quote:

Originally posted by realprimate:
QUOTE
The 1997-98 El Niño marked the first time in human history that climate scientists were able to predict abnormal flooding and droughts
UNQUOTE

So good on them. Things can only get better (er modelling wise!).

QUOTE...UNQUOTE. Numbers??? Predicted and Observed????

quote:

Of course, we could always tell the Chinese, the Indians, the Brazilians etc not to industrialise. This should preserve our world hegemony. And, of course, we could always tell the workers that they musn't indulge in cheap travel. That should keep the Maldives, Seychelles etc safe for naice people, so that they needn't mix with the riff-raff.

Yes, and that's why there is sort of a need to lead by example on this one. Maybe we are already too late. I'd rather be an ant just in case though...

We'll let you deal with the men and their AKs when they come asking for water.

quote:

Originally posted by Lucibee:
I'd rather be an ant just in case though...

Presumably not a flying ant though?!

JL, Here are some predictions for you.

Hope you can convert the graphic into numbers, and that 100 years is enough.

quote:

Originally posted by Lucibee:
JL, Here are some predictions for you.

Hope you can convert the graphic into numbers, and that 100 years is enough.

Excellent. Should be able to produce some sort of numbers from these three streams. Thanks.

quote:

Originally posted by JL(SFC58,AFCB):

Thanks. I'm not sure when the graph turns from being "post-dictions" into "pre-dictions", but the observation line doesn't look to have a particularly good correlation with any of the three scenarios for the later years.

I understand the graphs were published in 1988 and used in a presentation to the US Congress that year.

The major deviations are due to Pinatubo in 91 and El Nino in 98. Bearing those in mind, the 'B' graph shows a good match to the underlying trend.

quote:

Originally posted by Son of Mulder:

And if he'd included the temperature for 2006 and not had a dotted line but real number for 2005 then it would look more like it's following the blue (Scenario C for those reading in black and white).

http://en.wikipedia.org/wiki/Image:Instrumental_Temperature_Record.png

Looks to me like he backed 3 horses knowing that whatever happened he could find reasons for deviation from whichever.

Scenario C is his keeper prediction because if the actual temperature starts to go down as C predicts then he'll be able to say what a wonderful response the world has made to the AGW predictions.

You are misrepresenting the article. It dates from 2005 and included an estimate for that year based on the first 8 months, not a true measured result for the entire year.

Hansen made it perfectly clear what the three scenarious refer to and why he chose them. Why do you pretend otherwise?

quote:

Originally posted by TrueSceptic:
The major deviations are due to Pinatubo in 91 and El Nino in 98. Bearing those in mind, the 'B' graph shows a good match to the underlying trend.

Thanks. I'll have to try and convert the graphs into numbers. One man's "good match" is another man's "not much better than random" without them (and maybe even with them). But thanks for the source.

quote:

Originally posted by JL(SFC58,AFCB):

Thanks. I'll have to try and convert the graphs into numbers. One man's "good match" is another man's "not much better than random" without them (and maybe even with them). But thanks for the source.

I imagine that you can get the raw figures from NASA and elsewhere. Trying to get figures from graphs without frequent gridlines is a pain.

Truesceptic

quote:

You are misrepresenting the article. It dates from 2005 and included an estimate for that year based on the first 8 months, not a true measured result for the entire year.

Hansen made it perfectly clear what the three scenarious refer to and why he chose them. Why do you pretend otherwise?

But it's now 2007, we have 2 more years of actuals and currently running below Scenario C.
So I'm not trying to misrepresent the article - I'm adding an update which indicates that what seemed to be "dead on the money" as regards B is now starting to underrun (even scenario C).

quote:

Originally posted by TrueSceptic:
Trying to get figures from graphs without frequent gridlines is a pain.

Tell me!

quote:

Originally posted by Son of Mulder:

But it's now 2007, we have 2 more years of actuals and currently running below Scenario C.
So I'm not trying to misrepresent the article - I'm adding an update which indicates that what seemed to be "dead on the money" as regards B is now starting to underrun (even scenario C).

OK, you're adding to the graph a further 2 years. For consistency, are you getting the figures from the same source?

TrieSceptic

quote:

OK, you're adding to the graph a further 2 years. For consistency, are you getting the figures from the same source

I was just looking at the year on year deltas from the Wikipedia chart since the Hansen document doesn't have data for 2005/2006. That way you don't have to worry about what the baseline of the temperature anomolies is.

