An ionospheric F2 critical frequency database has been assembled to determine the variability of the F region as a function of local time, latitude, season, and geomagnetic activity. The database comprises observations from 75 ionosonde stations covering a range of geomagnetic latitude and includes 43 storm intervals. The database was previously used to develop the Storm-Time Empirical Ionospheric Correction Model (STORM). The mean and standard deviation have been evaluated by sorting the data by local time, season (five intervals centered on equinox, solstice, and intermediate intervals), latitude (four regions each 20¿ wide in geomagnetic latitude), and up to eight levels of geomagnetic activity. The geomagnetic activity index was based on a weighted integral of the previous 33 hours of ap and is the same as that used by the STORM model. The database covers a full solar cycle, but insufficient information was available to sort by solar activity without compromising the estimates of variability on the other sorting parameters. About half the data were contained in the first level of geomagnetic activity, between 0 and 500 units of filtered ap corresponding to Kp ≤ 2, and half above that level. When local time dependence was included in the binning, sufficient data were available to sort into two levels of geomagnetic activity, quiet (Kp ≤ 2+) and disturbed (Kp > 3-). For all latitudes and levels of geomagnetic activity, the lowest variability was typically found in summer (10--15%), and the largest variability occurred in winter (15--40%), with equinox (10--30%) lying between the solstice extremes. The exception was low latitudes at equinox, which had surprising low variability (10%), possibly because of the weak interhemispheric flow at this time of year. At middle and low latitudes, the variability tended to increase with geomagnetic activity in winter and equinox but remained fairly constant in summer. At high latitudes, the surprising result was that in all seasons, and in winter in particular, the variability tended to decrease, probably because of the increased upwelling of neutral molecular species and stronger chemical control of the ionosphere. The data have also been used to build a table of estimated variability suitable for inclusion in the International Reference Ionosphere or any other climatological model. For periods where data were scarce or nonexistent, an estimated variability was provided on the basis of expectations of the consequences of physical processes. This was necessary to fill in the table of values in order to develop a module suitable for inclusion in the International Reference Ionosphere.