Droughts are usually classified as meteorological, agricultural or hydrological, depending on the variables under investigation. The investigations vary from the purely descriptive through severity classifications to modelling and simulation. Drought predictions in real time do not seem to be useful as yet. Drought probabilities have value in the selection of strategies which try to minimize the negative effects of droughts. More work is needed, especially, on hydrological drought. The methods of analysis should become, in general, more sophisticated. Available techniques to optimize drought strategies should find wider practical applications. A national water plan for South Africa is needed.