SEARCH

SEARCH BY CITATION

Keywords:

  • Allocation;
  • Chlor-alkali process;
  • Coagulation and disinfection chemicals;
  • Life cycle inventory;
  • Multifunctionality;
  • Potable water treatment

ABSTRACT

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS AND DISCUSSION
  6. CONCLUSIONS
  7. Acknowledgment
  8. REFERENCES
  9. Supporting Information

Chemicals are an important component of advanced water treatment operations not only in terms of economics but also from an environmental standpoint. Tools such as life cycle assessment (LCA) are useful for estimating the environmental impacts of water treatment operations. At the same time, LCA analysts must manage several fundamental and as yet unresolved methodological challenges, one of which is the question of how best to “allocate” environmental burdens in multifunctional processes. Using water treatment chemicals as a case study example, this article aims to quantify the variability in greenhouse gas emissions estimates stemming from methodological choices made in respect of allocation during LCA. The chemicals investigated and reported here are those most important to coagulation and disinfection processes, and the outcomes are illustrated on the basis of treating 1000 ML of noncoagulated and nondisinfected water. Recent process and economic data for the production of these chemicals is used and methodological alternatives for solving the multifunctionality problem, including system expansion and mass, exergy, and economic allocation, are applied to data from chlor-alkali plants. In addition, Monte Carlo simulation is included to provide a comprehensive picture of the robustness of economic allocation results to changes in the market price of these industrial commodities. For disinfection, results demonstrate that chlorine gas has a lower global warming potential (GWP) than sodium hypochlorite regardless of the technique used to solve allocation issues. For coagulation, when mass or economic allocation is used to solve the multifunctionality problem in the chlor-alkali facility, ferric chloride was found to have a higher GWP than aluminum sulfate and a slightly lower burden where system expansion or exergy allocation are applied instead. Monte Carlo results demonstrate that when economic allocation is used, GWP results were relatively robust and resilient to the changes in commodity prices encountered during the study period, with standard deviations less than 6% for all chlor-alkali-produced chemicals reported here. Overall outcomes from the study demonstrate the potential variability in LCA results according to the allocation approach taken and emphasize the need for a consensus approach to water sector LCAs. Integr Environ Assess Manag 2013;X:000–000. © 2013 SETAC Integr Environ Assess Manag 2014;10:87–94. © 2013 SETAC


INTRODUCTION

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS AND DISCUSSION
  6. CONCLUSIONS
  7. Acknowledgment
  8. REFERENCES
  9. Supporting Information

A principal and long-standing goal for water utilities around the world is to provide safe drinking water to the community in accordance with the water quality parameters established in national regulations. More recently, international attention has turned to sustainable development (Brundtland 1987) that in principle requires that the water sector now also include economic, environmental, and social considerations into the design and operation of water treatment processes. In the case of potable water treatment facilities, chemicals can be the next most significant cause of greenhouse emissions after power supply (Lundie et al. 2004). As such, selecting chemicals for use in coagulation and disinfection processes ought to involve assessment not only of technical performance aspects but also the cost per unit of treated water and the potential impacts on human health, safety, and the environment.

Although debate continues about how to assess the impact of freshwater extraction (Kounina et al. 2013), environmental life cycle assessment (LCA) is a standardized (ISO 2006) and increasingly popular (Peters 2009) quantitative method suitable for comparative water systems analyses (Raluy et al. 2005; Schulz et al. 2012). Likewise, the environmental performance of alternative chemicals for a specific treatment objective can be assessed using LCA. Such a task requires a detailed understanding of the chemical manufacturing processes involved and a comprehensive evaluation of the methodological options available during modeling of the life cycle inventory (LCI) data (Frischknecht 2000). One of the most important chemical production processes associated with water and wastewater treatment is the chlor-alkali process, wherein chlorine gas, sodium hydroxide, and hydrogen gas are produced simultaneously in an electrolytic cell; this coproduction causes what is known as a “multifunctionality” problem in LCA methodology. The stepwise procedure presented in the International Organization for Standardization (ISO) standard 14044 for solving this problem (ISO 2006) has led to controversies associated with the modeling approach (e.g., system expansion vs allocation) (Weidema 2000; Weidema and Schmidt 2010) and the variability and relevance of the results when market information is used (e.g., prices) (Pennington et al. 2010; Pelletier and Tyedmers 2011).

