Input-Output-based Life Cycle Inventory: Development and Validation of A Database For The German Building Sector
Input-Output-based Life Cycle Inventory: Development and Validation of A Database For The German Building Sector
Input-Output-based Life Cycle Inventory: Development and Validation of A Database For The German Building Sector
y
n
, (2)
where
z
n
u
= cumulative emissions of airborne substance u (u =
CO
2
, CH
4
, etc.) for a product from sector n without
consideration of emissions from sectors to be elimi-
nated;
R = matrix that consists only of 1s except in rows and
columns for the commodities and industry sectors to
be eliminated, which are set to 0;
y
n
= coefcient of nal demand vector y; and
.* = element-by-element multiplication.
Validation
Approach
Validation of the IO-LCI data sets generated according to
equation (2) is performed by comparing them with the cor-
responding reference data sets of 106 building materials from
the Ecoinvent database. In this comparison, all of the statisti-
cal data used in further calculations as well as the data from
Ecoinvent refer to the same base year of 2003.
The Ecoinvent database contains several thousand generic
data sets for products or processes referring mainly to the
geographical area of Switzerland or (western) Europe. These
process-based data sets are updated regularly and are well doc-
umented (Frischknecht et al. 2004a, 2004b). There are two
types of data sets in the Ecoinvent database: unit processes and
system processes. Unit processes are single processes for which
input and output ows (elementary ows as well as product
ows from other unit processes) are specied in relation to the
functional ow. The results, which are the sums of the emis-
sions and resource inputs fromall the upstreamunit processes of
the product under consideration, are documented in a so-called
systemprocess. This is a product- or process-related data set that
contains all the directly and indirectly (i.e., cumulative) needed
resources and released emissions. Further details are provided
in the work of Frischknecht and colleagues (2004a, 2004b). To
compare these data sets to IO-LCI data sets that are calculated
according to equation (2), systemprocesses obviously have to be
used, as both data sets contain the cumulative emission outputs
of a product.
Validation is done by calculating the particular cumulative
emission data of IO-LCI data sets for the seven airborne emis-
sions and comparing these data with the cumulative emission
values from the corresponding system processes (M uller and
Schebek 2011). The quotients of relative deviations are cal-
culated for each building product according to equation (3):
q
u
=
e
u
p
u
, (3)
where
e
u
= cumulated emissions of gas u fromthe IO-LCI data set
(u = CO
2
, CH
4
, etc.); and
p
u
= cumulated emissions of gas u from the system process
(u = CO
2
, CH
4
, etc.).
Additionally, where q
u
= 1, the cumulative emissions for the
IO-based LCI data set and the system process are the same;
where q
u
> 1, the emissions from the IO-based LCI data set
are greater than the corresponding emissions from the system
processes, while where q
u
< 1 it is the other way around.
Results
The calculated quotients q
u
are illustrated as frequency dis-
tributions for all seven airborne emissions in gure 1. The spread
in the quotients for CO
2
emissions and NMVOCare quite sym-
metric, which cannot be observed for the other emissions. Yet
in contrast to CO
2
and NMVOC, there are more extreme de-
viations for other emissions, leading to some extreme outliers
upwards (maximum factor of 158.456 for NMVOC in contrast
to factor 11.735 for CO
2
) and downwards (minimum factor of
0.00018 for NMVOC in contrast to factor 0.00068 for CO
2
).
M uller and Schebek, IO-based LCI: Validation of German Database 507
RESEARCH AND ANALYSI S
0%
5%
10%
15%
20%
25%
30%
35%
40%
0 0.0625 0.125 0.25 0.5 1 2 4 8 more
up to up to up to up to up to up to up to up to up to than
0.0625 0.125 0.25 0.5 1 2 4 8 16 16
R
e
l
a
t
i
v
e
f
r
e
q
u
e
n
c
y
Emission quotients
0%
5%
10%
15%
20%
25%
30%
35%
40%
0 0.0625 0.125 0.25 0.5 1 2 4 8 more
up to up to up to up to up to up to up to up to up to than
0.0625 0.125 0.25 0.5 1 2 4 8 16 16
R
e
l
a
t
i
v
e
f
r
e
q
u
e
n
c
y
Emission quotients
0%
5%
10%
15%
20%
25%
30%
35%
40%
0 0.0625 0.125 0.25 0.5 1 2 4 8 more
up to up to up to up to up to up to up to up to up to than
0.0625 0.125 0.25 0.5 1 2 4 8 16 16
R
e
l
a
t
i
v
e
f
r
e
q
u
e
n
c
y
Emission quotients
(a)
(b)
(c)
Figure 1 Emission quotients relative frequency of (a) carbon dioxide (CO
2
) emissions; (b) methane (CH
4
) emissions; (c) ammonia (NH
3
)
emissions; (d) nitrous oxide (N
2
O) emissions; (e) nitrogen oxide (NO
x
) emissions; (f) sulfur dioxide (SO
2
) emissions; and (g) nonmethane
volatile organic compound (NMVOC) emissions. Subgure (h) shows the under/overestimation of emissions by input-output (IO)-based life
cycle inventory (LCI) data sets.
