Published online by Cambridge University Press: 03 October 2008
For Muguga, in Kenya, confidence limits of expected ‘twenty-day’ rainfall may be reliably derived through a logarithmic transformation of the skew rainfall data. Frequency distributions of rainfall totals received during specific crop-growing seasons are examined for three widely separated Kenya sites. It is shown that confidence limits for such seasonal rainfall cannot be obtained by summation of the corresponding limits for the successive twenty-day intervals included within the season. Twenty-day limits should not therefore be accumulated in seasonal comparisons of crop water requirements and precipitation. Such assessments of agricultural site potential must take account of the year-to-year variability of precipitation, but are best undertaken with actual rainfall amounts.