The majority of the LCI data on chemicals included in pre-allocated databases (Althaus et al. 2007) or in LCA studies of water filtration plants (Vince et al. 2008) is presented without further consideration of the possible impacts of fundamental methodological choices taken during the LCI modeling of their production. This is potentially problematic, because arbitrary LCI allocations may lead to incorrect LCA results and misinformed decisions based on those results (Reap et al. 2008). There are various means of solving the multifunctionality problem in LCA depending on the modeling approach chosen (Ekvall and Tillman 1997; Curran 2007). Additionally, when prices are used in economic allocation, a simple deterministic approach (Guinée et al. 2004) may be inadequate for assessing the robustness of the results—a source of uncertainty often overlooked in the literature. In Australia, this uncertainty is amplified further by the use of out-dated and foreign LCI data sets (Alvarez-Gaitan et al. 2011).

Using a comparative process-based LCA of coagulation and disinfection chemicals for potable water treatment, we present a comprehensive investigation of the impacts on LCA results stemming from the various approaches most commonly taken to solve system multifunctionality. To do this we collected the most up-to-date process and economic data available for the most important water treatment chemicals manufactured in Australia (Alvarez-Gaitan et al. 2011). We also present details of the LCI modeling using both system expansion and allocation, including a Monte Carlo simulation to assess the robustness of economic allocation when prices change. Two water treatment comparisons are presented for the global warming potential (GWP) impact category: ferric chloride versus aluminum sulfate for coagulation; and chlorine gas versus sodium hypochlorite for disinfection. In doing so, we aim to enhance our understanding of underlying methodological uncertainties associated with estimating the environmental impacts of chemical use in water treatment operations, while also providing insights into the uncertainties associated with basic chemicals use on the C footprint of other nonwater systems.

METHODS

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS AND DISCUSSION
  6. CONCLUSIONS
  7. Acknowledgment
  8. REFERENCES
  9. Supporting Information

Chemicals and inventory data

The chlor-alkali industry produces chlorine gas, sodium hydroxide, and hydrogen gas simultaneously by the electrolysis of brine using 3 different technologies: Hg, diaphragm, and membrane cells (EIPPCB 2013). This study deals only with membrane technology as it is the most commonly used technology in Australia. From these membrane outputs, many other chemicals are produced including sodium hypochlorite and ferric chloride that are important for the water industry. The other relevant manufacturing process is the production of aluminum sulfate, which leaves no residuals and uses little electricity directly; however, the upstream environmental impact for this chemical is expected to be important due to the use of sulfuric acid and aluminum hydroxide that are produced in energy and resource-intensive processes (EIPPCB 2005). Additional details of the manufacturing processes of the chemicals included in this study are given in the Supplemental Data A.

Extensive data-gathering activities were undertaken to create the process models for the production of chlorine gas, ferric chloride, sodium hypochlorite and aluminum sulfate. First, we sent questionnaires to the different Australian chemical producers. Second, site visits were undertaken to support data collection activities and for data verification purposes. Finally, we performed informal interviews with chemical producers to discuss topics specific to each production process; this yielded detailed information regarding raw materials, energy and transport. A sample questionnaire is included in the Supplemental Data B.

Data for background systems was obtained from a life cycle database (PE International 2012). This included data for the regional Australian power grid mix and ancillary materials used in the chlor-alkali plant (e.g., carbon dioxide, sodium metabisulfite). Salt used in the chlor-alkali plant was considered to be rock salt (PE International 2012) due to the lack of data on the production of solar salt; therefore, we expect that the impacts associated with this input will be higher than those associated with the actual production of solar salt. Sulfuric acid, used as a feedstock in the production of aluminum sulfate and for chlorine gas drying, was considered as a production mix (80% from sulfur combustion and 20% as a coproduct in nonferrous metals production) and sourced from South Korea. Emissions to soil, water, and air associated with the chlor-alkali plant and aluminum sulfate production were taken from the Australia's National Pollution Inventory (NPI) (DSEWPC 2011). The NPI is a publicly available internet database that keeps annually updated information of the amounts of 93 toxic substances released to the environment by more than 4000 Australian industrial facilities. The reference year for the data gathered was 2008 maintaining consistency with other related work based on input-output analysis (IOA) (Alvarez-Gaitan et al. 2013).

