“Recalibrating” Risk:
“Recalibrating” Risk:
Reactions to Three
Reactions to Three--Mile Island,
Mile Island,
Chernobyl and Fukushima
Chernobyl and Fukushima
Elisabeth
Elisabeth Pat
Paté
é--Cornell
Cornell
Elisabeth
Elisabeth Pat
Paté
é--Cornell
Cornell
Management Science and Engineering Management Science and Engineering
Stanford University Stanford University Duke University Duke University September 20, 2013 September 20, 2013
Three reaction levels to
Three reaction levels to
nuclear accidents (
nuclear accidents (post facto
post facto))
The public and the media (perceptions)
The industry: understanding the accident and
taking corrective measures (technical and taking corrective measures (technical and organizational)
The government: new regulations, systems’
improvements and requirements for risk analysis (PRA)
[Intergovernmental reactions (e.g., through the World
Outline
Outline
Common factors to the three situations
(Three-Mile Island, Chernobyl and Fukushima Daiichi)
For each of the three accidents: a brief
description and post facto reactions description and post facto reactions
Main lessons. “Recalibration” can mean either
new perception of the same risk or systematic risk updating based on new data. Discussion of the role of quantitative risk analysis to explain what happened and to move forward
Focus on the role of risk analysis and
Focus on the role of risk analysis and
damage assessment
damage assessment
If perception is reality, trying to inject some
reality into perception
– In evaluation of damage after an accident – In evaluation of the risks looking forward
The risk of no risk analysis: e.g., Fukushima
Future risks given current construction program.
Currently 68 nuclear plants under construction, e.g., 28 in China, 10 in Russia, 7 in India, 5 in South Korea (WANO). What to do about it?
A major safety problem
A major safety problem
How to help the world nuclear industry and
their governmental agencies to improve, based on the lessons from accidents?
No amount of Western rhetoric is likely to No amount of Western rhetoric is likely to
change decisions to build nuclear reactors
worldwide. Main reasons: a large and growing need for energy; pollution by burning carbon.
Three
Three--Mile Island, Chernobyl,
Mile Island, Chernobyl,
Fukushima: three different stories
Fukushima: three different stories
Different types of plants
– TMI: pressurized-water reactor (Babcock & Wilcox)
– Chernobyl: graphite-moderated RBMK. No secondary containment. Positive void coefficient (increase in steam
leads to increase in reactivity=> more heat, more steam in a leads to increase in reactivity=> more heat, more steam in a positive loop)
– Fukushima Daiichi: six boiling-water reactors (GE)
Different causes of accidents
– Three-Mile Island, Pennsylvania 1979: misunderstanding of
the system (e.g., poor design of the control panel)
– Chernobyl, Ukraine, 1986: a criminal decision
Several levels of specificity of PRA’s
Several levels of specificity of PRA’s
The common levels (apply to types of plants)
– Level 1 estimates the frequency of accidents that
cause damage to the nuclear reactor core => core damage frequency (CDF).
– Level 2 starts with Level 1 results, estimates the – Level 2 starts with Level 1 results, estimates the
frequency of accidents that release radioactivity from the nuclear power plant.
The site-specific level: (often the missing one)
Level 3 starts with the Level 2 radioactivity release accidents (source term), estimates consequences in terms of public injury and environmental damage
The difficulties of PRA
The difficulties of PRA
Does not hold on a tee-shirt: costly and requires
a lot of expertise that industry does not always have (more people talk about it that know how to do it…)
Generic models now available
Generic models now available
Site-specific models require a lot of expertise
(e.g., seismic hazard)
Second-level of uncertainty done when
necessary (methods exist and are applied) but complex and time consuming (Ref. for example: Jonathan Helton, SANDIA)
What has changed (and still changing)
What has changed (and still changing)
Maturation of technology: examples
– Few RBMK’s left (10, all in Russia.) Some
retrofitting (e.g., some secondary protection) Replaced by pressurized-water reactors
Replaced by pressurized-water reactors
– More generally: digital controls (e.g., of feed
water injection in the secondary circuits) and increase of reliability over analogs
Greater exchange of information about
Maturation of organizations
Maturation of organizations
Major organizational changes: the Institute of
Nuclear Operations (INPO) in the US. Industry self-regulatory body, created after TMI, has
serious influence on industry (INPO has “teeth”)
The World association of Nuclear Operators
(WANO). Weaker: no real power since many countries have different safety criteria often guided by local economics, politics and
competences (no teeth). But WANO provides mechanisms for sharing lessons learned.
