Response from Martin Sevior to rebuttal 2 from Jan Willem Storm van Leeuwen
June 2nd, 2006
My response is in blue, unedited text from Storm van Leeuwan (Storm) is in black.
See previous discussion here:
Construction of a nuclear power plant
The study Storm & Smith [Q6] describes in Chapter 3 the methods to estimate the energy requirements of construction of a reference nuclear power plant.
The construction energy requirements comprise:
• energy consumed at the construction site, including transport;
this component can be measured directly
• energy embodied in the construction materials, such as concrete, steel
and copper, but also in chemicals and other auxiliary materials:
that is, the energy consumed in the processes to obtain that materials
• energy needed to construct and maintain capital goods, such as machines and equipment
• energy embodied in services and human labour.
During the 1970's and 1980's the methodology of energy analysis has been developed, maturing to a useful tool to calculate the energy requirements of a good or economic activity with reasonable accuracy, see for example IFIAS 1974 [Q99], IFIAS 1975 [Q100], Roberts 1975 [Q101], Chapman 1975 [Q113], Chapman-1 1976 [Q104], Chapman-2 1976 [Q106], Roberts PC 1976 [Q105], Reister 1977 [Q97], Bullard, Penner & Pilati 1978 [Q102], Roberts PC 1982 [Q103], Constanza & Herendeen 1984 [Q119].
The construction of a nuclear power plant is an extensive and very complex activity.
Process analysis leads to a large underestimation of the total construction energy requirements, when labour and supporting activities of the construction are discounted, see e.g. Rombough & Koen 1978 [Q120]. This is the case in a number of energy analyses published in the past. Input/output analysis is well suited to large aggregated activities, like the construction of a nuclear power plant. Chapman 1975 [Q106] concluded:
“In principle this is an unsatisfactory procedure since the inputs to nuclear systems are likely to be uncharacteristic products of the sectors documented in the input-output tables. However there are grounds for believing that provided a product has a large vector of inputs, ie requires inputs from many other sectors of the economy, then the average energy intensity derived from the input-output table is fairly reliable.”
The I/O analysis may be simplified by using the general energy/gdp ratio of a particular year in a particular country to calculate the net energy requirement of a complex activity. The general energy/gdp ratio (or energy intensity) e is defined as the quotient of the total primary energy consumption of a country (in joules) and gross domestic product (for exemple in US dollars).
This simplification gives a fairly reliable value of the energy embodied in that activity, including energy costs of labour, services, subsidies, etcetera (Tyner, Constanza & Fowler 1988 [Q124]). This is affirmed by other studies, e.g. Rombough & Koen 1978 [Q120], Roberts PC 1982 [Q103], Bullard, Penner & Pilati 1978 [Q102], Constanza & Herendeen 1984 [Q119].
As Constanza & Herendeen put it:
“Embodied energy (calculated the way we suggest) is a good, non-trivial static correlate of the economic value of the relatively large aggregates of goods and services that make up the entries in the I/O tables.”
Certainly, the construction of a nuclear power plant is a large aggregate of goods and services. Nuclear technology can be considered being high-tech, on top of an extensive industrial and economic infrastructure of other high-tech production processes.
The studies of Rombough & Koen 1975 [Q120] and Bullard, Penner & Pilati 1978 [Q102] showed that the value calculated via a detailed I/O analysis is somewhat higher than the value found via the simplified method. Both studies concluded that construction of a power plant is somewhat more energy-intensive than the average economic activity.
It may be that IO LifeCycle Analyses (LCAs) are unsuited to circumstances where the cost of the activity is substantially larger than the average cost of the projects within the sector. Assigning the energy cost of labour to a construction activity inevitably leads to double-counting within the framework of the overall economy. Labour consumes energy in the transport and residential sectors which are accounted separately. That is, is apportioning the energy into different sectors of the economy should only be done once. Energy consumed by labour in transport, should be assigned to transport. Energy consumed by labour in residential dwellings is assigned to the residential sector.
If energy is also assigned to employment activities it has been counted twice.
The ISO (via the ISO14000 (http://www.iso14000.com/) certifications), EcoInvent (http://www.ecoinvent.ch/)and the Environment Product Declarations http://www.environdec.com (EPD) employ process-based LCA's. While process-based LCA's are acknowledged to lead to under-estimates of the energy cost, recent research shows that the extent of the under-estimation is at most 80%.