Solving allocation by system expansion (substitution)

Applying system expansion to solve the multifunctionality problem associated with the LCI modeling of chlorine gas requires the identification of the products and processes that are potentially suitable substitutes for chlor-alkali-produced hydrogen gas and sodium hydroxide in the Australian market. In the case of hydrogen gas, the principal commercial route for its production is steam reforming of natural gas, so initially this seemed an appropriate form of system expansion for this output. Further investigations, however, revealed that the rate of H production in the chlor-alkali industry is small and in practice does not represent a suitable opportunity for further downstream processing outside the facility. In some chlor-alkali plants, H is used for the production of hydrochloric acid and as a fuel to produce steam that is used in sodium hydroxide evaporation (Orica Watercare, personal communication, August 19 2011). Therefore, the most suitable use of hydrogen gas in this study is as a fuel for combustion and the ratio of substitution is 1 to 2.84 H to natural gas that is calculated from their higher heating values (Demirel 2012).

Sodium hydroxide is considered a coproduct of chlorine production that is sold mainly to the pulp and paper, alumina, and general chemical industries (Orica Watercare 2013). Although sodium hydroxide is crucial in the production of alumina (Santen 1998), in pulp and paper applications it may be substituted for sodium carbonate (Na2CO3) (FMC Corporation 2004). Outside these 2 main markets, sodium hydroxide might be also substituted with sodium carbonate in pH adjustment, acid neutralization, and flue gas desulfurization. Therefore, the function of the output of sodium hydroxide in this study is considered as an alkali suitable for acid neutralization, and the ratio of substitution is 1.325:1 sodium carbonate/sodium hydroxide based on their molecular weights and Equation (1) (FMC Corporation 2000):

  • display math(1)

An overview of the approach taken to model chlorine gas production using system expansion is presented in Figure 1. Once chlorine gas was modeled, the environmental burden of sodium hydroxide could be calculated by solving ENaOH from Equation (2):

  • display math(2)
image

Figure 1. System expansion (substitution) applied to the production of chlorine gas using GaBi 5 (PE International 2012).

Download figure to PowerPoint

Where E represents the environmental burden and the subscripts are for the whole electrolysis process, chlorine gas, sodium hydroxide, and hydrogen gas, respectively. The small contribution of hydrogen gas is estimated by using natural gas as a surrogate. After obtaining the figures for chlorine gas and sodium hydroxide using system expansion, it is possible to model the production of liquefied chlorine gas, sodium hypochlorite, and ferric chloride in GaBi 5 software (PE International 2012). It is important to clarify that the environmental burden of chlorine gas calculated earlier is “wet” chlorine that requires further drying and compression before being sent to liquefaction or sodium hypochlorite/ferric chloride production.

Ferric chloride is produced using chlorine gas, water, steel scrap, and spent pickle liquor (SPL), where steel scrap represents 93% of the Fe source. Chlorine gas was modeled as previously explained and the environmental burden of postconsumer Fe scrap and SPL require additional considerations for open loop recycling. The environmental burden of postconsumer steel scrap was taken from the most recent Worldsteel methodology report (Worldsteel Association 2011). In this document, the cutoff and end-of-life approaches are presented. The first approach considers postconsumer scrap “free” of any environmental burden from the upstream life cycles. This approach considers steel scrap as waste. In contrast, the second approach considers postconsumer steel scrap as a resource that substitutes primary steel production but with a lower environmental burden. We have applied the second approach to our modeling that is the approach supported by the metal industry (Atherton 2007).