Part 2: (a)Three
Part 2: (a)Three--Mile Island
Mile Island
Pressurized water reactor in Pennsylvania
Partial meltdown (1979); small amount of
radio-active gases and iodine in the atmosphere; no know victim.
know victim.
Accident sequence: failure in the non-nuclear
secondary system (loss of coolant); stuck-open valve that took a while to detect. H2 bubble?
Authorized release of some radioactive water in the Susquehanna river. 1B cleanup costs
Huge impact on the public
Huge impact on the public
A poorly known technology (too complex?)
A movie (“China syndrome”) that exacerbate
public intense fears
Miscommunication and bad information Miscommunication and bad information
A political climate in which there was little
incentive for a rational discussion.
Over-evacuation and presidential over-reaction Intellectual reaction: “Normal accidents” and
Huge impact on the industry
Huge impact on the industry
Decline in the number of nuclear reactors Decline in the number of nuclear reactors
(Source IAEA) (Source IAEA)
In the US
In the US
Cancellation of 51 nuclear reactor projects:
effectively stopped the growth of the
nuclear program (there are 103 at this time)
Already in 1975: stopped the re-processing Already in 1975: stopped the re-processing
of fuel => keep the wastes on site in pools
But increase in the safety of existing plants The Nuclear Regulatory Commission (NRC)
stepped in and the Institute for Nuclear Power Operations (INPO) was created
Changes initiated by the NRC
Changes initiated by the NRC
Upgrading and strengthening plant design and
equipment requirements (e.g., auxiliary feed-water system)
Focus on human performance (culture, training) Enhancing emergency preparedness
Setting programs for early identification of
potential problems
Expanding international activities to share info Supported (more or less) the industry’s effort to
The Institute for Nuclear
The Institute for Nuclear
Power Operations (INPO)
Power Operations (INPO)
Industry organization approved by NRC for
self-regulation, training and accreditation
Sets safety objectives, criteria and guidelines
Conducts regular evaluations of nuclear power
Conducts regular evaluations of nuclear power
plants, communicates the results to industry. INPO ratings influence insurance rates
Provides a forum for candid exchange of
information regarding operations and safety
“Recalibration” and safety
“Recalibration” and safety
improvements
improvements
Great uncertainties regarding nuclear safety at the
time, and political climate => large increase in risk perception
Measures taken after that, both from NRC and INPO
were beneficial to improvements of nuclear safety were beneficial to improvements of nuclear safety
TMI however was a large setback to the nuclear
power industry and decreased the use of an
alternative generation of electricity (instead: burning coal, oil and gas)
It also accelerated investments in renewable energies
(b) The most severe accident: Chernobyl
(b) The most severe accident: Chernobyl
RBMK (graphite moderated) in the Ukraine under
the jurisdiction of the Soviet Union
An experiment going wrong in an unstable system.
Operator error: disarming of safety systems and violation of all safety rules (04/86) to prove safety!
– Steam to the turbines was shut off (disconnection of – Steam to the turbines was shut off (disconnection of
protection systems)
– Power spike => rapid increase in steam pressure – A sequence of explosions severing coolant lines – Steam explosion of the reactors and the plant – Huge fire and graphite dispersion (fought with
Consequences and losses
Consequences and losses
31 deaths (Russian evaluation). About 150 later on.
