Environ. Sci. Technol. 2000, 34, 4789-4796
Life Cycle Inventory of Refinery Products: Review and Comparison of Commercially Available Databases CONCEPCIO Ä N J I M EÄ N E Z - G O N Z AÄ L E Z A N D MICHAEL OVERCASH* Department of Chemical Engineering, North Carolina State University Box 7905, North Carolina State University, Raleigh, North Carolina 27695-7905
Refinery products serve as the source for a significant portion of energy use and industrial chemicals. Assessing the variability and reliability of the life cycle inventory (LCI) data for the refinery process is an important issue for the acceptance of life cycle studies. The purpose of this research is to review and compare the LCI results for refinery products among several available databases, evaluating the level of variability and technical consistency among data sets. Another objective is to highlight the need for greater transparency and standardization in LCI databases. We found important links between the type or media of emissions and the unit processes found in typical refineries. The variability of estimated emissions to the atmosphere is approximately 50-150%, while variability in aqueous discharges is higher, approaching 1000%. Variability for solid emissions is on the order of 30%. This variability is believed to be related to the preparation and summary use of individual practitioner databases and not to the actual primary variations among refineries. Some variations clearly reflect technically incorrect data. Improvement of refinery LCI information can be achieved by the research and user communities devoting effort to produce credible, representative LCI data for this critical manufacturing unit. Greater transparency would significantly improve specific areas of LCI assessment, as many of the current databases are highly nontransparent, thus allowing errors to be undetected. The central role of refinery LCI data in life cycle assessment argues strongly for practitioners to develop more representative and less variable LCI information.
Introduction In the life cycle assessment field, the direct products of a refinery are important elements in constructing the life cycle inventory (LCI) of virtually all products. This importance is derived from the use of oil as process heating fuel, gasoline in transport, or petrochemicals as reactants. Thus understanding the LCI of a representative refinery is a common need in this field and in the evaluation of sustainability indices. The European evaluation of life cycle technology has identified the diversity and quality of LCI data as a major * Corresponding author telephone: (919)515-1315; fax: (919)5153465; e-mail:
[email protected]. 10.1021/es991140f CCC: $19.00 Published on Web 10/11/2000
2000 American Chemical Society
barrier and area for improvement (4). This is particularly true for the LCI of petroleum refining because of the ubiquitous inclusion of refinery products in virtually all LC studies. If LCI data representing the widely used products of a refinery have significant variation and uncertainty, then the conclusions for all LC studies have high variability. Thus, it is important to reduce the variability in information, particularly if it results from differences in reporting formats, assumptions, and even in the definition of terms. A critical review of refinery LCI information is a necessary step to improvement in LC research. Improved LCI information will allow more credible distinctions between technology or policy alternatives, the goal of decision-making in LC evaluation. The LCI of a refinery appears in a number of databases used in the LC field. Often users take an LCI generated in various papers or reports for products, such as plastics, solvents, automobiles, solar cells, etc., in which the refinery LCI is an embedded part. This exchange of LCI information makes it important to understand the underlying data for refinery products. It is generally true that the LCI databases for refineries were collected from primary data, but in subsequent use differences in interpretation can arise; differences that, without adequate explanation can lead to a loss in transparency. In a search of over 400 LC articles, not a single article comparing refinery LCI data or for that matter any other chemical LCI data comparisons were found. Thus, a review of refinery LCI data is unique. Since most LCI databases are proprietary and are not presented as primary data, a comparison can only be made of the published version. However, these published refinery LCI data are what most researchers are using in current LC studies.
Objectives The overall goal of this paper is to provide a comparison of LCI data from the perspective of a user or researcher, dependent on published LCI data. The specific objectives are as follows: (i) To prepare a common format, related to actual refining technology, for comparison of refinery LCI data. (ii) To utilize chemical engineering principles to extract further information. (iii) To establish the level of variability in estimates of major and minor emission from refineries. (iv) To examine the refinery LCI databases for technical discrepancies. (v) To highlight the need for greater transparency and coherence among LCI databases.