SPL can be regenerated into hydrochloric acid and ferric oxide but this is not the case for the steelworks where SPL is being sourced from in this study. As such, if it is not used in the production of ferric chloride, it needs to be first neutralized with calcium oxide where the metals precipitate as sludge and then further treated through filtration, dewatering, and rendering with Portland cement to make it suitable for landfill disposal. Therefore, we have given ferric chloride a credit for the estimated avoided use of these materials according with the requirements for stabilization of pickle liquor presented in the literature (Stanczyk et al. 1982).

Solving allocation by physicochemical causation

In the chlor-alkali plant, electricity, salt, ancillary materials, and emissions to soil, water, and air are related to each of the outputs of the membrane cells through the electrochemical reaction of Equation (3):

  • display math(3)

Chlorine gas, hydrogen gas, and sodium hydroxide are produced in fixed proportions according to this reaction. Mass or molar relationships can be used to describe how these outputs are allocated to the inputs in the membrane cells. Using the most commonly applied mass allocation approach (Boustead 2005; Althaus et al. 2007), the key data are if 1000 kg of chlorine gas is produced, then 1126 kg of sodium hydroxide (100% w/w) and 28 kg of hydrogen gas are necessary coproducts. In reality, these mass relationships must be calibrated with the mass balance of the chlor-alkali facility to obtain the real allocation factors. For example, based on the above stoichiometry and using the molecular weights of the species involved, we will require 1.649 tonnes of sodium chloride per tonne of chlorine gas produced. In reality, however, this value is higher (EIPPCB 2013) due to the purge of brine from the chlor-alkali membrane plant circuit to control impurities. The allocation factors used here are presented in the Results and Discussion.

If molar relationships are used, the basic data is that 1 mol of chlorine gas is produced along with 2 mols of sodium hydroxide and 1 mol of hydrogen gas. When this approach is applied (Leimkuhler 2010), only 25% of the energy and material usage is ascribed to the oxidation of chloride ion to chlorine gas in the anode, whereas the other 75% is associated with the reduction of hydrogen ions to hydrogen gas and the release of hydroxide ions into the solution in the cathode, giving this latter compartment a bigger share of the environmental burden.

Another approach recommended in The International Reference Life Cycle Data System (ILCD) Handbook (ILCD 2010) is the use of an enthalpy basis for allocating the amount of energy used in the production of the outputs in the membrane cell. This approach makes sense in principle, but the Handbook is not entirely clear on which enthalpy is recommended. The first possible option is using enthalpy of formation, but this thermodynamic property is zero for elements such as chlorine and hydrogen gas. Using this approach to the outputs of the membrane cell will result in allocating the entire burden of electricity usage to sodium hydroxide that would seem a rather dubious outcome. Another possibility is applying enthalpy of combustion to the outputs, but in this case the entire burden will be ascribed to hydrogen gas because neither chlorine gas nor sodium hydroxide is flammable. These problems seem to be solved by using other thermodynamic property, exergy.

Exergy represents quantitatively the potential work embodied in fuels and nonfuel materials alike (Ayres et al. 1998). Calculating this property for the outputs of the membrane process we obtained 1743 kJ per tonne of chlorine, 2156 kJ for sodium hydroxide, and 3162 kJ for hydrogen. This form of allocation illustrates the potential for improvement in those chlor-alkali plants where hydrogen is not being used and also displays the relevance of exergy as a valuable indicator to measure efficiency in the chemical industry (Dincer 2002; Ayres et al. 2011). Allocation factors obtained for the application of exergy are presented in the Results and Discussion.

Solving allocation by economic relationships

In the case of allocation using economic relationships, a 3-year average annual price (Guinée et al. 2004) during the period 2008–2010 was used to calculate the associated allocation factors (presented in the Results and Discussion). For sodium hydroxide and chlorine gas leaving the electrolyzer, we used the same 3-year average market price as for sodium hydroxide at commercial concentration and liquefied chlorine respectively. For hydrogen gas, the natural gas market price in the Australian state of New South Wales (NSW) for the respective year was used as a proxy (ACCC 2009). In this case the gross energetic value of the total amount of hydrogen produced was calculated and then the price per giga joule (GJ) applied.