[QUOTE]Originally posted by JL(SFC58,AFCB):
I'm not even sure I agree that. Parts of the Earth (probably the majority) is warming, but some parts are static and some are cooling.QUOTE]

You get your cherry picker (as time goes on it'll have to be more and more precise) and I'll get my snow shovel. (Found a use for it).

And what happened to our £5 bet on global temperatures? I never forget anyone who dodges a round...

quote:

Originally posted by JL(SFC58,AFCB):
[
Numbers??? Predicted and Observed????

JL. The numbers are serious.

http://www.msnbc.msn.com/id/4739094/

QUOTE
The 1997 El Nino, for example, caused an estimated $20 billion in damage worldwide, offset by beneficial effects in other areas,

1877 El Nino, meanwhile, coincided with a failure of the Indian monsoon and a famine that killed perhaps 40 million in India and China,

Here's some short term predictions.

http://www.johngrayrecyclers.org/ourblog/id49.html

QUOTE
The world is likely to experience the warmest year on record in 2007, the UK's Met Office says.
An extended warming period, resulting from an El Nino weather event in the Pacific Ocean, will probably push up global temperatures, experts forecast.

They say there is a 60% chance that the average surface temperature will match or exceed the current record from 1998.

The scientists also revealed that 2006 saw the highest average temperature in the UK since records began in 1914.

The global surface temperature is projected to be 0.54C (0.97F) above the long-term average of 14C (57F), beating the current record of 0.52C (0.94F), which was set in 1998
UNQUOTE

You've got to admit the underlying trend is relentlessly upwards.

PREDICTED

http://www.agu.org/meetings/wp06/wp06-sessions/wp06_OS14B.html

Hindcast results for the period 1982-2004 show a high predictability of ENSO. All past El Nino and La Nina events, including the strongest 1997/98 warm episode, are predicted successfully with anomaly skill scores above 0.7 at 12-month lead time.

ACTUAL
http://physicsweb.org/articles/world/11/8/8
QUOTE
For example, sea-surface temperatures near Peru increased by more than 5 ºC during the 1997/8 event
QUOTE

Doesn't sound like much of a prediction it sounds more like we saw what happened before and so we'll give a 60:40 probability that it'll happen again and be a bit higher than 1998 given that the starting base is higher than in 1997.

realprimate:

1 I'm happy with my £5 investment. Even money's reasonable odds to both of us. Mine's a lager

2 These predictions are remarakably unspecific. To fulfil the criterion of falsifiability, the test must be noticeably better than could be achieved by randon numbers, and these predictions don't seem (to me) to meet that requirement. I fully agree with Son of Mulder.

3 Hindcast results don't count!

quote:

Originally posted by JL(SFC58,AFCB):

3 Hindcast results don't count!

Why not? It's appropriate for testing models.

WIKI - A hindcast is a way of testing a model. Known or closely estimated inputs for past events entered into the model to see how well the output matches the known results.

An example of hindcasting would be entering climate forcings (events that force change) into a climate model. If the hindcast accurately showed weather events that are known to have occurred, the model would be considered successful.
WIKI

quote:

Originally posted by realprimate:

quote:

Originally posted by JL(SFC58,AFCB):

3 Hindcast results don't count!

Why not? It's appropriate for testing models.

WIKI - A hindcast is a way of testing a model. Known or closely estimated inputs for past events entered into the model to see how well the output matches the known results.

An example of hindcasting would be entering climate forcings (events that force change) into a climate model. If the hindcast accurately showed weather events that are known to have occurred, the model would be considered successful.
WIKI

Not in scientific method, they don't. If the model has been built on a set of data, then of course it will predict that data. The trick is to get the model to predict data which it doesn't "know". See any good book on simulations.

quote:

The trick is to get the model to predict data which it doesn't "know".

But how does it know that it doesn't know?

Also, if a model uses a specific set of data, it doesn't always follow that it will always be able to reproduce the results a few years down the line, whether it is a hindcast or forecast.

If you've left out an essential factor in the original set of data, and miscoded their interactions in the model simulation (unknowingly, of course), then obviously it won't end up where you expect it to - but that's how you test the model and realise that you've left out factor X.

And the model doesn't know whether it is running from 1970 to 1990 or from 2005 to 2025 - unless it is sentient and has a good memory! But presumably you'd say that the 1970/1990 model is no good because it existed in the past, and the 2005/2025 model is unrealible because it is trying to make a future prediction.

