ABSTRACT

ABSTRACT: Early warning is needed to help mitigate socio-economical and environmental impacts of droughts. Seasonal streamflow forecasts have been dominated by statistical methods in the past. Recently, dynamic physically based seasonal forecasts from global climate models have become available operationally and can be used to drive detailed hydrological models. Our forecast scheme for the Limpopo combines statistical methods for longer lead times with a distributed hydrological model forced with a seasonal metrological forecast for shorter lead times (<6 months). The statistical model is set up and tailor-made for prediction at stations of interest, it is straightforward and has little infrastructure requirements. The second approach provides a great array of hydrological information, offering flexibility to predict different indicators. It achieved higher and more stable skill scores than the statistical forecast. A combined system is feasible and supplements drought early warning systems.