Why is solar "prediction" in such a primitive state.

[michaelmozina]

A few years ago, it was predicted that the next solar cycle would be 30-50% *stronger* than the last solar cycle.

http://www.ucar.edu/news/releases/2006/sunspot.shtml

March 6, 2006

BOULDER—The next sunspot cycle will be 30-50% stronger than the last one and begin as much as a year late, according to a breakthrough forecast using a computer model of solar dynamics developed by scientists at the National Center for Atmospheric Research (NCAR). Predicting the Sun's cycles accurately, years in advance, will help societies plan for active bouts of solar storms, which can slow satellite orbits, disrupt communications, and bring down power systems.

Fast forward a few years.....

http://science.nasa.gov/headlines/y2009 ... iction.htm

May 29, 2009: An international panel of experts led by NOAA and sponsored by NASA has released a new prediction for the next solar cycle. Solar Cycle 24 will peak, they say, in May 2013 with a below-average number of sunspots.

"If our prediction is correct, Solar Cycle 24 will have a peak sunspot number of 90, the lowest of any cycle since 1928 when Solar Cycle 16 peaked at 78," says panel chairman Doug Biesecker of the NOAA Space Weather Prediction Center.

It seems to me as though solar "prediction" is more of an "art" than a "science" at this point. Why?

 

[MeteorWayne]

The simple answer is because we have only a decade or so of any kind of detailed observations of the sun. (other than sunspot counts, which while they go back a century or 3 tell us nothing about the interior dynamics)

There are no robust models that descibe the evolution of the interior of the sun (even yours )

We have a profound lack of detiled observations over more than a single half cycle.

All we have is empiical observations over a VERY short time period, which are not sufficient to create a workable model of the solar interior. I think that will take a few more cycles; i.e. another 50 years or so.

That's how science works

 

[michaelmozina]

The simple answer is because we have only a decade or so of any kind of detailed observations of the sun. (other than sunspot counts, which while they go back a century or 3 tell us nothing about the interior dynamics)

I would have to agree that we have a very limited number of years of satellite images to work with.

There are no robust models that descibe the evolution of the interior of the sun (even yours )

FYI, "mine" doesn't even assume that sunspot cycles are "strictly" driven by interior events.

We have a profound lack of detiled observations over more than a single half cycle.

All we have is empiical observations over a VERY short time period, which are not sufficient to create a workable model of the solar interior. I think that will take a few more cycles; i.e. another 50 years or so.

That's how science works

I do think it's appropriate to note that "mathematical models" are a dime a dozen, and often times utterly useless without accurate models and accurate data to work with. Whatever mathematical models they've been using to date seem to be rather crude and ineffective at actually "predicting" solar cycle events. Even the newer "predictions" are more "postdicted" from recent observation than "predicted" based on accurate modeling.

 

[Saiph]

It's incredibly hard to build a real model of the sun. I spent an entire semester learning the very rudimentary basics of how to model a star...and anything beyond that required major computer crunching power.

Throw in MW's correct observation that we don't have much in the way of observations, and there is very little to anchor the already very complicated and difficult model too...

Thus poor predictions.

 

[michaelmozina]

It's incredibly hard to build a real model of the sun. I spent an entire semester learning the very rudimentary basics of how to model a star...and anything beyond that required major computer crunching power.

I would have to agree.

Throw in MW's correct observation that we don't have much in the way of observations, and there is very little to anchor the already very complicated and difficult model too...

Thus poor predictions.

I suppose that is all true.

 

[MeteorWayne]

michael, if your theory has any specific quantitative predictions (with some physics justification), throw them out here now

Who knows, even if it just a random correct prediction, it will add some credibility

Your enthusiastic friend, Wayne

 

[michaelmozina]

michael, if your theory has any specific quantitative predictions (with some physics justification), throw them out here now

Not yet. I need more data to "predict" anything useful with my model.

Who knows, even if it just a random correct prediction, it will add some credibility

Nah, the standard model folks tried that already and look where it got them.

Your enthusiastic friend, Wayne

Let me ask you a simple question. WHAT type of data (based on current technologies) might help improve the solar forecast as it relates to sunspot activity and why would that type of data help in the prediction process in your opinion?

 

[MeteorWayne]

"Let me ask you a simple question. WHAT type of data (based on current technologies) might help improve the solar forecast as it relates to sunspot activity and why would that type of data help in the prediction process in your opinion?"

About 33 years of high resolution multi satellite data.

Wayne

 

[derekmcd]

It's those damned, pesky Chinese butterflies. There are, quite simply, too many random, unpredictable events to build a reliable model to predict short term patterns. The weather, here on Earth, isn't much different. We're getting better at our earthbound meteorological models (math and all)... but that's because we live here.

 

[michaelmozina]

It's those damned, pesky Chinese butterflies. There are, quite simply, too many random, unpredictable events to build a reliable model to predict short term patterns. The weather, here on Earth, isn't much different. We're getting better at our earthbound meteorological models (math and all)... but that's because we live here.

