Improving our understanding of flood forecasting using earlier hydro-meteorological intelligence

In recent decades, Taiwan has suffered from severe bouts of torrential rain, and typhoon induced floods have become the major natural threat to Taiwan. In order to warn the public of potential risks, authorities are considering establishing an early warning system derived from an integrated hydro-meteorological estimation process. This study aims at the development and accuracy of such a warning system. So it is first necessary to understand the distinctive features of flood forecasting in integrated rainfall-runoff simulations. Additionally the adequacies of a warning system that is based on extracting useful intelligence from earlier, possibly faulty numerical simulation results are discussed. In order to precisely model flooding, hydrological simulations based upon spot measured rainfall data have been utilized in prior studies to calibrate model parameters. Here, precipitation inputs from an ensemble of almost 20 different realizations of rainfall fields have been used to derive flood forecasts. The flood warning system therefore integrates rainfall-runoff calculations, field observations and data assimilations. Simulation results indicate that the ensemble precipitation estimates generated by a Weather Research Forecasting (WRF) mesoscale model produce divergent estimates. Considerable flooding is often shown in the simulated hydrographs, but the results as to the peak time and peak stage are not always in agreement with the observations. In brief, such forecasts can be good for warning against potential damaging floods in the near future, but the meteorological inputs are not good enough to forecast the time and magnitude of the peaks. The key for such warning system is not to expect highly accurate rainfall predictions, but to improve our understanding from individual ensemble flood forecasts.