Probably more later but impossible to track down (dispersion of population)=> no possibility of
epidemiological study (dispersion of people and too many confounding factors) but probably additional many confounding factors) but probably additional thyroid cancers and birth defects. UN assessment:
4,000 based on toxicology, linear, no-threshold models
Evacuation of a large zone and a whole town (Pripyat).
135,000 people displaced
Anxieties and degradation of social fabric
Public reactions
Public reactions
Huge but variable: realization of the facts (not
only the risks) of a massive explosion in a
nuclear reactor. Skepticism about loss reports
Large initial over-evaluations of the losses The Russians: slow to respond and
The Russians: slow to respond and
acknowledge losses. Closure of most RBMK’s under intense international pressures
Europe: fear of fall out and food contamination US: increase awareness of nuclear reactor
dangers (no RBMKs), of the dangers of poorly trained operators and risks of human errors
NRC reactions (
NRC reactions (
Source NRC
Source NRC
))
Determining facts, assessing implications forregulating U.S. nuclear power plants; longer-term studies (much less reaction than after TMI)
But the US does not have RBMK’s and this was a
clear case of violation of all safety rules and disarming of safety systems
of safety systems
Some changes in reactor design, construction and
maintenance
Mostly: review of procedures and controls for normal
operations and emergencies
Competence/motivation of management & staff Ensuring availability of backup safety systems
Increased use of PRA
Increased use of PRA
“Risk-informed regulations”: increased use of
PRA on all regulatory matters to complement deterministic methods and defense-in-depth
Three areas: nuclear reactors, wastes, and
materials safety materials safety
Implement safety goals with appropriate
considerations of uncertainties to reduce excessive conservatisms
A complement to deterministic method; not a
“Recalibration” of risk
“Recalibration” of risk
Another piece of evidence of catastrophic
risk of nuclear power plants accident
Initial overestimation of the losses and large
increase in risk perception
But a known dangerous type of reactor But a known dangerous type of reactor
(RBMK), unstable and without secondary containment
New assessment of the risk of quasi-criminal
behaviors and actions (challenging the
system beyond capabilities and shut down of safety features in spite of orders not to do it)
(c) Fukushima Daiichi nuclear plant
(c) Fukushima Daiichi nuclear plant
Six boiling-water nuclear reactors designed by GE
and operated by TEPCO
A magnitude 9 subduction-plate earthquake in the
Senriku area => 14-meter tsunami that struck the
plant and disarmed the diesel generators. Loss of all electric power and of all related safety features.
electric power and of all related safety features.
The fundamental problems: bad design criteria given
the site. A 5.7 m tsunami design criterion and an
undersea water coolant inlet that got clogged in the tsunami. Then: poor emergency management.
Partial meltdown and explosions. Failure to flood the
Losses and damage
Losses and damage
Most (all?) casualties associated with the event
were from the tsunami: up to 25,000 including injuries.
Additional casualties due to the reactors? Hard
to tell. West winds blowing towards the ocean to tell. West winds blowing towards the ocean => relatively low radiation to the population.
No death so far attributed to radiation exposure.
About 40 people injured.
But increased risk of thyroid cancer and birth
People’s reactions to Fukushima
People’s reactions to Fukushima
Huge! In part due to confusion (including in the
press and in academia) between the victims of the tsunami and the potential victims of the
nuclear reactor accident
Loss of confidence in the Japanese authorities Loss of confidence in the Japanese authorities Evacuation of a large zone
Destabilization of the area and its social fabric
(but mostly due to the tsunami)
Drastic reduction of the Japanese electric
Variable repercussions in the world
Variable repercussions in the world
A mix of political and technical reactions.
Generally a review of existing and new plants
Germany’s decision to close its nuclear power
plants based in part on risk perception (if not plants based in part on risk perception (if not actual change)
France not to do so, except for the aging ones
like Fesseiheim that was to be shut anyway
China, Russia, India and others not to cancel
but to revisit some features of their ambitious program. But is there any risk analysis there?
