1
00:00:00,159 --> 00:00:10,159
[MUSIC]

2
00:00:18,573 --> 00:00:21,005
So far, you've been provided with all of
the

3
00:00:21,005 --> 00:00:23,949
inventory data you've used in your bottled
soda LCA,

4
00:00:23,949 --> 00:00:26,573
which is helping you focus on learning the
basics

5
00:00:26,573 --> 00:00:30,300
of inventory analysis and on building out
your model.

6
00:00:30,300 --> 00:00:33,180
In practice, however, you'll often
encounter data

7
00:00:33,180 --> 00:00:36,290
gaps when compiling a unit process
inventory.

8
00:00:36,290 --> 00:00:38,990
For example, you may find that important
flow data

9
00:00:38,990 --> 00:00:43,530
are missing from existing inventory
sources, or even more commonly,

10
00:00:43,530 --> 00:00:48,080
that an inventory does not even exist for
one or more of your unit processes.

11
00:00:48,080 --> 00:00:49,890
What do you do when this happens?

12
00:00:49,890 --> 00:00:53,580
Of course, if you have direct access to
the actual processes

13
00:00:53,580 --> 00:00:57,200
you can try to collect primary data to
fill your data gaps.

14
00:00:57,200 --> 00:01:01,473
But in many LCA studies, we don't have
access to the real world processes.

15
00:01:01,473 --> 00:01:04,830
And so primary data collection isn't an
option.

16
00:01:04,830 --> 00:01:08,640
And even though today's commercial and
public LCA databases contain a lot

17
00:01:08,640 --> 00:01:11,610
of great inventory data, they do not come
close to being

18
00:01:11,610 --> 00:01:16,010
inclusive of many materials, processes and
products you may like to study.

19
00:01:17,030 --> 00:01:19,860
For example, perhaps you're studying a
specialty

20
00:01:19,860 --> 00:01:22,530
material that is not used widely, or

21
00:01:22,530 --> 00:01:24,890
perhaps you're studying a new material or

22
00:01:24,890 --> 00:01:28,300
unit process for which data don't yet
exist.

23
00:01:28,300 --> 00:01:30,970
Or, as is most common, you may simply be

24
00:01:30,970 --> 00:01:33,750
studying one of the many thousands of
materials and unit

25
00:01:33,750 --> 00:01:35,555
processes that are in use in the real

26
00:01:35,555 --> 00:01:38,510
world but haven't been subjected to an LCA
yet.

27
00:01:39,542 --> 00:01:41,440
As you've learned, a good life cycle

28
00:01:41,440 --> 00:01:45,510
inventory takes time, planning and money
to compile.

29
00:01:45,510 --> 00:01:48,410
So, existing inventories are limited.

30
00:01:48,410 --> 00:01:52,860
So, if you practice LCA, dealing with data
gaps comes with the territory.

31
00:01:54,120 --> 00:01:56,655
Today we'll discuss using analysis based

32
00:01:56,655 --> 00:01:59,930
estimation methods to generate inventory
data.

33
00:01:59,930 --> 00:02:03,010
By analysis based, what I mean is that we
used

34
00:02:03,010 --> 00:02:08,180
reasoned data analysis or engineering
methods to estimate inventory data.

35
00:02:08,180 --> 00:02:11,350
In other words, we don't make estimates
based on guess work

36
00:02:11,350 --> 00:02:15,280
or intuition, which can seriously reduce
the credibility of an analysis.

37
00:02:16,850 --> 00:02:19,330
Let me give you three quick examples of
analysis

38
00:02:19,330 --> 00:02:23,580
based estimation which you'll also
practice in your homework assignment.

39
00:02:23,580 --> 00:02:24,980
The first is an example of

40
00:02:24,980 --> 00:02:29,080
what some might call a black box inventory
where we assemble

41
00:02:29,080 --> 00:02:31,984
consistent inputs and outputs without
explicitly

42
00:02:31,984 --> 00:02:34,520
characterizing the functional
relationships between them.