An additional consideration when economic allocation is used is the uncertainty in results stemming from variability in commodity prices (Pennington et al. 2010). To illustrate this point, the prices of sodium hydroxide and chlorine gas as taken from industry sources (Sydney Water, personal communication, August 17 2011) are presented as relative indices in Figure 2. The robustness of the results are tested in this article using deterministic and probabilistic approaches. The deterministic approach consists of calculating the economic allocation factors accordingly for the 2006–2010 period for which we had data. The probabilistic approach was undertaken using a 500 iteration Monte Carlo in GaBi analyst that is an in-built feature of GaBi 5 software (PE International 2012) assuming that prices fit a normal distribution. For interested readers, details of this tool are given in the GaBi database and modeling principles 2011 document (PE International 2011). This tool also requires the user to input the standard deviation (expressed in percentage) of the parameters under analysis. Using the price data gathered from our industry partners for the period 2006–2010, these values are 18.3% for sodium hydroxide, 14.9% for chlorine gas, and 2.23% for hydrogen gas. The output of this simulation is a probability distribution of the GWP results with details of the arithmetic median, standard deviation, and percentiles.

image

Figure 2. Australian relative retail price indices for chlorine gas and sodium hydroxide. Relative price index (y-axis; arbitrary units) shown for each chemical during the 2006-2010 period as a fraction of the baseline year (2006).

Download figure to PowerPoint

RESULTS AND DISCUSSION

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS AND DISCUSSION
  6. CONCLUSIONS
  7. Acknowledgment
  8. REFERENCES
  9. Supporting Information

Regarding the individual contribution of the inputs to the overall picture in the chlor-alkali plant, the majority of greenhouse gas (GHG) emissions in these results (up to 87.5%) are associated with the use of electricity that, in the case of Australian-made chemicals, is generated in coal-fired power stations. The second most relevant source of GHG emissions is the use of sodium chloride (up to 5.1%) followed by the transportation of raw materials (up to 4.7%). The results for each of the membrane cell outputs in the chlor-alkali plant using different techniques to solve the multifunctionality problem and GWP as impact category are presented in Tables 1-4 and are discussed below. The differences between the various allocation approaches investigated here are striking and highlight the importance of including as many scenarios as possible when solving multifunctionality problems in LCA.

Table 1. System expansion applied to membrane cell outputs (1 tonne of chlorine gas, 1.126 tonnes of sodium hydroxide and 28 kg of hydrogen gas) and final GWP impact category results
ChemicalTotal unallocated GWP; CO2-eqSubstitution function/productSubstitution ratioGWP of Substitute kg CO2-eq/kgFinal GWP; kg CO2-eq
  1. GWP = global warming potential.

Chlorine gas n/an/an/a646
Sodium hydroxide (100% w/w)ieam1479-gra-00022900Acid neutralisation/Sodium carbonate1:1.31.42194
Hydrogen gas Fuel/Natural gas2.86:14.645
Table 2. Allocation factors using mass relationships applied to membrane cell outputs (1 tonne of chlorine gas, 1.126 tonnes of sodium hydroxide and 28 kg of hydrogen gas) and final GWP impact category results
ChemicalAllocation factor; %GWP; kg CO2-eq
  1. GWP = global warming potential.

Chlorine gas46.41339
Sodium hydroxide (100% w/w)52.31509
Hydrogen gas1.337
Table 3. Allocation factors using exergy relationships applied to membrane cell outputs (1 tonne of chlorine gas, 1.126 tonnes of sodium hydroxide and 28 kg of hydrogen gas) and final GWP impact category results
ChemicalAllocation factor; %GWP; kg CO2-eq
  1. GWP = global warming potential.

Chlorine gas24.7713
Sodium hydroxide (100% w/w)30.5880
Hydrogen gas44.81292
Table 4. Allocation factors using economic relationships applied to membrane cell outputs (1 tonne of chlorine gas, 1.126 tonnes of sodium hydroxide and 28 kg of hydrogen gas) and final GWP impact category results
ChemicalAllocation factor; %GWP; kg CO2-eq
  1. GWP = global warming potential.

