Quantification of urban metabolism through coupling with the life cycle assessment framework: concept development and case study

Cities now consume resources and produce waste in amounts that are incommensurate with the populations they contain. Quantifying and benchmarking the environmental impacts of cities is essential if urbanization of the world’s growing population is to occur sustainably. Urban metabolism (UM) is a promising assessment form in that it provides the annual sum material and energy inputs, and the resultant emissions of the emergent infrastructural needs of a city’s sociotechnical subsystems. By fusing UM and life cycle assessment (UM–LCA) this study advances the ability to quantify environmental impacts of cities by modeling pressures embedded in the flows upstream (entering) and downstream (leaving) of the actual urban systems studied, and by introducing an advanced suite of indicators. Applied to five global cities, the developed UM–LCA model provided enhanced quantification of mass and energy flows through cities over earlier UM methods. The hybrid model approach also enabled the dominant sources of a city’s different environmental footprints to be identified, making UM–LCA a novel and potentially powerful tool for policy makers in developing and monitoring urban development policies. Combining outputs with socioeconomic data hinted at how these forces influenced the footprints of the case cities, with wealthier ones more associated with personal consumption related impacts and poorer ones more affected by local burdens from archaic infrastructure.


Introduction
The contribution of cities to a number of global environmental pressures such as climate change, water stress, biodiversity loss and resource scarcity is recognized to be strong [1]. These pressures will likely increase vastly, if future urban consumptive patterns follow current trends, as the percentage of urban dwellers worldwide is predicted to swell from 50% currently to 70% of total population by 2050 [2]. Much of the rural-urban migration will occur in developing economies [3], in turn compounding the negative environmental implications with a predicted large growth in economies, generally higher standards of living and increased consumption [4]. Considering that humanity's current consumptive habits already surpass the planet's carrying capacity [5,6] it is paramount to quantify the contribution of urban areas, both mature and growing, in order to support appropriate policy interventions.
A range of methodologies designed to assess urban sustainability exist [7], however urban metabolism (UM) is one of a limited number of these to actively pursue the quantification of a city's environmental burden (another equally promising field being urban ecology [68]). UM refers to a broad range of quantitative methods that attempt to conceptualize urban areas as organisms, requiring goods and energy to maintain functionality and support growth, while emitting waste as a byproduct [8]. The framework is powerful in that it allows for an assessment of the material and energy requirements of a city's infrastructure that emerges from the sociotechnical subsystems (technical, cultural, institutional, economic, etc), even if the our current understanding of these subsystems and their interactions are murky [146]. In the last nearly 50 years the number of UM studies has been modest. A recent review by Kennedy cites at least 75 studies that implicitly or explicitly fall within UM's realm [10] with UM being applied at numerous different scales, from higher spatial resolution (neighborhoods [11]) to lower spatial resolution (cities [12] and city regions [13]). There has been little standardization of the tool, which remains a conceptual approach with large variations between studies regarding the materials, energy sources and pollutants included in the individual assessments. Furthermore, two main UM schools have arisen, one utilizing material flow accounting (MFA), the other non-mass based [14].
Developed shortly after UM-G1, UM-G2 is characterized by its attempts to move beyond mass. It interprets environmental loading predominantly using the emergy (embodied energy) concept [46][47][48][49][50][51][52][53][54][55][56][57] or on occasion, the ecological footprint (EF) method [58][59][60], and is usually performed under the urban ecology umbrella. UM-G2 addresses the shortcomings of UM-G1 in that both the emergy and EF metrics attempt to account for embedded environmental impacts of metabolic flows and the assimilation of some waste flows from cities [49,61]. Moreover, urban ecology (not EF) studies open the 'black box' and attempt to map the interconnections of urban subsystems. Despite these advances the UM-G2 methodologies still have gaps that hinder their ability to fully quantify the environmental effects of cities, in that (i) the ability of emergy and EF to adequately represent all relevant environmental impacts of all flows is limited, and their conversion factors are disputed [14,16] and (ii) the complexity of the emergy concept inhibits the communication of its results to policy makers, and consequently, its practical application [14].

