Validating Solanki 2004 solar activity reconstruction

Posted: July 22, 2011 by tchannon in Astrophysics, climate, Ocean dynamics, Solar physics

Solanki 2004 is a widely cited reconstruction of solar activity based on INTCAL98 (1998) 14C as a proxy. Nature abstract

I have used INTCAL09 (2009) as a basis and a trivially simple method to reconstruct the Solanki et al result, with differences and excluding the older portion of their result. (14C record deteriorates to coarse sampling about 9500BP).

All following plots are time left to right and adjusted to AD calendar.


Figure 1

Method and data used

Intcal09 is a 14C reference work presented as two 14C reference curves, one for atmospheric (vegetation) and the other for marine sediment.

An intractable problem with extracting a radiation proxy signal is large long term change which has no definitely known causal. This is usually compensated out using signal processing filtering or shape subtraction or statistical pseudo filtering.

On noticing the marine 14C response is muted I guessed there is natural gaussian filtering of the atmospheric original as material passes down to marine sediment and might be exploitable.

Assuming the authors of reference data have been fastidious on accuracy, somewhat likely, perhaps an old trick will work.

Difference the two datasets point by point. This does require rescaling of one or the other (because of attenuation by the marine process), which was done by trial and error using std deviation minimisation of the output as a guide. (goes though a minima)

Scale value 1.039


The above gave an interesting result which was then refined by lead/lag of one data against the other, atmospheric leading by about 20 years; from plot about 18 years would be closest but the result is not especially sensitive to change so it has been left simple.


Figure 2


For plotting the reconstruction was Y-axis normalised to the Solanki data and the X-axis left as-is, allowing the plot software to use different time steps: Solanki is sampled at 10 years and Intcal09 at 5 years.

Overall thumbnail of two input data and effect.


Figure 3


  • r2 = 0.63 no lag (using every other sample for reconstruct data)
  • r2 = 0.74 10 year lag

Visually the two data are self evidently the same, differing in detail.

Warning, huge linked image follows, web browser might need to scroll.


Figure 4


There is an obvious timing difference but there is likely to be argument over which is correct. A fundamental question is the delay between the ionising radiation creating 14C and incorporation into the proxy. Certainly there will be a time lag but I assume this is little more than a year for atmospheric.

There is very little time overlap between human direct solar records and proxy data. AA index, a magnetic measure is most often shown starting 1868 and sunspot number with varying degrees of reliability back no further than 1600. The Solanki record ceases 55BP which is 1950-55=1895AD

An initial look at the problem of tuning the 14C data to solar etc. data indicates there is no obvious solution and great ambiguity, hence I have left the result as-is. Discussion and correction is welcome.


I decided to check IntCal98, the data source for Solanki 2004 because I wondered why such short recent data was used. My guess was correct Solanki has withheld data, hiding the most recent data.

A bright red tail is then added to the plot and seems to be instrumental or observation data. Why is a mystery since it does not seem to add anything useful.


Figure 5


Figure 6

Here is the tail end of the Solanki data, 55BP is the last data.

  • 95 48.7 8.5
    85 55 12.7
    75 40 13.3
    65 27 9.4
    55 28.7 9.9

Abstract INTCAL98 stating “24,000-0 cal BP (Before Present, 0 cal BP = AD 1950)” here


As usual, ask if you want data.


NCDC/NOAA web page Solanki 2004

Radiocarbon IntCal09 web page


In response to Tallbloke (July 22, 2011 at 8:09 am) querying the reason for a discrepancy around year 1850 I add a comparative plot of data since 1400AD of IntCal98 and IntCal09.

Atmospheric is essentially identical. Marine combined with the differencing method is the cause of the difference.


Figure A1

  1. tallbloke says:

    Another great post, thanks Tim.

    The first thing I want to pick up on is this:

    “An intractable problem with extracting a radiation proxy signal is large long term change which has no definitely known causal.”

    One of the interesting aspects of that is that the centennial swings in the data remain at a roughly comparable size throughout the dataset, despite the large overall Y axis trend in the data.

    What does that say and what does it eliminate?

    What are the hypotheses put forward in the literature for the trend Tim?

  2. tallbloke says:

    Tim, did you check the INTCAL 98 data for the discrepancy around 1850?

  3. vukcevic says:

    Just forget about proxies, as I did since I looked into 10Be, which is suppose to be more reliable than C14.

  4. tallbloke says:

    Vuk: I think they have something valuable to tell us. Just because pet theories can’t be confirmed with them doesn’t mean they are useless.

  5. Tim Channon says:

    Tallbloke, added an addendum about 98 vs 09 to the article.

  6. vukcevic says:

    Hi tb
    Just because pet theories can’t be confirmed with them doesn’t mean they are useless.

    It’s actually opposite, it confirms my ‘pet hypothesis’, but it was not expected to do so. It confirmed that the data I used are reliable, and also that my data evaluation is sound. 10Be proxy will be used as one of the three principal supporting arguments.
    In a way I welcome the solid run of 10Be data for the period of my interest (1700-1950).10Be proxy however failed to reliable replicate the strength of the heliospheric magnetic field, what it was suppose to do.

  7. tallbloke says:

    VuK, yes, I’m with you now. I said something similar on a previous proxy thread, asking if it was a better geomagnetism proxy than solar proxy.

  8. Tim Channon says:

    I’ve decided to post in thread what could be an article.

    This is lobbing in the deeply bizarre, “impossible”.

    Authors are very serious, 23 of them, in Astroparticle physics 2010, not talking about 14C…

    “…we conclude that these results are consistent with the hypothesis that nuclear decay rates may be influenced by some form of solar radiation.”

  9. vukcevic says:

    I am sceptical about the results in the above paper. There is a few nTesla annual variation in the geomagnetic field (something to do with the tail of the magnetosphere in relation to the inclination of the heliospheric current sheet) which may not have been completely eliminated. Total elimination is not easy to achieve. There is a bit of paradox here: the weaker field is, more penetrating it is (more difficult to shield against)
    I occasionally worked in areas screened with mu-metal plates, and external field still got through. I am not certain but vaguely remember 1 mm thick panel reduces penetration about 6db which is not great deal. For effective shielding more layers are required.

  10. Tim Channon says:

    Magnetic is the obvious one and ought to be very easy to prove. Only needs a changeable field.

    I don’t believe it but dismissing completely is unwise.