43
00:02:35,520 --> 00:02:39,520
Such inventories can sometimes be
constructed using environmental statistics

44
00:02:39,520 --> 00:02:42,690
such as those issued by regional or
national government agencies.

45
00:02:43,720 --> 00:02:47,020
In my case I'll construct a black box
inventory

46
00:02:47,020 --> 00:02:50,420
for diesel fueled, light trucks in the
United States using

47
00:02:50,420 --> 00:02:53,940
national fleet, energy use and emissions
data.

48
00:02:53,940 --> 00:02:55,930
My data source is the US Department of

49
00:02:55,930 --> 00:03:00,320
Energy's Transportation Energy Data Book,
which is issued annually.

50
00:03:00,320 --> 00:03:01,950
The year of my data is 2005,

51
00:03:01,950 --> 00:03:06,140
and the geographical coverage is the
United States.

52
00:03:06,140 --> 00:03:10,670
The technology mix is the average of all
light diesel trucks in operation.

53
00:03:12,160 --> 00:03:15,830
Here are the data, which includes total
passenger miles traveled,

54
00:03:15,830 --> 00:03:17,880
diesel fuel use and emissions of

55
00:03:17,880 --> 00:03:22,230
CO2, carbon monoxide, particulate matter,
volatile organic

56
00:03:22,230 --> 00:03:29,350
compounds, nitrogen oxides, and sulfur
dioxide for the 2005 US light truck fleet.

57
00:03:29,350 --> 00:03:33,450
To construct the inventory, I'll choose a
product output of one passenger

58
00:03:33,450 --> 00:03:38,650
mile, and normalize all my input and
output data on this basis.

59
00:03:38,650 --> 00:03:40,990
My estimated inventory then looks like
this.

60
00:03:42,880 --> 00:03:46,960
Such an approach typically only works for
widespread technologies or

61
00:03:46,960 --> 00:03:52,100
entire industries for which all relevant
flow data are consistently tracked.

62
00:03:52,100 --> 00:03:55,450
Note that when using this approach, you
must be sure that it

63
00:03:55,450 --> 00:03:59,610
is acceptable in light of the data quality
requirements of your study.

64
00:03:59,610 --> 00:04:03,300
For example, this sort of estimated
inventory might be

65
00:04:03,300 --> 00:04:06,360
acceptable as a background process in a
LCA Model.

66
00:04:07,390 --> 00:04:08,070
Note also

67
00:04:08,070 --> 00:04:11,990
that we don't know the relationships
between inputs and outputs.

68
00:04:11,990 --> 00:04:16,685
For example, if we change the amount of
fuel used per passenger mile we

69
00:04:16,685 --> 00:04:21,700
don't know by how much emissions of VOC's
per passenger mile should also change.

70
00:04:21,700 --> 00:04:23,700
Hence, the name black box model.

71
00:04:24,840 --> 00:04:27,220
Another important consideration is that
one needs

72
00:04:27,220 --> 00:04:30,700
to establish that all relevant flows are
captured.

73
00:04:30,700 --> 00:04:33,350
In this case I'm only capturing combustion

74
00:04:33,350 --> 00:04:34,830
related air missions.

75
00:04:34,830 --> 00:04:38,380
So I need to document that any other
relevant flows such

76
00:04:38,380 --> 00:04:42,530
as waste engine oil discarded per
passenger mile are not considered.

77
00:04:42,530 --> 00:04:43,030
The

78
00:04:45,060 --> 00:04:49,010
second example is using verified emission
factors for combustion

79
00:04:49,010 --> 00:04:52,710
related unit processing such as furnaces,
kilns or boilers.

80
00:04:53,900 --> 00:04:56,870
Recall that I use this approach earlier in
the course, when I

81
00:04:56,870 --> 00:05:00,908
estimated the unit process inventory for
operating a residential hot water heater.

