The annual variation of geomagnetic activity is studied through the geomagn
etic indices aa, Dst and AE, according to different levels of intensity for
each of the indices. For thresholds that correspond from moderate to fairl
y intense storms (i.e., peak aa greater than or equal to 90), the distribut
ion follows the well-known pattern of a seasonal variation, with maxima aro
und the equinoxes and minima near the solstices. Deviations from this behav
ior are observed when the distribution refers to levels associated with the
occurrence of more intense storms. In particular, the annual distribution
of days with a geomagnetic index an greater than about 90, shows the occurr
ence of a peak in July. The contribution of very intense storms (aa greater
than or equal to 210) to the July peak, seems to be evenly distributed alo
ng the 12 solar cycles covered by this index. Furthermore, the indices Dst
and AE, although restricted to a much shorter interval of time (they have b
een recorded only since 1957), seem as well to show the existence of a peak
of occurrence for July. The study done for the indices Dst and AE gives so
me indication for the existence of another peak in November, also for thres
holds associated with intense storms. However, due to the lack of longer re
cords for these indices, the real existence of this peak in the geomagnetic
activity is questionable. A statistical analysis of the distribution of ev
ents according to the levels of intensity of the aa and Dst is also present
ed. From this analysis it is seen that the number of occurrences of storms
above a given level of intensity of those geomagnetic indices, can be appro
ximated by an exponential law. Furthermore, an estimation of the occurrence
of storms during a solar cycle as a function of the peak aa (or aa*) has b
een also done. (C) 2001 Elsevier Science Ltd. All rights reserved.