Well, I hear you but......

http://www.ucar.edu/news/releases/2006/sunspot.shtml

The scientists have confidence in the forecast because, in a series of test runs, the newly developed model simulated the strength of the past eight solar cycles with more than 98% accuracy. The forecasts are generated, in part, by tracking the subsurface movements of the sunspot remnants of the previous two solar cycles. The team is publishing its forecast in the current issue of Geophysical Research Letters.

The original claim here was 98% accuracy at predicting 8 previous solar cycles. It now seems however that this model's actual "predictive" power for future solar cycles is rather questionable. It seem hard to reconcile a 98% percent success rate with a complete about face as it relates to the strength of the solar cycle. I can see them missing the start date by a few months or a year, but the switch from 30-50 percent higher energy release to now predicting a *LOWER* energy release and fewer sunspots seems like a fairly large about face.

More importantly IMO is the fact that their mathematical model seems to be based *NOT* upon actual observation, but upon *assumptions* about the movement of plasma on the sun.

The sunspot process begins with tightly concentrated magnetic field lines in the solar convection zone (the outermost layer of the Sun's interior). The field lines rise to the surface at low latitudes and form bipolar sunspots, which are regions of concentrated magnetic fields. When these sunspots decay, they imprint the moving plasma with a type of magnetic signature. As the plasma nears the poles, it sinks about 200,000 kilometers (124,000 miles) back into the convection zone and starts returning toward the equator at a speed of about one meter (three feet) per second or slower. The increasingly concentrated fields become stretched and twisted by the internal rotation of the Sun as they near the equator, gradually becoming less stable than the surrounding plasma. This eventually causes coiled-up magnetic field lines to rise up, tear through the Sun's surface, and create new sunspots.

While SOHO can and has been used o track plasma movements directly under a sunspot that faces us (heliosiesmology techniques), it cannot observe the flow of plasma near the poles. I have no idea where they get the notion that plasma sinks near the poles to a depth of 200,000 KM unless this claim is based upon an "assumed model" of some sort rather than raw data from heliosiesmology studies.

 

[michaelmozina]

About 33 years of high resolution multi satellite data.

Wayne

Hmmm. I'm not sure that 33 years worth of SOHO, STEREO or Trace images would make it any easier for me personally to 'predict' the strength of a future solar cycle, but IBEX may provide some insights into the "environment" that the sun interacts with. IMO *that* type of "external" data might be helpful and useful, but I can't see anything in SOHO or STEREO data that would make it any easier for me to actually "predict" something about solar cycles many years into the future. I would say that sunspot activity can be loosely correlated to high energy emissions seen in 195A (any iron ion wavelength actually), but I currently can't think of a way to predict that high energy activity, let alone to predict the sunspot activity that often (but not always) accompanies these high energy events.

 

[derekmcd]

Is there point you are trying to make in the grand scheme of things? As with any predictive model of similar systems that should be deterministic, there are simply too many variable data points that allow for future guarantees. Simply because their model is off by a bit (that's actually still to be seen as it appears sunspots are beginning to appear), doesn't mean they don't understand how the sun works from a physics standpoint. It just means the data they input and how their model incorporates that data may be off.

 

[michaelmozina]

Is there point you are trying to make in the grand scheme of things? As with any predictive model of similar systems that should be deterministic, there are simply too many variable data points that allow for future guarantees. Simply because their model is off by a bit (that's actually still to be seen as it appears sunspots are beginning to appear), doesn't mean they don't understand how the sun works from a physics standpoint. It just means the data they input and how their model incorporates that data may be off.

I guess the primary point I was trying to convey is that solar "prediction" seems to be more of an "art" than a science, in spite of the claim of '98 percent accuracy" as it relates to modeling past events. In just a few years the mainstream went from predicting that the next solar cycle would be 30-50% *stronger* than the last one, to now claiming it will be one of the weakest cycles in 100 years. That's quite an "about face" in the predictions of mainstream models.

While I'm sure that there will be a "next" solar cycle, I see no logical way to "predict" it's relative strength (compared to the last one) based on available evidence, images and data. Evidently mainstream "methods" seem to be just as incapable of predicting such things with any sort of accuracy or we would not have seen such a "flip-flop" in just three short years.

The other key point I would make here is that while Wayne seems to believe that more satellite data might help make this prediction process more of a "science" and less of an art, I'm a skeptic. I don't really know what type of satellite data might help us determine this. *IF* there is an external component to this cyclical process *and* we can measure it's effect through IBEX or some other system, *THEN* we might be able to accurately predict cyclical changes. If however it's a strictly internal process as mainstream theory suggests, and it takes an extremely long time for energy releases in the interior to reach the surface, then it seems like it would be much harder to 'predict' cyclical releases with any sort of accuracy.