82
00:05:00,908 --> 00:05:05,097
I obtained air emission factors from the
US environmental protection

83
00:05:05,097 --> 00:05:10,138
agencies, AP 42 emission factor databases
which contains emission factors from

84
00:05:10,138 --> 00:05:12,978
many different combustion technologies
based

85
00:05:12,978 --> 00:05:15,500
on test data and engineering methods.

86
00:05:16,670 --> 00:05:18,700
One can also find emission factors from

87
00:05:18,700 --> 00:05:22,390
other environmental agencies and in the
engineering literature.

88
00:05:22,390 --> 00:05:25,690
Note that this approach typically only
captures air emissions.

89
00:05:25,690 --> 00:05:28,220
Which are the primary outputs of most
combustion processes.

90
00:05:29,250 --> 00:05:32,470
Other flows, such as disposal of recovered
ash or

91
00:05:32,470 --> 00:05:35,300
other pollutant abatement flows, would
have to be estimated

92
00:05:35,300 --> 00:05:37,050
from other sources.

93
00:05:37,050 --> 00:05:41,110
Thus, like our black box model the
emission factor approach applies

94
00:05:41,110 --> 00:05:46,670
to generic technology populations and is
thus best suited for background processes.

95
00:05:46,670 --> 00:05:49,750
However, the emission factor approach does
capture the

96
00:05:49,750 --> 00:05:54,170
functional relationships between fuel
inputs and air emission outputs.

97
00:05:54,170 --> 00:05:57,830
And thus, it is more flexible than a black
box approach.

98
00:05:57,830 --> 00:06:00,560
The third example is using an engineering
approach

99
00:06:00,560 --> 00:06:03,310
to quantify the functional relationship
between unit

100
00:06:03,310 --> 00:06:07,510
process inputs and outputs, using a
process model.

101
00:06:07,510 --> 00:06:10,380
For example, let's say I want to construct
a unit

102
00:06:10,380 --> 00:06:15,240
process inventory for an industrial
blower, used in a pneumatic conveyor.

103
00:06:15,240 --> 00:06:20,330
From basic energy engineering, I know that
the relationship between fan input power

104
00:06:20,330 --> 00:06:26,430
and fan output power is HP equals CFM
times PSI

105
00:06:26,430 --> 00:06:30,850
divided by 229 times the efficiency of the
fan.

106
00:06:30,850 --> 00:06:33,770
Where HP is fan horse power.

107
00:06:33,770 --> 00:06:37,130
CFM is the cubic feet per minute of air
flow.

108
00:06:37,130 --> 00:06:39,040
PSI is the pounds per square inch of

109
00:06:39,040 --> 00:06:42,280
air pressure and 229 is a unit conversion
factor.

110
00:06:43,330 --> 00:06:47,060
If I have some understanding of the
process I am analyzing I can

111
00:06:47,060 --> 00:06:51,450
estimate the CFM and PSI and use a typical
fan efficiency for the

112
00:06:51,450 --> 00:06:54,220
application from the engineering
literature.

113
00:06:54,220 --> 00:06:55,610
I can then convert to kilowatt

114
00:06:55,610 --> 00:06:58,430
hours of electricity required to provide
pressurized

115
00:06:58,430 --> 00:07:02,680
air at a given flow rate and pressure for
a given amount of time.

116
00:07:03,840 --> 00:07:06,540
This is an admittedly simple example.

117
00:07:06,540 --> 00:07:09,400
In practice, engineering estimation can
become

118
00:07:09,400 --> 00:07:11,640
more complicated for more complex
processes.

119
00:07:12,720 --> 00:07:16,370
The point is that if one can characterize
the underlying physics,

120
00:07:16,370 --> 00:07:20,800
or chemistry of a process in a way that
conserves mass, and energy it is

121
00:07:20,800 --> 00:07:24,020
often possible to construct engineering
based inventories

122
00:07:24,020 --> 00:07:26,060
when no other sources of data are
available.