The need for third generation UM
Though earlier UM methodologies provided a foundational framework for measuring the environmental impacts of cities, practitioners widely agree that current methodological faults need to be addressed [14,61,62]. Recently, Pincetl and colleagues [14,63,146] have suggested coupling UM with the life cycle assessment (LCA) framework to help mature the field of urban sustainability quantification. The benefits in coupling with LCA include (i) the ability of this technique to capture embodied environmental impacts of a metabolic flow applying a cradle to grave perspective, (ii) the quantification and communicability of model results in terms of numerous common and prescient environmental indicators, and (iii) an Figure 1. The different life cycle stages typically covered in an LCA each with its own associated environmental exchanges in terms of energy and mass requirements and waste and pollutant emissions. The gray box indicates that only the use stage is typically accounted for in traditional UM studies and the up-and downstream burdens tend to remain unaccounted for. advanced method with international standards, a large user base continuously improving LCA methodology as well as the availability of inventory data for many important flows entering and exiting cities [63]. To date the coupling of UM with LCA (third generation UM or UM-LCA) has only been performed in simplistic forms, as either a partial assessment subservient to an UM-G2 study [64,65] or to perform carbon accounting [67].
As such, this study endeavors a more comprehensive coupling of low resolution (city-scale) UM with LCA, to develop a UM-LCA model that make use of the full set of indicators available to LCA practitioners, and to utilize dedicated LCA modeling software. The aim of the study is not to develop a fully validated model, but to demonstrate the strength of this kind of UM-LCA modeling, and identify future research questions. The model is applied to five case cities in order to illustrate its applicability, robustness and ability to generate meaningful results catering to the identification of main impact sources in the urban metabolism and to the inter-city comparisons. Case cities with large variations in economic development and spatial makeup will be used in order to see if these differences yield clear distinctions regarding the scale, types and sources of environmental impacts produced. UM-LCA model outputs will be combined with basic economic data in order to explore this theme as well as to discuss the utility of UM-LCA to contribute with benchmarking indicators for SUD policy makers.

Method
In combining UM with the LCA framework the appropriate interface of the tools has to be identified. The basis of UM-G1 and UM-G2 studies is the determination of mass flows into the urban system over an annual period. In UM-G1, the city acts as a black box, with the interactions of the urban subsystems (see section 1) remaining outside of the model's scope [68]. Thus, UM-G1 views cities merely as users of the metabolic flows that they demand to maintain their operations and growth, and the goal of the model is simply the accounting of these demands.
In the life cycle thinking applied in LCA, the accounting performed in earlier UM studies can be envisioned as the use stage inventory analysis of the material and energy flows demanded to feed a city's metabolism. This form of mass and energy accounting fits well with the philosophy of life cycle thinking which conceptualizes a product or service system as consisting of several life cycle stages; raw material extraction, manufacturing, use and end of life (EoL) [69]. When coupled with UM, the LCA attempts to account for the environmental impacts of the other life cycle stages by summing and characterizing the environmental loading of inputs and emissions of all LCA stages. Thus by modeling the extraction of resources and their processing into the multiple flows entering the urban system (the supply chain upstream of the city) and also the EoL processes downstream of the use of a metabolic flow, the UM-LCA approach attempts to cover the total impacts of the urban system as shown in figure 1. Traditionally a process-based modeling of a product system was applied in LCA [69] but it has two main drawbacks. Having to cut off recursive loops (e.g. the production of the production equipment that produces the production equipment . . .) it inherently underestimates the environmental impacts of a modeled system compared to an LCA that bases its inventory analysis on economic input-output tables to account for interdependences between industries [67]. A second drawback originates in the sheer number of individual products that enter the urban system and the impossibility of modeling each of them using a bottom-up, process-based approach. Despite these shortcomings, the developed UM-LCA is process based, as the model can still provide useful (though less complete) results and it will allow insights into fundamental methodological issues that can then be applied to the future development of more complete models that overcome process-based limitations.