In response to this known deficiency, a hybrid approach of process and IO LCA's have been proposed to combine the strengths of both approaches. Where inputs to a process LCA are unknown, IO-tables can be used. This is the approach required by the EPDs and used by Vattenfall.
See Suh et al (Environmental Science and technology, Vol 38, 658, (2004))
The total energy requirements of construction cannot be measured directly, because of the sheer complexity of the construction activities: many different materials, activities and capital goods are involved. Therefore, the construction energy has to be estimated using the methods mentioned above and discussed in Chapter 3 of [Q6]. The results are given in Table 1.
It is not obvious that this is correct. The Vattenfall EPD, Donnes et al. (International Journal of Life Cycle Assessments, 10, P 10-23 (2005)) and the French ExternE assessments provide very good descriptions of the energy consumption of NPP production.
Table 1 Energy requirements, total and lifetime specific CO2 emission of
construction of a 1 GW(e) nuclear power plant. Source: [Q6].
low mean high
energy in PJ 40 80 120 R = 4.8
CO2 emission Tg 2.5 5.0 7.5
specific CO2 emission *, g/kWh 12 24 36
1 PJ = 1 petajoule = 1015 joule
1 Tg = 1 teragram = 1012 gram = 1 million metric tonnes
R = ratio thermal energy/electric energy
* averaged on lifetime 24 FPY
Energy requirements for construction = sum of electric and thermal (=fossil) energy in PJ. In the study [Q6] no primary energy units are used.
The large range of values is due by:
• uncertainty range in the data
• physical differences between individual nuclear power plants.
Energy requirements estimated in other known studies: see Table 11 from Chapter 3 of Storm & Smith 2005 [Q6].
A clue of some kind of a physical/chemical minimum of the construction energy can be estimated starting with the main construction materials: concrete and steel.
Table 2 Construction masses * of a 1 GW(e) nuclear power plant
steel concrete total
Gg Gg Gg
low 119 681 800
mean 150 850 1000
high 179 1021 1200
* excluding piping, wiring and other materials
1 Gg = 1 gigagram = 1000 metric tonnes
This table is based on Crowley & Griffith 1982 [Q229] and Shaw 1979 [Q230]. Uchiyama 2002 [Q205] cites a total construction mass of 1291 Gg.
Production of cement
Cement is made by heating CaCO3 (calciumcarbonate, limestone) with a siliceous material:
5 CaCO3 + 2 SiO2 -> (3CaO.SiO2) + (2CaO.SiO2) + 5 CO2
From this equation a stoichiometric ratio of 0.55 gram CO2 released per gram cement can be calculated, by the calcination reaction alone.
The specific energy consumption of concrete is 1.83 MJ/kg, according to IAEA-TecDoc-753 1994 [Q148]. If that amount of energy is generated by burning oil (75 g CO2/MJ), the specific CO2 emission is 137 g CO2/kg concrete. Assuming cement makes up 15% of high-density concrete - used in construction of nuclear power plants - the calcination process adds 83 g CO2/kg to the specific emission concrete. The total amounts to 220 g CO2/kg concrete.
It may not be a valid assumption that Oil is employed to provide the process heat required for cement production in Sweden. The EPD for concrete production from Buzzi Unicem (http://www.environdec.com/page.asp?id=130&menu=3,9,0&epdId=121) shows CO2 emissions to be 230 Kg/cubic meter. Concrete has a density in the range 1.8 - 2.4 tonnes per cubic meter, which implies CO2 emissions in the range 95 - 127 g CO2/Kg.
Production of steel
Stoichiometrically calculated CO2 release from blast furnaces (including slag forming) is about 2 g CO2 per g iron. Excluding coke production. IAEA-TecDoc-753 1994 [Q148] cites a value of 29.54 MJ/kg. Assumed this energy is mainly generated by burning coal (coke) with a specific CO2 emission of 92 g/MJ, the specific CO2 emission of steel production would be 2.7 kg CO2 per kg steel, somewhat higher than the stoichiometric minimum.
The Swedish steel industry achieves considerably better CO2 emission rates than this. The paper by Sandberg et al. (Scandinavian Journal of Metallurgy, Vol 30, 420-425 (2001), http://www.jernkontoret.se/informationsbanken/presentationer/pdf/sjm1o578.pdf) reports that Swedish steel production emits 0.3 - 1.47 tonnes CO2 per tonne steel depending on whether electric arc or blast furnace processes are employed.
