers generally study the influence of selected key process parameters that are regarded as sensitive or important to the results. Since the selection of the key process parameters is usually based on the knowledge or experience of the researcher (i.e., expert judgment), there could be considerable variation of the selected parameters across LCA studies (Laurent et al., 2014). For example, Kirkeby et al. (2006) and Evangelisti et al. (2014) evaluated the environmental performance of different food waste treatment technologies. In these two studies, four key parameters with potentially large impacts on the results were selected based on expert judgment, and only one parameter, the fugitive emissions of methane in the AD process, was identical in the two studies. Zhao and Deng (2014) only investigated the influence of the energy mix parameter without considering other process parameter in food waste LCA. It is generally agreed that the expert judgment based selection approach for the key process parameters may lead to overestimation or underestimation of the parameter importance, leading to wrong conclusions. In order to identify the key process parameters in consideration of all the process parameters involved in an LCA study, sensitivity analysis can be used to evaluate the influence of each process parameter to the result. The key process parameter indicates the greatest contribution to the environmental impact and hence provides a better understanding of the variability in the LCA result (Ning et al., 2013). Wolf et al. (2016) suggested that less data collection effort should be afforded for those parameters of minor importance in future studies. This approach was applied in LCA studies for wind power generation in order to identify the key process parameters when assessing the greenhouse gas (GHG) emissions (Padey et al., 2012). Regarding waste LCA studies, Eriksson and Baky (2010) identified the key process parameters for municipal solid waste management by using sensitivity analysis. Specifically for sewage sludge and food waste treatment, the key process parameters have not yet been identified in any published study. Uncertainty propagation quantifies the uncertainty of model results due to various process parameter uncertainties and aims to improve the robustness and reliability of the results by providing ranges of possible outputs. To date, very few LCA studies have applied uncertainty propagation with only 6% of waste LCA studies considered uncertainty propagation in their assessments (Xu et al.,
967
2015). The reasons can likely be attributed to the large effort required for the collection of uncertainty information on the parameters and the difficulty in determining uncertainty information for all process parameters in an LCA study (Wolf et al., 2016). Clavreul et al. (2012) proposed a general framework for uncertainty propagation by considering the key process parameters in waste LCA. Nevertheless, the selection process of key process parameters was not described in the study (i.e., how to determine the key process parameters). Determination of the key process parameters is important for possible improvement of treatment systems, as well as being used for uncertainty propagation. In order to identify a sustainable treatment method for sewage sludge and food waste, as well as quantify the parameter uncertainty involved in the LCA study, Hong Kong is chosen as a reference city in this study. Hong Kong has long been solely relying on the three strategic landfills for sewage sludge and food waste disposal. The local government introduced a new waste management policy entitled “Hong Kong Blueprint for Sustainable Use of Resources 2013e2022” (HKEB, 2013). The directions of the policy are to reduce the waste at source, as well as utilize the waste for sustainable uses such as recovering energy from waste treatment. To tackle the latter objective, the government aims to commission a couple of waste-related infrastructures for turning waste to energy, such as building a sewage sludge incineration facility (T-PARK), and organic waste treatment facilities (OWTF) for food waste treatment by anaerobic digestion (AD). In order to further raise the waste treatment capacity, the government proposes to use sewage sludge and food waste anaerobic co-digestion (coAD) in the existing sewage treatment works (STWs) (HKCEO, 2016). In the meantime, the relocation of three existing STWs is suggested to be feasible, according to the cavern development strategy in Hong Kong (CEDD, 2011). However, the treatment methods for the sewage sludge generated from the proposed cavern STWs have not yet been established, therefore, the evaluation of a sustainable waste treatment strategy is of paramount importance. With the use of AD and coAD in the future, a large amount of biogas will be produced in Hong Kong. It is a common practice to apply a combined heat and power (CHP) system for the biogas produced to generate both heat and electricity. The former can satisfy the heat load demand by the digesters and the latter can be used as a fuel source (USEPA, 2011). Meanwhile, since the heat cannot be efficiently transported over a long distance (Cromie et al., 2014) and the demand for heat in Hong Kong is limited, the heat generated from the CHP in existing STWs can only be used internally. If the heat is not fully utilized, the overall efficiency for the CHP is consequently greatly reduced (Cromie et al., 2014). As an alternative for biogas utilization, the use of a combined cycle gas turbine (CCGT) for upgraded biogas has gained more attention recently, achieving around 55% efficiency for electricity generation (Gutierrez et al., 2016). To the best of the authors’ knowledge, studies on CCGT are still sparse in regard to biogas utilization. In particular, for countries or cities with low heat demand, evaluation of CCGT for biogas utilization in generating electricity is essential. Based on the literature review, researches on identifying key process parameters and tackling the uncertainty in waste LCA are scarce while these are necessary for process improvement and to strengthen the result reliability. To fill these gaps, the major objective of this study is to identify the key process parameters and tackling the uncertainty for an LCA study for sewage sludge and food waste treatment in Hong Kong. To achieve this major objective, four specific objectives are included: (i) to determine the key environmental impact and analyze the process contributions of the various waste treatment scenarios for sewage sludge and food waste using LCA; (ii) to evaluate whether CHP or CCGT is more environmentally friendly for biogas utilization; (iii) to identify the
968
S.L.H. Chiu, I.M.C. Lo / Journal of Cleaner Production 174 (2018) 966e976
key process parameters using a proposed selection approach with sensitivity analysis; and (iv) to propagate the uncertainty of the key process parameters identified. 2. Materials and methods 2.1. Framework of the study Fig. 1 illustrates the process flow of the current study. This study evaluates the environmental performance of different treatment scenarios for sewage sludge and food waste with reference to the cavern development strategy in Hong Kong using LCA. With the LCA results, the key impact category, indicating the highest
contribution to the total environmental performance, is determined. Sensitivity analysis is then conducted in order to identify the key process parameters. In order to examine whether the selection approach by sensitivity analysis is statistically acceptable, multiple linear regression is conducted to determine if the selected key process parameters are statistically significant so as to affect the result. After that, the uncertainty information such as the probability distribution function and uncertainty range is further collected and assigned to the key process parameters. Finally, Monte Carlo simulation is conducted to propagate the parameter uncertainty involved in the study in order to obtain the probability distribution functions of different scenarios based on the environmental performance.