The current study implemented the conceptualized scheme in figure 1 following ISO 2006 standards [70] to develop the first full UM-LCA model. Significant metabolic flows were considered for the five case cities, and the appropriate up-and downstream processes from the direct metabolism were found for each city in line with a processbased LCA approach. The UM-LCA model was developed assuming that the urban areas were at steady state with regards to mass and energy, and therefore, accumulation of metabolic flows went unaccounted. Consequently, current technologies were used in accounting for the impacts of the EoL phase of the modeled flows. The product system modeling software GaBi 4 [71] was utilized to model the cities. Programs like GaBi allow LCA practitioners to account for exchanges between a system and its environment through built-in consistency checks and balances. The EcoInvent 2.01 [72] database was used to provide inventories of environmental exchanges (material and energy inputs, air, soil, water emissions, etc) for the modeled processes. However models were occasionally supplemented with processes from the built-in GaBi 4 PE professional database or custom built processes, when adequate processes were lacking elsewhere. Furthermore, best available representative processes were used to support the model (e.g. nuclear power generated in the United States was used to model nuclear power in Canada). EcoInvent 2.01, though expansive in the LCIs it provides, lacks precision modeling capabilities for many processes. For instance, all meat consumption (excluding fish and seafood) was modeled as sheep production as this was the only livestock production process available in the database. Sheep represents a compromise in the level of environmental impacts resulting from livestock production as it has less impact than beef, but more than chicken [139], the two most widely consumed animals in the case cities (one case excluded, Cape Town) [91]. Due to the vagueness of the metabolic data (e.g. 'plastics', 'aluminum', etc) many flows were only modeled through the material production phase; additional processing (e.g. converting aluminum into a beverage container) was ignored, thereby having a reducing effect on the results. Further information regarding the process choices used to model the UM flows as well as their predicted impacts on the model results can be found in the supplementary material (available at stacks.iop.org/ERL/8/ 035024/mmedia).
The developed model was based on attributional data for the LCIs, whereby only the direct effects of the systems were accounted in the impacts. Indirect impacts were not factored in (e.g. changes to agricultural land use in response to increased first generation biofuels consumption), and thus represents a cutoff rule for the current model.

Functional unit-a basis of comparison
In LCA the functional unit is the basis of comparison for two product systems in terms of a service or function that the product or process system provides. For instance LCA could model the different beverage packaging solutions that could be used to hold an equal amount of liquid or different travel methods that can be used to transport a person a given distance. Through a comparison of the predicted environmental impacts associated with having two different means provide the same services, the LCA methodology provides a measure of relative sustainability between different product or process choices [69].
Defining a functional unit for comparing urban areas provides a unique and complex challenge in that the systems (i) support different populations with different cultures, habits, diets etc, (ii) provide varying qualities of life to residents both between and within the cities and (iii) perform functions not only for themselves but for other geographic areas through export of manufactured goods. As these systems do not provide these functions at the same level it was decided to forego the traditional functional unit of an LCA. Instead the gross annual metabolic impacts from the cities were normalized to the per capita level (much like many previous UM studies-see [58] or [76]). Thus, the results will represent the impacts of a conceptual average citizen in the case cities, and comparisons between cities will only hint at gross differences in the quality of life of the residents and the methods by which the economies of the cities are supported. Weaknesses of this choice are discussed in section 4.1.

Indicators studied
LCA results can be communicated as life cycle inventories (LCI), midpoint environmental indicators, endpoint indicators or weighted impact scores [73]. Results quantify predicted (not actual) impacts from the model system(s), and are termed impact potentials (IPs). In this study, it has been chosen to communicate the LCA results through midpoint indicators in order to strike a balance between validity and communicability of the indicator set. The midpoint indicators use (characterization) factors to convert and aggregate system-environment exchanges (from the LCI, which provides an account of the exchange of raw materials/substances between the study system and the environment) and express them relative to indicators of pressures on the environment, material resources and/or human health (e.g. climate change, (stratospheric) ozone depletion, terrestrial ecotoxicity, decreased resource concentration etc). Generally, the further an indicator moves away from the LCI the more uncertainty and subjectivity posited within the indicator [74].
Various life cycle impact assessment (LCIA) methods for generating midpoints exist but for this study the ReCiPe 2008 method has been chosen, as it is the most recent, and contrary to other LCIA methods, ReCiPe allows for multiple cultural perspectives with which to assess the impacts [75]. Three perspectives are provided by ReCiPe based on the Cultural Theory of Risk [140]; Individualist, Hierarchist and Egalitarian. The individualist is least concerned about environmental impacts of humanity's actions, the egalitarian is most concerned, and the hierarchist takes a middle ground in terms of future environmental risk. These perspectives are then reflected in ReCiPe in the models used to convert the LCIs to midpoint IPs in terms of the severity of the potential impacts (e.g. the following timeframes are used in calculating global warming potential; individualist-20 years, hierarchist-100 years, egalitarian-500 years) [75].