123
00:07:27,060 --> 00:07:29,780
However, with this method it's critically
important

124
00:07:29,780 --> 00:07:32,310
that the engineering functions reflect the
real world

125
00:07:32,310 --> 00:07:34,850
equipment characteristics including
efficiency

126
00:07:34,850 --> 00:07:38,100
losses Thus, engineering estimates

127
00:07:38,100 --> 00:07:40,880
are best left to engineers with process
knowledge.

128
00:07:42,130 --> 00:07:43,250
Lastly,

129
00:07:43,250 --> 00:07:45,700
I want to point out that all of these
methods can

130
00:07:45,700 --> 00:07:49,870
also be helpful for checking unit process
inventory data throughout a study.

131
00:07:49,870 --> 00:07:53,910
As I mentioned previously, data validation
is an

132
00:07:53,910 --> 00:07:57,320
important step in the life cycle inventory
process.

133
00:07:57,320 --> 00:08:02,360
It is always a good idea to double check
all inventory data using logic,

134
00:08:02,360 --> 00:08:04,790
comparing the black box data, comparing to

135
00:08:04,790 --> 00:08:08,550
emission factors or comparing to
engineering models.

136
00:08:08,550 --> 00:08:11,480
For example, lets check the CO2 flow in

137
00:08:11,480 --> 00:08:15,260
my unit process inventory for diesel
fueled light trucks.

138
00:08:15,260 --> 00:08:18,070
Using an air emission factor for diesel
fuel combustion.

139
00:08:19,100 --> 00:08:22,530
According to the U.S. EPA, on average, a
liter of

140
00:08:22,530 --> 00:08:28,290
diesel fuel will release 2.7 kilograms of
CO2 during combustion.

141
00:08:28,290 --> 00:08:30,410
If I apply that emission factor to the
diesel

142
00:08:30,410 --> 00:08:33,980
fuel and CO2 data in my black box model,

143
00:08:33,980 --> 00:08:38,610
I can see that the CO2 flow value seems
reasonable for this dated fuel input.

144
00:08:40,260 --> 00:08:42,060
So we've now seen three ways we can

145
00:08:42,060 --> 00:08:46,140
address data gaps using analysis based
estimation approaches.

146
00:08:46,140 --> 00:08:49,910
However, there are important caveats that
I'll stress again.

147
00:08:49,910 --> 00:08:52,230
First, all approaches to addressing data

148
00:08:52,230 --> 00:08:55,470
gaps must be clearly documented and
reproducible

149
00:08:55,470 --> 00:08:56,950
so your audience can form their own

150
00:08:56,950 --> 00:08:59,870
opinions about the credibility of your
methods.

151
00:08:59,870 --> 00:09:02,920
Second, compiling primary data is clearly

152
00:09:02,920 --> 00:09:06,890
preferable to using estimation approaches,
whenever feasible.

153
00:09:06,890 --> 00:09:11,210
Third, any estimation method must meet the
minimum data quality

154
00:09:11,210 --> 00:09:15,200
requirements stated in the scope
definition of an LCA study.

155
00:09:15,200 --> 00:09:17,660
Otherwise, the goals of the study will not
be met.

156
00:09:18,850 --> 00:09:21,720
Despite these caveats, these methods can
often

157
00:09:21,720 --> 00:09:24,400
be helpful in the absence of better data.

158
00:09:24,400 --> 00:09:25,620
In my experience,

159
00:09:25,620 --> 00:09:27,680
using estimation to fill data gaps

160
00:09:27,680 --> 00:09:30,940
is especially helpful for generic
background processes.

161
00:09:30,940 --> 00:09:34,972
And, for determining if a given unit
process or flow qualifies for

162
00:09:34,972 --> 00:09:38,619
exclusion from a study on the basis of
stated cut off rules.

163
00:09:38,619 --> 00:09:46,860
[BLANK_AUDIO]