Methodology The data sets evaluated are shown in Table 1. Refinery LCI data were obtained from commercial databases, journal publications, or reports (Table 1). The broader sources of the specific refinery data could, in most cases, not be determined. The individual primary data from actual refineries used to derive the estimates were not available in these publications; hence, the underlying variability is not given. Only the variability at the user level (i.e., between studies) is evaluated in this paper. Again, when one draws on published sources to construct the LC of a higher order product (such as a tire or solvent), these user-level data provide the basis for estimation. In overview, a refinery is principally a collection of processes used to separate crude petroleum into product classes. This separation occurs principally by heating in unit processes. The distribution of products is routinely varied VOL. 34, NO. 22, 2000 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4789
4790 9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 34, NO. 22, 2000
TABLE 1. LCI Refinery Data Sets Compareda partition of energy
partition of gas emissions
partition of water emissions
av refinery products for South Europe av refinery products for South Europe av refinery products for North Europe av refinery products for North Europe naphtha, av Europe naphtha, av Europe naphtha, av Northern Europe av refinery products for Europe naphtha, av Europe
E, F
no
no
no
no
E, F
partition of solid waste
classification of solid waste
methodology for allocation
includes transport
ref
no
yes
mass
yes
5
industry
no
no
no
nc
nc
6
no
no
no
yes
mass
yes
5
PWMI Report 2 industry
no
no
no
no
no
nc
nc
6
no E, F, T no
P, F, T F, P, T, E no
no F, P, T, E no
nr F, P, T, E no
nr yes yes
nc nc nc
yes yes nc
9 10 7
no
no
no
no
no
nc
nc
7
no
no
no
nr
nr
nc
nc
6
E, F
F, P, T
F, P
F, P
yes
mass
yes
11
SIMAPRO diesel
naphtha for Italy, Netherlands, Austria, and UK diesel, av Europe
no
no
no
no
no
nc
nc
7
PEMS diesel
diesel, av Europe
no
no
no
nr
nr
nc
nc
6
EMPA diesel ECOPRO diesel FRANKLIN av European
diesel, av Europe diesel, av Europe av refinery products for Europe
no no F, T**
P, F, T no no**
no no no**
nr nr no**
nr nr no**
nc nc nc*
yes nc yes
9 10 8
data set label PWMI South European PEMS South European PWMI av European PEMS av European EMPA naphtha ECOPRO naphtha SIMAPRO naphtha BUWAL 250 SIMAPRO naphtha av Europe PEMS naphtha, av Europe Boustead naphtha, av Europe
product
primary source
PWMI Report 2 EMPA BUWAL 250 BUWAL 250 PWMI Report 2 ETH-ENET 30179 Okoinventar industry PWMI Report 2 ETH-ENET 30179 Okoinventar EMPA nc nc**
a *, probably mass fraction according to cross-checking of information presented as cited in a secondary source. **, as cited in a secondary source. nr, not reported. E, electricity. F, fuel. T, transport. P, process. nc, not clear as presented in database. mass, mass fraction.
(within some broad ranges) to meet consumer demand. Processes that crack or reformulate are used to change the product distribution along with the selection of crude oil sources. Despite the apparent complexity of the changing product mix, the distribution of refinery products must in all cases conform to mass balance principles. Thus for 1000 kg of input crude, the amount of saleable product is typically 940-970 kg with 30-60 kg of crude being used as energy for the refinery processes (average of refinery database estimates in this evaluation). In fact, excluding the use of crude as energy, the refinery is highly efficient with chemical losses of only ∼0.3-0.5% (1). Therefore, the crude oil input is close to the product output. In addition, water is a refinery input that then appears as the predominant component of wastewater. Use of a “per kilogram product basis” for the calculations thus provides a stable reference frame, even though the distribution of product can vary (e.g., between the amount of gasoline versus fuel oil). In current LC assessments, the refinery process and energy emissions are allocated to each product on the basis of the mass (kg) of product. Thus, 1 kilogram of one product has the same allocated emissions as a kilogram of another product, and so the distribution of products is less influential. Since a mass allocation technique is used in the present work, representation of emissions or energy on a kilogram of product, kilogram of naphtha, kilogram of diesel, etc. basis is equivalent. First, LCI data were collected from each data set regarding energy usage, raw materials, air emissions, water emissions, and solid waste. These values included precombustion data. The data were not all available or aggregated in the same categories, so an interpretative analysis was first needed. At the beginning, it was not clear how many of these data sets were actually independent; therefore, a comparison of the manner of presenting the information in each database was