The 'Hierarchist' perspective is the default perspective of the applied impact assessment methodology and it was used in this study, as it is based on the most common policy principles, when viewing the severity of an environmental concern in terms of time frame and other issues [75]. Of the eighteen midpoint impact categories available in ReCiPe this study utilized four: global warming potential (GWP), freshwater ecotoxicity (FE), particulate matter formation (PMF) and agricultural land occupation (ALO) that together give comprehensive coverage (air, land and water) of the  [11,33,[114][115][116][117][118] environmental impacts associated with both energy systems, chemicals use, transportation and production of biomaterials and in particular food. The latter two midpoint impacts were compared with economic measures to analyze decoupling trends. In one instance the per capita impacts for ALO and PMF were plotted against per capita GDP in order to see if relationships existed between economic factors and the types of IPs produced. Furthermore, citywide PMF IPs were divided by GDP to see relationships between the cities' economic output and their environmental pressures.

Case cities
The UM-LCA model was applied to five case cities: Beijing, Cape Town, Hong Kong, London and Toronto. A case study approach was chosen rather than a traditional sampling method in order to provide concrete and contextual insight into what the outcome such a model could be used for. The five cases were chosen on the basis of what Flyvbjerg [9] terms as a maximum variation strategy, so that a varied cross section of historical, political, social, economic, geographic and typological characteristics was represented, and the comparisons of the environmental performance of different types of cities could be elucidated. The choice of case cities was also influenced by data availability, in particular the presence of a relatively recent UM study from which metabolic flows could be extracted as input into the UM-LCA model.

Metabolic flows and data sources
The metabolic flows chosen for inclusion in the study were based on their importance in sustaining essential urban functions (food for residents, construction materials for growth/maintenance of building stock and infrastructure, energy for transport, buildings and industry, and other materials commonly consumed en masse). Metabolic flows were derived predominantly from earlier UM studies [11,31,33,55,58,60], and augmented with additional data sources, where necessary to ensure the same flows were covered for all case cities. Earlier UM studies from which data was mined generally covered the same groups of flows in the same level of detail to avoid biases between studies. Furthermore the years studied were within a reasonable timeframe so that large shifts in the nature of societies' metabolism were avoided. Where consumptive data was completely lacking, waste statistics were used to estimate annual consumption for a limited set of goods through certain cities. Figure 2 outlines the material and energy flows considered in the current study while the data sources are referenced at the bottom of table 1. The supplementary information (available at stacks.iop.org/ERL/ 8/035024/mmedia) provides inventories for the annual metabolic flows for all of the cities, including sources and estimation methods for all flows.

Results
Despite the UM-LCA model providing results in terms of midpoint indicators, those results presented in this section were chosen based on the ability to communicate (i) how the developed model improved upon the types of quantification of earlier UM methods, and (ii) the new types of analysis of cities that the UM-LCA model can provide.

Inter-model comparisons
For all of the cities modeled it was found that embedded flows up-and downstream of the cities' direct consumption represented a significant portion of the total mass and energy usage resulting from the cities' metabolic activities. For all the cities the volumes of total mass and energy flows accounted for using UM-LCA were thus significantly larger than those that would have been quantified using the UM-G1 method. To ensure that these findings were not a consequence  of the metabolic flows modeled in this particular case, a similar comparison was performed for mass consumption in Toronto using only those flows considered in the earlier base UM-G1 study. The model proved robust, with indirect flows dominating despite the different flow regime. Figure 3 presents the relative percentages of the direct and embedded mass and energy flows as determined by the UM-LCA models of the case cities, on the left for all cities and the on the right for the Toronto confirmatory test. In the comparison of direct to embedded mass flows water and air were not included as they dwarf the scale of the other metabolic flows in the model.
For all of the cities except Beijing embedded mass flows accounted for more than 60% of the total mass flows resulting from the cities' metabolic activities, indicating that a UM-G1 study using the same flows would have underestimated the total metabolic activities by a factor of approximately 2-3. Looking at the individual flows constituting the mass and energy flows, embedded mass accounted for a significant proportion for numerous high volume metabolic flows, such as gasoline (74%), diesel (88%), natural gas (52%) and meat (96%), explaining the discrepancy between UM-G1 and UM-LCA methods. Unlike the other cities, a majority of Beijing's total mass flows (75%) were accounted for directly, a result mainly attributed to its large concrete consumption, which consists of only 2% embedded mass.
