Chapter 2: The Run Up

The crew was nice enough to keep the low lights and the calmly atmosphere. They only allowed themselves one run through the isle to showcase the best products ever, that you could buy dirt cheap at 10000 feet up in the air and of course our tonight's special sandwiches.

Most of the passengers where at some of the stages of sleep. Here and there an overhead lamp shined on a passenger's magazine or  a book. There was also withisch reflections from computer screens on the face of a hard worker. Of course there where also the moving shadows of ...well, moving pictures from the screen of a dozing passenger. On the row seven, left, two men were talking and on their left, the empty seat was just a dark hole. One of them was a little overweight, hold by a too small blue suit on a light colored shirt with a difficult to ignore pink tie. The other, perhaps 10 years younger, seemed even younger with the short hair and a dark hoodie, below which a green t-shirt peered out at the waist, over the jeans. his face was serious and was listening to his travel mate with his hands resting on the tray. The conversation took pace at low volume, little more than whispering, which was totally drowned by the engine droning. on the opposite side of the isle, an old woman was deep asleep, occasionally letting go a snore.

The older man was telling a history on past tense. John Boone worked for a small governmental agency, which depended from the main US meteorological agency. His work there was an optimistic attempt to develop the next step (conceived as a big leap) on algorithms for weather prediction. Basically they had expanded their model to predict global weather. The big leaps on computation power allowed them to have a huge number crunching power. The strategy was to develop evolving algorithms which imitated the physics of the weather. The amount of input data was staggering, as much as the carry over information per iteration. Complexity was growing fast. They needed advanced maths, on statistics, algorithm optimization and networks

 The agency was small and not as resource rich as they needed, but got some algorithms working. The problem was that these algorithms were not predicting anything plausible. Results were discouraging so, as good engineers, they tried another approach. They wanted to correct the algorithms through other algorithms that searched for patterns in the weather, more than to implement detailed Physics, or, in this case, too detailed for the available computing power. Instead of very accurate predictions, they were looking for trends, whether they could predict trends. Things like this current cools down x degrees while this other has warmed up by y in so much time.... and hope to be able to make sense of all that. So they could feed the algorithms not only with numbers but with more relations between these numbers and reduce the amount of unknowns.

They had to get leads from many publications, local climate analysis and many other prediction models and create a network with them. The weather is, as you may know, a chaotic system, which is a name given to it not because looks like a kindergarten class after too much fresh apple juice, but because mathematically  is a complex system with more stable solutions than one, that is more possible solutions that fulfill the equations. The possible solutions are called an atractor and finding the atractor for the system they had created was short of chimeric, so they started to look whether it did look stable at all, that is, whether under similar inputs it produced very different results or it seemed to converge to similar solutions.

By 2001 they had advanced no step towards a general prediction of global weather or towards a robust weather model, but they had found something funny. They seem to be able to distinguish models that dealt stable solutions and models that had violent transitions before going for a solution. The more refined their model was, the more it seemed to be in the second group, which at the beginning sounded like " we are doing something wrong here".  nevertheless, since they did not know how the climate would be in 20 years time, their predictions where of little use. However they where rather conviced, that climate on earth was changing rapidly. All was very academic and complicated and they did not want to go publishing until they could check everything and show some hard evidence. At the time, the change of the earth climate was universally accepted in the community and everybody knew that polar bears were better off developing a skin as the elephant. Thus, they were looking for this hard evidence, by trying to get the same trends as the other models were predicting. They gave themselves a target of trend for 10 years. Yet they started with the year 1900. You see, it is much easy to predict something when you know the solution, and weather records on the XX century are not all that bad. At the end of 2003, the team was working hard on the trend theory and intended to use 100 years of data to feed into the algorithms