Computer Shopper - UK (2020-01)

JANUARY2020|COMPUTER SHOPPER|ISSUE383 105

very different paths. In particular,thatforecast gives no indication

whether it is sampling one of the rare cases where chaos leads to

limited predictability,even the next day,”Walters adds.

ENSEMBLEPERFORMANCE

One partial solution used by the Met Office involves so-called

ensemble forecasts. Ensemble forecasts are multiple forecasts

initialised at the same time,but with termsincluded to sample

uncertainty in the initial conditions and the forecast evolution.

This lets the team sample arange of possible outcomes.

“Not only does this help quantify the uncertainty in the future

forecast, but it allows us to produce probabilistic forecasts and

explore in detail the potential impacts of different alternative

outcomes,”Walters explains.

“Atlonger forecast ranges –say,six to 14 days –ensemble

forecasts are even more important, as the chaotic nature of the

atmosphere leads to areductioninpredictability,and each forecast

must be viewed as only one of apotential range of outcomes. At a

forecast range of about two weeks, there is atheoretical limit –again,

the consequence of chaos theory –beyond which there is very little

predictability in the structure of the atmosphere,and theforecasting

process forindividual weather systems breaks down.”

So ensembles allow ameasure of confidence to be placed on a

forecast, depending on the degree to which the weather is affected by

chaos theory at aparticular time,but is there any waytoactually

minimise the effect of chaos? Walters suggests there might be.

“Improved observations, and hence improved initial conditions,

can help,ascan improvements to our understanding of the

atmosphere,and improvements to the algorithms used in the

computer models,”hesays.

“But further improvements are allowed only by more powerful

computers, such as representing finer scales and including additional

complexity,suchasthe composition of the atmosphere,ormodelling

the evolution of the ocean stateaswell as the atmosphere.”

This sort of continual improvement to the modelling process

has led to the five-dayforecast of individual storm systems today

being as accurateasthe two-dayforecast was 30 years ago.All of

these improvements are within the limit of chaotic predictability,

however,and although further improvements will be made,chaos

will continue to provide alimit to the range at which we can predict

the evolution of the weather.

THEANTIKYTHERAMECHANISM

When it was found on awreck in the sea off the coast of the Greek island of Simi by sponge divers in

1900, it initially looked like nothing more than apiece of corroded bronze junk. However,there is

more to the Antikythera Mechanism –asit’snow called, named after the ship from which it was

taken –thanfirst appeared. The artefact remained largely ignored in the National Museum of

Archaeology in Athens until 1951, when it came to the attention of aBritish science historian

who,following X-rayand gamma rayimages, came to the astonishing conclusion that it

was a2,000-year-old analogue computer.

Containing 37 gears, the purpose of this, the world’s firstknown computer,was to

solve the models that define themovement of the heavenly bodies. In other words, its

function was to carryout astronomical simulations. In particular,itwas able to predict

the movements of the moon and the sun through the zodiac, to predict eclipses and

even to model the so-called irregular orbit of the Moon, aphenomena that had only

recently been discovered when this mechanism was built.

ABOVE:Even the world’s 27th fastest computer can’t make the Met Office’s

forecasts fully immune from the effects of chaos theory

RIGHT:The

sophistication of

the Antikythera

mechanism is quite

remarkable for

something that

was built in the

first century BC

Ph

oto

:S

im
on

Ha

mm

ett

,M

et

Offi

ce