508 Journal of Industrial Ecology
RESEARCH AND ANALYSI S
0%
5%
10%
15%
20%
25%
30%
35%
40%
0 0.063 0.125 0.25 0.5 1 2 4 8 more
up to up to up to up to up to up to up to up to up to than
0.063 0.125 0.25 0.5 1 2 4 8 16 16
R
e
l
a
t
i
v
e
f
r
e
q
u
e
n
c
y
Emission quotients
0%
5%
10%
15%
20%
25%
30%
35%
40%
0 0.063 0.125 0.25 0.5 1 2 4 8 more
up to up to up to up to up to up to up to up to up to than
0.063 0.125 0.25 0.5 1 2 4 8 16 16
R
e
l
a
t
i
v
e
f
r
e
q
u
e
n
c
y
Emission quotients
0%
5%
10%
15%
20%
25%
30%
35%
40%
0 0.063 0.125 0.25 0.5 1 2 4 8 more
up to up to up to up to up to up to up to up to up to than
0.063 0.125 0.25 0.5 1 2 4 8 16 16
R
e
l
a
t
i
v
e
f
r
e
q
u
e
n
c
y
Emission quotients
(d)
(e)
(f)
Figure 1 Continued.
M uller and Schebek, IO-based LCI: Validation of German Database 509
RESEARCH AND ANALYSI S
0%
5%
10%
15%
20%
25%
30%
35%
40%
0 0.063 0.125 0.25 0.5 1 2 4 8 more
up to up to up to up to up to up to up to up to up to than
0.063 0.125 0.25 0.5 1 2 4 8 16 16
R
e
l
a
t
i
v
e
f
r
e
q
u
e
n
c
y
Emission quotients
(g)
(h)
Underestimation by
IO-based LCI data sets
Overestimation by
IO-based LCI data sets
Figure 1 Continued.
The frequency distributions of NH
3
, SO
2
, and NO
x
showa shift
toward lower quotients, which means that these emissions in
most of the IO-LCI data sets are underestimated compared to
those of the system processes. In contrast, the N
2
O emissions
are drifting toward higher quotients and are therefore overesti-
mated by the IO-based LCI. This means that in most cases the
calculated N
2
O emissions for a building product in the IO-LCI
data sets are greater than in the corresponding system processes
and that the NH
3
emissions are lower.
The reasons that system processes are underestimated by
the IO data sets are assumed to be at least partially systematic
in nature: (a) Not all economic transactions are gathered in
Germany. For example, companies with fewer than 20 employ-
ees do not have to report to the Federal Statistical Ofce. In
the case of the building sector, 30% of all companies do have
less than 20 employees (Gr omling 2011). (b) Emissions are not
completely available for all processes due to the specic aims
of national inventory reporting (NIR; balancing of the green-
house gases according to certain grouped sources of emissions)
and the corresponding requirements. In contrast, there are ob-
viously different reasons for the overestimates by data sets and
for the specic shape of the frequency distribution, which have
to be investigated in more detail.
Reasons for Deviations Between
Process-based and Input-Output-based
Life Cycle Inventory Data Sets
Generic Differences Between Process-Based and
Input-Output-Based Life Cycle Inventories
Due to their different methodological approaches, the
generic differences between process-based and IO-based LCIs
may contribute to the deviations in the results. Some of these
differences were eliminated by the calculation methodology
that we used. This applies specically to the issue of system
boundaries. As mentioned before, the system boundaries of
IO-based LCI are much broader than those of process-based
LCI. The reason for this is that statistical data comprises all the
activities in a national economy. In contrast to most process-
based LCI, in IO-based LCI the corresponding emissions for
services such as trade, consultancy, advertising, and nance
are enclosed within the system boundaries. Although this
may be considered a generic advantage of IO-based LCI, we
excluded these sectors in the IO tables in order to match the
system boundaries to those of process-based LCI and thus to
be able to compare them (see equation (2)). Still some generic
differences remain in our approach.
510 Journal of Industrial Ecology
RESEARCH AND ANALYSI S
Differences in the Allocation of Emissions
IO-based LCI is grounded on monetary IOtables and related
models. The physical ows of goods between the sectors are
accounted for in monetary units. Cumulative emissions for a
single product are calculated from the nal demand of a whole
sector by allocating the share of emissions in relation to the
share of the monetary value of this product as related to the
monetary value of the whole sector. Incontrast, inprocess-based
LCI all processes are connected by physical ows measured
in physical units. Emissions are calculated by determining the
sums of the amounts contributed by all the processes within
the process chain in relation to the demand for the respective
process in physical units. It is only in the special case of multi-
output processes that allocation at the level of that process may
be done either according to monetary values or according to
physical values.
Geographical Boundaries
Figures for direct emissions in a national IO model are taken
from the reported national data. The same emission factors are
applied to imported goods. Consequently, imports are treated
as if they were produced domestically and as if production pro-
cesses were performed under the same conditions and at the
same level of technology as the domestic state of the art (Moll
et al. 2004). This may lead to an underestimation of the emis-
sions fromintermediate inputs produced inless developed coun-
tries in IO-based LCI. In contrast, the country-specic produc-
tion conditions and corresponding emissions are usually taken
into consideration in process-based LCI data sets (Ecoinvent
2011).