I give up!

quote:

Originally posted by JL(SFC58,AFCB):

3 Hindcast results don't count!

JL. A simplistic and sweeping statement. I know you want hard numbers but the model predictions are accurate enough to worry about. Indeed they might be too conservative.

http://environment.newscientist.com/channel/earth/mg195...less-we-act-now.html

QUOTE
Models based on the business-as-usual scenarios of the Intergovernmental Panel on Climate Change (IPCC) predict a global warming of at least 3 °C by the end of this century. What many people do not realise is that these models generally include only fast feedback processes: changes in sea ice, clouds, water vapour and aerosols. Actual global warming would be greater as slow feedbacks come into play: increased vegetation at high latitudes, ice sheet shrinkage and further greenhouse gas emissions from the land and sea in response to global warming.
UNQUOTE


Here’s some links below showing that hindcast results do count and are important for forecasting.

http://books.nap.edu/openbook.php?record_id=10087&page=13
QUOTE
Because the data exists for the prediction period, these hindcasts can be used to determine what the skill of prediction would have been over a long series of forecasts if they had taken place prospectively instead of retrospectively. Once skill in hindcast mode is demonstrated, coupled climate models can be used to predict the climate a season to a year in advance. Coupled climate models can be used to probe for predictability in the climate system on longer time scales when no observational data exists.
UNQUOTE

http://en.wikipedia.org/wiki/Climateprediction.net
QUOTE
Transient coupled model - This comprises an 80 year Hindcast and an 80 year Forecast. The Hindcast is to test how well the models perform at recreating the climate of 1920 to 2000 [8]. It was launched February 2006 under BBC Climate Change Experiment branding and later also released from the CPDN site. The forecast of what will happen in 2000 to 2080 follows the Hindcast and this is described as "experiment 3".
UNQUOTE

quote:

Originally posted by realprimate:

quote:

Originally posted by JL(SFC58,AFCB):

3 Hindcast results don't count!

JL. A simplistic and sweeping statement. I know you want hard numbers but the model predictions are accurate enough to worry about. Indeed they might be too conservative.

realprimate:

The parameters have already been calculated from past data, so it ought to be a “good fit” to that data. That is no test, however, of how well it will perform in the future, because you can calculate parameters for variables which have no bearing on the subject (or random numbers), and they can still fit.

As an illustration, I have constructed a model under which there are five influences on the average temperature in a year (data from Hadley, CET data set):


Multiply the number of years I had been married on 31 December of the year by 2.0546

Subtract the number of police recorded threats or conspiracies to murder (in ‘000s) for the year multiplied by 0.2531

Subtract the Consumer Price Index for the year multiplied by 0.5303

Add the average number of days from arrest to sentence for persistent young offenders multiplied by 0.0944

Add the opening FTSE 100 share index for 1st June (or next trading day thereafter) (in ‘000s) by 0.1374



This does wonderfully well in post-dicting the average temperature in a year; any modeller would almost be embarrassed at how well it performed.

However, I would be willing to wager quite a sizeable pot of gold that it won’t do so well for the next ten years!


That is why hind-casting doesn't count.

quote:

Originally posted by JL(SFC58,AFCB):

quote:

Originally posted by realprimate:

quote:

Originally posted by JL(SFC58,AFCB):

3 Hindcast results don't count!

JL. A simplistic and sweeping statement. I know you want hard numbers but the model predictions are accurate enough to worry about. Indeed they might be too conservative.

realprimate:

The parameters have already been calculated from past data, so it ought to be a “good fit” to that data. That is no test, however, of how well it will perform in the future, because you can calculate parameters for variables which have no bearing on the subject (or random numbers), and they can still fit.

As an illustration, I have constructed a model under which there are five influences on the average temperature in a year (data from Hadley, CET data set):


Multiply the number of years I had been married on 31 December of the year by 2.0546

Subtract the number of police recorded threats or conspiracies to murder (in ‘000s) for the year multiplied by 0.2531

Subtract the Consumer Price Index for the year multiplied by 0.5303

Add the average number of days from arrest to sentence for persistent young offenders multiplied by 0.0944

Add the opening FTSE 100 share index for 1st June (or next trading day thereafter) (in ‘000s) by 0.1374



This does wonderfully well in post-dicting the average temperature in a year; any modeller would almost be embarrassed at how well it performed.

However, I would be willing to wager quite a sizeable pot of gold that it won’t do so well for the next ten years!


That is why hind-casting doesn't count.

Lowest form of humour?