In its latest assessment report IPCC concluded that in order to get climate change under control world needs massive expansion of nuclear power, renewables, energy efficiency, and CCS. I am a numbers guy and therefore I was delighted when I found a useful database for many of the mitigation scenarios IPCC relied on in its latest report. There is a database for the scenarios and additional information and assumptions used on many scenarios can be found in another database. I found this very interesting since articles reporting on the scenarios often explain the underlying assumptions of the models poorly. I will focus now on how the modellers approached nuclear power. I didn’t have the patience to go through all scenarios and I focused on those with 450ppm CO2 target that contained all technologies optimally (allegedly). I found that quite a few modellers dealt with nuclear power in a way that left me wondering if their modelling is simply poorly disguised ideological propaganda.

Some main approaches used to influence how well nuclear power does in the models relative to variable renewables (wind and solar):

  1. In many models nuclear capacity increases massively. Hundreds and hundreds of reactors are constructed, but amazingly nobody learns anything! Capital costs for nuclear power are typically kept almost constant throughout the decarbonization pathways. On the other hand learning effects and technological evolution are assumed for other energy sources. For wind and solar power these are often assumed to be very dramatic and there are learning effects even for fossil fuels. So this tough love only seems to apply to nuclear power.
  2. Many models assume large cost reductions for wind and solar. In the end, this is not much more than a wishful guess.
  3. Some models assume anomalously large capacity factors for wind and solar. See for example, “Message Ampere2-450-FullTech-OPT” scenario. Capacity factors for wind are almost 40% while for solar power they use about 25-31% over the course of the century. Since real figures are more like half of the assumed figures, the model drastically underestimates the costs for wind and solar. (IMACLIM scenarios seem to do the same)
  4.  Some models (IMACLIM in particular) assume very low capacity factor for nuclear.  “IMACLIM Ampere2-450-FullTech-OPT” has a nuclear capacity factor of just 45% in 2100 while for wind and solar they have 36% and 38% respectively! This doesn’t just roughly double the cost of nuclear in these models, but also underestimates the costs for wind and solar.
  5. Some models (REMIND and MERGE-ETL) postulate a world running out of uranium together with no technology development for nuclear. This “peak uranium” then limits the role nuclear power plays in decarbonization.

Figure 1: Nuclear power in Remind Ampere2-450-FullTech-OPT scenario. Massive increase and then…

Let me discuss the sillyness of the last trick in more detail. Figure 1 shows what REMIND scenario got for nuclear power when all technologies were used “optimally”.  So massive increase in nuclear power until middle of the century and then rapid decline. Decline is caused by uranium supplies running out as soon as light water reactors with once-through fuel cycle have used 23 million tons of uranium. This is very strange for several reasons.

First, this number doesn’t seem to bear any clear connection to known uranium resources which are about third of this figure. Modellers probably felt that using known resources as an upper limit would have been too stupid to pass the laugh test.

Second, mineral resources have a habit of increasing together with demand since increasing demand stimulates increasing investment in exploration and technology development.  In the past one hundred years copper production has increased by an order of magnitude. All this time world has been “running out” of copper in about 40 years. Uranium is not especially rare element and there is no reason to believe we are running out of it anymore than we have for other metals such as tin which has about the same crustal abundance.

Third, from where does the assumption of no technology development come from? Wasn’t this supposed to be a scenario where all technologies are allowed? For nuclear power technologies that that improve the fuel efficiency by about two orders of magnitude are already known.

Fourth, why is there resource constraint only for nuclear power? The resource constraints are more severe for wind and solar power (and for bioenergy). In Figure 2 I show an image I picked up from a european study on critical metals for energy technologies. The elements with greatest supply risks are used in the construction of wind and solar power. (By the way, the only nuclear related element on the list is the low risk hafnium for control rods.) Figure 3 I picked up from a fairly recent Alonso et al. paper. Authors estimated that dysprosium (used in magnets) demand in renewables heavy mitigation scenarios is expected to be a whopping 2600% higher than projected supply already in 2035!


Figure 2: Critical metals for European “strategic energy technologies” according to European commission Joint research centre study.

Figure 3: Expected demand and supply for Dysprosium according to Alonso et al.

Figure 3: Expected demand and supply for dysprosium according to Alonso et al. (2012).

What would happen if we were to apply modellers approach for renewables? Let us just take silver as an example. Silver reserves are estimated at about 530000 tons. Let us assume that “real” resource is 4 times this (remember uranium resource was set at 3 times the known reserves) and that half of this can be used for photovoltaics. There are after all other uses for silver as well. Since 1GW of solar power requires about 80 tons of silver, this means that at maximum we can have about 13TW of solar capacity as opposed to almost 90TW cumulative capacity REMIND modellers extrapolated. Instead of being the largest contributor to the primary energy supply its contribution would fall into 5-10% range. The amount of silver required to construct the solar power in REMIND FullTech scenario is about 13 times larger than the estimated global silver reserves. Now can there be ways around these constraints? Probably there are and maybe we could use less silver, but using substitutes might imply higher costs and worse performance and furthermore, if one was not permitted to use already demonstrated technologies for nuclear power why should imaginary advances be permitted for other alternatives?

