The IAI has conducted annual surveys of industrial energy use since 1980 and other key environmental data such as perfluorocarbon emissions, fluoride emissions and bauxite residue volumes, since the late 1990s. Much of this data is published freely on the institute’s website (www.world-aluminium.org/statistics/). In addition to these regular surveys, the IAI has collected data specific to the development of life cycle inventories for the years 2000, 2005 and 2010. The survey forms to collect industry data that form the basis for this study (included as Appendices for reference in the Electronic Supplementary Material) were sent to statistical correspondents (both within and outside of the IAI membership) in early 2011 with a request for data for the calendar year 2010. Data was collected on a facility level with the data categories included in the LCI survey designed specifically to ensure that all relevant data were collected to cover the scope of this inventory. The data categories have been selected based on their environmental relevance specific to primary aluminium production or as they are widely acknowledged as industry measurements to monitor and report against environmental issues. The data collection and processing were monitored by a dedicated life cycle data review group that reported to the IAI Environment and Energy Committee.
Once the data was received from reporting companies, it was assessed internally. Quality checks were conducted by comparing a facility’s newly reported data to its previously reported values and the average values from the 2005 study for plants using similar technologies and processes. To ensure the integrity of the data, all values were checked individually and significant variations (±2 standard deviation) in reported data were queried with reporters and either confirmed or amended as appropriate. In addition, the data underwent third-party verification by an independent expert.
The collected data represented the direct inputs and outputs attributable to the processes at each facility. For indirect processes, such as production of ancillary materials, fuel and electricity, where specific data could not be supplied directly by the reporter, the background inventory datasets included within GaBi were used as a proxy. These background inventory datasets are sourced by GaBi from a number of reputable agencies such as European Aluminium, Plastics Europe and American Forest and Paper Association.
In the life cycle inventory, almost all averages were calculated as production-weighted mean values per tonne of relevant production output for those facilities that reported. There were some circumstances where this methodology did not accurately reflect specific process features, and so, alternative approaches were applied. Where there was an array of input/outputs per relevant process data (e.g. fuel mix), there was a need to count non-reported data points (zero values) so that a weighted average of a comprehensive array across the industry, and not just the average of a single criterion per production mass, was considered. This industrial weighted mean was used for seawater input, transport distances and fuel and power mix.
It should also be noted that as the ingot casting process excludes remelt or recycled aluminium, the LCI survey results for the ingot casting process yield a higher mass output than the electrolysis metal output. This was accounted for by adjusting the inputs and outputs from the survey average by a factor which was determined based on hot metal, alloying elements, total metal output and scrap sales. According to International Organization for Standardization (ISO) 14040 and 14044 (2006), this can be described as a situation of joint processes where a mass allocation approach is adopted. The absence of co-products for the unit processes considered in this study means that no other allocation methods were used.
Direct aluminium production process inventory data at the global level is published by the International Aluminium Institute per unit process (see Appendix I, Electronic Supplementary Material) and demonstrates, in part, the global aluminium industry’s dedication to report openly its environmental impacts. The data collected serves as a credible basis for subsequent life cycle assessments of aluminium products. With the integrity of such datasets heavily dependent on the coverage and representativeness of the data received from the surveys, there are limitations with this approach that must be acknowledged. These are discussed in later sections.
Impact categories—selection, classification and characterisation
Six midpoint environmental impact categories for this study were selected in line with the recommendations from the harmonisation study for LCA methodologies across the metals industry (PE International 2014). In addition, the newly developed environmental indicator in the form of water scarcity footprint (Buxmann et al. in this issue; International Organization for Standardization (ISO) 14046 2014) was also considered. The impact categories selected are presented in Table 1, and they represent the most frequently used for life cycle impact assessment.
In addition to these categories, a breakdown of the relative contribution to global warming potential (GWP) of industrial processes in the primary aluminium value chain was included, along with a breakdown of total primary energy transformed. The exclusion of impacts for land use, abiotic depletion potential (ADP), ecotoxicity and human toxicity are discussed in further detail in the Section 4.2 of this paper.
The Centre of Environmental Science (CML) methodology was selected to define the characterisation factors that convert the IAI inventory data to the common unit of the category indicator which allows determination of indicator results. This methodology is in line with the recommendations for LCIA methodologies in the metals industry and, in particular, for those with a global coverage (PE International 2014). The classification and characterisation of impact categories allow evaluation of their significance within the life cycle and over time. In this assessment, classification and characterisation were completed simultaneously. The classification process involves assigning inventory results to the impact categories listed in Table 1. Data used in the GaBi database for classification of the LCI results according to the impact categories is published by the following organisations: ISO, Society of Environmental Toxicology and Chemistry (SETAC), World Meteorological Organization (WMO) and Intergovernmental Panel on Climate Change (IPCC). The impact categories in Table 1 represent the accumulated impacts of the inputs and outputs of the system using category indicators. During the development of characterisation factors, consideration was given to a number of key influences including geography, population densities, chemistry, emission rates and other such technical characteristics that define the relationship between environmental flows and their potential impacts.
The modelling software used for this cradle-to-gate study was GaBi 6 (PE International 2013b). The GaBi model is built up through a hierarchical system which includes a series of plans and unit processes at its highest level. Figure 2 shows the plan for the IAI GaBi model for the GLO dataset and RoW dataset, within which the unit processes: bauxite mining, alumina production, electrolysis and ingot casting, are also visible. Each of these unit processes within the model is broken down into sub-systems which include a combination of direct industry data, i.e. IAI inventory, and background data for ancillary materials or processes, representing the inputs and outputs for each of the unit processes. These datasets are available in GaBi on a unit process level and can be interrogated down to the elementary flow level. The aluminium inventory data within GaBi has the added potential to be regionalised for LCAs due to the modelling of regional energy data. This is particularly important considering the significant impact of different energy mixes.
