A new reconstruction of sunspot activity variations from historical sunspot records using algorithms from machine learning – Are you up to that?

I received this email from Willie Soon today,

Dear friends and colleagues,

I am proud to attach this new paper:

“Sunspot Number Group: A New Reconstruction of Sunspot Activity
Variations from Historical Sunspot Recordings Using Algorithms from Machine Learning”

just appeared online at Solar Physics,

https://link.springer.com/article/10.1007/s11207-021-01926-x

Indeed, we think this paper is important in many different respects, including
even the rather obvious and revisionist efforts of some activists in recent times
10 years or more trying to modify the Group Sunspot Number (GSN) record with
The arguments and evidence are rather flawed as documented in this detailed paper.

If we are wrong, let the debate begin openly and objectively in the public and in science
globular. The rather ugly approach of the revisionists may not be obvious, but throughout
For the past 10 years, they’ve been systematically ignored and censored any constructive criticism
and the suggestions of our co-author Douglas Hoyt, a serious scholar of
reconstruction of sunspot activity records.

For some of us who call America home, the interesting aspect of this article is
to show the possibility of recovering the first sunspot drawings long lost
from Humphry Marshall’s Colonial America (1722-1801).

Sincerely,

Willie with colleagues Victor Velasco Herrera, Doug Hoyt and Judit Murakozy

ps: for some you may be interested for more details and discussion
Regarding the origin of this article, please consider these two talks

(1) Studying the role of the Sun in climate

(2) Study of sunspot activity cycles: Obstruction forecasting and forecasting

Here is the summary of the article.

abstract

Historical sunspot records and the building of comprehensive databases are among the most sought-after research activities in solar physics. Here, we revisit the problems and remaining questions about reconstructing the so-called group sunspot number (GSN) pioneered by D. Hoyt and colleagues. We use modern tools of artificial intelligence (AI) by applying different algorithms based on machine learning (ML) to the GSN records. The aim is to offer a new vision in reconstructing sunspot activity variants, i.e. Bayesian reconstructions, to obtain a complete probabilistic GSN record from 1610 to 2020 This new GSN reconstruction matches the historical GSN records. In addition, we make a comparison between our new probabilistic GSN record and the most recent GSN reconstruction generated by several solar researchers under various assumptions and constraints. Our AI algorithms were able to reveal various new underlying patterns and variant channels that could fully account for the complete GSN time variation, including periods with trace activity. extremely low or weak sunspots such as the Maunder Minimum from 1645 – 1715. Our results show that the GSN records are not exactly represented by the 11-year cycle alone, but other important periods for reproduction. To create a more complete history of GSN’s operation, it ranges from 5.5 years, 22 years, 30 years, 60 years and 120 years. Comprehensive GSN reconstruction using AI/ML may shed new insights into the nature and characteristics of not only the elementary 11-year sunspot cycles, but also the 22-year Hale polar cycles. in Maunder Minimum, among other results previously hidden so far. In the early 1850s, Wolf multiplied his original sunspot reconstructions by a factor of 1.25 to obtain Wolf’s standard sunspot number (WSN). Removing this multiplier, we find that the GSN and WSN differ by only a few percent between 1700 and 1879. When compared with the roughly international sunspot number (ISN) proposed by Clette et al. this. (Space Science. Rev. 186, 35, 2014), some differences are found and discussed. More sunspot observations are still needed. Our paper shows observers that have not been included in the GSN database.