Fig. 1. Process flow diagram of the current study.
S.L.H. Chiu, I.M.C. Lo / Journal of Cleaner Production 174 (2018) 966e976
2.2. System boundary and description The LCA is conducted with reference to ISO 14040 and 14044 (ISO, 2006a, 2006b). Six scenarios are proposed for sewage sludge and food waste treatment, representing all possible treatment scenarios for sewage sludge and food waste in Hong Kong. Scenario 1 is regarded as the baseline scenario as it represents the current treatment methods for these wastes in Hong Kong. A schematic
969
diagram of the various scenarios' system boundaries is shown in Fig. 2, and a detailed description of different waste treatment processes is given in Table 1. A 10:3 wet weight mixing ratio is adopted for coAD treatment, and for fair comparison, the same ratio for sewage sludge and food waste is applied in each treatment scenario. The functional unit in this study is defined as 350 t/d of sewage sludge and 105 t/d of food waste based on a 10:3 wet weight mixing ratio, taking into account for the estimated amount
Fig. 2. System boundary of the treatment scenarios for treating sewage sludge and food waste in accordance to the cavern development strategy in Hong Kong.
970
S.L.H. Chiu, I.M.C. Lo / Journal of Cleaner Production 174 (2018) 966e976
Table 1 Description of the feasible waste treatment methods in this study. Feasible treatment/disposal method
Waste
Description
West New Territories (WENT) Landfill
Food waste, fly ashes and bottom ashes from T-PARK
T-PARK (Sludge incinerator)
Sewage sludge
Organic waste treatment facility (OWTF)
Food waste
Anaerobic digestion
Sewage sludge
Anaerobic co-digestion
Sewage sludge and food waste
Waste is transported to the WENT landfill whenever landfill disposal is required. The landfill gas recovery rate is with reference to Woon and Lo (2014). Uncollected landfill gas is combusted in a flaring system and flared into the atmosphere. Complete combustion is assumed in this flaring process. Leachate is collected and pumped to a landfill leachate treatment plant. A generator fuelled by landfill gas is installed to provide heat and electricity for internal use. T-PARK has been operated since 2016 and is located at Tuen Mun, Hong Kong, which is next to the WENT Landfill. The maximum treatment capacity of T-PARK is 2000 t/d of sewage sludge (with 30% solid content). The heat energy generated from the incineration process is recovered and turned into electricity that can support the needs of the entire facility. Excess electricity is exported to the public grid (HKEPD, 2009a). OWTF adopts biological technologies, which consist of AD and composting to stabilize the food waste and turn it into biogas for energy recovery and useful compost products. In OWTF, the biogas produced is combusted to produce heat and electricity by the combined heat and power (CHP) system. The digestate is used to produce compost for land applications (HKEPD, 2009b). The compost produced avoids the application of artificial fertilizers. An anaerobic digester is built for treating the sewage sludge generated from the cavern STWs. The biogas generated during the AD process is combusted to produce heat and electricity by the CHP system. The digestate is transported and treated in T-PARK. The anaerobic digester, originally for sewage sludge AD in the cavern STWs, is used for coAD of sewage sludge and food. In this study, two biogas utilization methods for electricity generation are studied. The operational data of the anaerobic co-digestion process are collected from Yongyeon Wastewater Treatment Plant in South Korea (Ejlertsson and Magnusson, 2013). The same data are used in authors' previous study (Chiu et al., 2016) for the following reasons: (1) the plant is reconstructed for anaerobic co-digestion in 2010 in which the technology is compatible with current technology; and (2) Hong Kong and South Korea have a similar Asian diet and their food characteristics are similar. A mixing ratio of 10:3 by wet weight of the sewage sludge and food waste is used in this study, based on the South Korean plant. On the one hand, the biogas generated during the coAD process is combusted to produce heat and electricity by the CHP system in scenario 5. On the other hand, the biogas is utilized by the combined cycle gas turbine (CCGT) in scenario 6 to generate electricity. Biogas is first upgraded to 90% methane content (Woon et al., 2016) and then combusted by the CCGT system, with reference to Gutierrez et al. (2016). The remaining co-digestate is then transported to the T-PARK for combustion.