performed. The data sets present the energy usage information in energy units (MJ of fuel: gas, oil, coal) or mass units (kg of fuel) but rarely in both energy and mass units. To overcome this difference and to standardize the units, calculations of the corresponding mass or energy values for fuel usage were performed. The assumed heat values are 45 MJ/kg for oil, 30.3 MJ/kg for coal, and 54.3 MJ/kg for gas (39 MJ/m3) (2). The density value for natural gas is estimated to be 0.72 kg/ m3 (2). This allowed for a full comparison of mass (kg) and energy (MJ) terms through the use of these conversion factors. The next step in unifying the databases was to determine how much of the inputs and emissions reported in each data set was related to electricity production, fuel for heating (and transportation), and refinery processes. Only the ECOPRO naphtha and Boustead data sets explicitly describe this information, although some of the other data sets do provide partial explanations. For the rest, further calculations (given in the following sections) were required to estimate the distribution of energy use and pollutant emissions. These calculations were also found to be useful for a cross validation of the estimated energy and emissions factors. When the total electricity and other energy amounts were not reported, calculations were needed. The disaggregation of the LCI data was conducted with the following logic. Each refinery database reported the overall amounts of coal, oil, natural gas, hydro, and nuclear energy use. As an example, the energy data in SIMAPRO naphtha, average Europe are shown in Table 2. The energy includes the amounts for electricity, refinery fuel consumption for heating processes, and transportation (if reported). The emission factors must therefore consider pollutant emissions resulting from electricity consumption, combustion of fuel for heating, and emissions from the actual refinery processes.
TABLE 2. SIMAPRO Naphtha Energy Data energy
MJ/kg of refinery products
oil (nonfeedstock) coal gas hydro power nuclear
1.41 0.15 3.34 0.01 0.01
Since refineries do not have direct hydro or nuclear facilities, these energy sources could only represent the electricity used from the grid. Thus, these were used to back calculate the electricity consumption (MJ/kg of refinery product) for each life cycle database separately. Using the European electrical grid distribution, the approximate level of hydropower is 0.21 MJ of hydropower/MJ of electricity delivered (6, 7, 10). The refinery electricity required is estimated with the following relationship:
E ) H/fhydro,SIMAPRO
(1)
where E is the refinery electrical energy (MJ/kg of naphtha or other refinery product); H is the hydropower (MJ of hydropower/kg of naphtha), as reported in the database for the refinery (e.g., SIMAPRO); fhydro,Europe is the average factor for hydropower in electricity production in Europe ) 0.21 MJ of hydropower/MJ of electricity produced (see Table S1 of the Supporting Information). Therefore, the estimated electricity consumption is E ) (0.01/0.21) ) 0.048 MJ of electricity/kg of naphtha produced in a refinery. Hydro was selected to provide the basis for estimating electricity usage because the overall variability for hydropower use across the various LCI databases was lower (about 5-fold) than that for nuclear, for which the variation was about 100-fold. That is, if one calculates the electricity usage per kilogram of refinery product, as in eq 1, across all databases, the variation using hydro is 5-fold while that for nuclear is 100-fold. Such variation between electricity values when determined by hydro versus nuclear is an inconsistency about which LCI practitioners should be aware. In the case of a refinery, since electricity use is low, the impacts of this inconsistency on the overall energy use and emissions estimates are not large. Nevertheless, LCI databases for refinery products could achieve much higher transparency by reporting the source and type of electricity usage directly. Knowing the electricity use per kilogram of refinery product, the estimates of electricity-related emissions were made using the emission factors from each database (see Table S1 of the Supporting Information). The underlying variability of electric power generation emissions is not reported for the individual databases:
Li,elect. ) EFi,database
(2)
where Li,elect. is the electricity-related life cycle chemical i emission at a refinery, chemical i/kg of refinery product; E is the electrical energy used (MJ/kg of refinery product); Fi,database is the chemical i emission factor/MJ of European electricity production in SIMAPRO database, as an example (see Table S1 of the Supporting Information). Following the example above, the estimated carbon dioxide emissions for electricity production LCO2,elect. ) EFCO2,SIMAPRO ) (0.048 MJ of electricity/kg of naphtha)(1.33E + 05 mg of CO2/MJ of electricity) ) 6384 mg of CO2 related to electricity/kg of refinery product. Similar calculations for each chemical emission to air, to water, and as solid waste allow one to estimate the amount of electricity-related emissions that are attributable to a kilogram of refinery product. VOL. 34, NO. 22, 2000 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4791
