Existing stochastic models describing the probability distributions of the magnitude, duration, frequency, and timing of high flows resulting from a single flood-generating process are expanded to accommodate the more generalized case of mixed-generating processes. The proposed method is simple to apply and retains a theoretical and physical basis. The ability to predict high flows at any higher truncation level of interest without resorting to reanalysis of the historical data is presented. The methodology is applied to three basins in southwestern British Columbia that experience both snowmelt and rainfall generated floods. Good agreement between the data derived for the partial duration series of the study basins and the predicted distributions illustrates the general applicability of the technique.