Chlorine gas74.32155
Sodium hydroxide (100% w/w)21.8632
Hydrogen gas3.8113

When system expansion is applied, the substitution effects of the natural gas and sodium carbonate supply chains are very different (Table 1). In the case of natural gas, the avoided burden is 35 kg CO2-eq, which is used to substitute the production of 28 kg of hydrogen gas. In contrast and for the same functional unit, the substitution effect of sodium carbonate associated with the production of 1.13 tonnes of sodium hydroxide is 2229 kg CO2-eq. This much larger figure is associated with the large amounts of energy required in the Solvay process, including 2.2–2.8 GJ/tonne in the kiln and 7.5–10.8 GJ/tonne for all other process needs in plant (EIPPCB 2007). As a result, when the environmental burden of sodium hydroxide is resolved using Equation (2), most of the burden is ascribed to this chemical.

In Table 2, we present the GWP results and allocation factors associated with the production of chlorine gas using mass relationships. As expected, and in accordance with previous international data (PE International 2012), sodium hydroxide (100% w/w) was found to have a GWP similar to chlorine gas. The good availability of data for mass allocation contributes to the popularity of this approach, which is the most widely used in solving multifunctionality in LCI databases (Ekvall and Finnveden 2001; Althaus et al. 2007) and was chosen in the eco-profile of chlorine gas presented by Plastics Europe (Boustead 2005).

In Table 3, the results of applying allocation factors using exergy are displayed. Interestingly, less than a third of the GWP is allocated to sodium hydroxide, almost a quarter to chlorine gas and the remaining to the production of hydrogen. Although not as popular as the mass allocation approach, this technique has the benefit of being able to highlight how efficiently electricity is being used in the electrolysis of brine. This could be very useful in the evaluation of existing and new production processes involving electrochemical technologies.

In contrast to the data presented above, when economic relationships are used to solve multifunctionality, chlorine gas generally yielded a much higher GWP than sodium hydroxide or hydrogen gas (Table 4). There were, however, some notable challenges associated with the use of this approach. For example, retail and wholesale prices are often regarded as confidential that makes them difficult to publish or disclose to third parties. In addition, using generic reference prices might not reflect the dynamics of the local market (ICIS 2012). In this study, we were fortunate to have access to market price data from the same manufacturer supplying Australian-made chemicals to the water use associated with our case study. As an additional quality control measure, we also cross-checked the prices used here with those from previous studies conducted by the supplier. The allocation factors and GWP results for each chemical applying economic relationships (3 year average market prices for 2008–2010) are presented in Table 4.

Water filtration plant case study

For illustrative purposes, we applied the results of the above-described LCI modeling work for chlor-alkali chemicals to a typical water filtration plant (WFP) for the treatment of 1000 ML of noncoagulated and nondisinfected water. This plant, which for confidentiality reasons must remain anonymous, uses 494 kg per day of ferric chloride (42% w/w) for coagulation and 564 L per day of sodium hypochlorite (13% w/w) for disinfection. Details of the raw water quality at the WFP are included in the Supplemental Data A.

To use chlorine gas and aluminum sulfate for comparison with the chemicals currently used in the WFP facility, some additional considerations must first be taken into account. First, in the case of chlorine gas and sodium hypochlorite, we need to establish a link between the chlorine “available” in gas and in solution. According to previous studies (White 1999; Beavis and Lundie 2003), 5.12 kg of chlorine is required per ML of water to provide the same level of disinfection that is currently being provided by 41.0 kg per ML of sodium hypochlorite (13% w/w) during WFP disinfection. It is important to emphasize that this substitution is more complicated than our chemical disinfection relationship here and may include other considerations such as relative cost, health, and safety considerations (Travaglia 2004); however, these aspects are beyond the scope of the current study. Regarding coagulation, we have used equimolar dosages of Al and Fe (AWWA 2011) to compare the alum and ferric chloride coagulants, which, in the case of aluminum sulfate (47% w/w), will require approximately 713 kg per day to substitute the amount of ferric chloride currently used for coagulation in the plant. The GWP results for each chemical are presented in Figure 3 using system expansion (substitution) and mass, exergy, and economic partitioning.

image

Figure 3. GWPs (tonnes CO2-eq/1000 ML of water) of selected coagulation and disinfection chemicals based on alternative approaches to solving multifunctionality.

