Issues in flood forecasting: ungauged basins, extreme floods and uncertainty

Flood forecasts for any location, at times of extreme storms and with uncertainty estimates, embrace three issues that present important scientific challenges and the prospect of improved flood warning. Forecasting for ungauged basins using empirical regionalization of parameters of simplified models has met with mixed success and given little advance to our understanding of hydrological science. Prototype conceptual-physical area-wide model formulations, supported by terrain and soil property data, are trialled here and shown to have advantages for complex responding lowland catchments. Evidence is presented—using case study historical extreme storms and amplified forms of them—that demonstrates the potential value of distributed rainfall–runoff models for flood warning and for identifying flood-prone locations, especially for unusual or extreme storms and for locations that are ungauged. The challenges of model initialization, forecast updating and uncertainty estimation are discussed in relation to these area-wide models, future advances in ensemble rainfall forecasting and the benefits of risk-based decision-support for flood warning.