Some other methodological reasons may be important with
regard to their contribution to deviations. This may be true
specically for the sector afliation of building products, which
is done top-down in IO-based LCI, in contrast to the calcula-
tion of an individual production system in process-based LCI.
Furthermore, the quality of the underlying primary data from
NIR is of course decisive for the resulting IO-based LCI data
sets. The results of a systematic evaluation of the extent to
which these reasons contribute to deviations are presented in
the following.
Systematic Errors
Deviations between IO-LCI data sets and those of system
processes could be the consequence of a systematic error in
the calculation procedure, such as the monetary allocation of
emissions at a sector level to single products in IO-based LCI.
To evaluate the occurrence of systematic errors, the respective
correlations of the relative deviations betweenpairs of the seven
airborne emissions were investigated by statistical regression
analysis. For example, the correlation of CH
4
emissions to CO
2
emissions is shown in gure 2, where a high correlation of
relative deviations is observed. In the case of a systematic error,
correlations between all emissions should be expected to be in
the same range of magnitude.
The results for all seven airborne emissions are shown in
table 1. The coefcients of determination R
2
as well as the
correlation coefcients R show the relationship between two
variables (CO
2
as an independent variable on the x-axis, the
other gases as dependent variables on the y-axis). The result-
ing magnitude of the relationship ranges between a very high
correlation and a weak correlation according to benchmarks
recommended in the work of Brosius (1999). This shows that
a systematic error (i.e., one single factor such as the allocation
of emissions according to monetary units) cannot be the reason
for the deviations observed. Systematic error can consequently
be excluded as the general reason for a deviation.
Geographic Boundaries
Data sets in the Swiss Ecoinvent database often refer to
a specic geographic region, whereas German IO-LCI data
sets refer exclusively to Germany. This made it possible to
y = 1.65x - 0.28
R
2
= 0.79
1E-04
1E-03
1E-02
1E-01
1E+00
1E+01
1E+02
1E-04 1E-03 1E-02 1E-01 1E+00 1E+01 1E+02
q
u
o
e
n
t
C
H
4
quoent CO
2
Figure 2 Scatter plot of the deviations of methane (CH
4
) emissions relative to carbon dioxide (CO
2
) emissions on the product level.
M uller and Schebek, IO-based LCI: Validation of German Database 511
RESEARCH AND ANALYSI S
Table 1 Coefcients of determination and correlation coefcients
Magnitude of the
Variables relationship
(quotients of gases) R
2
R (Brosius 1999)
x: CO
2
; y: CH
4
0.79 0.89 Very high correlation
(R: 0.81)
x: CO
2
; y: SO
2
0.42 0.64 High correlation
(R: 0.60.8)
x: CO
2
; y: NMVOC 0.22 0.47 Average correlation
(R: 0.40.6)
x: CO
2
; y: NO
x
0.21 0.45 Average correlation
(R: 0.40.6)
x: CO
2
; y: NH
3
0.08 0.28 Weak correlation
(R: 0.20.4)
x: CO
2
; y: N
2
O 0.05 0.23 Weak correlation
(R: 0.20.4)
Notes: CO
2
= carbon dioxide; CH
4
= methane; SO
2
= sulfur dioxide;
NMVOC = nonmethane volatile organic compounds; NO
x
= nitrogen
oxides; NH
3
= ammonia; N
2
O = nitrous oxide; R
2
= coefcient of deter-
mination; R = correlation coefcient.
Table 2 Analysis of geographic relations
Number of Number of
IO-LCI data IO-LCI data
sets with sets with
Ecoinvent data Number of deviations > deviations <
set relates to data sets Percentage factor 10 factor 0.1
Switzerland
(CH)
36 34% 10 12
Germany (DE) 11 10% 0 2
Global (GLO) 4 4% 0 2
Scandinavia
(NORDEL)
1 1% 0 1
Europe (RER) 54 51% 12 23
Total 106 100% 22 40
Notes: IO-LCI = input-output life cycle inventory.
analyze if deviations are caused mainly by a lack of geographic
representativeness. To do this we evaluated the frequency of
the quotients of deviation in relation to the geographic attribu-
tion of the data sets. Table 2 presents the percentage of IO-LCI
data sets deviating either by a factor of 10 or more or by less
than a factor of 0.1 related to the geographic attribution of the
respective Ecoinvent data sets.
Most system processes in the Ecoinvent database reect Eu-
ropean relationships (RER), followed by data sets representing
Swiss relationships (CH). Twelve of the RER system processes
deviate from the IO-LCI data sets by an order of magnitude of
more than 10, and 23 by a factor less than 0.1, resulting in a
heterogeneous picture. Looking at the Swiss system processes,
the number of deviations downward equals the number of devi-
ations upward. System processes with a global or Scandinavian
geographic relation (GLO or NORDEL) only show deviations
upward (and therefore an underestimation of the corresponding
IO-LCI data sets). The number of data sets, however, seems too
low to support valid reasoning.
The results provide evidence that information about the
geographic relation of a system process cannot indicate if the
emissions of a corresponding IO-LCI data set can be expected
to be higher or lower than process-based data sets.