Table 3 Total CO2 emission from the production of the construction
materials iron and cement of a 1 GW(e) nuclear power plant,
and specific CO2 emission averaged on lifetime of 24 FPY
(210 billion kWh)
concrete steel total lifetime (24 FPY)
Gg CO2 Gg CO2 Gg CO2 g CO2/kWh
low 150 321 471 2.2
mean 187 405 592 2.8
high 225 483 708 3.4
It is hard to make a direct comparison because we do not know the mix of arc versus blast furnace steel used by Forsmark. The range would be 121 - 260 Gg CO2 emissions from concrete and steel in the low case.
The figures in Table 3 should be seen as little more than an indication:
• Energy requirements of mining sand, gravel, iron ore, coal and slag and transport of iron and concrete may be not included in these figures.
• Production of other materials, such as stainless steel, copper, etc., are not included either.
• The figures relate to raw materials only, without processing them into components of buildings or equipment.
The CO2 emissions due to concrete and steel could easily be overestimated given the quality of Swedish Industry. The Low value case appears broadly in agreement with the Vattenfall EPD.
Rombough & Koen 1974 [Q96] and 1975 [Q120] demonstrated in their studies that the embodied energy in the raw construction materials of a nuclear power plant makes up less than 5% of the total energy requirements of construction, as calculated via an elaborate I/O analysis.
This is result is in substantial disagreement with modern investigations which compare IO-LCA's to process-based LCA's. The work of Suh et al (Environmental Science and technology, Vol 38, 658, (2004)), show that processed based LCA's are at most 80% under-estimated, certainly not a factor a twenty. The Nordic EPD process recognizes this deficiency and requires unknown inputs to be approximated with generic data from IO tables.
The results of Jungbluth et al. (International Journal of Life Cycle Analyses 2004 1-11 ) and Schleisner (Renewable Energy 20 (2000) 279 - 288) which estimate the CO2 emissions of Wind Power show that Steel and concrete usage is some 5-8 times greater per KW-Hr of generated electricity for Wind Power compared to Nuclear Power. If this same factor of twenty were applied to Wind Power, it would have an energy payback time of 6 years rather than the 3-4 months commonly assigned to it.
The work of Donnes et al (International Journal of Life Cycle Assessments, 10, P 10-23 (2005)) is in broad agreement with the Vattenfall EPD and has been incorporated into the EcoInvent database.
The figures of Table 3 are included in the figures of the total and specific CO2 emission, mentioned in Storm & Smith 2005 [Q6] and in Table 1.
The work cited by Storm all dates from early studies from the 1970's in the field of Life Cycle Assessments. The field has since moved forward and plays an important role in the policy and planning of many projects. There are two main techniques for performing Life Cycle Assessments. Process-based assessments and IO-based assessments. The process based assessments are more useful for planning purposes since they can be used to identify the components that provides the largest contributions to the environment impact of the process. The Swiss Ecoinvent database, the EPD and the various ISO14000 standards all employ process-based LCA's. The deficiencies of the process-based LCAs are recognized but modern research shows that these provide at most an 80% correction to the environmental impact, not a factor of twenty. In addition the EPD employed by Vattenfall takes account of this by requiring unaccounted impacts to be identified and their effects estimated.
To my knowledge the figures of the Vattenfall EPD [Q152] are measured values. As pointed out above, the total energy requirements of construction cannot be measured directly.
I don't dispute the figures, nor the intention of Vattenfall. I dispute the use of the Vattenfall figures for purposes they aren't meant for.
How are the results of Table 3 to be reconciled with the 150 Gg stated by Vattenfall?
As noted earlier, the CO2 emissions from concrete and steel production could be over-estimated by Storm. Given the performance of the Swedish Steel industry and world best practice for cement production, CO2 emission rates in the range 120 - 260 Gg can be obtained from the table provided by Storm. However this question should really be directed to the authors and referees of the Vattenfall EPD.
Do I understand Sevior and Flitney correctly that they dismiss the figures found by ExternE UK [Q308] (see p 5 of their response)?
The UK analysis demonstrate the deficiencies of an unsophisticated IO analysis. The Sizewell-B reactor was very expensive, much more than the average construction project in the UK. This cost does not necessarily mean the energy cost was proportionally larger. The energy cost obtained from this analysis was much larger than the French and Belgian ExternE studies.
If so, that's because the authors of ExternE UK problably used a similar method to estimate the CO2 emission like ours?