Download figure to PowerPoint

Given the general preference of allocation for solving multifunctionality in LCA (Ekvall and Finnveden 2001), this case study provides other examples of how system expansion can be used as recommended in the standard (ISO 2006). Moreover, those LCAs using only mass or economic allocation risk overestimating the environmental burdens associated with the production of ferric chloride/chlorine gas and underestimating these for sodium hypochlorite from a system expansion perspective (Boustead 2005; Vince et al. 2008). Interestingly, the results from the application of exergy are similar to those when system expansion is applied for most of the chemicals with the exception of sodium hypochlorite.

The GWP of aluminum sulfate (12,920 kg CO2-eq/1000 ML of water) is presented here for comparison purposes but is not included in the tables and/or figures due to the nonmultifunctional nature of its production process. As shown in Figure 3, the GWP of ferric chloride is higher than that of aluminum sulfate when either mass or economic allocation is used; however, when system expansion or exergy allocation are applied, the GWP of ferric chloride is somewhat lower relative to that of aluminum sulfate. In the case of system expansion, the substitution of sodium carbonate by sodium hydroxide lowers significantly the overall burden associated with the use of chlorine gas. Similarly, when exergy allocation is applied, most of the environmental burden is associated with the production of hydrogen instead of chlorine gas or sodium hydroxide. Therefore, when chlorine gas is used in the production of ferric chloride, this reduced burden is reflected in the final outcome.

On the other hand, Figure 3 shows GWP results indicating that chlorine gas performs better than sodium hypochlorite using any of the modeling approaches to solve allocation. For the disinfection of 1000 ML of raw water, the ranges of GWP performance are 5289–13 250 kg CO2-eq for chlorine gas and 7904–15 482 kg CO2-eq for sodium hypochlorite, representing the full range of consequences of solving multifunctionality.

Robustness of the results when economic allocation is used

When the deterministic approach is used to test the robustness of economic allocation, results shown in Table 5 indicate that, for coagulants, aluminum sulfate performs better than ferric chloride in every single year. In contrast, chlorine gas presents a slightly lower GWP than sodium hypochlorite in any year included in the study. Both disinfectants obtained lower GWP figures in 2008 due to the sudden increase in the price of sodium hydroxide and relatively small incremental increase in the price of chlorine as shown in Figure 2.

Table 5. Results of GWPs (kg CO2-eq) for chlorine gas, sodium hypochlorite and ferric chloride during the disinfection and coagulation of 1000 mL of untreated water using a deterministic approach to economic allocation
Chemical20062007200820092010
  1. GWP = global warming potential.

Ferric chloride26 83326 87325 24527 38527 567
Chlorine gas13 28213 30712 31113 62013 731
Sodium hypochlorite13 73413 75012 95414 00514 094

The results from the Monte Carlo simulation (Table 6) revealed modest standard deviations of 5% and 6% for sodium hypochlorite and chlorine respectively and 4% in the case of ferric chloride. Despite water utilities in Australia and overseas having reported significant volatility in the price of chemicals during the period 2006–2010 (WRF 2009), this price variability seems to have had a relatively low impact on the final GWP results for the chemicals under study here. For the water treatment chemicals included in this study, the robust nature of this result should help to satisfy the commonly held concerns of various stakeholders regarding the validity of using market information to resolve multifunctional systems via economic allocation.

Table 6. Results of GWPs (kg CO2-eq) for chlorine gas, sodium hypochlorite and ferric chloride during the disinfection and coagulation of 1000 mL of untreated water using a probabilistic approach to economic allocation (Monte Carlo simulation; normal distribution; 500 iterations)
Descriptive statisticChlorine gasSodium hypochloriteFerric chloride
  1. GWP = global warming potential.