Differences Due to the Sector Afliation of Building
Products
The sector afliation of products may contribute to devi-
ations between data sets in several ways. First, the top-down
approach of emission allocation in IO-LCI makes it compli-
cated to discern single products within a sector. This problem
may be more severe in sectors producing a large number of het-
erogeneous products than in those with a limited number of
commodities. Second, different sectors indirectly reect the in-
uence of the geographic distribution of production processes.
In Germany, a big share of raw materials, basic commodities,
and energy intensive intermediates is imported. As pointed out
above, in IO-based LCI imports are taken into consideration
by using the domestic emission factors, whereas process-based
Ecoinvent data sets take into consideration whether upstream
processes take place at other geographic locations. This could
lead to an underestimation of emissions in IO-LCI data sets
precisely in those sectors that have a large share of imported
commodities or intermediates. The contribution made by the
sector afliation of building products was analyzed by compar-
ing data sets for various sectors that exhibit levels of deviations
greater than one order of magnitude.
In table 3, the relative deviations of emissions from system
processes to IO-LCI data sets are assigned to sectors. Most build-
ing products are produced in sectors 26.2 to 26.8, Manufacture
of other non-metallic mineral products, except glass and glass
products (28%), followed by sector 24 (except 24.4), Manu-
facture of chemical products, except pharmaceutical products
(19%), and sectors 20, Manufacture of wood and products of
wood, cork (except furniture) and plaiting materials, and 27.4,
Manufacture of basic precious metals and metals clad with
precious metals (each 12%).
Only a rather small number of data sets in sectors 14, 20,
26.1, 26.2 to 26.8, and 32 show deviations. The arrows in table
3 show the direction of deviation. An up arrow indicates that
the overestimation by IO data predominates and a down arrow
indicates that the underestimation by IO data predominates.
More than 50% of the data sets in sectors 24 (except 24.4),
25, 27.1 to 27.3, and 27.4 deviate more than one order of
magnitude. Products from certain sectors show a comparable
cluster of deviations. Of note is that deviations in sectors with
only very few data sets cannot be statistically interpreted.
These results support the assumption that in most cases the
calculations of emissions in IO-LCI data sets are too low in
sectors with a large share of imported goods (raw materials, in-
termediates, or nal goods). This holds true for sectors 27.1 to
27.3, Manufacture of iron and steel, ferrous alloys, and tubes,
and sector 27.4, Manufacture of basic precious metals and
512 Journal of Industrial Ecology
RESEARCH AND ANALYSI S
Table 3 Analysis of relative deviations considering all seven emission types sorted according to sectors
Number of Number of
Number IO-LCI IO-LCI
of data data sets with data sets with
Sector and products (NACE code) sets deviations > factor 10 deviations < factor 0.1
14 Other mining and quarrying products 11
*
2 4
20 Wood and products of wood, cork (except furniture) and plaiting materials 13 0 5
24 (except 24.4) Chemical products, except pharmaceutical products 20
*
11 6
25 Rubber and plastic products 2
*
2 1
26.1 Glass and glass products 8 1 2
26.226.8 Other nonmetallic mineral products, except glass and glass products 30 3 9
27.127.3 Iron and steel, ferro-alloys, and tubes 8 1 5
27.4 Basic precious metals and metals clad with precious metals 13 2 8
32 Radio, television, and communication equipment and apparatus 1 0 0
Total 106 22 40
Notes: *Deviation of emissions upward as well as downward.
NACE = Statistical Classication of Economic Activities in the European Community; IO-LCI = input-output life cycle inventory.
metals clad with precious metals. Sectors with a small share of
imported goods in prechain manufacturing are building materi-
als in the narrow sense of the word (sectors 14, 26.1, and 26.2
to 26.8) and perform differently. In relation to the total num-
ber of data sets in these sectors, the share of deviations greater
than one order of magnitude is below 50% (see correspond-
ing sectors in table 3). Many of these products are regionally
produced and are transported only short distances due to their
weight (Gr omling 2011). The production conditions (and the
emission factors used in NIR) are therefore geographically well
represented.
A specic result can be observed for the deviations in sector
24 (except 24.4), Manufacture of chemical products, except
pharmaceutical products. Nearly 50% of all deviations of one
order of magnitude or more (and therefore an overestimation
of emissions) can be found in this sector. The reason for this is
that important processes in this sector such as the production
of ammonia, nitric acid, and adipic acid lead to rather high
emissions of N
2
O in the German NIR (Strogies and Gniffke
2010). According to the underlying method, in IO-based LCI
these direct emissions are evenly allocated over all chemical
products, corresponding to the product prices, even if there are
no N
2
O emissions at all in a particular process.
Quality of Primary Data from National Inventory
Reporting
In German environmental accounting, the direct emissions
for sectors are calculated by the Federal Statistical Ofce by
taking the emission factors for these products from the Ger-
man NIR (Strogies and Gniffke 2010) and multiplying them
by the corresponding activity rates (statistical data for the
amount of fuels and energy used in each sector and the amount
of products manufactured). Emission factors as well as activ-
ity rates in NIR often do not take all the processes occur-
ring in a national economy into consideration (Warsen 2009)
because only the NIRs primary data (emission factors) are
well surveyed for certain energy-induced emissions, whereas
they are only partially surveyed for mainly process-induced
emissions.