Background inventory data for the following supplementary processes were used in the model: limestone production, caustic soda production, aluminium fluoride production, petroleum coke production, pitch production, electricity generation and supply, fuel production and supply and transportation. The background data within GaBi is sourced primarily from industry and is therefore considered technologically representative (GaBi Modelling Principles 2013) and up-to-date. In addition, all data are compliant with the guidelines issued by the International Reference Life Cycle Data System (ILCD), which serves, in part, as a measure of quality for such datasets being used in life cycle work.
The decision to model two datasets, GLO and RoW, was based on the fact that China’s primary aluminium production and demand have been essentially balanced for some time. Chinese imports of primary aluminium in 2010 were estimated at just over 0.2 mt, whilst exports were just under 0.2 mt (Antaike 2011). This is a comparatively low level of trade for a country where consumption in 2010 was close to 16 mt (IAI 2013b). The low level of external trade, generally poor data coverage and a significantly different power mix to the rest of the world means that modelling two datasets allows for distinct conclusions to be drawn about the environmental impacts for the globally traded aluminium market in 2010. The main differences between the datasets are from the energy mix and PFC emission data. These, in turn, are highly dependent on smelting technology and background electricity grid mixes. The availability of Chinese electricity consumption data, the most material influence on environmental impact, adds robustness to the Chinese data included as part of the GLO dataset. This is discussed further in the Section 4.2.
The energy intensive nature of the primary aluminium production process means that representative modelling of electricity supply systems and fuel mixes is critical to the accuracy and robustness of the output dataset. The electrolytic smelting process accounts for more than 95 % of total aluminium electricity consumption from cradle to gate, and the IAI has collected annual facility-level data on this input since 1980 (see Appendix II, Electronic Supplementary Material). The smelting electricity model developed in this study allows for the attribution of impacts, through the inclusion of background data, to regional industry-specific electricity mixes, rather than purely regional or national grid mixes, which for aluminium production is not always representative of the consumed power supply. This difference often stems from the aluminium smelter’s requirement for abundant, competitively priced electricity to sustain operations. As such, producers are typically located in areas with access to low-cost, reliable electricity generation where long-term contracts are fixed. Increasingly, there has also been a move towards greater integration upstream, and now, some aluminium smelters have integrated power plants that feed their electricity requirements. In periods of surplus, electricity they generate can also be sold to the grid. It is this intimate link with energy supply that means the energy mix of a region can differ notably from that which serves its aluminium industry. In this paper, the industry-specific mixes used are GLO 2010 and RoW 2010, but the electricity model developed by the IAI has been included in GaBi 6, so that practitioners are able to build their own regional datasets, based on the IAI’s regionalised power mix datasets (http://www.worldaluminium.org/statistics/primary-aluminium-smelting-power-consumption/) as seen in Tables 2 and 3.
The IAI’s annual energy survey (Appendix II, Electronic Supplementary Material) of aluminium smelters provided energy carrier data for the regional datasets created in GaBi, each of which was supplemented by regional background data specific to the IAI regional power mixes. The datasets are aligned with the following IAI statistical regions: Africa, Asia, China, North America, South America, Europe and Oceania. These statistical regions are intended to maintain anonymity for individual companies reporting data to the IAI by aggregating data based on geographic spread. Proxy data was used for regions with limited background data; for example, South African energy carrier background data (but not electricity mix) was used for ‘Africa’. The total impact of the electrolytic process and electricity consumption was calculated as the production-weighted average of impacts in all seven regions for GLO and six regions in the case of RoW.
The methodology adopted for thermal energy modelling was similar to that for electricity modelling. The impact of thermal energy input into the following unit processes was included: bauxite mining, alumina refining, anode production, paste production and ingot casting. A regional mix was constructed for each energy source, with the percentage share of each region again modelled (where necessary) on a relevant proxy background LCI dataset (e.g. Brazil for South America) present within the GaBi database. The global mix is a production-weighted average of the regional models, as for electricity.
In order to assess the contribution of the various processes to each impact category, the LCI data (IAI 2013a) is assigned to specific typologies using the GaBi software. The four typologies to classify the processes and material inputs and outputs within the system boundary include
Direct and auxiliary processes, encompassing material used in, or direct emissions associated with, the production of primary aluminium as well as the ancillary processes and materials such as caustic soda, lime and aluminium fluoride.
Transport which includes the movement of input material via road, rail or ship.
Electricity which includes the processes and materials needed to produce the electricity used directly in the production of aluminium.
Thermal energy which includes the processes and materials needed to produce the thermal energy used directly in the production of aluminium but excluding the pitch and coke for anode production.
This methodology enabled the contribution of the relevant processes, to be displayed within the LCIA results in Section 3.
Water scarcity footprint (WSFP) for the production of primary aluminium was calculated using an approach in accordance with ISO 14046 (2014), and the concept is explored in significantly more depth by Buxmann et al. (in this issue). Essentially, the methodology for single site WSFP analysis incorporates direct water consumption from production sites along the aluminium value chain; indirect water consumption of the different ancillary materials, fuel and electricity needed for the production process; and a local water scarcity index (Pfister et al. 2009). A generic water scarcity footprint per tonne of primary aluminium was then determined by summing the direct and indirect WSFPs of the plants and normalizing it to the reference flow of 1 kg of primary aluminium.