On the level of skill in predicting maximum sunspot number: A comparative study of single variate and bivariate precursor techniques

Abstract

Precursor prediction techniques have generally performed well in predicting the maximum amplitude of sunspot cycles, based on cycles 10–21. Single variate methods based on minimum sunspot amplitude have reliably predicted the size of the sunspot cycle 9 out of 12 times, where a reliable prediction is defined as one having an observed maximum amplitude within the prediction interval (determined from the average error). On the other hand, single variate methods based on the size of the geomagnetic minimum have reliably predicted the size of the sunspot cycle 8 of 10 times (geomagnetic data are only available since about cycle 12). Bivariate prediction methods have, thus far, performed flawlessly, giving reliable predictions 10 out of 10 times (bivariate methods are based on sunspot and geomagnetic data). For cycle 22, single variate methods (based on geomagnetic data) suggest a maximum amplitude of about 170 ± 25, while bivariate methods suggest a maximum amplitude of about 140 ± 15; thus, both techniques suggest that cycle 22 will be of smaller maximum amplitude than that observed during cycle 19, and possibly even smaller than that observed for cycle 21. Compared to the mean cycle, cycle 22 is presently behaving as if it is a + 2.6σ cycle (maximum amplitude about 225). It appears then that either cycle 22 will be the first cycle not to be reliably predicted by the combined precursor techniques (i.e., cycle 22 is an anomaly, a statistical outlier) or the deviation of cycle 22 relative to the mean cycle will substantially decrease over the next 18 months. Because cycle 22 is a large amplitude cycle, maximum smoothed sunspot number is expected to occur early in 1990 (between December 1989 and May 1990).