of sewage sludge from the cavern STWs (DSD, 2015). In addition, it is assumed that only the environmental impacts from the operational phase of the waste treatment facilities are considered, since they are regarded as the major environmental burdens (Gentil et al., 2010), while the environmental impacts of the construction and capital equipment are not included in this study. As mentioned in Cleary (2009), these items are considered as secondary environmental burdens and are relatively insignificant compared to the primary environmental emissions from the waste treatment process. Moreover, the environmental impact due to waste haulage is assumed to be negligible due to its low contribution, as observed in the authors’ previous waste LCA study in Hong Kong (Woon et al., 2016). The GHG and air pollutant emission factors for substituted electricity are collected from China Light & Power Hong Kong Company Limited (CLP, 2015) and those for substituted heat are from the Hong Kong and China Gas Company Limited (Towngas, 2015). CHP is applied for biogas utilization in scenarios 2, 3, 4, and 5. The heat generated is assumed to be used internally in the treatment plant as the demand for heat in Hong Kong is limited. However, it should be noted that successful applications of district heating or cooling systems in order to utilize the heat from CHP could be found in many countries. In order to compare the environmental performance of CHP and CCGT for biogas utilization, scenario 6 is proposed to apply CCGT to evaluate utilization of the biogas produced from the coAD treatment plant. It should be noted that all the key processes in scenario 6 are identical to scenario 5 except the use of CCGT replacing CHP. For scenario 5, the biogas produced is combusted to generate heat and electricity by the CHP system. The electricity generated is utilized by internal consumption and the excess electricity is exported to the public grid. Generally, 25% of the heat generated is utilized for the internal heat
demand by the anaerobic digesters (Woon et al., 2016). For scenario 6, the biogas is first upgraded to biomethane, with 90% methane content, for the application of CCGT to generate electricity (Cromie et al., 2014). The biogas is upgraded and purified by water scrubbing as it is the most suitable technology for biogas upgrading (Chiu and Lo, 2016). The life cycle inventory data on substrate composition and waste treatment processes are presented in Tables A.1 e A.7 in the supplementary information. The major assumptions and limitations of the scenario system boundary and data inventory include: (i) the data are mostly collected from local sources and they may not replicate global trends; (ii) only the environmental emissions from the operational phases are considered while the environmental emissions from construction and capital equipment are excluded as they are considered insignificant compared to the emissions from the operational phases (Cleary, 2009: Gentil et al., 2010); (iii) waste haulage is assumed to be negligible (Woon et al., 2016); (iv) heat produced from CHP is assumed to be used internally; and (v) the mixing ratio of sewage sludge and food waste in coAD is applied from a Korean plant data, and additional experiments are needed to determine the suitable mixing ratio of the substrates in Hong Kong. 2.3. Life cycle impact assessment SimaPro 7.2.4 software with ReCiPe version 1.04 (Goedkoop et al., 2009) is used for life cycle impact assessment. Four environmental impact categories are chosen as they are most commonly studied with reference to waste LCA studies, namely climate change, particulate matter formation, photochemical oxidant formation, and terrestrial acidification (Bernstad and la Cour Jansen, 2012). According to the ReCiPe method, climate change, photochemical oxidant formation and particulate matter
S.L.H. Chiu, I.M.C. Lo / Journal of Cleaner Production 174 (2018) 966e976
formation are categorized under the human health endpoint damage category by quantifying the midpoint impact to disabilityadjusted life years. Climate change and terrestrial acidification are categorized under the ecosystems endpoint damage category by quantifying the midpoint impact to the loss of species during a year. Among the midpoint impact categories studied, the midpoint impact with the highest contribution to the endpoint damage is identified as the key impact category. The key impact category is then used as a reference result for sensitivity analysis so as to identify the key process parameters. 2.4. Sensitivity analysis Sensitivity analysis is conducted for all process parameters in different scenarios based on the key impact category identified. All the process parameters involved in this study are presented in Table A.8 in the supplementary information. Sensitivity analysis can be classified into local sensitivity analysis (LSA) and global sensitivity analysis (GSA). LSA considers the output variability by changing the input parameter within a specific range. It is particularly useful in providing early approximation to identify the key parameters (Hemsath and Bandhosseini, 2015). Meanwhile, GSA differs from LSA in that all model parameters vary simultaneously when determining the parameter's sensitivity to overcoming the limitation of the linearity issue of LSA, and it requires specifying the input variability. Although GSA has its own merits in providing an overview of the influence of the inputs on the outputs, the conclusions from GSA should be drawn with care when the input variability is poorly known (Iooss and Lemaître, 2015). In addition, they are usually computationally expensive (Cariboni et al., 2007), and the choice of the method to use is driven by the computing cost in general (Saltelli et al., 2006). In the current study, LSA is chosen for sensitivity analysis so as to identify the key process parameters prior to uncertainty propagation. All the process parameters are assumed to be independent as their interactions are limited since they are mostly obtained from individual process. Sensitivity analysis is undertaken by varying one parameter at a time, while keeping the other parameters constant in order to determine the sensitivity of the parameters to variation in the input data variables. This is known as the one-at-a-time method (Morio, 2011). The minimum and maximum values of a parameter, which are considered as the amplitude of sensitivity analysis to reflect the degree of influence of the parameters on the assessment result (Boldrin et al., 2011), are collected from literature. By varying the process parameters with both minimum and maximum values, two sensitivity ratio (SR) values can be calculated, in which the higher SR value is chosen for the process parameter in order to reflect its sensitivity. To assess the influence of process parameters on the result, the SR is calculated for each process parameter using Eq. (1) (Clavreul et al., 2012). Since the SR can be a negative value for an avoided environmental impact result, the absolute value is applied for the SR with the aim of determining the sensitivity of the process parameter. A higher SR value indicates higher parameter sensitivity.
Final result Initial result Initial result SR ¼ Min or max process parameter Initial process parameter
971
solely determined by an exact value of SR, since there is no standard or reference for SR values. In addition, the SR value of individual process parameters can be different in each study due to the variation of the system boundaries. A more practical way to determine the key process parameters is by selecting parameters with high SR values. In this study, a selection approach based on the relative importance of the SR value is used to identify the key process parameters. The relative importance of the SR indicates the level of influence of the parameters to the result, representing the rank of the importance. It can be calculated by dividing an individual SR value with the total SR value of all the process parameters. The cumulative relative importance of SR can be obtained by summing up the relative importance of SR starting from the highest SR value. Once the cumulative relative importance of SR exceeds 50%, those process parameters used for the summation are identified as key process parameters. This selection process indicates that the selected parameters have contributed over half of the sensitivity among all the process parameters, while other parameters can be considered as trivial in regard to the result. In order to examine whether the proposed selection approach for the identification of key process parameters is reasonable, multiple linear regression is further conducted on the best scenario obtained based on the key impact category to investigate the significance of each process parameter to affect the result. Each process parameter in the best scenario is randomly varied within their possible ranges simultaneously with the degree of freedom of 30. It is noted that when the degree of freedom reaches 30, the t-distribution and the standard normal distribution are similar and the test result is regarded as reliable with sufficient number of trials (Anderson, 2013). The p-values obtained are then examined and those parameters with p-value lower than 0.05 are treated as significant according to the 95% confidence level, implying a statistically significant relationship with the result. The results from multiple linear regression can determine whether the key process parameters are statistically significant so as to affect the result, and therefore can support whether the proposed approach is statistically acceptable. Detailed calculation equations of the relative importance and cumulative relative importance of SR, and multiple linear regression are provided in Eq. A.1 and Eq. A.2 in the supplementary information. 2.6. Uncertainty propagation Monte Carlo simulation method is used to propagate the uncertainty of the selected key process parameters, which are represented by a stochastic variable with a probability distribution. By selecting a random sample from the probability distribution of the process parameters, Monte Carlo simulation calculates the impact results by repeating for 10,000 iterations. It has been built into commercial LCA software, and SimaPro 7.2.4 software is used to conduct the Monte Carlo simulation in the current study. The uncertainty information, which represents the input variability of each key process parameter, is then determined for the Monte Carlo simulation. The uncertainty information including probability distribution and uncertainty range are collected from the literature with detailed justifications provided in section 3.2.
