Use of an Ensemble Re-ordering Method for disaggregation of seasonal categorical rainfall forecasts into conditioned ensembles of daily rainfall for hydrological forecasting

Seasonal climate forecasts issued in discrete, tercile (i.e. above normal, near normal, and below normal) format, are not directly applicable in daily step simulation models for generating hydrological and crop yield forecasts. Hence, a model based on the Ensemble Re-ordering Method is developed to disaggregate seasonal rainfall forecasts into ensembles of conditioned daily rainfall in order to simulate the within-season rainfall characteristics for use in hydrological forecasting. The model was applied on the Mgeni catchment in KwaZulu-Natal, South Africa to test whether the forecasts that it generates are satisfactory for operational use in water resources and agricultural management strategies. Results from these tests illustrate the ability of the Ensemble Re-ordering Method to reproduce the transitional probabilities of rain days and dry days as well as the persistence of dry and wet spells which have statistical characteristics similar to those of actual rainfall data for most of the selected rainy months. The generated ensembles of rainfall sequences were then used as input to the daily time step, ACRU hydrological model in order to forecast monthly and seasonal streamflows. For the seasonal streamflow forecasts, the observed mean, standard deviation, skewness coefficient and coefficient of variation were simulated acceptably well in most of the seasons selected for testing. Monthly streamflow forecasts for a selected year, viz. 2004, also illustrate that the model captures the observed statistics well for most of the months of the year, except for the dry months from June to September. The Ensemble Re-ordering Method can be used with confidence to translate the skilful categorical rainfall forecasts into daily quantitative values for use as input into hydrological models to produce seasonal forecasts of hydrological and agricultural related variables.