Energy consumption through the cities showed a similar trend, with the embedded energy flows contributing between 48% (Toronto) and 76% (Hong Kong) of the cities' total life cycle energy requirements. Accordingly, UM-G1 models of the same systems would underestimate the total metabolic energy needs of the cities by factor between 2 and 4. Disagreement between the two UM methods are result of the inefficiencies in energy infrastructure [119], as well as the embedded energy in the resource extraction, manufacturing and transport of goods [120].

Inter-city comparisons of environmental performance
The five case studies illustrate that application of the UM-LCA model enables a more detailed insight into the way  [11] per capita, which agree with the current study to the degree that can be expected for such a method. For the other cities GHG accounting has only been performed on the energy consumption in the cities with: Beijing 11.8 tons [142], Cape Town 5.2 tons [143] and Hong Kong 5.3 [144] tons per capita, which are congruent with the GWP attributed to the building energy, electricity and transportation shown in figure 4 for those cities. By analyzing the GWP IPs it is possible to gain a more detailed insight into the contributing factors to the cities' GWP. This reveals a significant variation in GWP IPs across the five case cities. More interestingly, a pattern can be observed between the level of economic development and the main causes of greenhouse gas (GHG) emissions for the cities. Generally the more economically advanced cities (Hong Kong, London, Toronto) have GWP IPs whose origins appear to be related to the affluence and resulting private consumption of those cities' inhabitants. For GWP (and FE) IPs the potential impacts from food accumulated as biomass (determined as the difference between consumed and known disposal routes of food) has been ignored due to uncertainties in the fate and degradation processes.
For the wealthier cities, Toronto as an example, the largest contributing factors to the GWP are transport (27%) and building energy (24%), with waste disposal also playing a large role (21%) due to the city's composting activities that generate methane. GWP from transport stems primarily from the high usage rates of private automobiles in the city [121], while the building energy is a result of the reliance on distributed natural gas as a heating source, particularly of residential buildings [122]. In Hong Kong the environmental impacts of the residents' affluence appears in the IPs from food (27%), where the air transport of perishable seafood constitutes an important factor [112]. For London, it is once again primarily the consumption by residents (not industry) that drives the important contributors to the GWP IP [55]: electricity (26%), building energy (22%) and transport (18%).
Beijing and Cape Town both differ from the wealthier case cities in that the GWP IPs related less to volume of private consumption by citizens, but more to the archaic energy infrastructure and industrial activities of the cities. In Beijing it is the building energy (37%), predominantly from industrial coal boilers [123], and the consumption of goods (27%), mainly steel for construction, which are the largest factors in the IPs. In Cape Town, the coal-based electricity system is the largest single contributor to the GWP IP (37%), with composting also playing an important role (27%). Much of the electricity consumption can be attributed to industry [100], while the remaining IP from residential activities is less due to sheer volume consumed, but more symptomatic of the generating technologies.
The relation between environmental pressure and the infrastructural shortcomings of Beijing and Cape Town are further highlighted by FE IPs in figure 3. Beijing's 169.8 kg 1,4-DCB equivalents per capita (eq./cap yr.) dwarf the other case cities. Cape Town also has an elevated FE IP (61.5 kg 1,4-DCB eq./cap yr.) compared to the other case cities, despite lower consumption by residents and a smaller relative economy. Waste management causes more than half of the FE IPs for Beijing and Cape Town; in particular, the disposal of raw sewage to surrounding waters in both cities places a large burden on the local ecosystems. In Cape Town the impoverished informal settlements dump 12.5% of the city's household wastewater into surrounding creeks and rivers [99,124]. In Beijing 10% of residents were not connected to the water treatment system for the year studied, as the expansion of the city has outpaced the capacity of the municipal waterworks [125]. Furthermore, Beijing also had considerable contribution to FE from the mining activities upstream of the city's steel consumption.