(1)
Initial process parameter
3. Results and discussion 3.1. Process contributions to the key impact category
2.5. Identification of key process parameters Although the SR reflects the sensitivity of the process parameter to the result, the selection of key process parameters cannot be
As mentioned in Section 2.3, four midpoint impact categories are evaluated in this study and their relative contributions to the endpoint damage categories are presented in Table 2. Detailed results regarding the midpoint impact categories are presented in
972
S.L.H. Chiu, I.M.C. Lo / Journal of Cleaner Production 174 (2018) 966e976
Table 2 Midpoint impact category contribution to endpoint damage category in each scenario. Impact category
Unit
Scenario 1
Scenario 2
Scenario 3
Scenario 4
Scenario 5
Scenario 6
Climate change human health Photochemical oxidant formation Particulate matter formation Human healtha Climate change ecosystems Terrestrial acidification Ecosystemsa
DALYb DALY DALY DALY species-year species-year species-year
0.153 2.45 10-6 3.58 10-3 0.156 8.65 10-4 7.30 10-8 8.65 10e4
0.102 1.70 10-6 7.50 10-3 0.109 5.77 10-4 1.52 10-7 5.77 10e4
0.0176 1.23 10-7 4.06 10-4 0.0172 9.88 10-5 2.97 10-7 9.85 10e5
0.0334c 8.65 10-7 3.50 10-3 ¡0.0299 1.90 10-4 3.76 10-7 ¡1.90 10e4
0.0529 1.75 10-6 4.97 10-3 ¡0.0479 3.00 10-4 5.31 10-7 ¡3.01 10e4
0.0946 3.77 106 3.88 104 ¡0.0942 5.36 104 8.43 107 ¡5.37 104
a b c
Endpoint damage categories. DALY stands for Disability-adjusted life year. Negative value represents a positive impact to the environment.
Fig. 3. Process contributions by different scenarios to climate change impact category.
Table B.1 in the supplementary information. Regarding the endpoint damage categories for human health and ecosystems, scenarios 4, 5, and 6 have an overall positive environmental performance on human health and ecosystems in general, as a result of avoided emissions. It indicates the substitution effect of avoiding direct emissions from energy production using fossil fuels. Scenario 6 performs the best in both endpoint categories, with 52.3% more avoided impacts in human health, and 43.9% more avoided impacts in ecosystems, compared with scenario 5, which is the second best scenario. It indicates a better environmental performance is achieved with the use of CCGT over CHP for biogas utilization. Based on the contribution from the midpoint impact categories in Table 2, climate change has the highest impact in both endpoint damage categories and it is identified as the key impact category. Fig. 3 illustrates the environmental impacts related to climate change. A positive figure represents the emissions of waste treatment being greater than the avoided impacts resulting from resource utilization, and vice versa for a negative figure. Fig. 3 shows that scenarios 5 and 6 are the best two scenarios with 3.78 104 kg CO2e and 6.75 104 kg CO2e per functional unit, respectively. The avoided impacts are mainly contributed by the coAD process, bringing environmental benefits compared to those scenarios without this process. Apart from the coAD process in scenarios 5 and 6, the AD processes in scenarios 2, 3, and 4 also result in negative values, which indicates the avoided impacts are greater than the direct impacts of the process. In this study, a default value of 5% for fugitive CH4 emissions (IPCC, 2006) is applied for AD and coAD. The avoided impacts are contributed by the energy recovery system in generating heat and electricity, which considerably reduce the negative implications to the
environment. Meanwhile, it is interesting to note that the avoided impacts from scenario 4, which involve AD treatment for sewage sludge and food waste separately, exhibit similar reductions to scenario 5. This is because the compost produced from the food waste digestate is used as artificial fertilizer. For direct impacts, both the landfilling of food waste and incineration of sewage sludge or digestate at T-PARK create direct impacts on climate change. The landfill gas cannot be completely collected for the energy recovery system and flaring process. As a result, the uncollected landfill gas diffuses from the surface of the landfill and CH4 is released as a GHG to the atmosphere, resulting in climate change impacts. Regarding the sludge incineration in T-PARK, the most significant contribution during the incineration process is the N2O emission from the combustion of sewage sludge due to the high nitrogen content (Chiu et al., 2016). Since N2O is a potent GHG with a global warming potential of 298 (Forster et al., 2007), the incineration of sewage sludge makes a large contribution to climate change. Last but not least, the environmental impacts brought by sludge dewatering and food waste pre-treatment are relatively insignificant compared to other major processes.
