Sorry for the click bait title, though it is true, but not for the reason most people think. The new supercomputer is likely to increase the precision and reliability of forecasts, and doubtless industry and businesses will benefit. However I must admit that I won't benefit! I am a little scared of forecasts that are 'too accurate'. As it is we have inferior data to the metoffice anyway, but the concept of a MO discussion thread is still legit due to the convenience and detail the met doesn't disclose. However what happens when we live in a world when you will be able to say with 95% certainty that it will snow in 3 days time at 3:30pm in bridlington? All the fun is gone, all the speculation, a model output discussion becomes model output determinism. I am of course not concered about the long range, we could do with that improving, and chaos theory will always win out there, but are we on the twilight of medium and short range model discussion? Will there even be any point checking the models 4 times a day in a cold spell in January?

Originally Posted by: Quantum 

Q, you, your kids and your kid's kids will all be long gone before that happens, if it ever does, which I highly doubt. 

Originally Posted by: Matty H 

From a selfish point of view I'd still oppose the new computer, because I think the accuracy we have now is perfect and I wouldn't want forecasts to be any more accurate than this. Is it just me or have surprise troughs that come out of nowhere that you watch with surprised glee on the radar got rare to the point of non-existence? Even in my lifetime I remember several occurrences of trough surprises at school, and I miss them. And who wants to know they are going to get exactly 1 inch of snow, who wants 'just' one inch! I guess its like wrapping presents, somehow the wrapping makes the presents far more exciting, and the metoffice is gradually removing all the wrapping :(

Originally Posted by: Quantum 

Perhaps these dedicated weather computers may become self aware and develop the same level of intelligence as Michael Fish.

Think about that one.

Originally Posted by: idj20 

Perhaps these dedicated weather computers may become self aware and develop the same level of intelligence as Michael Fish.

Think about that one.

Originally Posted by: idj20 

Will he ever live that one down?.

Originally Posted by: AIMSIR 

During my studies in Reading, there was an important quote regarding how well we can ever expect to predict the weather in advance. I wish I could recall the exact wording or even who it was by - I would need to search hundreds of pages of notes to locate it.

Anyway, the message is that there is a fundamental limit to predictability that arises due to our inability to represent meteorological phenomena in models with 100% accuracy all the way down to the smallest of scales. Chaos in the atmosphere will ensure that this is never achieved.

With the initiation of convection, minuscule variations in heat and moisture are critical, but the resolution required to predict where those perturbations occur to within even a few km is beyond what reasonable computing power can achieve. This will always make the exact time and location of shower clouds impossible to predict beyond half a day or so, if even that. Instead, we will have to be content with predictions as to what broader-scale regions are favourable for shower development - not specific locations.

Broad-scale organisation makes convection more predictable, while frontal boundaries and the general area of precipitation are among the most predictable features of all - we may one day be able to give an arrival time to within 10 minutes or so with a high level of confidence. Embedded convection will remain a bit less predictable though, both in timing and location.

That's what I can recall from the Masters course in Reading, hopefully that's at least reduced your level of concern Quantum?

Originally Posted by: Stormchaser 

During my studies in Reading, there was an important quote regarding how well we can ever expect to predict the weather in advance. I wish I could recall the exact wording or even who it was by - I would need to search hundreds of pages of notes to locate it.

Anyway, the message is that there is a fundamental limit to predictability that arises due to our inability to represent meteorological phenomena in models with 100% accuracy all the way down to the smallest of scales. Chaos in the atmosphere will ensure that this is never achieved.

With the initiation of convection, minuscule variations in heat and moisture are critical, but the resolution required to predict where those perturbations occur to within even a few km is beyond what reasonable computing power can achieve. This will always make the exact time and location of shower clouds impossible to predict beyond half a day or so, if even that. Instead, we will have to be content with predictions as to what broader-scale regions are favourable for shower development - not specific locations.

Broad-scale organisation makes convection more predictable, while frontal boundaries and the general area of precipitation are among the most predictable features of all - we may one day be able to give an arrival time to within 10 minutes or so with a high level of confidence. Embedded convection will remain a bit less predictable though, both in timing and location.

That's what I can recall from the Masters course in Reading, hopefully that's at least reduced your level of concern Quantum?

Originally Posted by: Stormchaser 

Sorry for the click bait title, though it is true, but not for the reason most people think. The new supercomputer is likely to increase the precision and reliability of forecasts, and doubtless industry and businesses will benefit. However I must admit that I won't benefit! I am a little scared of forecasts that are 'too accurate'. As it is we have inferior data to the metoffice anyway, but the concept of a MO discussion thread is still legit due to the convenience and detail the met doesn't disclose. However what happens when we live in a world when you will be able to say with 95% certainty that it will snow in 3 days time at 3:30pm in bridlington? All the fun is gone, all the speculation, a model output discussion becomes model output determinism. I am of course not concered about the long range, we could do with that improving, and chaos theory will always win out there, but are we on the twilight of medium and short range model discussion? Will there even be any point checking the models 4 times a day in a cold spell in January?

Originally Posted by: Quantum 

Sorry for the click bait title, though it is true, but not for the reason most people think. The new supercomputer is likely to increase the precision and reliability of forecasts, and doubtless industry and businesses will benefit. However I must admit that I won't benefit! I am a little scared of forecasts that are 'too accurate'. As it is we have inferior data to the metoffice anyway, but the concept of a MO discussion thread is still legit due to the convenience and detail the met doesn't disclose. However what happens when we live in a world when you will be able to say with 95% certainty that it will snow in 3 days time at 3:30pm in bridlington? All the fun is gone, all the speculation, a model output discussion becomes model output determinism. I am of course not concered about the long range, we could do with that improving, and chaos theory will always win out there, but are we on the twilight of medium and short range model discussion? Will there even be any point checking the models 4 times a day in a cold spell in January?

Originally Posted by: Quantum