Tests of sunspot number sequences: 4. Discontinuities around 1946 in various sunspot number and sunspot group number reconstructions

TL;DR

This study investigates discontinuities around 1946 in various sunspot number sequences using multiple test data series and statistical methods, revealing the need for non-linear corrections to ensure data consistency across different reconstructions.

Contribution

It provides a detailed analysis of discontinuities in sunspot number records around 1946 and proposes non-linear correction factors for different sunspot number series to improve their consistency.

Findings

Pre-1946 sunspot numbers are underestimated in some series and overestimated in others.

Non-linear corrections are necessary for certain sunspot number reconstructions.

The study offers optimal correction factors to harmonize pre- and post-1946 sunspot data.

Abstract

We use 5 test data series to quantify putative discontinuities around 1946 in 5 annual-mean sunspot number or group number sequences. The series tested are: the original and new versions of the Wolf/Zurich/International sunspot number composite [$R_{ISNv1}$ and $R_{ISNv2}$] ; the corrected version of $R_{ISNv1}$ [$R_C$]; the backbone group number [$R_{BB}$]; and the group number composite [$R_{UEA}$]. The test data are: the group number $N_G$ and total sunspot area $A_G$ from the RGO photoheliographic data; the CaK index from re-analysis of MWO CaII K spectroheliograms; the group number from the MWO sunspot drawings, $N_{MWO}$; and ionospheric critical frequencies at Slough [$foF2$]. The test data all vary with sunspot numbers, in some cases non-linearly. Tests use both before-and-after fit-residual comparison and correlation methods, applied to intervals iterated to minimise errors and eliminate the effect of the discontinuity date. It is not assumed that the correction required is by a constant factor, nor even linear in sunspot number. A non-linear correction is required by $R_C$, $R_{BB}$ and $R_{ISNv1}$, but not by $R_{ISNv2}$ or $R_{UEA}$. The test datasets give very similar results in all cases. By multiplying the probability distribution functions together we obtain the optimum correction for each data series that must be applied to pre-discontinuity data to make them consistent with the post-discontinuity data. It is shown that, on average, values for 1932-1943 are too small (relative to later values) by about 12.3% for $R_{ISNv1}$ but are too large for $R_{ISNv2}$ and $R_{BB}$ by 3.8% and 5.2%. The correction applied to generate $R_C$ from $R_{ISNv1}$ reduces this average factor to 0.5% but does not remove the non-linear variation, and other errors remain uncorrected. A test is provided by $R_{UEA}$, which is identical to the RGO $N_G$ values over the interval used.