Within the Flood Management and Mitigation Program (FMMP) of the Mekong River Commission (MRC), flood forecasting is one of the core activities. Up to now, the forecasts are being made with a Streamflow Synthesis and River Routing (SSARR) model and with regression models. At this moment new model are being developed, based on a combination of URBS and Deltares' Flood Early Warning System (FEWS)
... [Show full abstract] with, amongst others, the objective of producing more reliable and more accurate output. In this paper the flood forecast accuracy and reliability of the current model is investigated and key parameters of the new system are examined. Herein, accuracy is defined as the deviation of a forecast from the observed value and reliability is defined as the number of forecasts that were actually disseminated compared to the number that should have been disseminated. The analysis of the 2006 and 2007 forecasts shows clearly that the longer the lead time, the lower the accuracy and the more upstream the forecast location, the lower the accuracy. The forecast reliability analysis shows an average failure percentage of about 20%. For the accuracy of the new models, the following key two parameters can be distinguished: Satellite Rainfall Estimates (SRE) inputs and ratings. The SRE are compared to ground data and it appears that the predicted depth is comparable, although there appears to be significant spatial/temporal variation. The ratings show uncertainty of up to several hundreds of millimetres. It can be concluded that there will always remain uncertainty in the forecast, mainly because of input uncertainty, but that this uncertainty is probably less with the new model system. The seriousness of the uncertainties in the input data is also related to the way the basin is modelled: as the objective at the MRC Regional Flood Management and Mitigation Centre (MRC-RFMMC) is producing forecasts for the Mekong mainstream, the error in the SRE appears to be acceptable as the preliminary results of the new Flood Forecasting System (FFS) show promising results. The forecast reliability is expected to increase because the new system is less dependent on missing data points.