What might we get if we remove this silly constraint from the model? Obviously I cannot repeat the exercise with the tools I have available, but we can get a rough estimate. Lets take the growth rate (4.8%) for nuclear power REMIND modellers established between 2020-2050 and just let it grow with the same rate until the end of the century. This is not extraordinary in the context of this model since for wind+solar the growth rate through the century was 7.6% even though capital costs are such the nuclear power seems to have a lower levelized cost of energy (5% discount) throughout the decarbonization pathway. I show the result in Figure 4. Nuclear power would end up dominating the energy supply.

I have a feeling that resource constraint was introduced specifically for this reason. Modellers first did their calculations without the constraint and ended up with a result that they found distasteful. They did not want to go on record with the scenario that might “rock the boat” or give people funny ideas. By introducing the resource limitation for nuclear power they could clip its wings and keep it supposedly as an option while limiting its role to the margin. In fact that strange 23 mton uranium resource limit seems to suggest that over the century LWR:s cannot produce more than maybe around 5% of the primary energy. I suspect that modellers worked backwards and set the resource limitation based on the maximum share of the energy supply they were ready to grant for nuclear power. Not cool.

Figure 4: There, I fixed it!

Figure 4: There, I fixed it!

Then there is PRIMES…sigh. This is a model I encountered few years ago as I was reading EU:s 2050 energy strategy. I remember glancing at the referee report and being troubled by the brief remark on page 6. Referee had asked about rather optimistic cost assumptions to which response was that if capital costs for wind are set higher then the future learning curve can be steeper. To me this suggested that modellers were perhaps fitting model to the fantasy. In the AMPERE database PRIMES scenarios for EU are also included. I was naturally most interested in the Ampere5-Decarb-AllOptions scenario which according to authors is a scenario “with all technological decarbonisation options available and used according to cost optimality; this scenario provides the least cost decarbonisation pathway for the EU.” Sounds interesting! However, as you look at the actual results you notice something weird. The capital costs assumed are such that nuclear (again) has the lowest LCOE throughout the decarbonization pathway. Despite this modellers claim that nuclear generation in EU will decline by 20% by 2050. How is this even possible?

Then I noticed a strange footnote on page 15: “PRIMES assumes that nuclear development has been significantly affected in the aftermath of the nuclear accident in Fukushima in March 2011. Both PRIMES and TIMES-PanEu impose national constraints regarding nuclear, such as countries’ decisions not to use nuclear power at all…” Please tell me that I am reading this wrong. They didn’t just exclude nuclear power from large parts of EU in their “all options” scenario for political reasons and then sell it as the cost optimal one?

I have now outlined several ways in which scenario modellers seem to suppress nuclear power from their reference scenarios where all options and technologies are supposedly on the table. This has also consequences for the other scenarios and comparisons between them. Since modellers suppressed nuclear power already in “the tech neutral” scenarios adding additional anti-nuclear policy, can be presented as not really having major cost consequences.

Figure 2: The box on the left has nuclear power in it and the box on the right had it removed. Amazingly it looks almost the same as the other empty box!

Figure 2: The empty box on the left has nuclear power in it and the box on the right had it removed. Amazingly it looks almost the same as the other empty box!

Since I am a bad boy I will conclude with some rough estimates on what would it take to replace (gasp!) solar and wind power at the end of the model scenarios with nuclear power that generates the same amount of electricity. I simply estimate the required nuclear capacity (90% CF) and use modellers assumptions about capital costs. Required yearly outlay is roughly total capital required divided by the lifetime of the plant. I will use 30 year lifetime for wind and solar and 60 years for nuclear. (Numbers are in billions of 2005$…I think.)

Model Wind+solar capital Nuclear capital (Wind+solar)/year Nuclear/year
Remind 450-FullTech-OPT 74540 62753 2485 1046
Message 450-FullTech-OPT 40620 64150 1354 1070
IMACLIM 450-FullTech-OPT 5680 5765 189 96
Primes Decarb-AllOptions (EU) 1430 826 48 14
Primes HIEFF-NoCCS-NoNUKE (EU) 1555 900 52 15

In all models the required yearly outlay (at 2100 or 2050 for PRIMES) for energy supply is dramatically lower if we replace wind and solar capacity with nuclear power. This despite the fact that MESSAGE and IMACLIM assumed unrealistically high capacity factors for variable renewables. It is remarkable than even though this kind of chicanery was going on behind many models, IPCC still ended up concluding that nuclear power must expand massively. This is perhaps partly because not all scenario builders were intellectually dishonest about this issue and some models ended up, for example, with ten fold increases in nuclear capacity. On the other hand I am afraid that all 450ppm scenarios are utterly unrealistic….and don’t get me started on their absurd bioenergy projections.

P.S. I spent some time copying the data I was interested in from the database. Interface seems a bit uncomfortable for that. Here is a link to some of the data I extracted.

P.P.S.  For laughs you might want to check IMACLIM model with 550 ppm goal and CCS excluded. Since the original one was very strongly dependent on CCS one would imagine that ruling it out would have interesting consequences for the energy mix. See what modelers assumed for the capital costs of nuclear here to suppress that out of control (critical?) nuclear growth early in the century.



Note added 6.10.2019: It would seem that the capital cost data for the AR5 scenarios has been removed from the database. At least I cannot find it anymore. Can someone explain where is it now?