The quality of the primary data can be evaluated by using the
regression factors presented in the Systematic Errors section
(above). For the most signicant climate-relevant gases, namely
CO
2
and CH
4
(IPCC 2007), the emission factors and activity
rates inNIRare well surveyed since CO
2
is mainly set free by the
combustion of fossil fuels and CH
4
is set free by the combustion
of natural gas or hard coal. Compared withthe absolute numbers
for all other reported airborne emissions, CO
2
is emitted in by
far the greatest amounts (Destatis 2007a).
The data situation for CO
2
and CH
4
is therefore rather good
in German environmental accounting. This can also be seen in
the good correlation between CO
2
and CH
4
. The rather small
number of emission factors in NIR for NH
3
, SO
2
, and NO
x
might lead to uncertainties in the data for sector-related direct
emissions, resulting in the calculated cumulative emissions with
IO-based LCI being underestimated.
Discussion
The systematic evaluation of differences between process-
based and IO-LCI data sets reveals that deviations cannot be
tracked to a systematic error and are not caused by one single
effect in the calculation methodology. Instead, sector-specic
effects seem to be a main reason for the deviations.
First, the use of domestic emission factors leads to greater
deviations in those sectors that have a large share of imports.
Examples are sectors 27.1 to 27.3, Manufacture of iron and
steel, ferro-alloys, and tubes, where deviations are large, and
sector 26.1, Manufacture of glass and glass products, where
they are small. Second, calculating the specic emissions of a
product with IO-based LCI leads to a blur of emissions that
are set free by a specic process, which, however, cannot be
discerned within a sector. The reason for this is that the level
M uller and Schebek, IO-based LCI: Validation of German Database 513
RESEARCH AND ANALYSI S
of detail in German statistical data is still rather aggregate. One
example is sector 24 (except 24.4), Manufacture of chemical
products, except pharmaceutical products, where nearly 50%
of all the deviations of one order of magnitude or more (and
therefore an overestimation of emissions) can be found. This
concerns important processes such as the production of ammo-
nia, nitric acid, and adipic acid, which leads to emissions of
N
2
O in rather large amounts. The direct emissions for sector 24
(except 24.4) are calculated by the German Federal Statistical
Ofce by taking the emission factors from the German NIR
(Strogies and Gniffke 2010) for these products and multiply-
ing them by the corresponding activity rates. In IO-based LCI
these direct emissions are evenly allocated over all chemical
products, even if there are no N
2
O emissions in the specic
process. A possible solution to this problem would be for Ger-
many to introduce more detailed national accounting as well as
environmental accounting comprising more than just 71 aggre-
gated sectors. A rst important step could be the disaggregation
of the chemical industry (sector 24 except 24.4), followed by
a disaggregation of the German economy into more than 71
sectors.
Furthermore, the underlying primary data from the NIR are
crucial for calculating IO-based LCI data. This is true for the
very basic fact that at present only seven emissions are avail-
able for use in IO-based LCI calculations. We demonstrated,
however, that the extent of quality assurance is different even
for these seven emissions.
It is obvious that the intensity of inspection can differ de-
pending on the necessity of the respective reporting and seems
to be highest for climate reporting, resulting in data of higher
quality. This of course affects the quality of the calculated IO-
based LCI data sets. A very strong relationship was observed
between CO
2
and CH
4
, and the values of the outliers (M uller
2011) were also not as extreme as those of other airborne emis-
sions from environmental accounting. This leads to the con-
clusion that the primary data for emission factors and activity
rates used to calculate the direct CO
2
and CH
4
emissions in en-
vironmental accounting are of high quality because they reveal
very good knowledge of the sector-specic application of fossil
fuels and therefore of the energy-related emissions.
Conclusions
In our case study, we provided data sets for building products
in Germany; data sets for other sectors in Germany may be gen-
erated in the same way. Our validation results show, however,
that there are large variations and deviations in IO-LCI data
sets compared to process-based LCI data, which of course is a
serious problem for the reliability of IO-based LCI.
Our results of a systematic evaluation of deviations between
IO-LCI data and process-related data indicate that our default
expectation is that for German IO-LCI data sets, CO
2
, CH
4
,
and NMVOC emissions are uniformly distributed; that SO
2
,
NO
x
, and NH
3
are predominantly underestimated; and only in
one case (i.e., for N
2
O) emissions are mainly overestimated as
to the sector of chemical products.
These ndings are contrary to the ndings from embodied
energy analysis, where up to 87%of so-called truncation was re-
ported (Crawford 2005, 2008). Crawford (2005, 2008) showed
that in most cases embodied energy in process data is lower
than for IOdata. Regarding cumulated emissions, these ndings
cannot be conrmed. Although we excluded service sectors in
IO tables, the results for cumulated emissions calculated with
the German IO model compared with system processes were
overestimated in many cases (e.g., see N
2
O emissions for most
products).
