Monthly and annual totals for the Penman estimate of potential evaporation have been reliably determined, for periods varying from 7 to 11 years, at 14 East African observatories. At each site, these annual and monthly totals are normally distributed, with coefficients of variation of the order of 5 and 10 per cent, respectively. Meteorological stations established for the sole purpose of supporting evaporation surveys should therefore be relocated at 5- or 6-year intervals so as to derive maximum benefit from the valuable instruments. The results of these variability analyses were applied in the evaluation of confidence limits for monthly and annual potential evaporation rates, and reference is made to recent compilations of such limits for 137 sites in Kenya and Tanzania.