Examination of individual cases in real-time or retrospectively often indicates skill in detecting the occurrence of a flood event and a reasonable evolution of water depth (at the scale of the calculation) and downstream movement of high water levels. A recently published study evaluating calculated flood occurrence from the GFMS against a global flood event database is reviewed. The statistics indicate that flood detection results improve with longer duration (> 3 days) floods and that the statistics are impacted by the presence of large dams, which are not accounted for in the model calculations. Overall, for longer floods in basins without large dams, the Probability of Detection (POD) of floods is ~ 0.7, while the False Alarm Rate (FAR) is ~ 0.6. Limitations in the flood calculations that are related to the satellite rainfall estimates include space and time resolution limitations and underestimation of shallow orographic and monsoon system rainfall. Recent cases from 2012, mainly over Asia, are discussed as examples of the utility of the output information and the importance of accurate rainfall input.
These calculations in their current form can provide information useful to national and international agencies in understanding the location, intensity, timeline and impact on populations of these significant hazard events. In addition to the website already mentioned, the present output information is also being made available via the Pacific Disaster Center (PDC) products, which are tailored for the global disaster management community. However, there is still much room for improvement of this type of system through improving the satellite rainfall estimates and using more sophisticated hydrology models and routing calculations. These coming improvements to the GFMS will be discussed in a companion paper at this meeting.