From the estimate of electricity use calculated above for each LCI database, the mass of oil, natural gas, and coal used for that electricity production (coal use is not typically large in refineries) was calculated (see Table S1 of the Supporting Information for the power grid used to back-calculate the mass of these fuels ascribed to refinery electricity use). Subtracting the amount of oil, natural gas, and coal used for electricity from the total oil, natural gas, and coal reported per kilogram of refinery product gives the amounts of these fuel types that are utilized for heating the various refinery unit processes. Using the typical combustion profiles of emissions from each of these three fuels (see Table S2 of the Supporting Information), we then separately calculate the amount of the total refinery emissions attributable to heating the various refinery processes (fuel combustion). For example, with SIMAPRO naphtha, the reported oil use per kilogram of refinery product is 1.41 MJ. The calculated oil use (expressed as MJ) attributable to the generation of electricity per kilogram of refinery product (based on earlier hydro information) is 0.017 MJ. Thus, the amount of oil consumed by heating (diesel combustion) is assumed to be 1.41 MJ - 0.017 MJ, or 1.39 MJ. On the basis of the oil heating value of 45 MJ/kg (2), the amount of oil used in heating refinery processes is 0.031 kg of oil/kg of refinery product (for the SIMAPRO naphtha database). Following a similar calculation, the energy and mass for refinery process heating is estimated to be 3.327 MJ/kg of refinery product for natural gas and 0.102 MJ/kg of refinery product for coal (which translates into 6.12E-2 kg of gas and 4.65E-3 kg of coal/kg of refinery product). Now, with the amount of oil, natural gas, and coal used for refinery heating, the individual chemical emissions attributable to heating were determined utilizing the emissions factors for combustion of each fuel type (see Table S2 of the Supporting Information; 7): n
Li,heating )
∑C F
(3)
j i,j
j)1
where Li,heating is the heating-related life cycle chemical i emissions at a refinery, kg of chemical i; Cj is the combustible fuel j (oil, natural gas, or coal mass, kg of fuel j/kg of refinery product); Fi,j is the emission factor for chemical i from burning fuel j (see Table S2 of the Supporting Information). For SIMAPRO naphtha data used as an example, the emissions of carbon dioxide due to heating were estimated as follows:
lcarbon dioxide, heating )
( ( (
)( )( )(
) ) )
0.0309
mg of CO2 kg of oil 3,760,000 + kg of naphtha kg of oil
0.0612
mg of CO2 kg of gas 3,188,339 + kg of naphtha kg of gas
0.0046
mg of CO2 kg of coal 2,840,000 ) kg of naphtha kg of coal 324,851
mg of CO2 w kg of naphtha
114% of total reported CO2 emissions (4) The percentage of each chemical emission per kilogram of refinery product was determined separately for electricity use and for heating use. Finally, the difference between total chemical constituent emission (as reported in each LCI database) and that for electricity plus heating-related emis4792
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 34, NO. 22, 2000
sions is the emission directly from the processing and refining of crude oil into the various refinery products. These are referred to as process emissions. Comparing the heating-related carbon dioxide emissions (see example), the estimated percentage of the total refinery carbon dioxide emissions from heating was determined as 114%. A percentage greater than 100% could be caused by rounding errors and different combustion profiles, but in either case, it indicates that most of the carbon dioxide emissions are produced during heating processes, with a small contribution from electricity production and practically none from processes, as expected for carbon dioxide. For the SIMAPRO naphtha example, the carbon dioxide emissions were determined as 2% from electricity, 114% from heating, and -16% from processes, which in practical terms can be translated to ∼2% from electricity, ∼98% from heating, and ∼0% from processes. These disaggregation methods were applied to each of the LCI databases found for refinery products. With this diversity of databases, one can examine issues of variability and technical inconsistencies. These comparisons were the objective of an in-depth review of the refinery product LCI databases. The various databases were analyzed to determine arithmetic averages, medians, and standard deviations for energy use and emission factors. These are reported in the following section.