In my view some scientific arguments would be appropiate here.
This appears to be a region of considerable uncertainty. Given the ISO14000 series of standards, EPDs and EcoInvent it is clear that the mainstream of LCAs employ process based analyses. Recent research comparing process to IO-based LCAs show at most an 80% discrepancy between the two approaches. Storms' observation of a factor 20 discrepancy is way outside the mainstream. It me it appears that the onus is on Storm and Smith to reconcile their predictions in the light of modern LCA research.
One could speculate that the origin of the discrepancy is due to the dollar cost of NPP construction being substantially higher than a typical construction job because of the very high value labour and capital costs of these NPP's. High value labour does not necessarily imply large energy costs. Banking and software development both consume high value labour but are not large energy consumption industries.
Sevior and Flitney incorrectly state that Storm & Smith used only two studies from the 1970s and ignored later studies.
In Table 2 of Chapter 2 of our study [Q6] we listed the known studies with estimates of the energy requirements of mining and milling. We explained our choices. As any reader may learn from Table 2 of Chapter 2 of our study, more recent studies than we used cite much higher figures. Sevior and Flitney seem still to misread our study.
It's not that we misread the study, it is that the study did not use these data. The only data used to construct the formulas employed for predicting the energy costs for Uranium mining were acquired in the mid-1970's.
Studies from the 1970s and 1980s are not necessarely outdated or invalid. If so, it should be proven in a scientific way.
Digging up rock from the ground seems not very liable to advances in technology, with respect to the specific energy consumption.
This is a big assumption. Mining technology improves with time like all other technologies. In any case the main component of the energy cost of uranium extraction of the Storm and Smith study is the energy cost of Milling, not Mining.
As with the Vattenfall EPD, the environmental assessments of Ranger and Olympic Dam mines are not necessarely energy analyses.
The method of Sevior and Flitney based on financial data appears to me full of hidden assumptions, bookkeeping problems, statistical pitfalls and uncertainties.
We compared the predicted energy cost of Uranium mining and milling for Ranger, Olympic Dam and Rössing to the energy consumption as reported. All are significantly over predicted (5 PJ, 60 PJ and 69 PJ vs 0.8 PJ, 5 PJ and 1 PJ respectively). In the case of Olympic Dam, the energy use includes a substantial copper smelter unrelated to uranium activities.
The energy consumption is predicted to be so large that is comparable to the energy consumption of a particular sub-section of the economy. In the case of Rössing, the over prediction is larger than the energy consumption of the entire country of Namibia.
We also made an assumption that the mines would not be operated if they were unprofitable. One can simply estimate the revenue generated by the reported quantity of Uranium sold and compare it to the cost of supplying the predicted energy. In the case of Rössing the predicted energy cost is an order of magnitude larger than the revenue obtained from Uranium sales.
Thus there are three independent estimates of the energy used for Uranium mining. All demonstrate large over-predictions from Storm and Smith.
All these are simple, straight forward calculations and comparisons. Our conclusion is that the equations from Storm and Smith that predict the energy cost of Uranium mining are unreliable for modern mines.
In my view Sevior and Flitney could contribute significantly to the discussion on mining and milling of uranium if they would be able to prove in a physical way which figures are right. The figures of Ranger, Olympic Dam, Rössing and any other mine, should be analyzed to make them fully comparable with the figures from the studies we listed. This implies knowledge of the physical flows of fuel, electricity, materials, etcetera, which are attributable to the mining and milling, directly and indirectly.
If the figures from the studies we used are not correct anymore, there must be an unambiguous physical explanation.
This is true but not relevant. What is more relevant is the energy cost of future mines. It is reasonable to assume that these will be at least energy efficient as current mines.
Remains the question of the dependency on the ore grade of the mining and milling operations.
The Olympic Dam mine is 97% efficient in it's extraction of Uranium from Ore which is above the Storm and Smith's predicted extraction efficiency. Many gold mines operate at Ore grades of 5 PPM which is 70 times smaller than Rossing. There is no need for Uranium mines to operate anywhere near these low concentrations for the foreseeable future though.
Finally Storm and Smith may be interested to review this paper that describes an experiment that extracted 1 Kg of Uranium from seawater via acquiculture.
Seko et al. Nuclear Technology, 144, 274-278 (2003) (available at the web link:
http://www.ans.org/pubs/journals/nt/va-144-2-274-278). A patent for this process has been granted. It can be obtained via the web from here: