Southern field close to flipping?

[Philalethes]

6 hours ago, Bean said:

Just a point to Jesterface23 , the highest daily Sunspot Number for SC25 (so far ) was recorded on the 19/01/2023 at a number of 206, any number for February is provisional at the moment and will not be a matter of record until SILSO release the monthly bulletin >    https://www.sidc.be/sunspots/bulletins/monthly/monthlybull202301.pdf

True, but they do also release estimates for February, even though I agree these would not be correct to use in this context.

What surprised me when looking at those estimates is that they currently have this as the highest daily SN for February:

Quote

2023 02 13 2023.119 196  13.4  28  35

In contrast I distinctly recall having seen this number well over 200 just a day or two ago (that estimate is for yesterday, 13-02-2023 as can be seen from the first columns), and I was under the impression that the number from this site is taken from the SILSO estimates. Perhaps they've updated that estimate since, or there's something I don't know about how the number is calculated.

Edited February 14, 2023 by Philalethes Bythos

[Jesterface23]

7 hours ago, Patrick P.A. Geryl said:

isn’t Solar max the month with the highest SN?

Not at all. Solar max/maximum is the highest of the 13 month smoothed monthly total sunspot number.

https://en.wikipedia.org/wiki/List_of_solar_cycles

If you aren't joking, just let that sink in for a bit.

Edited February 15, 2023 by Jesterface23

[Philalethes]

20 minutes ago, Jesterface23 said:

Not at all. Solar max/maximum is the highest of the 13 month smoothed monthly total sunspot number.

https://en.wikipedia.org/wiki/List_of_solar_cycles

If you aren't joking, just let that sink in for a bit.

And that's not even considering double peaks where activity is roughly equal either, which further complicates matters.

[Patrick P.A. Geryl]

3 uren geleden, Jesterface23 zei:

Not at all. Solar max/maximum is the highest of the 13 month smoothed monthly total sunspot number.

https://en.wikipedia.org/wiki/List_of_solar_cycles

If you aren't joking, just let that sink in for a bit.

Ok. Solar Max month. In cycle 24 this was February 2014. The 13 month smoothed average was in April 2014. 
This means the 13 month smoothed average will be in March 2023…

[Sam Warfel]

16 minutes ago, Patrick P.A. Geryl said:

Ok. Solar Max month. In cycle 24 this was February 2014. The 13 month smoothed average was in April 2014. 
This means the 13 month smoothed average will be in March 2023…

 No, we don’t know when the 13month smoothed average will peak, because we don’t know the future. Nor is that how a 13 month smoothed *average* looks. You can see the smoothed average SSN lags behind the daily or monthly SN. You can only tell with no less than 13 month’s hindsight when it’s peaked. 

Edited February 15, 2023 by Sam Warfel

[Patrick P.A. Geryl]

47 minuten geleden, Sam Warfel zei:

 No, we don’t know when the 13month smoothed average will peak, because we don’t know the future. Nor is that how a 13 month smoothed *average* looks. You can see the smoothed average SSN lags behind the daily or monthly SN. You can only tell with no less than 13 month’s hindsight when it’s peaked. 

There is another indicator: the flip of the average polar field

The 13 smoothed average was November 1979

Average flip: December 79

The 13 smoothed average was December 1989

Average flip: December 1989

So the 13 month smoothed average falls at the average flip or 1 month before…

 

[Patrick P.A. Geryl]

Op 3/2/2023 om 16:27, Philalethes Bythos zei:

You literally said you picked by hand one function that matched the best out of dozens. And nothing to prevent overfitting has been done. You could also use virtually any machine learning algorithm to "automatically" fit the given parameters to the function you want even more perfectly, but this would still be overfitting and have zero predictive value whatsoever. That is why such algorithms typically reserve a significant fraction of the data as a test set (hold-out), and make the model from the rest. In this case the sample size is so low anyway that even this would be of questionable validity.

At this point I'm getting more and more convinced that you have very little understanding of data science, and would urge you to contain your claims to this thread, at least until you're prepared to address the seeming lack of proper use of statistical methods.

I hope I'm wrong, and that you're just way smarter than everyone else (without doing much to prove that, though), but it's getting harder and harder to believe that narrative.

You say I am overfitting. Why didn’t you look at the two other calculations? No so called overfitting was done!

One is accurate to 10 percent and the other to 13 percent…

Go again to my link and look at them… Instead of writing an epistle that says nothing about the outcome…be so kind to do some basic math… 

[Jesterface23]

7 hours ago, Patrick P.A. Geryl said:

Ok. Solar Max month. In cycle 24 this was February 2014. The 13 month smoothed average was in April 2014. 
This means the 13 month smoothed average will be in March 2023…

Overall we will need to wait about 3 more years to know when we have hit the monthly maximum SN. Saying January 2023 is the monthly maximum is currently just a prediction, not reality. To say solar maximum will be 2 months after the monthly maximum based on the prior solar cycle is just digging more holes. Some past solar maximums have occurred months to over a year before or after the monthly maximum.

We're stuck with the data we've got and some we have a lot of, and some we don't.

[Philalethes]

8 hours ago, Patrick P.A. Geryl said:

You say I am overfitting. Why didn’t you look at the two other calculations? No so called overfitting was done!

You can't know this until you've actually managed to predict the future correctly.

[Patrick P.A. Geryl]

11 minuten geleden, Philalethes Bythos zei:

You can't know this until you've actually managed to predict the future correctly.

I want to know  opinions about these

1. The model with a 13.9 percent accuracy gives 155 for cycle 25

2. The model with a 9.1 percent accuracy gives 123 for cycle 25

3. The model with a 5.9 percent accuracy gives 132 for cycle 25

4. Cycle 24 was 145 365 day smoothed average measured

So what would be the most likely outcome?

Edited February 15, 2023 by Patrick P.A. Geryl

[Archmonoth]

11 hours ago, Patrick P.A. Geryl said:

There is another indicator: the flip of the average polar field

The 13 smoothed average was November 1979

Average flip: December 79

The 13 smoothed average was December 1989

Average flip: December 1989

So the 13 month smoothed average falls at the average flip or 1 month before…

 

 

I am no mathematician, nor have any degrees or certifications, but many of the data points used in an average mean are going to be close to the 13-month average. 

 

The 13-month average requires 13 months to establish what that average is. The 13-month average was the SAME as those dates you posted, because it includes dates similar to those months. 

 

A single month is not a 13-month average, but a single month can have the same number of sunspots as the average. You don't know what the 13-month average is until you have 13 months. 

 

I am saying the same thing in 3 different ways in the hopes one of the methods will make sense. 

 

[Patrick P.A. Geryl]

peer reviewed article.

gives also a low SSN for cycle 25🤩
Maximal growth rate of the ascending phase of a sunspot cycle for predicting its amplitude | Astronomy & Astrophysics (A&A)


https://www.aanda.org/articles/aa/full_html/2022/07/aa43509-22/aa43509-22.html

[Sam Warfel]

5 hours ago, Patrick P.A. Geryl said:

peer reviewed article.

gives also a low SSN for cycle 25🤩
Maximal growth rate of the ascending phase of a sunspot cycle for predicting its amplitude | Astronomy & Astrophysics (A&A)


https://www.aanda.org/articles/aa/full_html/2022/07/aa43509-22/aa43509-22.html

We’ve already reached much higher numbers than 110, and will almost undoubtedly have a maximum far in excess of 110, probably above 200

[Bean]

1 hour ago, Sam Warfel said:

We’ve already reached much higher numbers than 110, and will almost undoubtedly have a maximum far in excess of 110, probably above 200

"We’ve already reached much higher numbers than 110, and will almost undoubtedly have a maximum far in excess of 110, probably above 200"

The last SSN ( that's Smoothed Sunspot Number) was for July 2022 @ 86.6  so no we have not reached much higher numbers than 110 !

[Sam Warfel]

2 hours ago, Bean said:

"We’ve already reached much higher numbers than 110, and will almost undoubtedly have a maximum far in excess of 110, probably above 200"

The last SSN ( that's Smoothed Sunspot Number) was for July 2022 @ 86.6  so no we have not reached much higher numbers than 110 !

In SSN yeah, but based upon SN performance since July 2022 it’s a very reasonable guess that it will exceed 110

[Philalethes]

9 hours ago, Patrick P.A. Geryl said:

peer reviewed article.

gives also a low SSN for cycle 25🤩
Maximal growth rate of the ascending phase of a sunspot cycle for predicting its amplitude | Astronomy & Astrophysics (A&A)


https://www.aanda.org/articles/aa/full_html/2022/07/aa43509-22/aa43509-22.html

Here's a few other papers predicting the amplitude of SC25 along with excerpts of their predictions:

Prediction of solar cycle 25 using deep learning based long short-term memory forecasting technique:

Quote

Our projected prediction of cycle 25 using the proposed LSTM model, says that it will be stronger than cycle 24 and weaker than cycle 23. The solar cycle 25 will peak with an amplitude of sunspot number at 171.9 ± 3.4 and will be 47 % stronger than cycle 24. The solar cycle 25 will reach its peak in August 2023 ± 2 months.

Predicting Solar cycle 25 using an optimized long short-term memory model based on sunspot area data:

Quote

The predicted peak amplitude of SSA [monthly sunspot area] for Solar Cycle 25 was obtained as 2562.5 with LSTM+, and the maximum value of SSN was calculated as 213 based on the relation between SSA and SSN, which would be stronger than that of Solar Cycle 24. And according to the prediction result, Solar Cycle 25 would reach the peak around January 2025.

Forecasting Solar Cycle 25 with Physical Model-Validated Recurrent Neural Networks:

Quote

For Solar Cycle 25 the ESN algorithm forecasts a peak amplitude of 131 ± 14 sunspots around July 2024 and indicates a cycle length of approximately 10 years. The MESN forecasts a peak of 137 ± 2 sunspots around April 2024, with the same cycle length. Qualitatively, both forecasts indicate that Cycle 25 will be slightly stronger than Cycle 24 but weaker than Cycle 23. Our novel approach bridges physical model-based forecasts with machine-learning-based approaches, achieving consistency across these diverse techniques.

Predicting the 25th solar cycle using deep learning methods based on sunspot area data:

Quote

Our results show that the 25th solar cycle will be 55% stronger than Solar Cycle 24 with a maximum sunspot area of 3115±401 and the cycle reaching its peak in October 2022 by using the LSTM method. It also shows that deep learning algorithms perform better than the other commonly used methods and have high application value.

 

34 minutes ago, Sam Warfel said:

In SSN yeah, but based upon SN performance since July 2022 it’s a very reasonable guess that it will exceed 110

I'm not sure it's that easy to say just yet. For e.g. the SSN of October 2022 to exceed 110, which is one of the more favorable months I could find, it would require roughly an average monthly SN of ~167.8 for the next three months (although the fact that the last month of the SSN calculation is halved skews it a little bit, still a decent approximation). Alternatively, it would require roughly an average monthly SN of ~107.8 for the next 10 months to get the SSN for May 2023 above 110. If current activity levels keep up, then it can easily happen, but we're not quite there yet.

[Sam Warfel]

1 hour ago, Philalethes Bythos said:

Here's a few other papers predicting the amplitude of SC25 along with excerpts of their predictions:

Prediction of solar cycle 25 using deep learning based long short-term memory forecasting technique:

Predicting Solar cycle 25 using an optimized long short-term memory model based on sunspot area data:

Forecasting Solar Cycle 25 with Physical Model-Validated Recurrent Neural Networks:

Predicting the 25th solar cycle using deep learning methods based on sunspot area data:

 

I'm not sure it's that easy to say just yet. For e.g. the SSN of October 2022 to exceed 110, which is one of the more favorable months I could find, it would require roughly an average monthly SN of ~167.8 for the next three months (although the fact that the last month of the SSN calculation is halved skews it a little bit, still a decent approximation). Alternatively, it would require roughly an average monthly SN of ~107.8 for the next 10 months to get the SSN for May 2023 above 110. If current activity levels keep up, then it can easily happen, but we're not quite there yet.

Understood, thanks for breaking it down for me. 

Edited February 16, 2023 by Sam Warfel

[Philalethes]

3 hours ago, Sam Warfel said:

Understood, thanks for breaking it down for me. 

That being said, I still think there's some truth to your statement, because the uptick in activity has been significantly more rapid than for SC24 (as many have noted), and this is indeed reflected in the activity not just going back to July 2022, but all the way back to January 2022. My personal guess would also be that we're not yet at maximum, and that the SN will keep climbing, ultimately surpassing a SSN of 110; but because there's still a chance for activity to rapidly drop and stay low if this turns out to actually be the maximum (or the first of the two peaks), I wouldn't quite bank on it yet.

Essentially we're at a rather critical juncture now, because unless such a rapid drop in activity occurs very soon, it becomes more and more likely that the activity levels will be sufficient to surpass 110. The current average SN for February is ~130.1, which does at the very least not indicate any such drop yet.

[Sam Warfel]

1 hour ago, Philalethes Bythos said:

That being said, I still think there's some truth to your statement, because the uptick in activity has been significantly more rapid than for SC24 (as many have noted), and this is indeed reflected in the activity not just going back to July 2022, but all the way back to January 2022. My personal guess would also be that we're not yet at maximum, and that the SN will keep climbing, ultimately surpassing a SSN of 110; but because there's still a chance for activity to rapidly drop and stay low if this turns out to actually be the maximum (or the first of the two peaks), I wouldn't quite bank on it yet.