Reasons for the general overestimation of Australian IO
data sets could be that Crawford (2005) used rather old data
from the mid-1990s taken from an Australian material energy
intensity database (see Grant [2000]), which is probably in-
complete and outdated regarding all the prechains of a building
product.
Possible generic reasons for underestimation of emissions
calculated with the German IO model have been discussed
above. They are rooted in the methodology of IO and NIR,
which does not include all economic transactions in Germany
and excludes some process-based emissions due to the specic
aims of NIR. In addition to these generic reasons, the underes-
timation of emissions also occurs precisely in those sectors that
have a large share of imported raw materials or commodities
because of the assumption that domestic emission factors can
also be employed for imports. In contrast, overestimation seems
to occur for reasons specic to one sector and one emission.
This was demonstrated for N
2
O emissions within the chemical
sector.
Not surprisingly, the quality and level of detail of the under-
lying primary data are of crucial importance for the quality of
IO-based LCI data sets. Our results show that different levels of
quality assurance of data for specic emissions, due to reporting
obligations, clearly inuence the general level of scattering of
IO-based LCI data sets.
The consequences for the improvement of IO-based LCI can
be derived from this. In the rst place, of course, IO-based LCI
will benet from the ongoing improvement and expansion of
national reporting data. Data for more elementary ows in air,
water, and soil should be provided to expand the environmental
accounting tables. A supplemental database for Germany could
be the Pollutant Release and Transfer Register (PRTR), which
is accessible via the Internet and makes data available for a total
of 91 substances (UBA 2012). Analogous emissions registers
exist in other countries. The value of such registers for assessing
industry classes has been demonstrated (Wursthorn et al. 2010).
Beyond this, two issues seem of major importance for IO-
based LCI. The rst is that the further disaggregation of IO
tables would lead to a better correspondence between goods
and more specic sectors (and also the corresponding emis-
sion factors and prechains). Second, country-specic emissions
factors for imports would allow a more realistic calculation of
indirect emissions fromimports. Further researchshould be con-
centrated on combining the macroeconomic statistical and the
environmental data of several countries ina global environmen-
tal IO model (Narayanan and Walmsley 2008; Tukker 2011).
514 Journal of Industrial Ecology
RESEARCH AND ANALYSI S
In summary, a high degree of uncertainty has to be taken
into account if IO data sets are used today. They do, nonethe-
less, provide a way to generate a rst estimate for data gaps
where a basic understanding of the possible reasons and the
areas of deviation helps to make use of this information feasible
in a sensible way. The primary impact of the use of IO-based
data sets for building products will be to foster a comprehensive
assessment of the contribution that the construction phase
and its upstream processes make to the overall environmental
performance of a building. Applicability by practitioners will
be facilitated by inclusion of IO-based data sets in information
technology tools specically designed for architects and
planners, which can easily be done, as we have shown for the
example of the planning tool LTE OGIP (M uller 2011). In
addition, this also enables coupling to economic information.
Last but not least, the generic advantage of IO-based LCI as a
low-budget, fast, and feasible means for creating generic data
sets for LCA will make data sets easier to update. Additional
information resulting from broadening the scope of emissions
in accounting can be expected to be available in the coming
years. This may stimulate the practical use of IO-based data
sets by LCA practitioners to handle data gaps.
References
BMVBS (Bundesministerium f ur Verkehr, Bau und Stadtentwick-
lung). 2001. Leitfaden Nachhaltiges Bauen [Code of practice for
sustainable building construction]. www.nachhaltigesbauen.de/
leadmin/pdf/PDF_Leitfaden_Nachhaltiges_Bauen/Leitfaden.
pdf. Accessed March 2012.
BMVBS (Bundesministerium f ur Verkehr, Bau und Stadtentwick-
lung). 2011. oekobau.dat [Database oekobau.dat]. www.
nachhaltigesbauen.de/baustoff-und-gebaeudedaten/oekobaudat.
html. Accessed March 2012.
Brosius, F. 1999. SPSS 8.0: Professionelle Statistik unter Windows [SPSS
8.0: Professional Statistics in Windows]. Bonn, Germany: MITP.
Carnegie Mellon. 2011. Available EIO-LCA models. www.
eiolca.net/Models/index.html. Accessed March 2012.
Crawford, R. H. 2005. Validation of the use of input-output data for
embodied energy analysis of the Australian construction industry.
Journal of Construction Research 6(1): 7190.
Crawford, R. H. 2008. Validation of a hybrid life cycle inventory analy-
sis method. Journal of Environmental Management 88(3): 496506.
Destatis (Statistisches Bundesamt). 2005. Produzierendes Gewerbe
Produktion im Produzierenden Gewerbe 2004. Fachserie 4/
Reihe 3.1 [Industry - production in industry 2004. Series No. 4 /
3.1]. Wiesbaden, Germany: Statistisches Bundesamt.
Destatis (Statistisches Bundesamt). 2007a. Umweltnutzung und
Wirtschaft Tabellen zu den Umwelt okonomischen
Gesamtrechnungen 2007 [Utilization of the environment
and economy - tables of environmental accounting 2007].
Wiesbaden, Germany: Statistisches Bundesamt.