Richer case cities, despite higher levels of economic activity and consumption, have lower FE IPs than Beijing and Cape Town. Furthermore, FE IPs appear to be embedded in the goods consumed in the cities and do not occur locally. Generally the wealthier case cities tend to minimize local pollution issues while the less developed study cities remain affected by pollution issues that their wealthier counterparts have long ago dealt with. Figure 5 expands on this point by displaying the linkages between economic development and different types of IPs. Air pollution in the form of PMF is seen to be high for the level of wealth of the residents in Beijing and Cape Town. This can be regarded as a consequence of the outdated coal technology utilized for electricity, heat and industrial production in both cities [100,126]. Wealthier case cities tend to mitigate air pollution issues through the use of cleaner technologies in energy generation, consistent with the theory that local environmental pressures are inversely related to the wealth of the potentially affected population [127,128]. While Cape Town has similar per capita levels of PMF with wealthier cities, the PMF IP per unit of GDP shown in figure 4 reveals that, to date, only weak economic decoupling has occurred. The relatively low PMF levels in Cape Town are more a result of poverty than policy implementation; thus economic growth in Cape Town would likely result in larger marginal increases of PMF impacts compared to the wealthier cities unless compensated by the introduction of cleaner technologies.
Conversely, looking at ALO IPs for the case cities, there is a positive correlation between the growing wealth of the residents and the increase in environmental pressure from the cities' metabolic activities. The increasing ALO can be predominantly linked to the larger role that meat and dairy products play in the diets of residents in wealthier cities. As the wealth of the residents in the case cities increases, so does the tendency to export environmental loadings of metabolic activities.

Discussion
The hypothesis of the current study is that the UM approach can be embedded within the process-based LCA framework, yielding a hybrid UM-LCA model that can provide a more complete measurement of the environmental pressures exerted by a city. The UM-LCA model was developed and applied to five case cities showing that (i) UM-LCA can be successfully applied to cities where the data exists, (ii) the embedded impacts from a city's metabolic activity can be large and are ignored by UM-G1 studies, and (iii) the model output allows for the identification of where in a city's supply chain environmental pressures are highest and hints at how infrastructure and socioeconomic factors relate to these. However, UM-LCA remains methodologically immature, with the current study revealing some of the basic barriers to its successful application inherent both within the model and the supporting data.

Appraisal of the UM-LCA framework
The impetus for coupling UM and LCA was the growing awareness amongst UM practitioners that both UM-G1 and UM-G2 methods failed to either properly encompass the full impacts of city's metabolism or communicate results in an effective manner [15,64]. Gauging the developed UM-LCA's ability to address these deficiencies will provide an appraisal of its success and judge whether the UM-LCA method warrants further research effort. Furthermore, an assessment of the difficulties encountered in developing UM-G3 provides a platform to discuss ways of advancing UM-LCA and from which future research can build.
Results of the UM-LCA quantify the extent to which the metabolic activities of cities have been underestimated by the widely applied UM-G1 method. Both mass and energy flows resulting from the modeled systems were found to be grossly higher than what would have been accounted for if only direct consumption through the cities had been considered, confirming earlier suspicions [15,63,64]. The considerable mass and energy amounts embedded within consumed products explained this discrepancy for many substantial flows common to all urban areas (e.g. fuels and food). The case study hinted at an increased need for UM-LCA as an assessment tool, as a city's building stock matures, since cities not undergoing significant construction activity had the majority of mass flows embedded within consumed goods. UM-LCA's expansion beyond direct consumption has also begun to quantify the extent to which cities have departed from historical ecological self-sufficiency [47,129] and have become dependent on their hinterlands for resources and waste assimilation. Modeling the exported environmental loadings of cities will become essential as emerging economies (taken here as those cities, such as Beijing and Cape Town, with Human Development Indices below 'high' or 0.758 according to the latest index report [15]) advance and correct infrastructural deficits, thereby disconnecting urban regions from their ecological footprints, as was exhibited by the wealthy case cities in their abatement of local air and water pollution, and as observed in other cases [130,131].
Apart from the inclusion of embedded metabolic flows in the models, UM-LCA exhibited new methods to communicate the assessment findings. The model moved beyond mass and energy as abstracted proxies of environmental loading and expressed results in terms of more common environmental indicators more easily understood by actors across different scientific disciplines. Quantification through UM-LCA also provides environmental policy makers with a potential benchmarking tool for tracking temporal shifts in a city's sustainability resulting from policy interventions. Combination of the IPs with economic data generated additional measures of relative economic decoupling for the case cities. Decoupling is viewed by many as essential in the shift towards a sustainable society [132], particularly in emerging economies where much of the urban population growth is expected [4]. UM-LCA decoupling indicators have the potential to aid in tracking economic decoupling in these cities. For instance, decoupling of industrial energy consumption is currently utilized as a proxy for assessing climate change abatement in Chinese cities [133], but with UM-LCA, officials have the potential to advance beyond proxies to using GWP IPs as a monitoring and source tracking tool. However the limitation should be kept in mind that these improved indicators can only express relative sustainability at present. Coupling UM-LCAs improved quantification with the Earth's carrying capacity, perhaps by means of the planetary boundaries approach [5], might provide measures of absolute sustainability, overcoming this current model shortcoming [138].
