STATISTICAL FORECAST OF DROUGHTS

Abstract

The droughts are analyzed by the third asymptotic distribution of smallest values which contains a shape parameter λ, a location parameter θ and a lower limit, the minimum drought ε. If ε = 0 the two remaining parameters are estimated from the sample mean [xbar] and the standard deviation s. If the minimum drought ε is positive it is postulated that the estimate should be smaller than the smallest observed drought x ₁. An estimate for λ is obtained from the quotient ([xbar] −x ₁)/s. The estimation for θ and ε become linear functions of the mean [xbar] and the smallest drought x ₁. Since this theory gives an excellent fit to the observations it can safely be used for forecasting.