Thursday, May 5, 2011

Black swan event?: Fuel pollution in Ben-Gurion Airport

Israel is a one-airport country  (there are several small ones for internal flights, but with the exception of some charter flights, all international flights go through Ben-Gurion Airport, near Tel-Aviv, given this fact, anytime that this airport does not function (e.g.due to strike) it becomes a major news item.  Today all aircrafts were grounded due to the fact that the fuel that were inserted to the aircrafts was polluted, due to a pollution that has been discovered in the fuel facility in the airport, many flights were cancelled, and other delayed were filled with a small amount of fuel that will enable them to fly to one of the nearest airport (mostly in Cyprus, some in Jordan) and refuel, the picture above taken from the YNET news site show passengers that are stranded in the airport,     Is such event a black swan event ?   seems so, but surprisingly this is the second instance of this event, such event happened before at the same airport a few years ago, and took a day to recover.     

From event processing point of view the interesting questions about such events are:
  • Can black swan events be predicted?  first time that they happen?  second time that they happen? - this actually similar to the question - can we instantiate Merphy's laws -- infer everything that can go wrong? 
  • How can rules and actions about  treating black swan events can be set, if we cannot even imagine the event's nature? 
The earthquake in Japan made thinking about black swan events more popular,  the challenge is to advance the thinking in that area.      


Msx said...

I've been working on using CEP for security issues (infrastructure protection). The way I try to deal with such "black swan event" is based on the effect, and not the cause. Thus, the question is no more to prevent such events, but to react as better as possible.
That doesn't solve totally the problem, but it delays the time where the system faces to an totally unknown context.
As the risk assessment is mainly based on the combination of two measures, a probability of occurrence and an impact assessment, we use CEP rules to reduce the probability of occurrence (by preventing the white swan events) but also to reduce the impact (better management of the consequences of the black swan events). The limit of such an approach is that we "expect" that the black swan events share consequences with those identified...

Msx said...

To the question "can we instantiate Murphy's laws?", the partial answer we've been developing is to structure the different types of event with a taxonomy (actually, an ontology). Thanks to that (and the notion of inheritance), we are able to write few rules capturing many cases, without losing the detail of the detected case (given by the received events). This works event better thanks to the multiple inheritance provided by an ontology.

Marco said...

From Nassim Taleb's definition of a Black Swan Event:
"The non-computability of the probability of the consequential rare events using scientific methods (owing to the very nature of small probabilities)"

So no, even CEP will not help to predict a Black Swan, because by definition, one of its main properties is that it is _not_ predictable.

However, Msx is right when he says, CEP can be used to assess and with this insight reduce the impact of very unlikely events.