Results Assessment Based on Comparative Concepts. As LCI studies are typically conducted, few projects have collected all of their data independently. More typically, projects must import other specific LCI results that were collected by other investigators. Thus, in utilizing different LCI results, the underlying modules and estimates, such as for refinery production of fuels or feedstocks, are based on results from a mixture of databases. Although 15 LCI refinery data sets are indicated in Table 1, these may only be representative of five fully independent sets of data. For example, data used in SIMAPRO (7) were taken from the PWMI report (5) and from BUWAL 250 (10). Data from PEMS (6) were taken from the PWMI Report (5) and from the ETH-ENET Eco-inventory (9). Nevertheless, even in data sets with the same primary source, the level of detail presented varies, and it is usually difficult or impossible to trace original data sources. The electricity required in most of the databases (reported directly or estimated using hydro) lies between 0.01 and 0.075 MJ/kg of refinery product, with a median of 0.05 MJ/kg of refinery product. However, the estimated electricity for the naphtha and diesel data sets seem to be higher than the electricity for the rest of the refinery type data sets (average of 0.224 MJ/kg of refinery product). The feedstock needed to produce a kilogram of refinery products is relatively constant. The exceptions are EMPA naphtha and EMPA diesel, which present a higher estimate for the consumption of feedstock oil (see Table S3 in the Supporting Information). Regarding the fuel used for heating refinery processes (Figure 1), the PEMS naphtha, PEMS diesel, and EMPA naphtha databases indicate a significantly larger amount of oil and coal use. The higher use of oil does not appear to be in place of natural gas usage since the total fuel used is still larger for these three LCI data sets. For the rest of the databases, it appears that the oil and coal required for heating is larger for diesel than for naphtha (Figure 1). If this is true, then direct allocation according to mass of product was not used for these products, again an opportunity for transparency. SIMAPRO diesel is the exception (Figure 1), but this can be explained because the data presented in these databases for diesel are the same data used for the average
FIGURE 1. Comparison of oil, natural gas, and coal requirements reported in the data sets analyzed (feedstock not included). refinery product, which is consistent with mass allocation principles. These different fuel usage patterns explain the results for CO2 emissions shown in Figure 2. The magnitude of the variation in estimated carbon dioxide emission factors shown in Figure 2 is significant. EMPA, SIMAPRO, ECOPRO, and FRANKLIN are widely used databases. On a percentage basis, the variation in CO2 emissions is from 50% to 150% of the average or median. In LCI studies for which the CO2 emissions from refinery products are important, sensitivity analyses would need to consider these differences and resulting uncertainty in estimated emissions. The air emission factors for other pollutants (dust, CO, SOx, NOx, hydrocarbons, and methane) follow the same pattern of variation as that determined for the fuel requirement and the CO2 emissions (see Figure S1 of the Supporting Information). The largest variations in chemical emission factors among the LCI databases were for SOx and methane (150% for SOx and 70% for methane). Thus, the intermingling of different LCI databases for refinery products may lead to significantly different estimates for emissions of SOx (leading to different estimates for acid rain impact) and methane (leading to different estimates for global warming influence). Refinery waterborne emissions are generally much lower than air emissions. Waterborne emission factors were determined for chemical oxygen demand (COD), biochemical oxygen demand (BOD), nitrates, oils, hydrocarbons, and chloride, suspended solids, total dissolved solids, and individual ionic species. The data sets for diesel, PEMS naphtha, and EMPA naphtha exhibit comparatively larger water emissions, especially for oil (see Figure S2 of the Supporting Information for a subset of these results). This cannot be completely attributed to a larger fuel requirement for diesel since the proportions among the different chemical
parameters are not the same. Therefore, it can be concluded that more aqueous process emissions are assumed to occur for diesel in comparison to naphtha. As has been discussed previously, the methodology for allocation of emissions can produce very different results in a refinery (3). In general, the variations among databases for water emissions from the refinery were much larger than those for air emissions. The reason for this cannot be established just from the data as presented in the LCI databases. Part of the reason for the higher variation may be related to the generally lower amount of the water emissions (mg of chemical/kg of refinery product) than air emissions, resulting in a greater possibility of error. The water emission factors for chloride, suspended solids, dissolved solids, and sodium have such high variability (over 1000%) as to question whether such data are of much value in current impact analysis. Alternatively, the water and air emissions argue strongly for a revaluation of the important refinery LCI to achieve a less variable representative LCI. Further insight into the reliability of LCI emission factors can be gained by disaggregating the different sources of the emissions. The magnitude of the disaggregated of the emissions produced by heating, electricity generation, and processes make it clear that most of the air emissions are produced during heating and electricity generation, while most of the water emissions and solid waste are produced from refinery processes (Figure 3). Most of the emission distributions shown in Figure 3 produce results that are expected and consistent with the actual technology used in a refinery. However, the estimates shown for NOx are a notable exception. Refinery processes, as distinct from heating and electricity uses, produce no significant NOx emissions. Yet, when disaggregating the sources of NOx, most of the LCI databases predicted that the VOL. 34, NO. 22, 2000 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4793
FIGURE 2. Comparison of the total carbon dioxide emissions reported in the data sets analyzed.