Destatis (Statistisches Bundesamt). 2007b. Volkswirtschaftliche
Gesamtrechnungen Input-Output Rechnung 2003. Fachserie 18 /
Reihe 2 [Economic accounting - input-output calculation 2003.
Series No. 18 / 2]. Wiesbaden, Germany: Statistisches Bundesamt.
Ecoinvent. 2011. Ecoinvent database. www.ecoinvent.ch/. Accessed
August 2011.
EPA (U.S. Environmental Protection Agency). 2012. Green building.
www.epa.gov/greenbuilding/pubs/about.htm. Accessed March
2012.
Frischknecht, R., N. Jungbluth, H.-J. Althaus, G. Doka, R. Dones, T.
Heck, S. Hellweg, R. Hischier, T. Nemecek, G. Rebitzer, and
M. Spielmann. 2004b. Overview and methodology. Ecoinvent re-
port no. 1. D ubendorf, Switzerland: Swiss Centre for Life Cycle
Inventories.
Frischknecht, R., N. Jungbluth, H.-J. Althaus, G. Doka, R. Dones, T.
Heck, S. Hellweg, R. Hischier, T. Nemecek, G. Rebitzer, and
M. Spielmann. 2004a. The Ecoinvent database: Overview and
methodological framework. International Journal of Life Cycle As-
sessment 10(1): 39.
Grant, T. 2000. Australian material inventory database of life cycle assess-
ment value for materials, Melbourne, Australia: Royal Melbourne
Institute of Technology.
Gr omling, M. 2011. Ein volkswirtschaftliches Portr at der
deutschen Baustofndustrie [Economic gures on the con-
struction material sector]. www.baustofndustrie.de/root/img/
pool/downloads_2011/branchenportrait_2011.pdf. Accessed
March 2012.
Guin ee, J. B., M. Gorr ee, R. Heijungs, G. Huppes, R. Kleijn,
A. de Koning, L. van Oers, et al. 2002. Handbook on
life cycle assessment: Operational guide to the ISO standards,
1st ed. Berlin, Germany: Springer.
Hendrickson, C. T., L. B. Lave, and H. S. Matthews. 2006. Environ-
mental life cycle assessment of goods and services. Washington, DC,
USA: Resources for the Future.
Holub, H. and H. Schnabl. 1994. Input-output-rechnung: Input-output-
tabellen. [Input-output accounting: Input-output tables.] Munich,
Germany: Oldenbourg.
Huijbregts, M. A. J., G. Norris, R. Bretz, A. Ciroth, B. Maurice, B. von
Bahr, B. Weidema, and A. S. H. de Beaufort. 2001. Framework for
modelling data uncertainty in life cycle inventories. International
Journal of Life Cycle Assessment 6(3): 127132.
IPCC (Intergovernmental Panel on Climate Change). 2007. Cli-
mate change 2007: Synthesis report summary for policymakers.
www.ipcc.ch/pdf/assessment-report/ar4/syr/ar4_syr_spm.pdf. Ac-
cessed March 2012.
ISO (International Organization for Standardization) 2006a. Environ-
mental management Life cycle assessment Principles and frame-
work (ISO 14040:2006). Berlin, Germany: Beuth.
ISO (International Organization for Standardization). 2006b. Envi-
ronmental management Life cycle assessment Requirements and
guidelines (ISO 14044:2006). Berlin, Germany: Beuth.
Kohler, H. 1999. Stoffstr ome und Kosten in den Bereichen Bauen und
Wohnen [Mass ows and costs in the areas of building and habi-
tation]. Berlin, Germany: Springer.
Kreissig, J. and M. Binder. 2007. Aktualisieren, Fortschreiben und Har-
monisieren von Basisdaten f ur das Nachhaltige Bauen [Update and
harmonizationof basic data for sustainable building construction].
Leinfelden-Echterdingen, Germany: PE International.
Lenzen, M. 2000. Errors in conventional and input-output-
based life-cycle inventories. Journal of Industrial Ecology 4(4):
127148.
Lenzen, M., and G. Treloar. 2002. Differential convergence of life-
cycle inventories towards upstream production layers. Journal of
Industrial Ecology 6(34): 137160.
Leontief, W. W. 1970. Environmental repercussions and the economic
structure: An input-output approach. Review of Economics and
Statistics 52(3): 261271.
M uller and Schebek, IO-based LCI: Validation of German Database 515
RESEARCH AND ANALYSI S
Lichtenvort, K. 2004. Systemgrenzenrelevante
Anderungen von Fluss-
mengen in der
Okobilanzierung [System-relevant changes in ele-
mentary ows in life cycle assessment]. Ph.D. dissertation, Tech-
nische Universit at Berlin, Berlin, Germany.
Loerincik, Y. 2006. Environmental impacts and benets of informa-
tion and communication technology infrastructure and services,
using process and input-output life cycle assessment. Ph.D. dis-
sertation, Ecole polytechnique f ed erale de Lausanne, Lausanne,
Switzerland.
Miller, R.E. and P. D. Blair. 1985. Input-output analysis: Foundations
and extensions. Upper Saddle River, NJ, USA: Prentice-Hall.