3.2. Identification of key process parameters by the proposed selection approach Considering climate change impact as the key impact category, sensitivity analysis is then conducted for all the process parameters in order to calculate their respective SR values and relative importance. As mentioned in Section 2.5 in the methodology, when the cumulative relative importance of the SR value exceeds 50%, those process parameters that are used for the summation are
S.L.H. Chiu, I.M.C. Lo / Journal of Cleaner Production 174 (2018) 966e976 Table 3 Process parameters required to get the cumulative importance of SR value over 50%. Key process parameters
Scenario 1 Lower heating value of sewage sludge Energy recovery efficiency (sewage sludge incineration plant) Scenario 2 CHP efficiency (electricity) (AD/ coAD plant) Lower heating value of methane Percentage of methane in biogas Scenario 3 Lower heating value of sewage sludge Energy recovery efficiency (sewage sludge incineration plant) CHP efficiency (electricity) (AD/ coAD plant) Scenario 4 Lower heating value of methane CHP efficiency (electricity) (AD/ coAD plant) Percentage of methane in biogas Scenario 5 Lower heating value of methane CHP efficiency (electricity) (AD/ coAD plant) Energy recovery efficiency (sewage sludge incineration plant) Scenario 6 CCGT efficiency Lower heating value of methane Percentage of methane in biogas
Relative importance of SR value (%)
Cumulative relative importance of SR (%)
33.8
67.6
33.8
19.3
50.3
19.3 12.5 21.5
56.0
20.9 13.6
24.0 23.1
62.2
15.1 21.6 20.5
56.2
14.1
21.1 20.7 18.9
60.7
identified as key process parameters and are presented in Table 3. For detailed values of SR of all the process parameters, they can be found in in Table B.2 in the supplementary information. Multiple regression analysis is further conducted on scenario 6, which is the best scenario obtained based on the key impact category. The regression result is presented in Table B.3 and parameters of the lower heating value of digestate, energy recovery efficiency of sewage sludge incineration plant, biogas production rate, percentage of methane in biogas, lower heating value of methane, CCGT efficiency, and energy consumption for biogas upgrade attained p-values lower than 0.05, which implies that these parameters are significant to the result according to 95% confidence level. It is found that the key process parameters identified by the proposed selection approach are also regarded as significant parameters based on the multiple linear regression result. It indicates the proposed approach allows statistically acceptable selection of the key process parameters. Regarding the key process parameters identified, the percentage of methane in the biogas and the biogas production rate from the food waste depend on the characteristics of the waste. These parameters can be improved by applying appropriate pre-treatment technologies to the organic wastes, such as mechanical and thermal pre-treatment (Chiu and Lo, 2016). In order to maintain a good performance of the energy recovery efficiency of the sewage sludge incineration plant, CHP and CCGT efficiencies for electricity recovery, these energy generation facilities need to be frequently cleaned and maintained (DEFRA, 2013). The lower heating value of methane has relatively less fluctuation. The identification of the key process parameters is beneficial to future LCA studies of organic waste and extra efforts should be made to collect more reliable data for these key parameters in order to provide a more accurate result.
973
The uncertainty information including the probability distribution function and the uncertainty range are then determined for these key process parameters for uncertainty propagation as given in Table 4. The determination of the probability distribution function for process parameters can be classified into two groups. For the first group, the distribution function can be determined by statistical methods if extensive data are available (Huijbregts et al., 2003). However, it is generally impossible to differentiate distribution functions when sample sizes are very small (Penman, 2001). For the other group, an appropriate distribution can be selected by expert estimation when little information is available (Sonnemann et al., 2003). In the current study, the distribution function of the parameters is based on the expert judgment approach due to insufficient data. In order to determine the distribution function for the key process parameters selected, the relevant literature has been reviewed. For the energy recovery efficiency of the sewage sludge incineration plant, CHP and CCGT efficiencies for electricity of the AD or coAD plant, a lognormal distribution is suggested according to Galvez-Martos and Schoenberger (2014), as only positive values are possible. For the heating value of sewage sludge and methane, a lognormal distribution is used since it fits the dataset based on national inventory reports, IEA data and available national data for fuel (IPCC, 2006). A normal distribution is assigned to the percentage of methane in biogas and the biogas production rate of food waste with reference to Clavreul et al. (2012). The probability distribution function for empirical data is normal or lognormal in the current study, in line with Penman (2001). A 95% confidence interval is applied in the study and this means that there is a 95% probability that the confidence interval will contain the true population mean, indicating the range that the parameter would frequently fall into in most of the situations. For lognormal distribution, the 95% confidence interval is defined by dividing or multiplying the mean value with the squared geometric standard deviation. For normal distribution, the 95% confidence interval is approximately defined by subtracting twice the standard deviation and adding twice the standard deviation (Goedkoop et al., 2016). Three levels of the uncertainty range are proposed in order to reflect the parameter uncertainty: 50%, 25%, and 10% for high, medium, and low uncertainty of the key process parameters, respectively. The uncertainty of the lower heating value of sewage sludge is high since data on biomass as fuel are not well developed (IPCC, 2001). The percentage of methane in biogas and the biogas production rate from food waste are assumed to have medium uncertainty, as they are dependent on the characteristics of the waste. The energy recovery efficiency of the sewage sludge incineration plant, CHP and CCGT efficiencies for electricity of AD or coAD plant, and lower heating value of methane are subject to lower uncertainty as the technology is well developed and there is sufficient data to support using the lower heating value of methane. 3.3. Results with uncertainty propagation Fig. 4 illustrates the result of the probability distribution function and cumulative probability distributions of scenarios 1 to 5 under uncertainty. The probability distribution function graph aims to show the distribution of climate change values whereas the cumulative probability graph allows the identification of percentiles corresponding to the climate change impact presented on the horizontal axis. The statistical indices (i.e., mean, standard deviation, and possible range in the 95% confidence interval) of each scenario are presented in Table B.4. In Fig. 4a, the results show that scenarios 1, 2, and 3, the probability of having a higher climate change impact is notably greater than that of scenarios 4 and 5. Comparing scenarios 4 and 5, scenario 5 attains a better
974
S.L.H. Chiu, I.M.C. Lo / Journal of Cleaner Production 174 (2018) 966e976
Table 4 Probability distribution and uncertainty range of the key process parameters. a
Key process parameter
Unit
Probability distribution
Mean
Uncertainty range
GSD or SD
Energy recovery efficiency (Sewage sludge incineration plant) CHP efficiency for electricity (AD/coAD plant) Lower heating value of sewage sludge Lower heating value of methane Percentage of methane in biogas CCGT efficiency
%
Lognormal
20
±10%
1.05
Galvez-Martos and Schoenberger, 2014
%
Lognormal
36
±10%
1.05
Galvez-Martos and Schoenberger, 2014
MJ/kg dry basis
Lognormal
10
±50%
1.22
IPCC, 2006
kWh/m3
Lognormal
10
±10%
1.05
IPCC, 2006
%
Normal
60
±25%
7.5
Clavreul et al., 2012
%
Lognormal
55
±10%
1.05
Galvez-Martos and Schoenberger, 2014
a
References for probability distribution
GSD stands for geometric standard deviation, which is used for lognormal distribution. SD stands for standard deviation, which is used for normal distribution.