FIGURE 3. Comparison of the contribution to air and water emissions. largest contribution (60-80%) is from refinery processes (Table 3). An explanation for this might be that the reported 4794
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 34, NO. 22, 2000
emission factors for nitrogen oxides derived from fuel burning are not consistent with the reported data for the refinery,
TABLE 3. NOx Emission Distribution for the Databases Analyzed total value (mg of NOx/kg of refinery product) PWMI South European PEMS South European PWMI Average European PEMS Average European EMPA naphtha ECOPRO naphtha SIMAPRO naphtha BUWAL 250 SIMAPRO naphtha, av Europe PEMS naphtha, av Europe Boustead naphtha, av Europe SIMAPRO diesel PEMS diesel EMPA diesel ECOPRO diesel FRANKLIN av European average median SD averageb a
c, calculated; g, given.
b
3000 3000 2900 2900 2665 2900 2900 2900 2888 2900 2900 2864 2001 2660 3987 2892 2900 376
av from electricity (%)
av from heat (%)
av from processes (%)
0.53 0.17 2.76 0.08 4.15 0.44 0.42 0.42 4.06 7.17 0.09 3.66 3.55 2.30 0.48 2.02 1.42 2.14 4.96
15.85 15.98 17.57 18.22 94.53 18.67 18.22 16.58 65.93 92.82 17.99 54.00 94.85 37.59 14.39 39.55 18.45 32.07 94.07
83.62 83.85 79.67 81.70 1.32 80.89 80.89 83.00 30.01 0.00 81.92 42.34 1.60 60.11 85.13 58.43 80.28 33.96 0.97
distributiona c c c c g g c c c g c c g c c
Average of boldface data.
FIGURE 4. Comparison of the solid waste reported in the data sets analyzed. therefore underestimating the emissions caused by fuel burning. Alternatively, the NOx emissions from fuel burning in support of process operations may in these databases been assigned to those processes instead of the heat generation. Three of the databases did report a majority of the NOx emissions attributable to burning fuel for heating (Boustead, EMPA naphtha, and EMPA diesel). For the NOx emission distribution shown in Figure 3, the average of Boustead, EMPA naphtha, and EMPA diesel was used (instead of the average of all LCI databases). The solid waste data presented are extremely irregular throughout the data sets. In some databases the waste is classified at a relatively high level of detail, in some it is aggregated into only a few classes, and some do not even report solid waste (Figure 4). Even when the databases use the same primary source of information, the classification
of solid waste is often rearranged, and there is not a clear definition of the kind of waste included. The least variation (30%) in solid waste emission factors among LCI data sets occurs when all types of solid waste were simply added together to estimate the total solid waste generated. This is done assuming that there is no reporting of solids in more than one category (i.e., double counting) and hence the sum is a reflection of the total. As a technical consideration, no percent moisture is reported; hence, it is very unclear what mass (dry solids or wet mass) is actually represented by these solid waste estimates. For now, it is best to utilize only the total solid waste from the databases in characterizing refinery products for LCI assessment. Assessment Based on Technical Issues. An examination of the emissions reported in all the LCI databases for a refinery can also be used to assess the underlying technical validity VOL. 34, NO. 22, 2000 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4795
