Meridional winds can be determined from measurements of the peak heigh
t of the ionospheric F2 layer (h(m)), the peak density (N-m), or the t
otal electron content (N-1). At night, ionospheric changes follow the
wind with a time constant of 30-40 min. During the day, this increases
to 50-100 min for h(m) and 3-6 h for N-m and N-1. Thus, peak-height d
ata are most suitable for the direct calculation of atmospheric winds,
and daytime results should be advanced by 1 h. The wind calculations
require an estimate of the peak height h(o) at zero wind. h(o) is clos
e to the servo result (h(s)) at night, if the servo constant c is incr
eased by 25% to agree with current theory. During the day, however, th
e long time constants prevent the F layer from reaching equilibrium be
fore sunset. h(o) is well below h(s) from sunrise until afternoon, giv
ing serious errors in any results based on servo theory. Results from
a full ionospheric modelling program are used to obtain analytic expre
ssions which reproduce the true, zero-wind peak height with an accurac
y of a few kilometres, for all times of the day, all seasons and solar
flux, and all latitudes in the useful range of 20-60 degrees geomagne
tic. The wind W required to produce a given change in h,, varies close
ly as sin(1.4)I, where lis the magnetic dip angle, and the variation o
f h(m) with W is accurately reproduced by a modified servo equation. U
se of these results with accurate peak height data should give horizon
tal winds with an accuracy of about +/- 25 m/s. Peak heights derived f
rom scaled ionospheric data (M3000F2 and foF2) have an accuracy of typ
ically 10-20 km, giving overall errors of about 40 m/s in calculated w
inds.