Why doesn't the exact same model work out the same?
From Unofficial BOINC Wiki
Contents |
[edit] General
The simple answer is probably 'Maths Libraries'. A fuller answer needs to discuss chaos.
[edit] Chaos
Most people believe the weather is chaotic. This is famously expressed as the "Butterfly Effect". It is often ascribed to Lorenz. In a paper in 1963 given to the New York Academy of Sciences he remarks:
- One meteorologist remarked that if the theory were correct, one flap of a seagull's wings would be enough to alter the course of the weather forever.
By the time of his talk at the December 1972 meeting of the American Association for the Advancement of Science in Washington, D.C. the sea gull had evolved into the more poetic butterfly - the title of his talk was :
- Predictability: Does the Flap of a Butterfly’s Wings in Brazil set off a Tornado in Texas?
It is worth noting that the flap of a butterfly's wing is as likely to prevent a Tornado as create one. Thus the long term frequency of Tornados is not altered nor are the places that are most prone to tornados altered. Thus climate is not altered by a butterfly wing flap.
Climate Models can be shown to be chaotic. This means that a tiny disturbance can result in very different weather. For a while the systems will stay very similar but the differences will gradually increase until at some point the weather starts acting very differently.
[edit] Are Climate Models repeatable down to the tiniest detail?
Yes they are deterministic so there is no random element built in. Participants have tried this backing up a model and running it from the same point twice on the same computer and got identical output files.
[edit] So why would they fail to be repeatable?
Participants have also tried passing a backup to a different computer and then got different output files. How can that happen if the system is deterministic?
If two different computers use different maths libraries, they may calculate the answer to a complex formula using different simpler formulas. This could cause different rounding in the least significant digit. That is only a tiny difference but if that happens then that tiny difference will grow until chaos kicks in and makes the weather very dissimilar.
[edit] Also See
- Demonstration of Chaos effects in a climate model
- Climateprediction.net Project Index
- Climateprediction.net Science Index
- Climateprediction.net Science
- An Introduction To The Science Behind Climateprediction.net
- Ensemble
- Explanation of the Nature Journal - First CPDN Results
- Cold Equator Models
- Why doesn't the exact same model work out the same?
