Extended-Range Probabilistic Forecasts of Ganges and Brahmaputra Floods in Bangladesh

Abstract

The authors have developed a new extended-range flood forecasting system for large river basins that uses satellite data and statistically rendered probabilistic weather and climate predictions to initialize basin-scale hydrological models. The forecasting system overcomes the absence of upstreamflow data, a problem that is prevalent in the developing world. Forecasts of the Ganges and Brahmaputra discharge into Bangladesh were made in real time on 1?10-day time horizons for the period 2003?08. Serious flooding of the Brahmaputra occurred in 2004, 2007, and 2008. Detailed forecasts of the flood onset and withdrawal were made 10 days in advance for each of the flooding events with correlations at 10 days ≥0.8 and Brier scores <0.05. Extensions to 15 days show useable skill. Based on the 1?10-day forecasts of the 2007 and 2008 floods, emergency managers in Bangladesh were able to act preemptively, arrange the evacuation of populations in peril along the Brahmaputra, and minimize financial loss. The particular application of this forecast scheme in Bangladesh represents a ?world is flat? approach to emergency management through the collaboration of scientists in Europe (generating global ensemble meteorological and climate forecasts), the United States (developing and producing the integrated flood forecasts), and the developing world (integrating the flood forecasts into their disaster management decision-making protocol), all enabled by high-speed Internet connections. We also make suggestions of how scientific and technical collaborations between more developed and developing nations can be improved to increase their prospects for sustaining the technology adoption and transfer.