Fig. 4. (a) Probability distribution function in climate change impact; and (b) cumulative probability distribution in climate change impact of scenarios 1, 2, 3, 4, and 5.
performance in general. Yet, it might be possible that the climate change impact of scenario 4 is “more negative” than scenario 5 on some occasions. By referring to the cumulative probability graph in Fig. 4b, scenario 5 should at least avoid 2.56 104 kg CO2e with a 95% confidence interval, whilst the cumulative probability of scenario 4 in achieving this reduction is 31%. Based on the cumulative probability curve, the climate change performance of scenario 5 is always better than scenario 4 in most situations. The environmental performance of CHP and CCGT for biogas utilization is further evaluated and compared by examining scenarios 5 and 6. In both scenarios, the same amount of biogas is consumed, except that the biogas is required to be first upgraded to biomethane for the CCGT system in scenario 6. Sensitivity analysis is applied to scenario 6 after the assessment to determine its climate change impact. Fig. 5a shows the results of the probability distribution function and Fig. 5b presents the results of the cumulative probability distributions corresponding to the climate change impact of scenarios 5 and 6 under uncertainty. From the results, it is notable that scenario 6 attains better environmental
performance on climate change than that of scenario 5. The probability of scenario 6 achieving at least 5.32 104 kg CO2e avoided emissions is 95%, whilst it is below 5% for scenario 5, indicating a significant advantage of CCGT over CHP. Given that the CCGT system does not generate heat for internal consumption in the current study, and external heat is required for the anaerobic digesters, the avoided environmental impact is still greater in the CCGT scenario due to the higher efficiency in generating electricity. Particularly in the Hong Kong situation, the production of electricity is carbonintensive as coal contributes approximately 50% of the current fuel mix, whilst the heat generated by city gas is produced via naphtha and natural gas, serving as a cleaner fuel in terms of greenhouse gas emissions.
4. Conclusions This study applies LCA to evaluate the environmental performance of different treatment scenarios for sewage sludge and food waste with reference to the cavern development strategy in Hong
S.L.H. Chiu, I.M.C. Lo / Journal of Cleaner Production 174 (2018) 966e976
975
Fig. 5. (a) Probability distribution function in climate change impact; and (b) cumulative probability distribution in climate change impact of scenarios 5 and 6.
Kong. In addition, the key process parameters are identified and the uncertainty in this LCA is also tackled. Regarding the environmental performance of different treatment systems, the endpoint damage results indicate that scenarios 5 and 6, which involve coAD as the major treatment method, are more advantageous over other possible treatment methods for these wastes. Among the midpoint impact categories studied, climate change is the key impact category with the highest contribution to environmental performance. It is found that coAD treatment provides the greatest environmental benefits while landfilling results in the greatest environmental burdens in the waste treatment scenarios. The climate change impact category then serves as the reference result for sensitivity analysis in the proposed selection approach for the identification of the key process parameters. The parameters of electricity efficiency in different waste treatment facilities are the common key process parameters influencing the environmental performance, and therefore more attention should be paid on these process parameters in order to improve the environmental performance for cleaner production. Multiple linear regression is further conducted in order to justify whether the proposed methodology for key process parameters identification is statistically acceptable. The results show that the key process parameters identified from the proposed methodology are also regarded as significant parameters from the multiple linear regression result, at the 95% confidence level, indicating the parameter selection by the proposed approach is statistically acceptable. After uncertainty propagation, it is observed that scenarios 5 and 6 with coAD treatment perform better than other waste treatment scenarios in general. Regarding the use of CHP and CCGT for biogas utilization, scenario 6 has a 95% probability of achieving at least 5.32 104 kg CO2e avoided emissions while the probability of the second best scenario (i.e., scenario 5) achieving the same avoided emissions is below 5%, indicating a significant advantage of CCGT over CHP for biogas utilization in region with low heat demand, like Hong Kong. Based on the results from this study, it is concluded that coAD with CCGT application is the best treatment option for sewage sludge and food waste in accordance to the cavern development strategy