The abundance of indicators provided during the LCIA phase of the UM-LCA allows practitioners to quantify the relative performance of different cities over a multitude of environmental impact types/categories (e.g. air pollution, water pollution, land use, etc). It was through this expanded indicator set that the relationship between economic development of the case cities and the exporting of their environmental loadings became clear. Furthermore, different UM-LCA studies can be compared so long as the equivalent sets of mass flows and LCIA methods are employed (the latter being easy to apply using dedicated product system modeling software). Convenient comparisons of relative environmental performance is not merely a point of interest but also of practical use in SUD; cities can look towards peers with superior environmental performance in a sector and can adapt successful practices to the local context.
The ability for UM-LCA to identify key metabolic contributors to midpoint IPs provides information to policy makers regarding hotspots in a city's metabolism that could benefit from intervention or in relation to systematic environmental conscious re-design of urban areas. The breakdown of GWP and FE IPs in this study revealed the bias towards impacts from private consumption and infrastructural deficits in wealthy and emerging cities respectively. In this sense the new method has also revealed the limitations of SUD policy, in that directly reducing the private consumption of urban residents is outside the domain of municipal policy makers, and is a difficult agenda to push in the current consumer driven, growth-based economic paradigm [131]. This echoes the findings of Heinonen and Junnila that per capita GHG emissions in Finnish cities were more correlated with the wealth of residents than the density of the cities (a popular SUD attribute) [67]. Moreover, even if reductions in manufacturing emissions occurred through cleaner production methods, the sheer volume of increased consumption is likely to eclipse the marginal benefits achieved [131].
Even with its positive aspects, the UM-LCA is only useful in the correct spatial context, taking into account the multi-level governance of urban areas. The scale of the city-scale of the current study may not be representative of the drivers and subsystems affecting metabolic flows [145]. For instance the raw-sewage releases in Cape Town are symptomatic of the informal settlements [124], and addressing this issue may be better understood through further research (UM, socioeconomic or otherwise) at the scale of the informal settlements. However, knowing the scale at which to study environmental impacts requires an initial guess, and UM-LCA can provide a quantitative compass directing researchers towards those flows that emerge as significant environmental pressures due to the subsystems embedded in the black box of the model. With this knowledge in hand, more detailed analysis can determine at what scale (sub-urban, urban, or supra-urban) is appropriate for understanding the causes and benchmarking those flows.
Inherent issues in UM studies are matters of definition, for example, where a city's geographic boundaries are delineated and what flows are to be included. Moreover some additional concerns have been introduced through the coupling of UM and LCA. Of paramount importance in the UM-LCA is how the functional unit is defined. As designed in this study the basis of comparison between the case cities does allow for comparisons of their performance, but in an unrefined manner. Firstly some of the metabolic flows (e.g. energy and material for manufacturing) are attributable to the demands of populations outside the city with the case city benefiting from the economic activity of production but not the services the manufactured goods provide. Allocating these types of shared burdens is inherently done for materials by the MFA methodology on which the UM-LCA relies (which only quantifies material accumulation in the cities) but is ignored for the energy consumed in the cities. Thus, the current model wrongly allocated some energy impacts to the case cities, an issue that might be rectified through economic allocation using trade data on manufactured goods in a city, though this information is currently in short supply. In inter-city comparisons the different standards of living between residents in the cities has not been accounted for in the in this study, though metabolic flows have been shown by others to reflect variations in lifestyle [67]. This is not inherently at odds with the current study aim of understanding on a gross level the impacts associated with an average citizen. However, through normalization of IPs to the per capita level, relegates inter-city comparisons to the realm of abstraction, rather than a solid representation of the impacts from the compared cities' actual residents. Targeted and higher resolution UM flows might improve the method in this regard, for instance by applying the model to neighborhoods of similar income in different cities, or conversely, studying the IPs from different income neighborhoods in the same cities. These studies might better disentangle the different drivers for impacts as well as provide the contextual benefits discussed above, thus better informing policy development.