of LCI information. Using the median values for the chemical constituent emissions to air, water, and as solids, a technical analysis was undertaken. For the larger air emissions, the results seem to indicate no chemical or technical inconsistency (such as a disproportionate ratio of SOx to CO2). The only inconsistency was found for metals: estimated metals emissions to the air are significantly lower than would be expected from combusting the same amount of oil and natural gas (based on the combustion emission factors estimates from ref 7; reported in Table S2 of the Supporting Information). Since the total amount of metals (principally from burning oil for heating) is very low even with the corrected emission factors from Table S2, the influence of this inconsistency may not have a great impact on life cycle refinery metal emissions estimates. In contrast, a number of technical inconsistencies were discovered for the water emission estimates. First, the median total dissolved solids (TDS) was found to be very much lower (14 mg/kg of refinery product) than the summation of the typically reported ions that comprise the actual TDS value according to the chemical protocol (12) (Ca2+ + Na+ + Cl+ SO4-- + K+ + Mg2+ ) 3600 mg/kg of refinery product). Second, the total organic carbon emission factor was 10-30 times as large (TOC around 500 mg/kg of refinery product) as the reported chemical oxygen demand (COD around 14 mg/kg of refinery product). Also the oil emission (around 1000 mg/kg of refinery product) was 10-70-fold larger than the reported chemical oxygen demand (COD around 14 mg). These results are again not technically possible based on the actual analytical test performed to measure these parameters. Independent search of direct refinery wastewater data indicates the TOC:COD is about 0.4 (2) and not the 10-30 reported in the databases. It is not possible with these technical inconsistencies to determine which parameters are actually correct, but the inconsistency does illustrate the need for more technical review of such databases in order to maintain credibility.
sets and graphs comparing water-borne emissions and some airborne emissions reported in the data sets (3 tables and 2 figures) (10 pages). This material is available free of charge via the Internet at http://pubs.acs.org.
Acknowledgments
(12) APHA-AWWA-WPCF. Standard Methods for the Examination of Water and Wastewater, 15th ed.; American Public Health Association: Washington, DC, 1980; pp 93-94.
Basic life cycle research support has been generously provided by Pfizer Inc. and SmithKline Beecham.
Supporting Information Available Tables showing the LCI data for electricity and thermal energy production, oil required as feedstock reported in the data
4796
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 34, NO. 22, 2000
Literature Cited (1) Amoco. Project Summary of Amoco-USEPA Pollution Prevention Project at Yorktown, VA; Amoco Environment, Health, and Safety: Chicago, IL, 1995; 126 pp. (2) Kim, S. Average of over ten published sources. NCSU internal report, Chemical Engineering Department, 1999, 15 pp. (3) Furuholt, E. Resour. Conserv. Recycl. 1995, 14, 251-263, (4) Clark, G.; de Leeuw, B. Int. J. Life Cycle Assess. 1999 4 (4), 184187. (5) Boustead. I. Eco-Profiles of the European Plastics Industry; Report 2, Olefin Feedstock Sources; A Report for the European Center for Plastics in the Environment (PWMI): Brussels, May 1993; pp 10-13. (6) PIRA International. PEMS Version 4.4; Surrey, England, 1998. (7) PRe´ Consultants. SIMAPRO Version 4.0; Amersfoort, The Netherlands, 1998. (8) Franke, M.; et al. Tenside Surfactants Deterg. 1995, 32 (5), 384396, (9) Frischknecht R.; Hofstetter P.; Knoepfel I.; Done R.; Zollingr E.; et al. ESU-ETHZ, 1994: O ¨ koinventare fu ¨ r Energisysteme; Laboratorium fu ¨ r Energiesysteme, Gruppe Energie-Stoffe-Umwelt, ETH Zu ¨ rich/PSI: Villingen, 1994; ESU-Reihe 1/94. (10) EMPA, Swiss Federal Laboratories for Materials Testing and Research. ECOPRO 1.5; St. Gallen, 1996. (11) Boustead, I. Naphtha Life Cycle Inventory Information; Boustead, I., Ed.; Boustead Consulting, Horsham, W. Susex, UK: 1999; pp 1-7.
Received for review October 5, 1999. Revised manuscript received July 31, 2000. Accepted August 3, 2000. ES991140F