in Hong Kong. This study demonstrates an approach to identify the key process parameters and tackle the uncertainty in LCA. The identification of key process parameters is critical for environmental sustainability as these parameters are considered as hotspots with high influence to the environmental results. Besides, tackling the LCA uncertainty is also necessary as it provides stochastic result rather than deterministic result, hence the decision-makers can determine the likeliness to achieve particular environmental performance, especially when there are other selection criteria for consideration. Last but not least, this study only considers the environmental results as the performance indicator while modern waste treatment should also consider other important performance indicators such as life cycle costing and life cycle energy efficiency. They should be further studied with the consideration of their uncertainty involved by using the approach demonstrated in this study. Appendix A. Supplementary data Supplementary data related to this article can be found at https://doi.org/10.1016/j.jclepro.2017.10.164. References Anderson, A., 2013. Business Statistics for Dummies. John Wiley & Sons, United States. Assefa, G., Frostell, B., 2004. Social impact assessment within life cycle technology assessment. In: 4th SETAC World Congress. http://urn.kb.se/resolve?urn¼urn: nbn:se:kth:diva-88625 (assessed 07.08.17). Bernstad, A., la Cour Jansen, J., 2012. Review of comparative LCAs of food waste management systems e current status and potential improvements. Waste Manage 32 (12), 2439e2455. Boldrin, A., Neidel, T.L., Damgaard, A., Bhander, G.S., Møller, J., Christensen, T.H., 2011. Modelling of environmental impacts from biological treatment of organic municipal waste in EASEWASTE. Waste Manage 31 (4), 619e630. Cariboni, J., Gatelli, D., Liska, R., Saltelli, A., 2007. The role of sensitivity analysis in ecological modelling. Ecol. Model. 203, 167e182. CEDD, 2011. Enhancing Land Supply Strategy: Reclamation outside Victoria Harbour and Rock Cavern Development. Civil Engineering Development Department, Hong Kong. Chiu, S.L.H., Lo, I.M.C., 2016. Reviewing the anaerobic digestion and co-digestion
976
S.L.H. Chiu, I.M.C. Lo / Journal of Cleaner Production 174 (2018) 966e976
process of food waste from the perspectives on biogas production performance and environmental impacts. Environ. Sci. Technol. 23 (24), 24435e24450. Chiu, S.L.H., Lo, I.M.C., Woon, K.S., Yan, D.Y.S., 2016. Life cycle assessment of waste treatment strategy for sewage sludge and food waste in Macau: perspectives on environmental and energy production performance. Int. J. Life Cycle Assess. 21 (2), 176e189. Clavreul, J., Guyonnet, D., Christensen, T.H., 2012. Quantifying uncertainty in LCAmodelling of waste management systems. Waste Manage 32 (12), 2482e2495. Cleary, J., 2009. Life cycle assessments of municipal solid waste management systems: a comparative analysis of selected peer-reviewed literature. Environ. Int. 35, 1256e1266. CLP, 2015. Sustainability Report 2015. China Light & Power Company, Hong Kong. https://www.clpgroup.com/en/Sustainability-site/Report%20Archive%20% 20Year%20Document/SR_Full_2015_en.pdf (assessed 07.08.17). Cromie, T., Brown, J., Watters, P., Foley, A., Smyth, B., Rooney, D., Glover, S., Orozco, A., Groom, E., Irvine, C., 2014. Alternative Uses of Biogas Group Report. A Northern Ireland case study. Appendix 3.2, Biogas action plan for Northern Ireland. Northern Ireland. DEFRA, 2013. Incineration of Municipal Solid Waste. Department for Environment Food & Rural Affairs, United Kingdom. DSD, 2015. Sewage Treatment Works Information Sheet. Drainage Services Department, Hong Kong. http://www.dsd.gov.hk/EN/Publicity_and_ Publications/Publicity/Leaflets__Booklets___Fact_Sheets/index.html (Accessed 04.11.16). EC, 2010. International Reference Life Cycle Data System (ILCD) Handbook e General Guide for Life Cycle Assessment e Detailed Guidance, first ed. European Commission - Joint Research Centre - Institute for Environment and Sustainability, Luxemburg. EUR 24708. EN. EEA, 2013. EU Bioenergy Potential from a Resource-efficiency Perspective. Report no. 6/2013. European Environment Agency. EEA, 2017. Municipal Waste Management across European Countries. European Environment Agency. https://www.eea.europa.eu/themes/waste/municipalwaste/municipal-waste-management-across-european-countries (Accessed 03.08.17). Ejlertsson, J., Magnusson, B., 2013. Maximizing Biogas Production at a South Korean Biogas Plant, Scandinavian Biogas (South Korea). Eriksson, O., Baky, A., 2010. Identification and testing of potential key parameters in system analysis of municipal solid waste management. Resour. Conserv. Recy 54 (12), 1095e1099. Evangelisti, S., Lettieri, P., Borello, D., Clift, R., 2014. Life cycle assessment of energy from waste via anaerobic digestion: a UK case study. Waste Manage 34 (1), 226e237. Forster, P., Ramaswamy, V., Artaxo, P., Berntsen, T., Betts, R., Fahey, D.W., Haywood, J., Lean, J., Lowe, D.C., Myhre, G., Nganga, J., Prinn, R., Raga, G., Schulz, M., Van Dorland, R., 2007. Changes in atmospheric constituents and in radiative forcing. In: Climate Change 2007: the Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Gentil, E.C., Damgaard, A., Hauschild, M., Finnveden, G., Eriksson, O., Thorneloe, S., Kaplan, P.O., Barlaz, M., Muller, O., Matsui, Y., Ii, R., 2010. Models for waste life cycle assessment: review of technical assumptions. Waste Manage 30 (12), 2636e2648. Galvez-Martos, J.L., Schoenberger, H., 2014. An analysis of the use of life cycle assessment for waste co-incineration in cement kilns. Resour. Conserv. Recy 86, 118e131. Goedkoop, M.J., Heijungs, R., Huijbregts, M., De Schryver, A., Struijs, J., Van Zelm, R., 2009. ReCiPe 2008da Life Cycle Impact Assessment Method Which Comprises Harmonised Category Indicators at the Midpoint and the Endpoint Level, first ed. (Report I: Characterisation, First edition. Netherlands). Goedkoop, M.J., Oele, M., Leijting, J., Ponsioen, T., Meijer, E., 2016. Introduction to Consultant, Netherlands. LCA with SimaPro. PRe