Moll, S., J. Acosta, and A. Villanueva. 2004. Environmental implications
of resource use Insight from input-output analysis. Copenhagen,
Denmark: European Environment Agency, European Topic Cen-
tre on Waste and Material Flows.
M osle, P. 2010. Green building trends in Europe. London, UK: Ecobuild
London.
M uller, B. 2011. Weiterentwicklung und Validierung der Meth-
ode der Input-Output basierten Sachbilanz f ur deutsche
Geb aude okobilanzen [Enhancing and validating the method of
input-output based life cycle inventory in life cycle assessment of
buildings]. Ph.D. dissertation, Technische Universit at Darmstadt,
Darmstadt, Germany.
M uller, B. and L. Schebek. 2011. Closing data gaps in LCI based on
environmental IOA: A case study for German building products.
Poster presentation at the 2011 Conference of the International
Society for Industrial Ecology, Berkeley, California, 8 June.
Narayanan, G. and T. L. Walmsley. 2008. Global trade, assistance, and
production: The GTAP 7 data base. West Lafayette, IN, USA:
Center for Global Trade Analysis, Purdue University.
Rebitzer, G. 2005. Enhancing the application efciency of life cycle
assessment for industrial use. Ph.D. dissertation, Ecole polytech-
nique f ed erale de Lausanne, Lausanne, Switzerland.
Strogies, M. and P. Gniffke. 2010. Submission under the United
Nations Framework Convention on Climate Change and the
Kyoto Protocol 2010 National inventory report for the German
greenhouse gas inventory, 19902008. www.umweltbundesamt.
de/uba-info-medien/mysql_medien.php?anfrage=Kennummer&
Suchwort=3958. Accessed March 2012.
Suh, S. and G. Huppes. 2005. Methods for life cycle inventory of a
product. Journal of Cleaner Production 13(7): 687697.
Suh, S., M. Lenzen, G. J. Treloar, H. Hondo, A. Horvath, G. Huppes,
O. Jolliet, U. Klann, W. Krewitt, Y. Moriguchi, J. Munksgaard,
and G. Norris. 2004. System boundary selection in life-cycle in-
ventories. Environmental Science & Technology 38(3): 657664.
Treloar, G. 1997. Extracting embodied energy paths from input-output
tables: towards an input-output-based hybrid energy analysis
method. Economic Systems Research 9(4): 375391.
Tukker, A. 2011. EXIOBASE 1.0: Its creation and rst results. www.
lcaforum.ch/portals/0/df45/DF45--11%20Arnold%20Tukker.
pdf. Accessed October 2011.
UBA (Umweltbundesamt). 2012. PRTR-Protokoll: Protokoll uber
Schadstofffreisetzungs und verbringungsregister [Pollutant
release and transfer registry]. www.home.prtr.de/download/
PRTR-Protokoll.pdf. Accessed March 2012.
UKGBC (UK Green Building Council). 2012. Sustainable Built Envi-
ronment. www.ukgbc.org. Accessed March 2012.
Warsen, J. 2009. Validierung von Stoffussdaten in der
Okobilanz
durch Daten aus dem offentlichen Berichtswesen. Ph.D.
dissertation, Technische Universit at Darmstadt, Darmstadt,
Germany.
Weidema, B. P., A. M. Nielsen, K. Christiansen, G. Norris, P.
Notten, S. Suh, and J. Madsen. 2005. Prioritisation within
the integrated product policy. Copenhagen, Denmark: Dan-
ish Ministry of the Environment, Environmental Protection
Agency.
Wursthorn, S., W.-R. Poganietz, and L. Schebek. 2010. Economic-
environmental monitoring indicators for European coun-
tries: A disaggregated sector-based approach for mon-
itoring eco-efciency. Ecological Economics 70(3): 487
496. http://dx.doi.org/10.1016/j.ecolecon.2010.09.033. Accessed
March 2012.
WorldGBC (World Green Building Council). 2012. www.worldgbc.
org/worldgbc/about/. Accessed March 2013.
About the Authors
Bodo M uller was a doctoral student at Karlsruhe Institute
of Technology (KIT), Institute of Technology Assessment and
Systems Analyses (ITAS) in Karlsruhe, Germany, where he
performed the investigations presented in this article. He re-
ceived his Ph.D. in 2011 at the Technische Universit at Darm-
stadt, Darmstadt, Germany, and is currently working as Senior
Manager for Sustainability Evaluation for a chemical company.
Liselotte Schebek is a professor at the Technische Universit at
Darmstadt, and holds the chair in industrial material cycles in
the Faculty of Civil Engineering and Geodesy. She was also af-
liated with the KIT ITAS at the time this article was written.
Supporting Information
Additional supporting information may be found in the online version of this article at the publishers web site:
Supporting Information S1: This supporting information (table S1) is an excerpt of the 284 input-output life cycle
inventory (IO-LCI) data sets that were calculated either to ll data gaps in
Okobau.dat or to be used for validation purposes
with the Ecoinvent data.
Table S1 Input-output life cycle inventory (IO-LCI) data sets calculated for validationand for lling data gaps in
Okobau.dat
(excerpt).
516 Journal of Industrial Ecology