Floods occur in monsoon delta regions regularly during the rainy season. However, occasionally they occur with an intensity and duration that is devastating to the society and economy in delta countries such as Bangladesh. There are multiple reasons why floods occur in Bangladesh ranging from excessive rainfall over the large catchment areas of the Ganges, Brahmaputra and the Meghna, to excessive high sea-level heights associated with the negative phase of the Indian Ocean Zonal Mode (dipole).

We show that long-term predictability exists for flooding even in an area as small as Bangladesh. This is because a delta region is an integrator of precipitation predictability that exists over the much larger major river catchment areas, and the sea-level variation has low frequency variability on interannual time scales. A scheme for long term forecasting of Bangladesh floods on the time scale of months is presented using a combination of models, including the ECMWF seasonal forecast model. It is also shown that it is possible to enhance the short term forecasts of Bangladesh floods (currently 1 day) to multiple days using ECMWF precipitation forecasts in conjunction with data representing the state of the ocean at the head of the Bay of Bengal. Both the long-term and short-term flood forecasting schemes are designed