Friday, 29 January 2016

One in thirty four million?

A couple of States deputies have raised questions over why Jersey had not had iodine tablets issued as they had in areas in France near nuclear reactors.  In response CM Gorst is reported to have said he had research  giving the probability of a nuclear accident at Flamanville impacting on Jersey was ‘one in every 34 million years’. 

Without sight of the report or knowing exactly what it was assessing the chances of, I cannot categorically say that is wrong.  Such things are critically dependent on the assumptions  and inputs into the calulation.  I will however state I think that figure is rather unlikely.  My first point would be to call on CM Gorst to release this research.  Of course he won't do that if it is covered by 'commerical confidentiality'  in which case I would imagine it comes from a nuclear industry indsider, possibly even EDF itself.  That would harldy constititute  fair and impartial research.  You might want to ponder why the commerical sensitivity of a business is thought to more important than the safety of the whole Island's population too.

Even without resorting to research I could recall three nuclear incidents that had they happened at Flamanville would have impacted Jersey - Three Mile Island, Chernobyl and Fukushima. Sheep controls on farms in Wales and Cumbria were only lifted in 2012, 25 years afer Chernobyl spewed the fallout cloud that caused the problem. In fact out of roughly 400 nuclear plants operating worldwide there have been at least 30 meltdown accidents.  Any one of those if it happend in Normandy would affect us. Arguable eachone of them where ever they happen wil affect us eventually. There were other serious incidents too , but I cannot be sure they would have impacted us.  So in the sixty five years or so of commercial plant operations we have had 30 incidents among 400 plant. 

One of the factors that does make a difference is the age and the designof the plant.  Newer plant very pronbabaly are rather safer tha tthe old ones.  As it happens while I was preparing this piece , a new item apperaed raising concerns over some 1950's plant restarted in Belgium recently. Often the stats are done using the newest plant safety and construction criteria.  The existing reactors at Flammanvile are froom the 1980s, but  new facilites are being constructed.  There have been problems at flammanville on a number of occassion, but none so serious  (yet!) to meet the criterion of affecting us badly (who sets those crieria?) .  There is at least one significant concern that has been raise in repsectof  the plant in Normandy - the Blayais incident in 1999.  Rather as in Fukushima, if an event occurs that is beyond the parameters believed at construction,  then the building is outside of it designn parameters.  We know storms and surges are stronger now than they were in the 1980's and still increasing.

The United States Nuclear Regulatory Commission (NRC) usually gives the odds of a nuclear meltdown anywhere worldwide as one in a million years.  That is a similar ballpark to the figure Sen Gorst gives.  On the other hand the same NRC only requires reactors to be build to a 1 in 10,000 year core damage frequency !  And yet in Japan, where they are rather sensitive to such things they  reckon the probability of a major nuclear accident has reduced from 1 in 40 years post Fukushima to now 1 in 80 years !  That's some discrepancy.  Again it all depends critically on the assumptions you make and what you define as major.

A few other points to bear in mind.  Just because the probability of  some event is low does not mean it won't happen rather soon. A one in a million years may sound safe, but it could still happen tomorrow.  When doing risk assessements, not only should you reckon the probability, but also the impact.  In general I am happier facing a high probability event that isn't fatal, than a low probability of extinction.  The calculation of the probability of nuclear failure  is difficult because there are few incidents.  Estimating and extrapolating from a small data set is fraught with problems.    Attempting to do it by looking at the design and assessing failure modes and risks might be better , but then you have to accept that if you miss a possible black swan event like the size of the earthquake/tsunami that casused the Fukushima situation, you have a problem.

One in 34 million years.  I don't beleive it any more than I believe I will win the UK national lottery.  And just like that lottery my intention is to not buy a ticket at all, and thereby guarantee not losing .

No comments:

Post a Comment