Analysis of synoptic data from the Vector Spectromagnetograph (VSM) of the Synoptic Optical Long-term Investigations of the Sun (SOLIS) and the NASA/NSO Spectromagnetograph (SPM) at the NSO/Kitt Peak Vacuum Telescope facility shows that the reversals of solar polar magnetic fields exhibit ele-ments of a stochastic process, which may include the development of specific patterns of emerging magnetic flux, and the asymmetry in activity between northern and southern hemispheres. The presence of such irregu-larities makes the modeling and prediction of polar field reversals extremely hard if possible. In a classical model of solar activity cycle, the unipolar magnetic regions (UMRs) of predominantly following polarity fields are transported polewards due to meridional flows and dif-fusion. The UMRs gradually cancel out the polar mag-netic field of the previous cycle, and rebuild the polar field of opposite polarity setting the stage for the next cycle. We show, however, that this deterministic picture can be easily altered by the developing of a strong cen-ter of activity, or by the emergence of an extremely large active region, or by a ‘strategically placed’ coronal hole. We demonstrate that the activity occurring during the current cycle 24 may be the result of this random-ness in the evolution of the solar surface magnetic field.

Context. We apply our recently developed method to reconstruct synoptic maps of the photospheric magnetic field from observations of chromospheric plages and the magnetic polarity of sunspots. Here, we apply the method to an extended time interval from 1915 to 1985. Aims. Systematic magnetographic observations of the solar photospheric magnetic field were initiated as recently as the 1970s and the lack of earlier observations limits our ability to study and understand the long-term evolution of the Solar global field. This study is aimed at creating synoptic maps of magnetic fields for the pre-magnetograph era and using these maps as input for modern simulation models to investigate the long-term (centennial) evolution of the Sun's global magnetic fields. Methods. We reconstructed active Solar regions by identifying chromospheric plages from Ca II K line synoptic maps and assigning magnetic polarities based on the observed polarity of sunspots. We used a surface flux transport (SFT) model to simulate the evolution of the photospheric magnetic field from the reconstructed active regions. We used the potential field source surface (PFSS) model to determine the amount of open magnetic flux from the reconstruction and from magnetographic observations. We also reconstructed the coronal field during two eclipses and compared the result with eclipse drawings. Results. We successfully reconstructed the photospheric magnetic field from 1915 to 1985. The number and total magnetic flux of the reconstructed active regions shows a realistic cyclic behavior that mostly follows the evolution of the sunspot number, even on relatively short timescales. The polar field strengths of cycles 19 and 20 do not reflect the evolution of the sunspot number very accurately, which may be related to problems related to the calcium data during cycle 19 and the long data gap during cycle 20. The polarity of polar fields and the amount of open field both at high and low latitudes all demonstrate the expected cyclic behaviour. The agreement of the modeled coronal structure with eclipse drawings in 1922 and 1923 is fair.