10th percentile12 50013 10025 700
50th percentile (median)13 30013 80027 000
Arithmetic mean13 10013 80026 900
90th percentile14 10014 50028 300

CONCLUSIONS

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS AND DISCUSSION
  6. CONCLUSIONS
  7. Acknowledgment
  8. REFERENCES
  9. Supporting Information

Using the global warming potential of some common water treatment chemicals as a case study, we have undertaken an in-depth assessment of the robustness of several methodological approaches used to solve multifunctionality in the chlor-alkali production process. On the basis of GWP, we have shown that the environmental performance of chlor-alkali-produced coagulants and disinfectants is sensitive to the approach used to solve system multifunctionality, with the scale of this variability somewhat chemical-specific.

In the case of coagulants, the more commonly applied approaches, such as mass or economic allocation, give an overwhelming preference toward aluminum sulfate. On the other hand, the use of exergy allocation or system expansion changes the environmental performance ranking to favor ferric chloride. For disinfectants, the robust order of preference regarding GWP performance suggests that chlorine is the preferred disinfectant to sodium hypochlorite, regardless of the underlying uncertainties associated with the choice of allocation procedure. Although there was a clear performance advantage for chlorine use under system expansion and mass- and exergy-based allocation (2.6-, 1.6-, and 1.5-fold lower GWP,1 respectively), the performance advantage of chlorine was greatly diminished using economic allocation (≈3% difference).

The overall findings of this research highlight the need for an agreed approach for evaluating water sector scope 3 emissions to standardize GHG emissions inventories and level the playing field for water utilities currently striving to improve the environmental performance of their operations. Without such an approach, water treatment engineers run the risk of making misinformed decisions and realizing suboptimal outcomes from well-intentioned operational measures aimed at reducing the carbon footprint of potable water services delivered to the community.

We would tend to recommend, consistent with ISO 14044 (ISO 2006), that system expansion is preferable to mass allocation, although it is somewhat more demanding on the analyst, for comparative analysis that affect a major customer of the chemicals industry, it is appropriate to take the perspective offered by this method. We found it was relatively easy to perform compared with economic allocation. Although some analysts reject economic allocation for theoretical reasons, our experience suggests it is to be avoided due to practical difficulty. Mass and exergy based allocation are more practical. We found the recommendation of the ILCD Handbook (ILCD 2010) regarding the use of enthalpy to be confusing at best, and recommend using exergy instead on enthalpy in the case of chlor-alkali chemicals.

Acknowledgment

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS AND DISCUSSION
  6. CONCLUSIONS
  7. Acknowledgment
  8. REFERENCES
  9. Supporting Information

This research was undertaken as part of an Australian Research Council (ARC) Linkage Project (LP0991017). The primary author was also supported by an ARC funded Australian Postgraduate Award Industry PhD scholarship. Greg Peters was supported by Formas grant 2012-1122 and the MISTRA Future Fashion program.

REFERENCES

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS AND DISCUSSION
  6. CONCLUSIONS
  7. Acknowledgment
  8. REFERENCES
  9. Supporting Information

Supporting Information

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS AND DISCUSSION
  6. CONCLUSIONS
  7. Acknowledgment
  8. REFERENCES
  9. Supporting Information

All Supplemental Data may be found in the online version of this article.

FilenameFormatSizeDescription
ieam1479-sm-0001-SuppData-S1.docx188K

Supplement Figure S1. Flow diagram of the liquid aluminium sulphate manufacturing process (EIPPCB 2005)

Figure S2. Flow diagram of the chlor-alkali membrane cell process (EIPPCB 2013)

Figure S3. Flow diagram of the production process of ferric chloride using scrap iron and chlorine (EIPPCB 2005).

Figure S4. Flow diagram of the production process of ferric chloride using scrap iron, hydrochloric acid and chlorine (EIPPCB 2005).

Figure S5. Flow diagram of the sodium hypochlorite in a continuous production process (EIPPCB 2013).

Figure S6. Simplified flow diagram of production of ferrous chloride (SPL) in a continuous production process (Hasler and Stone 1997)

ieam1479-sm-0001-SuppTab-S1.xlsx57KTable S1. Raw water parameter values of the water filtration plant used in the case study (Sydney Water 2011)

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.