Bayesian Methods for Reconstructing Sunspot Numbers Before and During the Maunder Minimum

The Maunder Minimum (MM) was an extended period of reduced solar activity in terms of yearly sunspot numbers (SSN) during 1610?–?1715. The reality of this “grand minimum” is generally accepted in the scientific community, but the statistics of the SSN record suggest a need for data reconstruction. The MM data show a nonstandard distribution compared with the entire SSN signal (1610?–?2014). The pattern does not satisfy the weakly stationary solar dynamo approximation, which characterizes many natural events spanning centuries or even millennia, including the Sun and the stars. Over the entire observation period (1610?–?2014), the reported SSN exhibits statistically significant regime switches, departures from autoregressive stationarity, and growing trends. Reconstruction of the SSN during the pre-MM and MM periods is performed using five novel statistical procedures in support of signal analysis. A Bayesian–Monte Carlo backcast technique is found to be most reliable and produces an SSN signal that meets the weak-stationarity requirement. The computed MM signal for this reconstruction does not show a “grand” minimum or even a “semi-grand” minimum.