Indeed, at the level of a single observation by a single observer, there is inevitably a degree of subjectivity. However, as sunspots are simple well-defined features, the range of subjectivity remains limited. The rms dispersion between simultaneous daily observations by many separate observers is about 8%. Part of this dispersion is also due to random variations in local conditions (weather, seeing), indicating that the actual impact of human subjectivity is even lower.
The main source of subjectivity are: - the limit for the smallest sunspots: Here the subjectivity range is bounded by the existence of a sharp lower limit for the smallest sunspots, namely 2500km (~2 arcsec) corresponding to the average size of solar photospheric granules. This also limits the effect of different telescope apertures, once the aperture is larger than ~70mm. - how groups are split, when they are complex and densely packed on the solar disk: This factor of subjectivity only dominates during periods of high solar activity. However, its impact is limited by the fact that such complex situations involve only a small fraction (< 5 %) of all sunspot groups.
Now, the sunspot number is derived by a global statistics over many observers. The daily random disparity among observers is then behaving like a detector noise, equivalent to the noise in any measuring equipment. It is efficiently reduced by the statistrical averaging over many observers. On the other hand, the systematic bias of an observer (i.e. tendency to over/undercount, or over/under split) is included in his k personal scaling coefficient. This coefficient is determined on a monthly basis for each observer relative to the Locarno pilot station and is used to bring each station to the same scale as the entire network. What is important here is that each observer remains stable in his/her counting practice to keep his/her k coefficient as constant as possible over years. This global process thus largely eliminates the subjectivity of each individual observation. This is confirmed by the current comparisons with other indices obtained by modern instruments (flux measurements, automated image-based counts). They give very high linear correlations, typically above 95%.BACK