 

[MeteorWayne]

Michael, you should love the claim of 98% accuracy. It's based on pure empirical hindcasting.

And of course, the article is more than 3 years old.

The fact is that we understand some of the physics of the sun's interior, but by no means all. And certainly not in enough detail to have a realistic physical model. Heck, we don't even have that for the earth's magnetic field generating geodynamo and we live right here!
Like the sun, that is limited by incomplete physics, incomplety known initial conditions, and less than unlimited computer capacity. It will always be that way, even after life has been cooked off of the earth.

We learn more in science every day, but we will never learn it all

 

[michaelmozina]

Michael, you should love the claim of 98% accuracy. It's based on pure empirical hindcasting.

That is the problem IMO.

And of course, the article is more than 3 years old.

I suspect the whole reason for revising the forecast is directly related to what we've learned in those three years, and the fact the beginning of the solar cycle is so late.

The fact is that we understand some of the physics of the sun's interior, but by no means all. And certainly not in enough detail to have a realistic physical model. Heck, we don't even have that for the earth's magnetic field generating geodynamo and we live right here!
Like the sun, that is limited by incomplete physics, incomplety known initial conditions, and less than unlimited computer capacity. It will always be that way, even after life has been cooked off of the earth.

We learn more in science every day, but we will never learn it all

I hear you. I guess the part that "irked" me so much about the first article is the claim of 98% accuracy at postdicting the last 8 cycles. That figure seems entirely contrived since I'm sure it was modified to fit those cycles. Their model has certainly been less than accurate at predicting the start of the next cycle. Now the whole concept of the next cycle being stronger than the last one is being abandoned altogether? That wholesale change in energy output seems rather dubious. It's one thing to miss start dates of the cycle. It's quite another thing to miss the strength factor altogether and essentially "guess" heavily in both directions. It suggests to me that their "model" for making such predictions is simply the wrong model entirely.

Math alone can't tell us the whole story, it requires a *PHYSICAL UNDERSTANDING* of the actual processes in play. It seems to me that they aren't even using the right physical model in the first place or there would not be such a need to revise these figures. It will be another 3 years until we reach the peak of the next cycle, so I have no doubt that these numbers will be revised again based on what we learn between now and then.

As long as we're just "guessing", I'm guessing the next solar cycle is just now starting, it will end late, and the overall number of sunspots will be close to the last cycle.

 

[MeteorWayne]

Well, the UCAR 3 year old article you posted was based on empirical hindcasting. What they stated was accurate, they were able to hindcast 98% of the accuracy. It's similar to the process for tropical forecasting. They search through the parameters and find those that correlate the best with the results. While there is a physical basis for the predictions, even though there are better ones, the ones they use are the ones that are most useful in predicting the actual behaviour. That is PURE empiracism.

Neither is based on purely a physics model. They are just trying to find the predictors and predictands that perform the best.

As you said, we do not have anywhere near enough of a detail physical model for the sun to make accurate predictions for the future. I believe I already stated that, so I don't know why you are arguing with me.

Your guess is just a guess. If you have a universal model that can make predictions, then go ahead and post them.

If not, your guess is no better, and probably worse than those who have some empirical math to back them up.

 

[michaelmozina]

Well, the UCAR 3 year old article you posted was based on empirical hindcasting. What they stated was accurate, they were able to hindcast 98% of the accuracy. It's similar to the process for tropical forecasting. They search through the parameters and find those that correlate the best with the results. While there is a physical basis for the predictions, even though there are better ones, the ones they use are the ones that are most useful in predicting the actual behaviour. That is PURE empiracism.

Neither is based on purely a physics model. They are just trying to find the predictors and predictands that perform the best.

As you said, we do not have anywhere near enough of a detail physical model for the sun to make accurate predictions for the future. I believe I already stated that, so I don't know why you are arguing with me.

Your guess is just a guess. If you have a universal model that can make predictions, then go ahead and post them.

If not, your guess is no better, and probably worse than those who have some empirical math to back them up.

I am definitely *NOT* arguing with you about the need for additional data. I certainly agree with you on that point.

I simply question whether another 22 years or so worth of SOHO/TRACE/STEREO/Hinode images and data would necessarily improve things as it relates to sunspot forecasting. It's clear to me for instance that high energy emissions like we observe in the 195A wavelength are in fact "related to" sunspot activity. However, even when there is observed concentrated activity in 195A, there is no guarantee of a sunspot in that location. The reverse does not seem to be true however. There are no sunspots when there is no activity observed in the higher energy wavelengths.

Take this last sunspot for instance. It is clearly located in the same general region as the 195A activity as we might expect, but many such concentrated active regions that have been observed in 195A do not generate similar sunspots. The amount/level of activity is evidently important, but how do we "predict" 195A activity, let alone the level of that activity? I do in fact agree with you that much more data is required. I guess I'm unclear what *kind* of data might help us forecast something like the strength of the next solar cycle.