Gutierrez, E.C., Xia, A., Murphy, J.D., 2016. Can slurry biogas systems be cost effective without subsidy in Mexico? Renew. Energ 95, 22e30. Heijungs, R., 1996. Identification of key issues for further investigation in improving the reliability of life-cycle assessments. J. Clean. Prod. 4 (3e4), 159e166. Hemsath, T.L., Bandhosseini, K.A., 2015. Sensitivity analysis evaluating basic building geometry's effect on energy use. Renew. Energy 76, 526e538. HKCEO, 2016. Policy Address of Hong Kong. Hong Kong Office of the Chief Executive. http://www.policyaddress.gov.hk/2016/eng/index.html (Accessed 04.11.16). HKEB, 2013. Hong Kong Blueprint for Sustainable Use of Resources 2013-2022. Hong Kong Environmental Bureau, Hong Kong. HKEPD, 2009a. Sludge Treatment Facilities Environmental Impact Assessment Report. Hong Kong Environmental Protection Department, Hong Kong. HKEPD, 2009b. Organic Waste Treatment Facilities Phase 1 Environmental Impact
Assessment Report. Hong Kong Environmental Protection Department, Hong Kong. HKEPD, 2017. Monitoring of Solid Waste in Hong Kong Waste Statistics for 2015 (Hong Kong). Huijbregts, M.A., 1998. Application of uncertainty and variability in LCA. Int. J. Life Cycle Assess. 3 (5), 273e280. Huijbregts, M.A., Gilijamse, W., Ragas, A.M., Reijnders, L., 2003. Evaluating uncertainty in environmental life-cycle assessment. A case study comparing two insulation options for a Dutch one-family dwelling. Environ. Sci. Technol. 37 (11), 2600e2608. Humbert, S., Loerincik, Y., Rossi, V., Margni, M., Jolliet, O., 2009. Life cycle assessment of spray dried soluble coffee and comparison with alternatives (drip filter and capsule espresso). J. Clean. Prod. 17 (15), 1351e1358. Iooss, B., Lemaître, P., 2015. A review on global sensitivity analysis methods. In: Uncertainty Management in Simulation-optimization of Complex Systems (Pp. 101e122). Springer, US. IPCC, 2001. Good Practice Guidance and Uncertainty Management in National Greenhouse Gas Inventories. Intergovernmental Panel on Climate Change, Switzerland. IPCC, 2006. IPCC Guidelines for National Greenhouse Gases Inventories. Intergovernmental Panel on Climate Change, Switzerland. ISO 14040, 2006a. Environmental Management e Life Cycle Assessment e Life Cycle Assessment Principles and Framework. International Standards Organization, Geneva, Switzerland. ISO 14044, 2006b. Environmental Management e Life Cycle Assessment e Requirements and Guidelines. International Standards Organization, Geneva, Switzerland. Kirkeby, J.T., Birgisdottir, H., Hansen, T.L., Christensen, T.H., Bhander, G.S., Hauschild, M., 2006. Evaluation of environmental impacts from municipal solid waste management in the municipality of Aarhus, Denmark (EASEWASTE). Waste Manage. Res. 24 (1), 16e26. Laurent, A., Clavreul, J., Bernstad, A., Bakas, I., Niero, M., Gentil, E., Hauschild, M.Z., 2014. Review of LCA studies of solid waste management systemsePart II: methodological guidance for a better practice. Waste Manage 34 (3), 589e606. Morio, J., 2011. Global and local sensitivity analysis methods for a physical system. Eur. Phys. J. 32 (6), 1577. Ning, S.K., Chang, N.B., Hung, M.C., 2013. Comparative streamlined life cycle assessment for two types of municipal solid waste incinerator. J. Clean. Prod. 53, 56e66. Padey, P., Blanc, I., Le Boulch, D., Xiusheng, Z., 2012. A simplified life cycle approach for assessing greenhouse gas emissions of wind electricity. J. Ind. Ecol. 16 (s1), S28eS38. Penman, J., 2001. Good Practice Guidance and Uncertainty Management in National Greenhouse Gas Inventories. Intergovernmental Panel on Climate Change, Switzerland. Righi, S., Oliviero, L., Pedrini, M., Buscaroli, A., Della Casa, C., 2013. Life cycle assessment of management systems for sewage sludge and food waste: centralized and decentralized approaches. J. Clean. Prod. 44 (0), 8e17. Saltelli, A., Ratto, M., Tarantola, S., Campolongo, F., 2006. Sensitivity analysis practices: strategies for model-based inference. Reliab. Eng. Syst. Safe 91, 1109e1125. Singh, R.P., Singh, P., Araujo, A.S., Ibrahim, M.H., Sulaiman, O., 2011. Management of urban solid waste: vermicomposting a sustainable option. Resour. Conserv. Recy 55 (7), 719e729. Sonnemann, G.W., Schuhmacher, M., Castells, F., 2003. Uncertainty assessment by a Monte Carlo simulation in a life cycle inventory of electricity produced by a waste incinerator. J. Clean. Prod. 11 (3), 279e292. Towngas, 2015. The Hong Kong and China Gas Company Limited. Sustainability Report. Hong Kong. https://www.towngas.com/getmedia/73b36b28-8281476d-ac00-8d8ea6bd93cc/Sustainability-Summary-Report-2015.pdf.aspx? ext¼.pdf (assessed 07.08.17). USEPA, 2011. Protocol for Quantifying and Reporting the Performance of Anaerobic Digestion Systems for Livestock Manures. United States Environmental Protection Agency AgSTAR Program, USA. Wolf, P., Groen, E.A., Berg, W., Prochnow, A., Bokkers, E.A., Heijungs, R., de Boer, I.J., 2016. Assessing greenhouse gas emissions of milk production: which parameters are essential? Int. J. Life Cycle Assess. 22 (3), 441e455. Woon, K.S., Lo, I.M.C., Chiu, S.L.H., Yan, D.Y.S., 2016. Environmental assessment of food waste valorization in producing biogas for various types of energy use based on LCA approach. Waste Manage 50, 290e299. Xu, C., Shi, W., Hong, J., Zhang, F., Chen, W., 2015. Life cycle assessment of food waste-based biogas generation. Renew. Sust. Energ. Rev. 49, 169e177. Zhao, Y., Deng, W., 2014. Environmental impacts of different food waste resource technologies and the effects of energy mix. Resour. Conserv. Recy 92, 214e221.