Reaction of the USNRC
Reaction of the USNRC
New emphasis on site-specific risks: Level 3
PRA, which industry finds expensive (and there are few experts)
Huge effort in terms of emergency
preparedness and response. More detailed preparedness and response. More detailed analysis of load combination
Probable slow down of the nuclear program,
but mostly for economic reasons (inexpensive natural gas -$4 per million btus) -at least as
The risk of no risk analysis
The risk of no risk analysis
Back to Fukushima: the deterministic choice of
tsunami design criterion (5.7m), based mostly on the 1960 event from Chili (ε raise in 2002)
Interviews at NRC and EPRI: no PRA for Fukushima.
Japanese excuse: “too complicated”; “data were too Japanese excuse: “too complicated”; “data were too old” => totally ignored the long-term record
Yet, INPO states that TEPCO estimated that
p(Ts.>6m)<10-2 in next 50 yrs (based on what?)
First full probabilistic risk analysis of earthquakes
and tsunamis was published by Epstein in April 2011 (Tokyo Institute of Technology. Ninokata laboratory)
A partial history of
A partial history of subductionsubduction--plate plate earthquakes and tsunamis along the
earthquakes and tsunamis along the SanrikuSanriku and Sendai coasts (Epstein)
and Sendai coasts (Epstein)
Year Magnitude Interval in years
869 8.6 1611 8.1 742 1611 8.1 742 1793 8.2 182 1896 8.5 103 1933 8.1 37 1960 8.5 27
The Epstein study and results
The Epstein study and results
Two Bayesian analyses of tsunami>8m starting with
flat priors and including 6 events in 1,171years.
– The first assuming ergodocity – The second not
Results: Ergodic model ~ 1/730 per year, ~4.3% in 30
years. Non-ergodic model: ~ 1/430 per year, ~ 7% in 30 years
Note: the regulation in Japan for large, early release
factors is 10-6/year, smaller than both calculated bounds => an organizational, cultural, economic problem?
“Recalibrating” the risk
“Recalibrating” the risk
The Japanese invoked an “Act of God” and “Black
swans”. Considered closing their nuclear program but had to change their mind. Still struggling with the site
The German “Ethics commission report”: “The risk The German “Ethics commission report”: “The risk
has not changed but the perception has”. And “risk is the product of probability and consequences”
(insufficient for a good PRA)
My concern: are nations such as China that
construct nuclear power plants in seismic/tsunami areas doing a serious site-specific risk analysis and using it in their design?
Conclusions
Conclusions
The three accidents triggered different reactions and “recalibrations” because
– They were of fundamentally different natures: a lack
of understanding of clarity of some mechanisms at TMI, a criminal “experiment” at Chernobyl, an absurd choice of design criterion at Fukushima Daiichi.
choice of design criterion at Fukushima Daiichi.
– There were increased levels of experience,
competence and global response in the last 40 years
– Economic circumstances have changed (decrease in
costs of other energy sources)
– New perception of the environmental and health risks
If perception is reality, we need to
If perception is reality, we need to
inject some reality in perception!
inject some reality in perception!
PRA is an imperfect method (no “prediction”) but
provides invaluable information reflecting state of knowledge
It does not hold on a tee-shirt: risk is NOT the
product of probability and consequences of failure product of probability and consequences of failure (importance of distribution and especially tail of it)
In the end, risk taking and acceptance choices are
political decisions. But some comparisons may put these choices in perspective (understanding that thousands of deaths, one at a time, from coal or oil burning are more acceptable than nuclear accidents)
“Recalibrating” risk
“Recalibrating” risk
In general, there was an overestimation of the losses
(and great fears fanned by the media) at least in the first years following the three accidents.
Risk/loss amplification phenomenon: a poorly known
technology, complex and not easy to understand.
Great societal disruptions followed the three accidents
Great societal disruptions followed the three accidents Delayed, poorly known potential effects (cancers)
PRA helps assessing and communicating uncertainties
and can make plants safer, especially the 68 under construction … provided that it is done
And especially for these, exchange of information