Another issue raised through the UM-LCA tool is how to model accumulated durable goods ('stocks') in cities. Researchers have already attempted UM models that account for changes in material stock assuming average durables lifetimes [37] and Kennedy has developed a mathematically rigorous methodology for tracking UM flows [134]. The difficulty in UM-LCA arises in trying to model the EoL phase of these accumulated stocks, as the uncertainty about the technology increases as the temporal scope of the model increases. For instance, what waste management strategies will be used in handling building materials that will remain in a city's stock for decades or centuries? The current study falsely assumes a steady state for EoL processes, but it is impossible to determine whether that introduces more error than trying to predict and model the future. However, this assumption has likely lead to an overestimation of IPs for durable flows, as some of these will either remain in the cities' material stock indefinitely or long enough for cleaner EoL processing to be developed. The use of uncertainty and/or sensitivity analysis may assure assessors to the importance that EoL processes play in the stability of the overall model results.
External to the UM-LCA methodology but essential to the strength of the results is data availability. As has been a theme in previous UM studies, finding reliable figures for material consumption in the case cities was difficult. Previous studies have overcome data gaps by utilizing trade statistics to perform mass balances on geographic areas [37,58], while others have made use of national household consumption surveys as a proxy for average consumption by city residents [33]. However, these sources are lacking for many cities [10,14], particularly those in emerging economies where much of the urban population growth, economic development and consumption are expected to occur. Crude consumptive estimates gleaned from the limited data available in many cases do not suffice to produce robust UM-LCA models. For instance, the method of estimating consumption from waste statistics may underestimate the true metabolic volumes as the true consumption (material accumulated and output) is not accounted. Headway is being made in this direction, such as projects by the World Bank to collect UM data for a number of emerging world cities [10]. But currently, gaps in consumption data persist as a primary shortcoming in properly benchmarking SUD. In the current study these data gaps led to the exclusion of numerous important metabolic flows (e.g. clothing, copper, asphalt, etc) which may have lowered the results.
A final note regarding UM-LCA modeling is the utilization of process-based LCA, which is methodologically simple, but does not account for the interdependences amongst industries and therefore potentially underestimates IPs [66]. Economic input-output LCA (EIO-LCA) can account for inter-industry relationships using national economic inventories; however, they rest on the basic assumption that all manufacturing occurs within the country modeled, which is clearly at odds with the globalized nature of trade. To accommodate this hybrid-process-based EIO-LCAs have been developed to incorporate the strengths of both methods; (i) increased completeness of material and energy flows with the EIO-LCA and (ii) technological and geographic representativeness of the process-based LCA. The current model most likely underestimates those inter-industry material and energy dependences that strengthen the EIO-LCA, and therefore, it can only be considered a step towards the superior hybrid EIO-LCA.

Conclusion
The conceptual UM-LCA model has the potential to provide an improved assessment tool for urban environmental loading beyond earlier UM methods through inclusion of the environmental pressures embedded in the goods that cities consume and by offering a clear set of communicative indicators. This initial foray into UM-LCA has highlighted a number of future research needs, including: (i) a better definition of a functional unit for cities, (ii) a need for a method to model durable goods accumulated in urban systems and (iii) the further development of the model towards the more robust hybrid EIO-LCA approach using consequential life cycle inventories. UM-LCA has the ability to generate a robust accounting of a city's emergent environmental burdens; as a black box model, however, it cannot shed light on the complex mechanisms and institutional drivers that lead to the metabolic flows a city demands. As such, UM-LCA requires socioeconomic, political and ecological observations in order to provide a truly holistic understanding of the metabolism of cities [14,135,145]. Minx and others have taken in this direction by calling for the inclusion of metabolic drivers and quality of life indicators as part of a more complete UM framework [138]. Moreover, the problem of integrating UM-LCA results into the SUD policy implementation process needs to be addressed, as quantitative data has historically taken a backseat in urban policy development and monitoring [136,137].
Conclusions regarding the relative performance of the case cities should be taken as prima facie considering the shortcomings in supporting data for the model. Increasing the number of studies linking urban economic activities and environmental pressure would help validate the findings here. Significant data shortcomings have plagued other UM studies, and though efforts are being made in some municipalities to collect UM data under the auspices of a number of projects [14]. The near ubiquitous lack of reliable consumption data for cities serves to highlight the uncertainty in current sustainability assessments of urban areas, and claims made by cities regarding sustainability should be viewed with caution while this data remains in absentia.