Essentially we're at a rather critical juncture now, because unless such a rapid drop in activity occurs very soon, it becomes more and more likely that the activity levels will be sufficient to surpass 110. The current average SN for February is ~130.1, which does at the very least not indicate any such drop yet.

Well we are looking at a very blank eastern side of the sun. As usual, I’d expect some quiet periods, but I’m certainly hoping that this wasn’t max, or first peak! I want a strong SC!

That’s why I was intrigued by Scott’s comment that it seems like SC25 may have a single peak. I’d like to see how that’s calculated, since it’s pertinent to the discussions of Solar max timing and strength if it’s a single-peak or double-peak SC. 

[Patrick P.A. Geryl]

6 uren geleden, Sam Warfel zei:

In SSN yeah, but based upon SN performance since July 2022 it’s a very reasonable guess that it will exceed 110

All models based on the SNN have a major flaw: the accuracy is low. This is the reason they always give large ranges. In our article we found that the high resolution sunspots are very good comparable with the solar flux, while the ISN was off by more then 10 percent. 
it is suggested that the measurement fault for the 10.7 Solar flux since the beginning is maximum 2 percent, while Clette admits there is a serious discrepancy for the ISN. 

[Sam Warfel]

1 hour ago, Patrick P.A. Geryl said:

All models based on the SNN have a major flaw: the accuracy is low. This is the reason they always give large ranges. In our article we found that the high resolution sunspots are very good comparable with the solar flux, while the ISN was off by more then 10 percent. 
it is suggested that the measurement fault for the 10.7 Solar flux since the beginning is maximum 2 percent, while Clette admits there is a serious discrepancy for the ISN. 

How is that relevant to the discussion?

[Jesterface23]

2 hours ago, Patrick P.A. Geryl said:

In our article we found that the high resolution sunspots are very good comparable with the solar flux, while the ISN was off by more then 10 percent.

The WSO's data is limited on it's own, given the average solar cycle is ~11 years. The high res sunspot count doesn't even have a complete cycle of data yet and the use for scientific data is limited.

28 minutes ago, Sam Warfel said:

How is that relevant to the discussion?

Pretty sure that is well overdue, but there has been a lot learned

[Patrick P.A. Geryl]

1 uur geleden, Sam Warfel zei:

How is that relevant to the discussion?

My model is based on the 10.7 Solar flux which has a low measurement fault. It is known that around 1980 a large discrepancy occurred between the Solar flux and the SSN. Therefore I get a small fault by calculating the strength of the upcoming cycle, while that is impossible to achieve with the SSN.

in our accepted article the difference goes from -7.4 for cycle 20 to 7.1 for cycle 24. These are means. Above that there are many months with differences above 15 percent…How do you think you can make a good prediction with such a differences?

It is known that a daily difference in the SN can be more then 30 percent by different stations. There is a factor on the SIDC website that covers this.

EISN info | SILSO


https://www.sidc.be/silso/eisninfo

Edited February 17, 2023 by Patrick P.A. Geryl

[Patrick P.A. Geryl]

1 uur terug, Jesterface23 zei:

The WSO's data is limited on it's own, given the average solar cycle is ~11 years. The high res sunspot count doesn't even have a complete cycle of data yet and the use for scientific data is limited.

Pretty sure that is well overdue, but there has been a lot learned

I seriously disagree. The relation between the high resolution sunspots and the solar flux is pretty linear from high to low. That can’t be said from the ISN. Huge differences at the low, because they see almost no sunspots while we see a lot in high resolution.

Same at the high…

This is logic because if you have 50 percent to 600 percent more sunspots, the fault diminishes considerably….

[Patrick P.A. Geryl]

7 uren geleden, Philalethes Bythos zei:

That being said, I still think there's some truth to your statement, because the uptick in activity has been significantly more rapid than for SC24 (as many have noted), and this is indeed reflected in the activity not just going back to July 2022, but all the way back to January 2022. My personal guess would also be that we're not yet at maximum, and that the SN will keep climbing, ultimately surpassing a SSN of 110; but because there's still a chance for activity to rapidly drop and stay low if this turns out to actually be the maximum (or the first of the two peaks), I wouldn't quite bank on it yet.

Essentially we're at a rather critical juncture now, because unless such a rapid drop in activity occurs very soon, it becomes more and more likely that the activity levels will be sufficient to surpass 110. The current average SN for February is ~130.1, which does at the very least not indicate any such drop yet.

The high resolution 1K dropped to 144 today. This is a factor for low ISN values below 100 for the next days…